Sustainable Energy Technologies and Assessments 73 (2025) 104097

Contents lists available at ScienceDirect

Sustainable Energy Technologies and Assessments

journal homepage: www.elsevier.com/locate/seta

Review article

Artificial intelligence and machine learning models application in biodiesel

optimization process and fuel properties prediction
Muhammad Arif a,f , Adel I. Alalawy b, Yuanzhang Zheng c, Mostafa Koutb d, Tareq Kareri e,
El-Sayed Salama f,* , Xiangkai Li a,*
a
MOE, Key Laboratory of Cell Activities and Stress Adaptations, School of Life Sciences, Lanzhou University, Lanzhou 730000, Gansu Province, PR China
b
Department of Biochemistry, Faculty of Science, University of Tabuk, Tabuk 71491, Saudi Arabia
c
Discovery Biology, CuriaGlobal Inc., NY, USA
d
Department of Biology, Faculty of Science, Umm Al-Qura University, 715, Makkah, Saudi Arabia
e
Department of Electrical Engineering, College of Engineering, Najran University, Najran 11001, Saudi Arabia
f
Department of Occupational and Environmental Health, School of Public Health, Lanzhou University, Lanzhou 730000, Gansu Province, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: Inefficient transesterification, low-quality fuel properties, and high resource consumption are the bottlenecks
Biodiesel associated with conventional biodiesel production. The current research trends include the application of arti­
Artificial intelligence ficial intelligence (AI) and machine learning (ML) to optimize the biodiesel process for improved yield and fuel
Machine learning models
quality. Previous reviews discussed the applications of ML in the optimization of transesterification parameters
Optimization
and fuel properties. However, there is a lack of deep discussion on feedstock selection, optimization, process
Predictions
monitoring, and cost analysis. The challenges during biodiesel production, ML model selection, and assessment
of plant and animal lipid potential for biodiesel under different conditions using AI tools are reviewed. All the
parameters that affect biodiesel yield and fuel properties through ML, the efficiency of different models, and
pilot-scale techno-economic analyses are also discussed. Biodiesel production from animal and plant lipids
showed high yield potential ranging from 78-99 %. ML models demonstrated higher efficacy in trans­
esterification optimization to attain > 90 % yield. Various AI models exhibit a predictive efficiency range (R2 =
0.85 to 0.99) for yield and fuel qualities. Economic analyses reveal that the choice of feedstock and catalyst
significantly impacts final production costs. ML and AI approaches exhibit the potential for improving the bio­
diesel process.

Introduction and uncertainty of biodiesel process parameters through imputation and

correlation [6]. The mathematical and computational models offer an
Advanced machine learning (ML) technology has solved the complex exploration opportunity to increase biodiesel yield and fuel quality
challenges (such as reactor configuration, stability, and operation) of improvement [7]. Tremendous research has been carried out on bio­
biodiesel production [1,2]. The ML models analyze multivariate data diesel sustainability (including yield, fuel properties, economic viability,
(such as reaction temperature and time, catalyst types, and concentra­ engine efficiency, and emission). Integrating ML in the biodiesel process
tions) for the biodiesel process to optimize and predict the yield during can significantly improve the reactor configuration and optimize oper­
transesterification [3]. Several mathematical models (including the ational parameters (catalyst, temperature, time, and solvent oil ratio).
Takagi-Sugeno model and simple regression) are reported in the litera­ This technology enables monitoring and control of the process through
ture on biofuel generation. However, few are available associated with data sensing and predicts feedstock yield and composition to address
ML and artificial intelligence (AI) for biodiesel quality estimation and energy and environmental challenges [8].
prediction yield [4,5]. The research trends in the biodiesel process pri­ The artificial neural network (ANN) is a mainly employed model in
marily focus on AI technology for large-scale data computation and the production process of biodiesel. The model showed high accuracy
visualization. It would offer a solution for data complexity, nonlinearity, and prediction efficiency even with data having challenges (such as

* Corresponding authors.
E-mail addresses: [email protected], [email protected] (E.-S. Salama), [email protected] (X. Li).

https://doi.org/10.1016/j.seta.2024.104097
Received 2 September 2024; Received in revised form 14 November 2024; Accepted 15 November 2024
Available online 29 November 2024
2213-1388/© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

uncertainty, incompleteness, noise, and missing values). ANN and its performances could be used in different applications (such as feedstock
integrative models could potentially process the non-linear and multi­ selection, biodiesel conversion, and process monitoring) [17]. Only a
variate data using a multilayer perceptron architecture [9,10]. The ANN few studies have explored the integration of ML models for sustainable
comprises three layers which include an input layer for variables, one or biodiesel production. Various ML models showed certain applications
more hidden layers for processing, and output layers for results. It em­ and limitations, and specific models are applied based on the data
ploys the backpropagation learning approaches to optimize the weight available (Table 1).
and reduce the prediction error [11]. The models are classified as uni­ ML used data from several stages of biodiesel processing to highlight
directional or bidirectional propagation based on the design, and they the hidden features such as the influence of fatty acid (FAs) composition,
may be supervised or unsupervised learning models based on their free fatty acids (FFAs) concentration, and moisture content on yield.
learning mode. The ANN optimization involves data collection and These features also show the correlation among variables such as pal­
preprocessing, training and testing, selecting topology, and applying mitic acid, stearic acid, and linolenic acid with cold flow plugging
ANN models for validation and simulation [12]. To reduce the relative properties (CFPP), cetane number (CN), and oxidation stability (OS)
error between actual and predictive values, different tuning procedures [18]. The optimization of transesterification parameters and FAs com­
(such as training and adjusting synaptic weights passed through syn­ positions can significantly improve both yield and fuel performance and
apses) are used in ANN models. After model optimization, the perfor­ facilitate decision-making based on biodiesel data. The ML model
mances are assessed through various statistical matrices such as mean evaluates different qualities of feedstocks (such as oil composition,
absolute error (MAE), correlation coefficient (R2), mean square error density, and viscosity) and predicts their potential for biodiesel gener­
(MSE), and root mean square error (RMSE) [13]. The trained model is ation, optimizing the feedstock ratio. Models (such as SVR, RF, eXtreme
then used to monitore different transesterification factors (such as gradient boosting (XGB), and ANN) predict the optimal catalyst con­
temperature, catalyst, and solvent molar ratio) and for the fatty acid centrations and type, time, agitation, solvent oil ratio, and pressure that
methyl esters (FAMEs) yield [14,15]. enhance transesterification and decrease undesirable byproducts (soap
Other models such as particle swarm optimization (PSO), bee algo­ formation) [19]. The ML algorithms can monitor important process
rithms, response surface methodology (RSM), genetic algorithms (GA), parameters (temperature, water level, and pH) in real time. They predict
and ant colony optimization, are primarily used during various steps of faults/deviations and reaction completion and provide information
biodiesel production [16]. The AI models (such as adaptive neuro-fuzzy about process stability and performance. Biodiesel properties (such as
inference system, ANFIS, or ANN combined with neuro-fuzzy, GA, and viscosity, density, CFPP, and CN) can be improved by using ML models
RSM, and partial least squares, PLS) showed high prediction based on FAs compositions [20].

Table 1
Reported machine learning models that are often used during biodiesel production.
Model/ Type Description Advantages Disadvantages References
algorithm

Linear − Support the relationship between the dependent and − Easy to understand − A poor correlation between [161]
regression independent variables − Identify the important variables is established
variables of the model − Overfitting tendency is
high

Logistic ¡Classify the problem as having outcomes (discrete) − Low variance − Fail to classify multi-class
regression –No need for training data
− Easily apply − Not suited for correlated
− Gives probability of attributes
output − Tendency for model to
overfit

Decision tree − A classifier model of data that shapes and organizes the − Easy interpretation − Failed in generalization of [162]
data like a tree − Applied on less data data well
− Suit for both A complex tree is formed if
categorical and max depth is not defined
numerical data

Random forest − Takes the average of many decision trees. Each tree is − A highly valid model is − Slow to output predictions [163]
weaker than the whole decision tree, but combining them generated. relative to other algorithms
improves overall performance − Train rapidly − Difficulties in predictions
− More valid for small − Highly biased for
datasets correlated
− Overfitting is less. prediction

Gradient − Result in high predictive accuracy using multi-model –No need for data pre- − Outlier sensitive [164]
boosting predictions processing − Training needs more time.
− Missing data handling. − Cannot extrapolate
− Support different loss − Less valid for categorical.
functions − Harder to optimize
− Predict fast

Neural − NNs are interconnected Neurons that show the − Self-learner − Have so many layers [165]
networks relationship using different algorithms to mimic human − Interpret complex data. − Black box solutions
(NNs) neurons − Easy to use − Hard to refine
− Take a long time to train

2
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

Plant oils, animal fats, and microorganisms (bacteria, yeast, and wastes from leather tanneries), and algae [23]. The choice of feedstocks
algae) lipids are highly viscous, reducing the engine’s performance if influences the cost, yield, and quality of biodiesel. Over 350 oil-
directly used. These lipids could be transesterified to generate diesel-like producing plant crops are identified as potential sources for biodiesel
fuel. The non-edible plant oil and animal fats have several advantages production worldwide [24]. The edible feedstocks contribute > 80 % of
for biodiesel production over others (such as lower cost, no food the total biodiesel. However, these feedstocks have food security and
competition, oxidation stability (OS), and higher CN value) [21]. high-cost problems [25]. Environmental factors (such as weather and
However, animal fats (tallow and lard) have several challenges with water variation, crop diseases, and soil fertility) affect feedstock and oil
high sulfur content, acid values, water content, and cold flow properties yield [26]. The transition from edible feedstocks to non-edible (such as
[22]. Plant-derived oil has a low yield due to its complex cell wall jatropha, castor, mahua, and waste cooking oil) showed several chal­
structure that needs pretreatment, a higher portion of polar lipids, and lenges. These challenges include toxic compounds, heavy metals, plastic
toxic compounds (erucic acid and phorbol ester). Two approaches, particles, and sulfur, which need pretreatment before conversion,
feedstock blending (consistent properties) and ML model prediction increasing biodiesel costs [27]. Feedstocks with > 0.25 % water and > 3
(conditions optimization with improved yield) is used to address the % FFAs content decreased the yield and catalyst efficiency and increased
challenges of plant and animal lipids (Fig. 1). product separation cost [19].
The motivation of this review is to compile and summarize recent Feedstock selection for biodiesel generation is dependent on regions,
research findings on the applications and predictive efficiency of ML accessibility, and economic status of the country [24]. Worldwide
models (such as ANN, ANFIS, RSM, and RF). It focuses on trans­ countries such as Canada (Canola), Brazil and USA (soybean), Indonesia,
esterification optimization to improve biodiesel yield using various Malaysia (coconut and palm oil), Germany, Italy, Finland (Rapeseed),
feedstock (plant oil and animal waste fat). The challenges in feedstock and India (Karanja and jatropha) used feedstocks for biodiesel [28].
selection (such as FAs compositions, water content, and FFAs concen­ Biodiesel feedstock contains various FAs from C14-C24, predominantly
tration) through ML models have been addressed. It also evaluates the C16-C18. The variation of FAs is dependent on various factors (such as
performances of various AI models for fuel property prediction based on genetic and geographical origin, as well as harvesting seasonal differ­
FAs composition. This review discusses the challenges associated with ences). The physicochemical properties of feedstocks (such as moisture
ML models such as data availability, model complexity, and general­ content and FAs compositions) result in iodine value (IV) and kinematic
ization issues. The techno-economic analysis of the pilot-scale biodiesel viscosity (KV) variation [29]. The variation in saturated, unsaturated,
production system is summarized for acid, base, and enzyme catalytic and chain length is mainly due to compositional changes in feedstocks.
conversion of various feedstocks. The high content of unsaturated fatty acid results in high IV exhibiting
corrosive and oxidation properties [30]. Oils that have high content
Biodiesel production process FFAs and moisture typically take a long time to process (4–12 h), cata­
lyst concentrations (4–10 %), and high reaction temperature
Feedstock selection (100–160 ◦ C) [31]. Changes in lipid content influence trans­
esterification. High concentration of FFAs content in feedstocks (such as
Biodiesel is derived from diverse feedstock including plant oil (edible WCO or animal fats) reduces biodiesel yield and increases product
and non-edible), animal fats (tallow, chicken fat, dairy scum, and meat separation cost during alkaline transesterification. Research shows that

Fig. 1. Major challenges in the conversion of lipid-rich feedstocks produced from animals and plants to biodiesel.

3
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

biodiesel yield decreases from 90 % at 0.3 % FFA concentration to below The decalcification approach involves the supplementation of eth­
50 % at 6 % FFA level [32]. Water content in the lipid affects both ylenediaminetetraacetic acid (EDTA) for crude biodiesel purification
chemical catalysts and enzymes. It can lead to soap formations through [44]. EDTA forms a complex with calcium and increases its solubility in
hydrolysis and reduce the catalyst availability and biodiesel yield [33]. water. However, this method used for calcium separation is not effective
for alcohol separation. Membrane technology also offers a viable
Transesterification approach for biodiesel separation, with ceramic membranes preferred
over organic members due to thermal stability [45]. The membranes
Transesterification includes non-catalytic and catalytic methods to have a pore size (12–400 mm) that inhibits the triacylglyceride and al­
convert various oils into biodiesel, which is similar to petroleum diesel. lows biodiesel [46]. The primary advantages of the ceramic membrane
The reaction required one mole of triglyceride and three moles of include its ability to separate the sodium soap and methanol, sample
alcohol solvent to generate three moles of biodiesel and 1 mole of separation process, and management, scalability, and low energy re­
glycerol [34]. Several parameters influence the process, including re­ quirements. However, over the time, the need for membrane replace­
action time and temperature, catalyst types and concentrations, solvent ment can increase the cost of the biodiesel production [47]. Thus,
types and amounts, feedstock purity, and reactor configurations. successful biodiesel purification and separation will improve the engine
Transesterification through the catalyst is majorly applied during oil performance and could be used as direct fuel instead of blending.
conversion due to easy handling, operation, high efficiency, and yield.
Heterogeneous catalysts are preferred over homogenous ones as they Quality control
can be easily separated from the product and recycled many times [35].
Optimization of other operational parameters (such as temperature, The produced biodiesel must be within the standard range for suc­
alcohol-to-oil ratio, reaction time, and stirring speed) is also vital for cessful commercialization and market value. The physical properties
higher biodiesel yield. Appropriate stirring speed enhances the oil and chemical composition analysis determined the biodiesel quality
contact with the catalyst, maintains the reactor temperature constant, [48]. National standards provide some analytical methods for biodiesel
and improves biodiesel conversion. Methanol-to-oil ratio adjustment testing. The main purpose of these standards is to define specific prop­
ensures complete transesterification [36]. Feedstock pretreatment de­ erties and limit ranges of biodiesel to maintain engine performance and
creases FFAs and moisture content and improves efficiency. Heteroge­ good mileage [49]. The American Society for Testing and Materials
neous catalysts are easily separated from biodiesel and reused without (ASTM) needs to revise and update all quality tests for biodiesel prop­
losing efficiency [35]. A two-step process is preferred for animal and erties to ensure compliance of the public and companies. Characteriza­
plant lipids with high FFAs content that is more energy efficient tion of the feedstock and biodiesel is required to assess the qualities of
compared to a single-step process at higher temperatures [37]. the fuel [50]. The critical parameters include metals (sodium, potas­
Animals-derived fats (such as chicken gizzard, lard, and tallow) and sium, calcium, and magnesium), non-metals (phosphorus and sulfur),
plant oil contain different FA compositions transesterified at different alcohol control, water content, flash point, cloud point, acid number,
conditions (such as reaction temperature, time, and alcohol ratio) to glycerin content, and OS test are needed to evaluate biodiesel [51]. An
generate biodiesel. The efficiency of various catalysts (such as acids, acid test is carried out during the biodiesel storage to measure the bio­
bases, and enzymes) is compared. Multiple concentrations of catalysts diesel quality as the carboxylic acid is formed during the oxidation of
are being used to yield biodiesel > 85 % (Table 2). Animal fats collected FFAs. The presence of water facilitates the FAMEs hydrolysis that pro­
from slaughterhouses predominantly contain palmitic, stearic, oleic, duces the FFAs [52].
linoleic, and linolenic acids. They are transesterified using a KOH Catalysts should be preferred which are not affected by several fac­
catalyst (6.5 %) and methanol ratio (3:1) at 60 ◦ C for 1.5 h, resulting in a tors (such as insoluble components, moisture, alcohol, and particulate
99.2 % biodiesel yield [38]. Acids and alkaline catalysts produced bio­ matter). Polar compounds, sulfur, phosphorus, sodium, and calcium
diesel in a short time. However, environmental safety, high reaction need to be evaluated in the feedstocks [39]. The FFAs, OS index,
temperature, corrosion, saponification, and product impurities are still saponification value, CN, IV, KV, and density are worthy properties of
unresolved challenges. The lipase catalytic process has disadvantages biodiesel that define fuel quality [53]. Different waste cooking oils (14
such as high cost and slow reaction conversion [39]. Further studies are samples) were assessed for sulfur content analysis, and 13 samples had
required on catalyst recovery and recycling on a large scale and devel­ higher sulfur concentrations compared to the standard (<10 mg/kg).
oping eco-friendly catalysts with high conversion of biodiesel. Trans­ The possible reason for the high sulfur content contributed to the
esterification can affect biodiesel’s separation cost and yield (Fig. 2). cooking oil added from the sulfur-containing food products [54]. The
high sulfur content in the biodiesel produced sulphuric acid and sulfate
Purification and separation during combustion in the engine, which causes corrosion and sulfate
particulate emissions [55]. Appropriate feedstock selection and quality
Techniques used for the purification in biodiesel process are influ­ tests are important to avoid engine and environmental issues.
enced by various factors, including types of impurities, biodiesel vol­
ume, and the associated processing costs [40]. The washing process is Storage and distribution
essential for removing excessive alcohol, glycerol, soap, and catalyst
residuals, resulting in a purified product that meets the standards for the Degradation, corrosion, and contamination are the major problems
application. To date, the purification technique for biodiesel refinement related to the distribution and storage. Biodiesel storage for a long time
includes wet and dry washing, membrane, and precipitation [41]. The is essential. However, it is sensitive to oxygen and produces short-chain
distillation technique removes unreacted alcohol and water from the FAs, peroxides, and gums that could reduce the engine efficiency.
crude biodiesel. After distillation, water is used to resin the impurities Various other parameters (such as air and heat, which may enhance the
from the biodiesel [42]. Wet washing is a simple and effective approach degradation and produce aldehydes and peroxides) affect the fuel
to refine the biodiesel to remove glycerol and extra methanol. However, quality over time [56]. Storing tanks exhibit > 80 % corrosion issues
this method has disadvantages (such as high usage of water, trace from moderate to severe, which is one reason for the storage tank fail­
amount of water left remains and required large tank for working). Dry- ure. The leakage of the storage tank and pipes is due to poor mainte­
washing refers to a wet-washing technique to avoid water loss [40]. The nance and corrosion assessment. Steel is mainly used to store biodiesel
adsorbent is added to crude biodiesel, allowing it to agitate for 20 min at as it is resistant to corrosion and economically feasible [57]. Fewer
35 ◦ C, and the impurities are attached to sorbent. The adsorbent can be studies have revealed that corrosion types and the complicated behav­
reused after washing with magnesia solution (2 %) [43]. iours of fuel tanks cause changes in the physical properties of the fuel.

4
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

Table 2
Comprehensive overview of the feedstock characterization and the parameters involved in the transesterification parameters without the application of machine
learning models.
Feedstock Fatty acid profile Catalyst Raw material Reaction Biodiesel Major References
used (%) used ratio condition yield finding

Animal feedstocks
Chicken Palmitic acid KOH (heterogeneous MeOH to oil T = 50–70 ◦ C 94 % Acid value was lower than 1 %, and [166]
gizzard (C16:0) = 1.45, catalyst) ratio 3:1–9:1 Time = 4–6 h no pretreatment of the oil was
linoleic acid Stirring speed carried out
(C18:2) = 57.97, = 400–600
arachidic acid rpm
(C20:0) = 28.90
Chicken fats Palmitic acid KOH catalyst (0.85 wt%) MeOH to T = 60 ◦ C N/A High viscosity, high density, and low [167]
(C16:0) = 22.30, chicken fat Time = 1h volatility
oleic acid oil molar Stirring speed FFA = 0.53 %
(C18:1) = 42.47, ratio 6:1 = 850 rpm
linoleic acid
(C18:2) = 19.30

Animal fats Palmitic acid KOH catalyst (6.5 wt%) MeOH to fat T = 60 ◦ C 99.2 % Biodiesel quality from the reactor [38]
from the (C16:0) = 26.8, oil molar Time = 1.5 h meets the requirements of EN 14214
leather stearic acid ratio 3:1 (yield = 96.5 wt%)
industry (C18:0) = 27.6,oleic
acid
(C18:1) = 36.5
Pork fat Palmitic acid Alumina-doped CaO MeOH to fat T = 65 ◦ C 92.5 % Catalyst was recycled for 4 times [168]
(C16:0) = 18.66, catalyst (8 wt%) oil molar Time = 1.5 h Kinematic viscosity was 7.186 mm2/
linoleic acid (C18:2) ratio (9:1 s
= 11.67,arachidic
acid
(C20:0) = 12.32
Fleshing Palmitic acid H2SO4 (1.2 % wt%) MeOH to fat T = 60 ◦ C 95 % High acid value of 25 mgKOH/g and [169]
wastes from (C16:0) = 33.49, ratio 6:1 Time = 2h FFA value of 12.5 %. Two-step
leather stearic acid Stirring speed conversion was preferred.
tanneries (C18:0) = 13.05, = 350 rpm
oleic acid
(C18:1) = 44.09

waste fish fat ​ KOH catalyst (0.97 wt%) MeOH to fat T = 60 ◦ C 98.1 % cosolvent is economically feasible as [170]
ratio 9.1:1 Time = 0.5 h the consumption of alcohol, oil, and
Cosolvent reaction time was reduced
ratio = 29.1 %
Waste chicken Palmitic acid A. terreus lipase soluble MeOH to fat T = 42 ◦ C Soluble lipase High acid value (6.56 mg KOH/g) [171]
fat oil (C16:0) = 17.96, (1 %) and ratio 6:1 Time = 36 h (78 %) while
stearic acid (C18:0) Fe3O4_PDA_Lipase (6 %) Fe3O4-PDA-
= 20.85,oleic acid Lipase 90.6 %
(C18:1) = 42.92,
linoleic acid
(C18:2) = 16.54
Plant feedstocks
Palm kernel Lauric acid (C12:0) CH3COOH (15 % wt%) MeOH to oil T = 60 ◦ C 98 % High acid value (8.4 mg KOH/g) of [172]
oil = 45.6, myristic acid ratio from Time = 0.5 h oil
(C14:0 = 19), 15:1
palmitic acid
(C16:0) = 7.3,
Jatropha oil N/A KOH (0.55 % wt%) MeOH to oil T = 65 ◦ C 96.98 % 74.77 % biodiesel yield was higher [173]
ratio from 6:1 Time = 1h than without acid pretreatment
Stirring speed
= 400 rpm
Pongamia Palmitic acid KOH (1 % wt%) MeOH to oil T = 60 ◦ C 95.4 % Higher free fatty acids content [174]
pinnata seed (C16:0) = 9.66, ratio 6:1 Time = 2h (18.58 %) was found and
oil oleic acid (C18:1) = Stirring speed esterification was carried out with
52.4,linoleic acid = 750 rpm H2SO4
(C18:2) = 16.53,
Frying palm Palmitic acid Free lipase from MeOH to oil T = 30 ◦ C 94.6 % 16.6 % water content was reported [175]
oil (C16:0) = 42.1,oleic C. antarctica lipase A ratio 7:1 Time = 22 h in oil
acid (5.5 %)
(C18:1) = 42.7,
linoleic acid
(C18:2) = 9.2
Tall oil fatty oleic acid (C18:1) = H2SO4 (0.5 % wt%) MeOH to oil T = 60 ◦ C 96.76 % Biodiesel yield was 96.7 % by [176]
acids 57.72,linoleic acid ratio 15:1 Time = 1h homogenous catalyst H2SO4 in 1 h,
(C18:2) = 35.41, Stirring speed While 90.24 % was produced in 4.7 h
= 300 rpm using a heterogeneous catalyst
(Ambersep BD 19)

T = Temperature, N/A = Not available, wt% = weight percent of oil.

5
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

Fig. 2. Conversion of a variety of lipidic feedstocks to biodiesel through a systematic and sequential process.

The properties of biodiesel change over time while it is stored in a tank. Application of AI in biodiesel production
[58]. During long-term storage, oxidized products (such as peroxides are
formed. Oxidation leads to the formation of FFAs, which increases the Feedstock selection and optimization
acid value of the fuel and causes corrosion. [59]. During biodiesel syn­
thesis, mono/di-glycerides absorb the water, trigger FAMEs hydrolysis, It is not economically feasible to characterize the feedstock each time
and cause corrosion. The physical appearance also changes during before the biodiesel conversion cycle as this would require excessive
storage, representing the contamination and degradation of the bio­ resources and time [53]. However, ML models have the potential to
diesel [60]. Storing tanks of different compositions (stainless steel and address these obstacles effectively. For instance, Soybean yield was
galvanized steel) are exposed to biodiesel, water containing bio/diesel estimated based on ML models using meteorological information (tem­
(5 %), and without water at 25 ◦ C for 50 days. It shows the corrosion rate perature, precipitation, and crop evapotranspiration) as input variables
in stainless steel tanks is 0.5 times > in biodiesel related to diesel. In to predict soybean yield. The average predicted yield was in the ex­
contrast, diesel (3.5 times) has a higher corrosion rate in galvanized steel pected range, similar to historical data (2.73 kg/ha) [63]. ML Models
than biodiesel [58]. An antioxidant additive reacts with free radicals to can be trained to assess the oil content and impurities in the feedstocks,
restore stability and ensure biodiesel quality, which results in stable which helps the producer select suitable feedstocks for biodiesel [64].
compounds that do not participate in the oxidation reaction [61]. The The AI-based optimization determined the ideal ratio of mixing different
addition of tertiary butylhydroquinone can increase stability by avoid­ feedstocks, which results in the final biodiesel with better fuel properties
ing oxidation reactions [62]. such as high energy density and low emissions [65]. The ML technology

6
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

in the agriculture sector improves the yield of oily crops and resource Process optimization
management [66]. Various classification and regression ML models such
as Bayesian, support vector machines (SVM), ANN, multilinear regres­ Process optimization in biodiesel production refers to systematically
sion (MLR), instance-based models, decision trees (DT), and clustering adjusting several operational factors (such as catalyst concentration,
models. These models facilitate the analysis of agricultural data, temperature, solvent-to-oil ratio, reaction time, pressure, and stirring
enabling better decision-making to optimize crop yields and resource speed) to achieve the highest possible conversion of bio-oil into bio­
utilization [67]. Utilizing machine learning tools on oilseed crops is vital diesel while minimizing resource consumption and energy input [74].
for predicting yield, detecting diseases, and managing water and soil Several studies have demonstrated that various factors (such as catalyst
effectively. This refers to sustainable crop cultivation and reduced bio­ concentration, methanol ratio, temperature FAs composition) simulta­
diesel production costs [68]. The oil and seed yield of rapeseed were neously showed complex relationships (linear and non-linear) with
predicted using RF (R2 = 0.94 and 0.91) and ANN (R2 = 0.85 and 0.83), biodiesel yield. The conventional method of transesterification optimi­
respectively [69]. The oil content in Carum copticum was better pre­ zation through a trial and error approach is time-consuming in attaining
dicted by ANN (R2 = 0.90) than MLR (R2 = 0.75). The high performance the desired conditions (such as temperature, catalyst concentration,
is due to ANN’s ability to process non-linear and multivariate data more types, and solvent molar ratio) for higher biodiesel yield [6].
effectively than MLR [70]. ML models streamline such obstacles through the prediction and
Biodiesel derived from single feedstock oil often has poor OS and optimization of the conditions for transesterification from historical
cold flow properties, high acid values, and lower heating values [19]. data. K-nearest neighbor (KNN) and decision tree (DT) models were
The biodiesel produced from feedstock with higher saturated fatty acids trained on input data (such as catalyst concentration, methanol-to-oil
content is more viscous and inefficient in colder regions. ML model ratio, temperature, and time) to predict biodiesel yield. Both models
analyzes the data of feedstock properties (FA composition, FFAs content, showed high accuracy and prediction (R2 > 0.92) [75]. Trans­
moisture content in oil, and oil density) to predict the best blending ratio esterification conditions are optimized through application ML ap­
for better fuel properties such as CN and high heating values [71]. proaches and statistical models, including ANN, ANFIS, SVR, RF, XGB,
Several feedstock oil mixing approaches are applied to improve bio­ and RSM, and represent (R2 > 0.95) (Table 3). WCO was converted to
diesel quality. Palm oil and Karanja oil were mixed in almost 1:1, biodiesel (~95.7 % yield) using ANN and RSM at optimized conditions
resulting in higher OS (9.5 h) and showing that OS (21.47 %) and cetane such as water content (12 %), reaction time (20 h), lipase (35 %), and
value (4.2 %) were improved as related to Karanja oil [72]. An equiv­ alcohol to oil ratio (6:7) [76]. Input variables (such as reaction duration,
alent ratio of castor oil and waste fish oil (1:1) increases the biodiesel molar ratio of alcohol/oil, and temperature) were used for the output
conversion yield by up to 97 %, with the resulting fuel properties that variables optimization using the ANFIS model with ALIFMO and NSGA-
meet international standards [73]. However, further research is needed II, which results in biodiesel yield (>96.5 %) with predication (R2 = 1.0)
to develop models for predicting feedstock availability and automated [77]. Oil from different feedstocks, such as animal fats, plant oils, WCO,
qualitative analysis. Integrating ML models (such as classification and and mixed oil, produces biodiesel with up to 99 % yield. Various ML
regression) during biodiesel processing can enhance yield accuracy and models are used to predict biodiesel yield from different feedstocks
which minimize the error loss function (Fig. 3). under various conditions. Advanced ML models (such as LSSVM, ANFIS,

Fig. 3. A schematic representation of parameters from various stages of the biodiesel production process used in various predictive models.

7
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

Table 3
Comparative investigation of the key parameters that influence transesterification and utilizing various machine learning models.
Biodiesel ML Study Conditions for Output Hidden Model model Biodiesel Comment/ References
Feedstocks Models aim input variables variable layers predictive accuracy yield highlights
capability

Animal-derived lipid feedstock

Waste goat RSM Optimization of Alcohol: oil = FAME 2 RSMpred = RSM = 96.7 % Silica gel adsorbed [177]
tallow and FFA conversion 32:1 yield 93.27 % 0.990 water facilitates
ANN via RSM and T = 70 ◦ C (96.7 %) RSMexp = ANN = higher FFA
ANN H2SO4 = 60 % 96.7 % 0.9966 conversion FAME
Time = 2.5 h ANNexp =
96.7 %
ANNpred =
99.8 %
Chicken fat Kinetic Ultrasonic Alcohol: oil = FAME NA R2 = 0.83 for RMSE = 98.2 % Lipase interaction [178]
and model amplitude and 4:1 yield the effect of 6.60 for with fat and methol
white reaction T = 40 ◦ C (92.6 %) enzyme and catalyst followed the Ping
grease parameters H2SO4 = 6 % molar ratio RMSE = Pong BiBi model
optimization by Time = 0.33 h R2 = 0.91 for 5.99 for
kinetic model Ultrasonic ultrasonic ultrasonic
amplitude = 40 amplitude amplitude
% and cycles
Ternary GA Waste blending Alcohol: oil = Predicted N/A R2 = 0.98 RMSE = 98.31 % Methanol: oil was a [179]
blended optimization 6:1 98.61 1.2347 significant feature
oil through GA T = 61 ◦ C Blending enhanced
Catalyst = 2 % feedstock
Time = 2.45 availability
Plant-derived lipid feedstock
Palm kernel ANFIS To develop Alcohol: oil = ROI, PBT, 5 R2 values For ANFIS N/A ANFIS based model [153]
oil and techno-economic 1:6–8 NPV, and ANFIS model ROI = 1.9 gave the best
ANFIS models for T = 55–65 ◦ C PC ROI = 0.99 × 10-5 performance
predicting Cat qty = 4–8 % PBT = 0.99 PBT = 6.2
parameters Time = 1–3 h NPV = 0.99 × 10− 6
to evaluate the Catalyst PC = 0.99 NPV = 4.9
feasibility of PKO calcination T = × 10− 6
800–900 ◦ C PC = 5.1 ×
10− 6
Castor oil ANN Optimization of Alcohol: oil = FAME 2 ANNexp = ANN = 1.2 76.85 % ANN predictions [180]
reaction 25: 1 yield = 76.85 % × 10− 2 deviation (±5.58 %)
parameters and T = 60 ◦ C 76.85 %
kinetic model H2SO4 = 3 %
Time = 4h
Reutealis RSM Box-Behnken Alcohol: oil = FAME N/A R2exp = 0.98 N/A 97.78 % An infrared-assisted [181]
trisperma design used for 8:1 yield = R2pred = 0.98 reactor and H2SO4
oil reaction T = 64 ◦ C 97.78 % were used to lower
optimization KOH = 2 % the acid value
Time = 1–2 h
Waste RSM RSM-RSM and Alcohol: oil = RSM-ELM 15 R2pred = 0.97 1.19 95.5 % Hybrid models such [182]
cooking and RSM-SVM 6.1:1 yield of for RSM- as RSM-ELM (2.74
oil SVM methods use T = 70 ◦ C 96.86 % ELM %) and RSM-SVM
optimization and NaOH = 1 % RSM-SVM R2pred = 0.95 (3.34 %) deviate
prediction Time = 1–2 h yield of for RSM- from the
Mixing = 94.4 %. SVM experimental yield
709.42 rpm
Soybean oil ANN Optimizing the Heterogeneous FAME 300 R2 = 0.99 for N/A 83.88 % Actual and predicted [75]
conditions ANN catalyst yield ANN values are very close
(KNN) condition to a trained model
Alcohol: oil =
10.6:1
T = 24.80–90 ◦ C
Cat qty = 3.4 %
Time =
0.83–4.3 h.
Soybean oil BPNN Evaluation of Na2CO3@BFD BPNN 4 N/A BPNN 98.50 % Na2CO3@BFD [132]
GA- various models blast furnace GA = GA = catalyst has
BPNN for optimization dust 98.90 4.3291 activation energy
SSA- via BPNN Alcohol:oil = BPNN BPNN (65.73 kJ/mol),
BPNN GA-BPNN 6–18:1 SSA = SSA = which was lower
SSA-BPNN T = 45–85 ◦ C 98.34 2.432 than that of solid
Cat qty = 3–11 BPNN = BPNN = base catalysts
% 100.43 9.2 × 10-1
Time = 0.5–2.5
h
Jatropha- ANN Conversion of Alcohol: oil = FAME 1 R2 = 0.99 N/A 98.92 % Comparing the [183]
algae oil blend oil using 6–12 yield = predicted values and
blend ANN technique T = 35–45 ◦ C 97.05 the actual values
for optimization Cat qty = 0–2 % with low error
Time = 1–3 h showed the ANN
model is reliable

8
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

FFA = Free fatty acids, FAME = Fatty acids methyl ester, ANFIS = Adaptive neuro-fuzzy inference system, ANN = Artificial neural network, SVR = Support vector
regression, RF = Random forest, XGB = Extreme gradient boosting, KRR = Kernel ridge regression, T1FLS = Type-1 fuzzy logic, T2FLS = Type-2 fuzzy logic, RSM =
Response surface methodology, LASSO = Least absolute shrinkage and selection operator, WC = Waste cooking oil, WPO = Waste palm oil, WAF = Waste animal fats,
T = Temperature, ROI = Return on investment, PBT = Payback time, NPV = Net present value, PC = Production capacity, Cat qty = Catalyst quantity, exp =
Experimental value, pred = Predicted value.

and XGB) exhibited high prediction accuracy (R2 = 0.99) in various biodiesel from WCO, reducing reaction time at room temperature
feedstocks and outperformed traditional methods like RSM (Table 4). [53,83]. ML models are combined with advanced statistical models and
Findings of recent studies further supported the effectiveness of ML data sensing software for real-time monitoring (changes in reaction
model predictions and optimizations of the several stages of biodiesel temperature and biodiesel and glycerol production) and controlling
production. Developed ML models such as MLR, RF, XGB, and SVR that (such as reaction temperature and water content) during biodiesel
predict the fuel properties from FAs composition [1]. An ANN model production. Advanced measuring techniques (including chromato­
accurately predicts hybrid biodiesel fuel properties (such as KV, density, graphic and spectroscopic) for quality analysis of the final product are
calorific value, and flash point) from FAs compositions of mixed and also used to make decisions (Fig. 4).
single oils and achieves (R2 = 1) for mixed feedstocks [78]. The XGB The ANFIS, ANN, and RSM models are applied to calculate the en­
model outperformed (R2 = 0.99) in biodiesel yield optimization ergy ratio (output/input energy) for the hydrodynamic cavitation
compared with MLR, RF, and DT. The model predicts the best reaction reactor automatically controlled, the average energy ratio achieved
conditions, such as temperature, catalyst concentration, methanol to oil (1.205), the predictability accuracy for ANFIS (R2 = 0.98), ANN (R2 =
ratio, calcination temperature, and reaction time, were 55.9 ◦ C, 5.95 %, 0.96), and RSM (R2 = 0.99). The optimum conditions, such as circula­
15:1, 600 ◦ C, and 77 min, receptively, for higher biodiesel yield (98.5 %) tion time (2 min), rotational speed (8000 rpm), pump flow rate (1.4 L/
[6]. min), and solvent molar ratio (4), result in 91.87 % biodiesel yield [84].
The temperature of the microwave reactor used for biodiesel generation
Real-time monitoring and control from WCO was controlled through LabVIEW software combined with
the GA-ANFIS controller. This approach precisely adjusted the micro­
Biodiesel industries require rapid adjustment for real-time moni­ wave power levels to optimize the reaction [85]. The PSO is used to
toring, optimization, and control of the production process, which can adjust the reaction rate constant of four different mechanism models
be possible through AI models [79]. ML determined the yield and used in the transesterification of soybean oil. After optimization, the
established the complex relationship between input and output vari­ progress of the transesterification process was monitored using an ANN
ables, also predicting the parameters that affect the process and the with a self-organizing map model. This approach monitors the reaction
desired quality of the biodiesel [80]. ML models manage the system progress using molar concertation of the compounds, catalyst concen­
process effectively in response to variations in factors and have the trations, and reaction duration. The mean error values (4.3–8.5 %) be­
capability of comprehensive data analysis from production [81]. For tween the actual and predicted values derived from the PSO model [86].
real-time analysis, data is usually derived from each step (such as reactor The ultrasonic reactor connected LabVIEW software and intelligent
operational conditions, process controlling system, and input material) sensors to enhance the biodiesel yield with reduced input energy. Molar
[82]. Different ML models are established during production to maintain ratio, reaction time, frequency, temperature, and power are input vari­
the system, optimize, detect errors, and make predictions. ML ap­ ables, and biodiesel yield is output using ANFIS, RSM, and ANFIS-GA
proaches such as MLR, principal component analysis, DT, K-nearest models. The yield was accurately predicted by the ANFIS (R2 = 0.94),
neighbor classifier (KNN), SVM, GA, ANN, and RF are used in the RSM (R2 = 0.92), and ANFIS-GA (R2 = 0.97) [87]. The ML approach can
monitoring of the biodiesel process [8]. The ANFIS controls and moni­ effectively monitor the biodiesel process (such as reaction, yield, and
tors the process of a pilot-scale microwave reactor that produces fuel quality).

Table 4
Recent studies have focus on usng machine learning models to predict the biodiesel yield.
Feedstocks Predictive Aim Input variables Output Prediction Biodiesel Advantage/ References
models conditions (prediction) Capability yield (%) limitation

Canola oil LSSVM To assess Alcohol: oil = FAME yield R2 = 1 92.87 GA-LSSVM model [184]
LSSVM model 20–40 (92.87 %) was more precise than
efficiency forbiodiesel T = 270–400 ◦ C MLPNN modeldeveloped
yield prediction Time = 3–20 min
Canola oil Takagi- To optimize the Alcohol: oil = 100 % R2 = 0.99 100 PSO-ANFIS model was [185]
SugenoANFIS biodiesel yieldusing the 20–40 more accuratethan the
ANFIS tuned by PSO T = 270–400 ◦ C MLPNN
Time = 3–20 min
Macroalgae oil XG Boost To maximize the Alcohol: oil = 89.6 % R2 = 0.99 89.6 Most robust fit for the [186]
model biodiesel yield 5.95:1 dataset compared to
T = 55.9 ◦ C MLR, RF, DT
Time = 77 min
Kernel oil RSM Comparative T = 65 ◦ C Time = RSM = 94.82, R2 = 0.98 for 93.48 ANN has better [187]
ANN optimization of yield 2.5 h Alcohol: oil ANN = 93.21 % RSM predictive ability than
using RSM and ANN = 7:1 R2 = 0.99 RSM
Catalyst = 2.5 %
Waste Soybean RSM Comparison of Alcohol: oil = 1:2 RSM = 30 %, R2 = 0.98 for N/A ANN model indicates [188]
cooking oil ANN different ML models for T = 60 ◦ C ANN = 90 % RSM higher predictive ability
higher biodiesel yield Time = 6 h R2 = 0.99 for than RSM
Catalyst = 1.5 g ANN

ANFIS = Adaptive neuro-fuzzy inference system, ANN = Artificial neural network, RF = Random forest, RSM = Response surface methodology, LASSO = Least
absolute shrinkage and selection operator, WCO = Waste cooking oil, WPO = Waste palm oil, WAF = Waste animal fats, T = Temperature, ROI = Return on investment,
PBT = Payback time, NPV = Net present value, PC = Production capacity, T = Temperature (◦ C), LSSVM = Least squares support vector machine, PSO = Particle
swarm optimization, N/A = Not available, MLPNN = Multilayer perceptron neural network.

9
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

Fig. 4. An overview of biodiesel production process data used in different analytical tools and machine learning models.

Biodiesel quality control and assurance algal oil-based biodiesel due to the unique compositional characteristics
of algal oils [91]. ML models perform better with specific types of oil
ML approaches to predicate the fuel qualities with high accuracy and compositions, such as oils with more linear relationships between
less error [88]. The FAs compositions from carbon 8 to carbon 24, de­ composition and biodiesel quality (CN). They might be well-suited to
gree of saturation, and blending of different oils are used to estimate the simpler models like linear regression. In contrast, more complex com­
prediction of CN, density, IV, cloud point (CP), and KV. The generic positions require deep learning approaches (such as ANN, ANFIS, and
models show more accurate and precise predicted values and less rela­ SVR) [92]. The MLR was trained on the same FAs data for CN and CFPP
tive error, and ANN and ANFIS are the most frequently used ML models prediction. The model performance was different in CN (R2 = 0.88) for
(Table 5). The PLS, RF, and SVM approaches compare the biodiesel CFPP (R2 = 0.78) [93]. ML models, such as linear regression, are usually
spectroscopic data and predict the fuel properties (such as CFPP and preferred for CN and density prediction due to linear correlation [94].
KV). The results exhibited that the SVM algorithm provides the optimum However, Multiple non-linear regression models enhanced CN pre­
predicted value for the CFPP (0.9 ◦ C). In comparison, PLS regression dictions by capturing complex relationships among fuel properties,
demonstrates the best result for KV at 40 ◦ C [89]. Different oils are outperforming traditional MLR approaches in accuracy [95]. The FFAs
mixed proportionately to improve the biodiesel properties (such as KV, have complex relations with biodiesel yield; usually, complex ML
density, CN, and OS) using a simplex lattice mixture design model. The models could perform well [96].
optimized ratio for Karanja: palm (51.6:48.4 v/v) results in KV, density, The FAs composition, calorific value, and viscosity of oils highly
CN, and OS 3.854 mm2/s, 860 kg/m3 56.189, 9.565 h, respectively [72]. affect the model prediction capability during biodiesel production.
The ANN can accurately predict glycerol removal from biodiesel (R2 = Various ML models perform differently for multiple data. For example,
0.99) using deep eutectic solvents (ammonium or phosphonium) [90]. models like SVM, ANN, RF, and ANFIS are more effective for datasets
ML models can effectively reduce risk, improve operational perfor­ with complex interactions among features. In contrast, MLR models suit
mance, and attain sustainability. more direct relationships between FAs and fuel properties [97]. Oil
Different oil types (such as vegetable oils and animal fats) have composition affects model selection due to the complex relationships
distinct FAs compositions, demonstrating that models optimized for between FAs and fuel properties. The model performance identifies
vegetable oil-based biodiesel often showed low performance for the critical input variables and the data dimensions. The MLR predict the

10
 Table 5
A summary of the prediction efficiency of various artificial intelligence models for fuel qualities.
Feedstocks Aim Input Output Hidden Actual Predicted Statistical indices Difference in Comment References

M. Arif et al.
features features layers values values predicted and actual
value

Mixed oil Prediction of biodiesel LCSF-DU, MFC and SCSF- OS, KV, 1–14 OS = 3.27–20 For ANN R2 (ANN = RMSE CV and KV were best [101]
properties based on the FAME DUm and CV CV = 37–41.43 OS = 0.992–0.999 and ANN = 2.21 × 10-2 predicted for SCSF-DUm
composition using ANN and KV = 2.2–5.83 5.77–20.29 RT = 0.99–1.0) − 1.973 approach while OS was
random tree (RT) CV = and RT = 1 × 10-5 predicted for the LCSF-
36.82–41.37 − 3.68 × 10-2 DU approach.
KV =
2.36–5.73
For RT
OS = 3.27–20
CV =
37–41.43
KV = 2.2–5.83

Plants and animals Fuel properties 14 FAME including Cetane 14 25.46–66.13 N/A R2 = 0.853 (PLS) % error = 4.50 (PLS) ANN-adapted PLS was [100]
lipid data prediction based on FAME C10:00-C24:00 number R2 = 0.993 % error = 1.06 more robust than the PLS
obtained from the composition via ANN-adapted (ANN-adapted (ANN-adapted PLS) approach for biodiesel
literature PLS and PLS approach PLS) CN prediction
Data obtained from Prediction of biodiesel fuel Cn (14.10–17.86), dn KV and N/A KV = KV = R2 = 0.989 for Mean absolute % ANFIS related to IV and [189]
the literature properties using ANFIS (0.21–1.54), T IV 1.35–7.94 1.72–7.94 KV error kV have higher
(293.15–373.15 ◦ C) for IV = IV = R2 = 0.996 (IV) 0.39 for KV predictive ability and
KVMU 82.25–137.84 82.17–137.52 0.25 for IV accuracy
(16.31–79.14 %), PU
(15.92–141.47 %), dn
(0.95–1.59) for IV
Refined canola oil Production of biodiesel from C16:0 = 13 %, CP, 1 CP (15), CP (13.5), R2 = 0.98 for CP N/A CFP was dependent on [190]
(RFO), WFO and RFO and compare C18:0 = 5.48 %, PP, PP (12), PP (11), R2 = 0.94 for PP FA composition by
wasteFrying oil their fuel properties to C8:1 = 38.2 %, CFPP CFPP (14) CFPP (10.9) R2 = 0.96 for applying the ANN model.
11

(WFO) literature using ANN C18:2 = 27.2 %, for WFOCP for WFOCP CFPP SFA was expected to hurt
, and 103 studied Modeling C18:3 = 7.7 % (− 3.5), (− 3.6), CFP
samples PP (− 10), PP (− 9.5),
CFPP (− 7.5) CFPP (− 6.8)
for RFO for RFO
Cotton oil biodiesel Modeling of density and KV Diesel Density, 1 Density = N/A N/A ANN has KV (0.19) ANN has higher [191]
blend with a ratio using the ANN and LR Biodiesel KV 835.8–886.9 density (0.02) MAPE prediction and accuracy
20–75 % approaches B20 KV = while LR has KV than the LR
B30 2.76–4.76 (16.87) density

Sustainable Energy Technologies and Assessments 73 (2025) 104097

B40 (0.19)
B50 MAPE
B70
Fatty acid profile of Generalization of the RF and C8 to C22 CN and N/A CN = 54.8 ​ RF model CN value RMSE = Causal analysis and ALE [93]
> 200 dataset XGB model CFPP CFPP = 0.7 R2 = 0.88 (CN) 2.54 and 2.37 for RF determined the optimal
R2 = 0.91 (CFPP) and XGB, FA profile via
XGB model respectively. optimization
R2 = 0.89 (CN) CFPP value
R2 = 0.87 (CFPP) RMSE = 2.03 and
2.39 for RF and XGB,
respectively.
Sunflower, peanut, Predict biodiesel’s KV saponification value = Density N/A Density = N/A R2 = 0.954 Average relative PU/MU as input variable [192]
coconut, copra, density, FP, higher heating 197.08––265.22, IV = HHV 864–875.78 (density), R2 = error (%) 0.084 increased accuracy and
beef tallow, value (HHV), and OS on the 0–134.98, KV HHV = 0.76 (density), 0.61 R2 value for prediction
rapeseed walnut bases of PU, saponification, KV = 2.64–4.75 FP 37.83–39.85 (HHV), R2 = (HHV), 2.7 models
oil and IV using MLR and PU/MU = 0–3.8 OS FP = 0.955 (KV), 1.85
124–181.5 (KV), R2 = 0.957 (FP), 4.4
OS = (FP), R2 = 0.941 (OS)
0.105–1.81 (OS)

LCSF-DU = Long chain saturation factor and degree of unsaturation, MFC = Modified fatty acid methyl ester compositions, SCSF-Dum = Straight chain saturation factor and modified degree of unsaturation, CV = Calorific
value (MJ/kg), KV = Kinematic viscosity (mm2/s), OS = Oxidative stability (h), Cn = Carbon number, dn = Number of double bonds, MU = Mono unsaturated (%), PU = Poly unsaturated fatty acids (%), SFA = Saturated
fatty acids, Mw = Molecular weight T = Temperature (◦ C), Mw = Molar weight, CP = Cloud point (◦ C), PP = Pour point (◦ C), CFPP = Cold filter plugging point (◦ C), IV = Iodine value (mg of I2/100 g), LSSVM = Least
squares support vector machine, PSO = Particle swarm optimization, N/A = Not available.
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

(fuel properties such as CN and density) of the biodiesel from FAs (such reliability compared with the conventional PLS approach (R2 = 0.85)
as lauric, myristic, stearic, oleic, linoleic, and linolenic acid). The [100]. This suggests that ANN has more validity and reliability in pre­
developed model was tested for experimental data and showed pre­ diction accuracy when compared with actual and predictive values.
dictions for high fuel properties. In comparison, the exact model was Calorific value estimates the inherent energy and is vital for the accep­
ineffective when it was tested for KV (R2 = 0.69) and high heating value tance of biodiesel as a viable fuel. The prediction value of different FAs
(R2 = 0.62) [92]. compositions is calculated from ANN representing MFC (R2 = 0.99),
SCSF-DUm (R2 = 0.97), and LCSF-DU (R2 = 0.84) [101]. The role of ML
technology in predictive analysis is to function like traffic guidelines.
Predictive maintenance Following such guidelines minimizes the risk of accidents. Similarly,
prediction maintenance assisted in foreseeing potential problems such
The prediction potential of advanced AI and ML technologies in as compositional challenges, process optimization, yield, and fuel
biodiesel production significantly reduces the risk of failure, enhances properties before escalating into significant issues. The predictive
product quality, and decrease processing time. Predictive analysis in the capability of ML models (such as ANN and MLR) for transesterification
biodiesel industry is primarily formulated through regression techniques and estimating various biodiesel qualities (such as CN, IV, KM, density,
[98]. The MLR, SVR, and XGB are employed to evaluate the comparative and OS). Predictability could be improved by selecting the most relevant
efficiency of model for predicting the CN value. Among these, XGB is the input features in the data engineering process and model tuning to attain
effective predictability of the trained model (R2 = 0.99) and test model maximum performance and accuracy (Fig. 5).
(R2 = 0.89) are close to each other, indicating that the model is not only
accurate but also can be generalized, which is a crucial factor attained Challenges in machine learning
through ML-based models [93]. ML models consist of inherent param­
eters known as hyperparameters which can be optimized using various Data quality and availability
techniques (such as grid search, manual search, and Bayesian optimi­
zation) to attain the desired output [99]. The ANN-adapted PLS model The data availability challenges have been addressed through
predicted the CN value (R2 = 0.99) and showed high accuracy and

Fig. 5. Integration of feature selection and data engineering strategies for optimizing machine learning models in biodiesel production.

12
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

various strategies for biodiesel production. Imputation and elimination Energy efficiency
significantly addressed the missing values in the data [102], while
outliers were either removed or substituted for more typical data [103]. Various ML models (such as ANN, SVM/SVR, ANFIS DT, and MLP)
To improve model performance, data scaling techniques (min–max are applied to predict energy consumption. Recently, hybrid and
scaling or z-score normalization) are applied to input variables [104]. ensemble approaches have gained attraction due to their enhanced
The multi-fidelity data approach combining literature and experimental validity and accuracy in energy predictions [116]. The grey wolf opti­
data has been used to enhance the effectiveness of prediction and mizer (GWO) and RSM are compared to maximize the biodiesel pro­
address challenges associated with limited data availability for ML duction from waste sunflower oil by transesterification, resulting in a
models in biodiesel production [105]. biodiesel yield of 97 %, indicating the precision and accuracy of both
Several studies used a limited data range (30–60 datasets) during models [112]. In contrast to the optimal conditions identified by RSM,
training and testing ML models, which resulted in overfitting issues in the ANFIS-GA hybrid model achieved a biodiesel yield that was 2.27 %
the model. Based on the data given to the ML algorithm (ANN model) a higher while reducing reaction duration by 3.27 %, energy consumption
certain amount of data (60–80 %) is selected for the model development by 2.98 %, solvent requirement by17.44 %, and catalyst required by
[106]. However, these problems can be avoided via various techniques 3.70 %. These results show that the ANFIS-GA model has a better per­
(such as out-of-bootstrap, leave-one-out, and k-fold cross-validation) formance compared to RSM in biodiesel production [117]. It recom­
[93]. Control variables such as solvent-to-oil ratio, reaction temperature mends that integrating models could result in higher biodiesel
and duration, and catalyst concentration are widely studied in biodiesel production and rapid decision capability in optimizing energy uses.
production through ML [107]. This refers to fuel quality analysis, which
can be affected by independent variables including FAs profile, FFA, and Scalability
water content [108]. Still, no such studies suggest a model combining
the optimization process and fuel properties prediction. No available Several studies reported the ML model application for optimization
research discusses the application of ML to the fuel properties of bio­ during the biodiesel process at the lab scale and showed promising re­
diesel and the impurities left after production. Hence, upgrading the sults. Still, there is insufficient data available focused on large-scale,
biodiesel industry through ML approaches, optimization processes, and real-time monitoring, control, and optimization of biodiesel produc­
fuel characterization studies should include the fate of the impurities at tion [118]. ML technology has the capability of cost-effective biodiesel
an early stage. production by improving fuel quality and yield through optimization
[119]. The reliability and accuracy of several ML models depend on the
Model complexity data given; sometimes, it is insufficient and incomplete, which may
affect the model’s validity [120]. The diversity of data points (input
The complex data are used in the optimization of biodiesel produc­ variables) in the biodiesel process (such as feedstock characteristics,
tion using ML approaches that integrate diverse datasets (such as feed­ reaction factors, and environmental parameters) should be carefully
stock characterization, reaction conditions, environmental factors, and selected to ensure they are relevant and not collinear. Avoiding irrele­
FFAs composition). This integration allows the models to address non- vant input variables is crucial for achieving reliable results in the model.
linear relationships within the data, which results in increased model [121]. Chromatographic analysis (such as HPLC, LC-MS, and GC) for
complexity [109]. Feature selection focuses on variables affecting the fuel properties estimation is sometimes expensive and time-wasting,
yield and fuel quality to reduce model complexity. The correlation with limited applications on a commercial level. Therefore, sensor
analysis can avoid multicollinearity problems. The DT and RF models software based on ML technology could be designed to measure the
can identify and remove less important variables [110]. important parameters of the biodiesel production process [122].
There are some challenges in the ANN model about the weighting The Internet of Things (IoT) applications have allowed real-time
procedure, which include a lack of information for understanding the process monitoring and optimization [123]. For example, a hybrid
assigning of values to input variables. The adjusting process of the particle swarm optimization (PSO-ANN) model connected to the IoT
synaptic coefficients between several neurons in an ANN model using temperature sensors optimizes the temperature of the operational pa­
input variables as training data is known as weighting. This adjustment rameters (such as esterification kettle, dry pots, and distillation tower’s
is essential for the model to learn and make accurate predictions. Due to temperature) to improve biodiesel production and reduces the cost
the complexity of this process and the difficulty in interpreting how [124]. The QCBIoT prototype device exhibited the potential of IoT in
inputs are transformed into outputs, ANN models are often referred to as quality control, predicting biodiesel pH and density with high accuracy
black boxes [109]. The ANN approach simulates biodiesel production (average errors of 1.82 % for pH and 2.78 % for density) [125]. Big data
under different processes enhancing strategies, assigning a high value to analytics has further improved model predictions and accuracy. Various
a process that might cause uniqueness loss [111]. However, the inte­ ML models (such as ANN, ANFIS, and PNN) could be trained on big data,
gration of modern AI might be able to solve the black-box mysteries and resulting in improved model performances such as (predictions, model
lead to more discoveries shortly [112]. The output of the models in ML generalization, and accuracy). These technologies, such as AI, IoT, and
technology might be regenerated for an application due to the stochastic Big data analytics, contribute to sustainability and efficiency, allowing
nature of the parameters (such as weight and bias). The application of process control accurately, data-driven optimization of operational pa­
the ANN model is limited to training-based data due to its low extrap­ rameters, and real-time quality monitoring for higher yields and reduced
olation ability, which can be combined with energy conservation energy uses in biodiesel production [1].
equations and thermodynamic principles. Such a combination is called
hybrid models, which are more reliable and valid for biodiesel produc­ Future directions for artificial intelligence
tion [113]. Some studies do not have adequate information about the
ANN models’ topologies and weighting, which are impractical to apply Sustainability in biodiesel production
in real-time simulation [114]. The developed ANN is undesirable as
absences/a few statistics parameters are discussed. Another important Modern technologies (such as AI, IoT, and big data) upgraded the
aspect of the available data is limitation and noise, which can disrupt the biodiesel industrial system by improving process efficiency (fewer by-
validity of complex ML models. It is essential to highlight the significant products such as soap formation) and optimized the parameters (such
strategies for data augmentation and preprocessing.[115]. The complex as temperature, catalyst load, and methanol ratio). These technologies
models may require computational expertise, affecting their availability are used in the design, processing, product manufacturing, and simu­
and scalability in the biodiesel industry. lation, resulting in sustainability and efficiency [126]. A two-stage

13
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

automated inspection system was integrated into the biodiesel produc­ process optimization in biodiesel production, accounting for variations
tion process from animal lipids. The purchasing cost of the feedstock is in lipid content from various feedstocks.
44.32 % of the total cost, and the percentage of impurities in biodiesel
can be reduced by consuming minimum energy [127]. The IoT (smart Predictive maintenance
concept) improved efficient production with high quality and made the
operational process flexible [128]. Such an intelligent concept will The IoT device data are used to train AI models for maintenance
connect the processing unit with AI using sensors to improve produc­ prediction and identify faults in the operational process of biodiesel.
tivity and performance, thus attracting the attention of researchers and These devices collect real-time sensor data (such as temperature and
entrepreneurs [129]. water levels), which are then analyzed by AI and ML models to detect
Integrating the AI model shows effective results and suggests that anomalies and predict faults (overheating and leakage) [110]. ML
biodiesel will be a sustainable fuel in the future. RSM and ANN predict models are trained and sensitive to deviations from normal operations.
the optimal conditions (such as a temperature of 65 ◦ C with 1 % KOH The models compare actual sensor data with predicted patterns to
catalyst and a 10:1 methanol-to-oil ratio for 1.5 h) for biodiesel con­ identify damage that indicates a fault in the machine or process ab­
version (98 %). The ANN (R2 = 0.99) shows better performance normalities. The ML models recommend optimization or changes in
compared to the RSM (R2 = 0.98) at optimal conditions [130]. Opera­ important process factors by analyzing real-time sensor data (IoT)
tional conditions (such as oil type, catalyst type and concentration, [139]. A fault detection system combined with AI was developed for the
methanol-to-oil ratio, reaction time, and reaction temperature) were marine oil separation system by analyzing vibration patterns. The sys­
used as input variables in ANN and ANFIS models to predict biodiesel tem converted the vibrations to frequency data and used this data with a
conversion (99.87 %) with high accuracy (R2 = 0.99) [131]. The back combination of GA-ANFIS to predict early warning signs of potential
propagation neural network model (BPNN) is optimized through a faults [140].
sparrow search algorithm to predict biodiesel yield. The optimal con­ The ANN model was trained with corrosion data from different
dition for the transesterification (methanol/oil ratio 18, catalyst 7 % and sources with input features (such as metal, saturated, monounsaturated,
65 ◦ C for 1.5 h) and predicted yield (99.5 %) with an MAE of 0.92 while and polyunsaturated fatty acids, blend biodiesel/diesel, temperature,
the experimental yield (100 %) shows little differences, the model could total acid number, and the time to exposure). The model demonstrates
improve biodiesel conversion efficiently [132]. high performance (R2 = 0.96) in corrosion rate prediction and suggests
that metal type and total acid number critical input variables are highly
Advanced control systems involved in corrosion rate [6]. Mechanical faults are often not detected
by conventional methods. An innovative AI system integrated with ANN
A key application of ML models is controlling the biodiesel produc­ models has shown the ability to optimize multiple input variables and
tion system. Three different models, the adaptive controller, ANFIS, and predict faults (such as tube leakage), up to 5 min earlier compared to
logic fuzzy, are studied to simultaneously maintain the pilot scale traditional methods, potentially preventing significant damage [141]. A
reactor that produces biodiesel from WCO. The microwave generates neural network (GA-BP) is employed to predict optimal values for re­
heat that is controlled to maintain a constant temperature in the reactor action parameters, catalyst concentration (8.82 %), solvent-oil ratio
needed for transesterification. ANFIS is more sensitive compared to (16.79), and temperature (65.12 ◦ C), resulting in 99.15 % yield of bio­
conventional controllers to temperature, which recommends that ML diesel [142]. ML-based predictive maintenance can reduce downtime,
models are an attractive approach for controlling the system [83,106]. efficiently utilize resources (such as catalysts and alcohol), improve
The development of a controller combined with the GA-ANFIS model biodiesel quality, and extend.
can update itself based on real-time error information in contrast to
conventional controllers, optimizing the power supply [133,134]. Pro­ Blending and quality assurance
cess analytical techniques are used to collect spectral and concentration
data during transesterification. Multivariate curve resolution- Biodiesel should meet the criteria established by the international
alternating least squares (MCR-ALS) is employed to analyze this data. standard (EN 14,214 and ASTM D6751) for easy sale in the market
Methanol and NaOH are used to assess the FAs components based on [143]. The quality of the fuel is determined by several factors such as
their spectral characteristics and kinetics for the biodiesel with (R2 = contaminants from the feedstocks, incomplete material (such as catalyst
0.99) [135]. The concentration of various glycerides (mono, di, and tri), and alcohol) from transesterification, and molecular composition of the
FAMEs, solvent, and glycerol in biodiesel derived from jatropha is FAs [144]. ML technology can identify the desirable feedstock for bio­
measured by a Kalman filter as a virtual sensor. This technique could be diesel through the FAs composition optimization with optimal fuel
applied in real-time as it needs variables (such as pH and temperature) properties [145]. There are still challenges in developing an ideal bio­
[136]. diesel composition that meets diverse criteria, such as reliable engine
performance, environmental safety, and being suited for colder regions
Catalyst and enzyme optimization [146]. ML models can identify suitable biodiesel feedstocks based on
their characteristics (such as oil content and FAs composition), which
Several ML approaches can optimize complex parameters to achieve are important to enhance yield and fuel quality [147]. ML approaches
higher biodiesel production while operating under limited resources (such as KNN and PLS) can detect adulteration in the blending process of
[137]. The ML models predict and optimize catalyst efficiency under biodiesel-diesel blends (B0-B20). Ultrafast gas chromatography com­
different lipid compositions. A comparative study using RF, SVR, XGB, bined with chemometric methods allows for effective analysis of
and kernel ridge regression models assessed various catalysts under blended fuel compositions and ensures compliance with standard
various transesterification conditions. A study showed that the homog­ guidelines.[148].
enous acid catalyst achieved a higher yield (87 %) in 10 min than the
enzyme-mediated process, which requires 300 min for a high yield. The Economic sustainability
XGB model demonstrated the highest accuracy (R2 = 0.98) in predicting
biodiesel yield [6]. The ANN model predicted a higher yield (81.60 %) at Techno-economic analysis
optimized catalyst concentration (4 %) and solvent ratio (13) with
relative (RMSE error = 1.01 × 10-2), exhibiting that ANN is the most The techno-economic analysis of biodiesel involved the costs of
accurate model compared with LASSO and SVR [138]. These ML ap­ feedstock, catalyst, labor, electricity, plant maintenance, and biodiesel
plications show the potential for enhancing catalyst selection and production. The case studies of biodiesel production on a pilot scale

14
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

from various plants and WCO were compared using different catalyst

WCO = Waste cooking oil, SCO = Spent cooking oils, N/A = Not available, TEC = Total equipment cost, Qt = Quantity, CapEx = Capital expenditures, SIC = Specific investment cost, RMC = Raw material cost, PBC =
Reference
acids/bases/enzymes, resulting in lower < 1 $/L of biodiesel (Table 6).
The key obstacles to economic biodiesel production are variations in

0.93/kg [193]

0.99/kg [194]

[150]

[195]
feedstock cost, raw material market cost (methanol, catalyst), and bio­
diesel selling price, depending on region. It is difficult to estimate profit

0.868/

0.750/

1.047/

0.448/

0.502/
(US$)
and loss every time for the biodiesel production process. The ML models

0.41
PBC

/kg

kg

kg

kg

kg

kg
(such as ANN, RSM, and ANFIS) optimized biodiesel parameters,
enhanced the conversion rate, and reduced production costs [149].

O and P cost

174800.93
The selection of catalysts (acids, bases, and enzymes) significantly

116,350

460,075

187,104

158,878

748,700

835,000
74,183
affects both the yield of biodiesel production and the overall cost. Three

(US$)
different catalysts (H2SO4, KOH, and lipase) were evaluated for the
biodiesel production cost from WCO. The operational parameters (such
as reaction temperature of 50 ◦ C for acid and base, 65 ◦ C for lipase, oil to

Labor cost

1,540,000

1,820,000
(US$/yr)

110,880

110,880

112,176

750,000

700,000

600,000
methanol ratio of 1:4, and reaction time of 6 h) under optimal conditions
with the biodiesel conversion efficiency of 100 %. However, the raw
material cost for KOH ($0.57/kg), H2SO4 ($0.48/kg), and (lipase $0.88/

Electricity cost
L) varied significantly. These variations in catalyst cost affect the final
price of the biodiesel alkaline ($0.868/kg), acid ($0.750/kg), and lipase

4,158,000

7,980,000
(US$/ton)

251,238
33,229

23,355

72,000

66,000

48,000
($1.047/kg) [150]. This means that the selection of catalysts could in­
fluence the final production cost, and acid catalysts are considered
economically feasible. The ML models provide more efficient and cost-

0.24 US$/L

0.48 US$/L

0.88 US$/L
0.9 US$/kg
effective solutions. The ANN and ANFIS models are composed of

material

0.57 US

0.43 US

0.45 US
different interconnected nodes (artificial neurons) that can process

$/kg

$/kg

$/kg
Raw

N/A
cost
complex input variables (factors) and capture non-linear relationships
between inputs and outputs. An ANN model consists of 4–7-13–1 (4 oils

SIC (US

847.81

719.92

336.14
types as input variables, 7 and 13 are hidden layers, and 1 output var­

$/ton)

N/A

N/A

N/A

N/A

N/A
iable) with sigmoid transfer functions that can predict the optimum
conditions for blending of various oils resulting in lower free acid value

2803676.44

6,312,897 12,652,060

7,487,000 29,468,832

8,350,000 32,865,600
of oils and used for biodiesel production [151]. The economic feasibility
CapEx cost

2992010.2

1,871,036 6,782,506

1,588,783 5,759,338

2,689,145
of biodiesel can be improved by ML techniques to save time and re­
(US$)

sources and predict optimum conditions. Artificial bee colony search

methods (ABCSM) and RSM models provide insights into the profitable
indices for biodiesel generation from palm kernel oil. The ABCSM output
TEC (US

695,656

695,656

741,833
parameters include batch size oil (13482.48 kg), production cost
$)

($3102448/yr), and return on investment (ROI) (39.8 %). The regres­

sion models show high accuracy (R2 = 0.97) for ROI and (R2 = 0.99) for
production

production capacity (PC), which recommends that ML models can

Biodiesel

99.70 %

98.88 %

monitor the techno-economic feasibility of biodiesel efficiently [152].

100 %

100 %

100 %
95 %

99 %

95 %
Comparative assessment of the economic feasibility of biodiesel production on a large scale.

The ANN and ANFIS models are applied to predict palm kernel oil’s total
cost and profitable value. Process optimization involves input of

48,000 ton/yr,
different parameters (such as alcohol/oil ratio, temperature, catalyst
29.99 ton/yr

6000 ton/yr
1094.5 tone

(0.19 $/kg)
Catalyst Qt

(0.25$/kg)

3,024,000
(0.27$/L)

concentration, and time). The output techno-economic matrices, which

include ROI, net present value (NPV), payback time (PBT), and PC) are 102,000
(cost)

10 %
1%

5%

5%

evaluated for economic feasibility and model predictability. The ANFIS

1,

8,
model showed higher accuracy than ANN with RMSE values for ROI
(7.39 × 10-7), PBT (5.32 × 10-7), NPV (5.89 × 10-7), and PC (5.89 × 10-
Oil: solvent

1:7.51200
(cost/ton)

4:61 (300

3:22 (300

4:61 (249

1:121200

7
Biodiesel production cost, O and P = operation and mantainence.
). It also suggests that ANFIS and ANN models could effectively predict
($/ton)

($/ton)
$/ton)

$/ton)

$/ton)

techno-economic parameters [153]. NPV is considered a vital output

1:4

1:4

1:4

parameter among economic matrices due to its comprehensive nature in

assessing a project’s financial viability over time [154]. The NPV metric
temperature

is sensitive to changes in biodiesel costs, feedstock costs, and operational

Reaction

efficiency, making it a more reliable measure than ROI and PBT. A

60 ◦ C

80 ◦ C

60 ◦ C

60 ◦ C

60 C

50 ◦ C

53 ◦ C

65 ◦ C
◦

positive NPV shows financial feasibility, as demonstrated in case studies

where biodiesel projects were deemed viable based on NPV calculations
[155,156]. The current cost of biodiesel is still higher compared to pure
Feedstock cost Catalyst

H2SO4

H2SO4

Lipase
NaOH

NaOH

Lime
KOH

KOH

diesel. However, the environmental aspect (pollution emission control)

type

cannot be ignored.

Government policy and research development

0.380/kg

0.380/kg
0.75/kg

0.25/kg

0.75/kg

0.36/L

0.36/L

0.36/L
(US$)

The high cost of biodiesel and limited production capacity are major
challenges to industry expansion [157]. Government policies (such as
loan assurances, public investments in technological companies, and
vegetable
feedstocks
Available

non-repayable subsidies) can improve the economic viability of bio­

Table 6

Virgin

WCO

WCO

WCO

WCO

WCO

diesel. Biodiesel blending policies (such as B5, B10, B20, and B35) in­
oil

SCO

SCO

crease market participation and lower oil prices [158]. Tax reductions

15
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

have proven effective in reducing biodiesel costs [159]. These subsidies References
are short-term strategies to encourage market entry, research funding is
crucial for long-term viability. The research grant aims to reduce high [1] Jin X, Li S, Ye H, Wang J, Wu Y, Zhang D, et al. Investigation and optimization of
biodiesel production based on multiple machine learning technologies. Fuel
biodiesel costs by investigating new non-edible feedstock with high oil 2023:128546. https://doi.org/10.1016/j.fuel.2023.128546.
content. Additional research can improve oil yield through nutrient [2] He X, Wang N, Zhou Q, Huang J, Ramakrishna S, Li F. Smart aviation biofuel
optimization, crop management, and the development of stress-tolerant energy system coupling with machine learning technology. Renew Sustain Energy
Rev 2024:113914. https://doi.org/10.1016/j.rser.2023.113914.
varieties. Developing advanced production technologies to produce [3] Liu Y, Sayed BT, Sivaraman R, Alshahrani SM, Venkatesan K, Thajudeen KY, et al.
biodiesel with improved quality and efficiency [160]. The combined Novel and robust machine learning model to optimize biodiesel production from
efforts of government policy and research funding are expected to algal oil using CaO and CaO/Al2O3 as catalyst: Sustainable green energy. Environ
Technol Innov 2023:103018. https://doi.org/10.1016/j.eti.2023.103018.
decrease the biodiesel price and make it a more competitive fuel option. [4] Xing Y, Zheng Z, Sun Y, Agha Alikhani M. A review on machine learning
application in biodiesel production studies. Int J Chem Eng 2021:1–12. https://
Conclusion doi.org/10.1155/2021/2154258.
[5] Naveed MH, Khan MNA, Mukarram M, Naqvi SR, Abdullah A, Haq ZU, et al.
Cellulosic biomass fermentation for biofuel production: Review of artificial
The biodiesel production process is improved through ML and AI intelligence approaches. Renew Sustain Energy Rev 2024:113906. https://doi.
model applications. Feedstock variation (including FAs, FFAs, and org/10.1016/j.rser.2023.113906.
moisture content) and operational parameters (such as catalysts, tem­ [6] Sukpancharoen S, Katongtung T, Rattanachoung N, Tippayawong N. Unlocking
the potential of transesterification catalysts for biodiesel production through
perature, alcohol oil ratio, and reaction time) effectively influenced machine learning approach. Bioresour Technol 2023:128961. https://doi.org/
biodiesel yield. The complex relation of reaction parameters is opti­ 10.1016/j.biortech.2023.128961.
mized and predicted for biodiesel yield and fuel properties using [7] Osman AI, Nasr M, Farghali M, Rashwan AK, Abdelkader A, Al-Muhtaseb AaH,
Ihara I, Rooney DW. Optimizing biodiesel production from waste with
advanced ML models (such as ANN, RF, SVR, XGB, RSM, and ANFIS) computational chemistry, machine learning and policy insights: a review.
with R2 = 0.90). Feedstock selection and catalyst type significantly Environ Chem Lett 2024:1005–71. https://doi.org/10.1007/s10311-024-01700-
impact the final biodiesel. Integrating IoT and big data technologies y.
[8] Shelare SD, Belkhode PN, Nikam KC, Jathar LD, Shahapurkar K, Soudagar MEM,
enhances production efficiency and process optimization and facilitates et al. Biofuels for a sustainable future: Examining the role of nano-additives,
real-time monitoring. Different feedstocks accurately predicted the economics, policy, internet of things, artificial intelligence and machine learning
techno-economic cost for biodiesel, and the production costs could be technology in biodiesel production. Energy 2023:128874. https://doi.org/
10.1016/j.energy.2023.128874.
decreased by using WCO, animal fats and efficient catalysts. The ANN [9] Haryanto A, Telaumbanua M. Application of artificial neural network to predict
and ANFIS models are comparatively more suited to biodiesel produc­ biodiesel yield from waste frying oil transesterification. Indonesian. J Sci Technol
tion. However, efficient hybrid models and better data availability can 2020:62–74. https://ejournal.kjpupi.id/index.php/ijost/article/view/111.
[10] Kavitha S, Ravi YK, Kumar G, Nandabalan YK. Microalgal biorefineries:
further enhance the predictive capabilities of ML applications in bio­
Advancement in machine learning tools for sustainable biofuel production and
diesel production. value-added products recovery. J Environ Manage 2024:120135. https://doi.org/
10.1016/j.jenvman.2024.120135.
CRediT authorship contribution statement [11] Uslu S. Optimization of diesel engine operating parameters fueled with palm oil-
diesel blend: Comparative evaluation between response surface methodology
(RSM) and artificial neural network (ANN). Fuel 2020:117990. https://doi.org/
Muhammad Arif: Conceptualization, Visualization, Formal anal­ 10.1016/j.energy.2023.129110.
ysis, Writing original draft, Investigation, Data curation. Adel I. Ala­ [12] Varuvel EG, Seetharaman S, Joseph Shobana Bai FJ, Devarajan Y,
Balasubramanian D. Development of artificial neural network and response
lawy: Writing, review & editing, Visualization, Investigation. surface methodology model to optimize the engine parameters of rubber seed oil
Yuanzhang Zheng: Visualization, Investigation, Validation. Mostafa – Hydrogen on PCCI operation. Energy 2023:129110. https://doi.org/10.1016/j.
Koutb: Investigation, Validation, Visualization, Writing, review & energy.2023.129110.
[13] Vinoth Arul Raj J, Praveen Kumar R, Vijayakumar B, Gnansounou E,
editing. Tareq Kareri: Data curation, Writing, review & editing, Bharathiraja B. Modelling and process optimization for biodiesel production from
Investigation. El-Sayed Salama: Conceptualization, Supervision, Nannochloropsis salina using artificial neural network. Bioresour Technol 2021:
Writing, review & editing, Resources, Project administration, Funding 124872. https://doi.org/10.1016/j.biortech.2021.124872.
[14] Gupta N, De R, Kodamana H, Bhartiya S. Batch-to-batch adaptive iterative
acquisition, Formal analysis, Validation. Xiangkai Li: Conceptualiza­ learning control-explicit model predictive control two-tier framework for the
tion, Supervision, Writing review & editing, Formal analysis, Data control of batch transesterification process. ACS Omega 2022:41001–12. https://
curation. doi.org/10.1021/acsomega.2c04255.
[15] Panahi HKS, Dehhaghi M, Guillemin GJ, Gupta VK, Aghbashlo M, Tabatabaei M.
Biodiesel production systems: real-world reactor technologies and processes.
Declaration of competing interest Sustainable Biodiesel 2024:91–118. https://doi.org/10.1016/B978-0-12-820361-
3.00004-8.
The authors declare that they have no known competing financial [16] Jha SK, Bilalovic J, Jha A, Patel N, Zhang H. Renewable energy: Present research
and future scope of Artificial Intelligence. Renew Sustain Energy Rev 2017:
interests or personal relationships that could have appeared to influence 297–317. https://doi.org/10.1016/j.rser.2017.04.018.
the work reported in this paper. [17] Said Z, Sharma P, Thi Bich Nhuong Q, Bora BJ, Lichtfouse E, Khalid HM, Luque R,
Nguyen XP, Hoang AT. Intelligent approaches for sustainable management and
valorisation of food waste. Bioresour Technol 2023:128952. https://doi.org/
Acknowledgments 10.1016/j.biortech.2023.128952.
[18] Cui Z, Huang S, Wang M, Nie K, Fang Y, Tan T. Improving the CFPP property of
This work is supported by the Fundamental Research Funds for the biodiesel via composition design: An intelligent raw material selection strategy
based on different machine learning algorithms. Renew Energy 2021:354–63.
Central Universities (lzujbky-2024-ey12). We sincerely thank the au­ https://doi.org/10.1016/j.renene.2021.02.008.
thors and publishers of the articles from which the Figures and [19] Yaşar F. Comparision of fuel properties of biodiesel fuels produced from different
Tables were adapted for use in this article. oils to determine the most suitable feedstock type. Fuel 2020:116817. https://doi.
org/10.1016/j.fuel.2019.116817.
[20] Corral-Bobadilla M, Lostado-Lorza R, Sabando-Fraile C, Íñiguez-Macedo S. An
Data availability artificial intelligence approach to model and optimize biodiesel production from
waste cooking oil using life cycle assessment and market dynamics analysis.
Energy 2024:132712. https://doi.org/10.1016/j.energy.2024.132712.
No data was used for the research described in the article.
[21] Najeeb J, Akram S, Mumtaz MW, Danish M, Irfan A, Touqeer T, et al. Catalysts
2021;171. https://doi.org/10.3390/catal11020171.
[22] Mata TM, Cardoso N, Ornelas M, Neves S, Caetano NS. Evaluation of two
purification methods of biodiesel from beef tallow, pork lard, and chicken fat.
Energy Fuel 2011:4756–62. https://doi.org/10.1021/ef2010207.
[23] Munir M, Ahmad M, Rehan M, Saeed M, Lam SS, Nizami AS, et al. Production of
high quality biodiesel from novel non-edible Raphnus raphanistrum L. seed oil

16
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

using copper modified montmorillonite clay catalyst. Environ Res 2021:110398. [48] Jahirul M, Rasul M, Brown R, Senadeera W, Hosen M, Haque R, et al.
https://doi.org/10.1016/j.envres.2020.110398. Investigation of correlation between chemical composition and properties of
[24] Anwar M. Biodiesel feedstocks selection strategies based on economic, technical, biodiesel using principal component analysis (PCA) and artificial neural network
and sustainable aspects. Fuel 2021:119204. https://doi.org/10.1016/j. (ANN). Renew Energy 2021:632–46. https://doi.org/10.1016/j.
fuel.2020.119204. renene.2020.12.078.
[25] Habib MS, Tayyab M, Zahoor S, Sarkar B. Management of animal fat-based [49] Mokhtar M, Sukmono A, Setiapraja H, Ma’ruf M, Yubaidah S, Haryono I,
biodiesel supply chain under the paradigm of sustainability. Energ Conver Rochmanto B, Soewono RT, Adhi Sukra KF, Thahar A. Towards nationwide
Manage 2020:113345. implementation of 40% biodiesel blend fuel in Indonesia: A comprehensive road
[26] ul ain Rana Q, Rehman MLU, Irfan M, Ahmed S, Hasan F, Shah AA, Khan S, test and laboratory evaluation. Biofuel Res J 2023:1876–89. https://doi.org/
Badshah M. Lipolytic bacterial strains mediated transesterification of non-edible 10.18331/BRJ2023.10.3.2.
plant oils for generation of high quality biodiesel. J Biosci Bioeng 2019:609–17. [50] Adhikesavan C, Ganesh D, Charles Augustin V. Effect of quality of waste cooking
https://doi.org/10.1016/j.jbiosc.2018.11.001. oil on the properties of biodiesel, engine performance and emissions. Cleaner
[27] Sharma P, Usman M, Salama E-S, Redina M, Thakur N, Li X. Evaluation of various Chem Eng 2022:100070. https://doi.org/10.1016/j.clce.2022.100070.
waste cooking oils for biodiesel production: A comprehensive analysis of [51] McCormick RL. Assessment of BQ-9000 biodiesel properties for 2022. Golden, CO
feedstock. Waste Manag 2021:219–29. https://doi.org/10.1016/j. (United States): National Renewable Energy Laboratory (NREL); 2023.
wasman.2021.10.022. [52] Figueiredo EdS, Vieira EdC, D’Elia E. Comparing a novel voltammetric method
[28] Singh D, Sharma D, Soni SL, Sharma S, Kumar Sharma P, Jhalani A. A review on with a standardized method for quality control of biodiesel. J Ind Eng Chem
feedstocks, production processes, and yield for different generations of biodiesel. 2015:353–61. https://doi.org/10.1016/j.jiec.2014.09.003.
Fuel 2020:116553. https://doi.org/10.1016/j.fuel.2019.116553. [53] Brahma S, Nath B, Basumatary B, Das B, Saikia P, Patir K, et al. Biodiesel
[29] Rambabu K, Bharath G, Hai A, Show PL, Banat F, Manickam S. Enhancing production from mixed oils: A sustainable approach towards industrial biofuel
biodiesel production from waste date seed oil through ultrasonic-assisted production. Chemical Engineering Journal Advances 2022:100284. https://doi.
optimization: a sustainable approach using non-edible feedstocks. Chemical org/10.1016/j.ceja.2022.100284.
Engineering and Processing-Process Intensification 2023:109601. https://doi. [54] Pölczmann G, Tóth O, Beck Á, Hancsók J. Investigation of storage stability of
org/10.1016/j.cep.2023.109601. diesel fuels containing biodiesel produced from waste cooking oil. J Clean Prod
[30] Rajasozhaperumal G, Kannan C. Influence of fatty acid composition on the 2016:85–92. https://doi.org/10.1016/j.jclepro.2015.08.035.
tribological performance of methyl esters under boundary lubrication regime. [55] Tsoutsos T, Tournaki S, Gkouskos Z, Paraíba O, Giglio F, García PQ, et al. Quality
Arab J Sci Eng 2023:3581–97. https://doi.org/10.1016/j.cep.2023.109601. characteristics of biodiesel produced from used cooking oil in southern europe.
[31] Guo L, Xie W, Gao C. Heterogeneous H6PV3MoW8O40/AC-Ag catalyst for ChemEngineering 2019;19. https://doi.org/10.3390/chemengineering3010019.
biodiesel production: Preparation, characterization and catalytic performance. [56] Cardeño F, Lapuerta M, Rios L, Agudelo JR. Reconsideration of regulated
Fuel 2022:123352. https://doi.org/10.1016/j.fuel.2022.123352. contamination limits to improve filterability of biodiesel and blends with diesel
[32] Maniam GP, Lim HS, Hussin NM. Effect of free fatty acid on transesterification of fuels. Renew Energy 2020:1243–51. https://doi.org/10.1016/j.
waste cooking oil. Current Science and Technology 2023:57–63. renene.2020.06.079.
[33] Zhang P, Chang Z, Wang D, Blamo BJ, Li W. Effect of water content on product [57] Komariah LN, Arita S, Prianda BE, Dewi TK. Technical assessment of biodiesel
distribution of base-catalyzed transesterification. Waste Biomass Valoriz 2016: storage tank; A corrosion case study. Journal of King Saud University - E
95–102. https://doi.org/10.1007/s12649-016-9584-3. Engineering Sciences 2023:232–7. https://doi.org/10.1016/j.
[34] Nahas L, Dahdah E, Aouad S, El Khoury B, Gennequin C, Abi Aad E, et al. Highly jksues.2021.03.016.
efficient scallop seashell-derived catalyst for biodiesel production from sunflower [58] Chandran D, Raviadaran R, Lau HLN, Numan A, Elumalai P, Samuel OD.
and waste cooking oils: Reaction kinetics and effect of calcination temperature Corrosion characteristic of stainless steel and galvanized steel in water emulsified
studies. Renew Energy 2023:1086–95. https://doi.org/10.1016/j. diesel, diesel and palm biodiesel. Eng Fail Anal 2023:107129. https://doi.org/
renene.2022.12.020. 10.1016/j.engfailanal.2023.107129.
[35] Oyekunle DT, Barasa M, Gendy EA, Tiong SK. Heterogeneous catalytic [59] Dugala NS, Goindi GS, Sharma A. Evaluation of physicochemical characteristics
transesterification for biodiesel production: Feedstock properties, catalysts and of Mahua (Madhuca indica) and Jatropha (Jatropha curcas) dual biodiesel blends
process parameters. Process Saf Environ Prot 2023. https://doi.org/10.1016/j. with diesel. Journal of King Saud University-Engineering Sciences 2021:424–36.
psep.2023.07.064. https://doi.org/10.1016/j.jksues.2020.05.006.
[36] Slamet S, Dewi ASC. Synthesis of biodiesel from waste cooking oil using [60] Tabish A. Corrosion behaviour of biofuel. petroleum and chemical industry.
heterogeneous CaO catalyst: Effect of stirring speed. AIP Conference Proceedings. International 2018:1–19.
AIP Publishing; 2018. 10.1063/1.5064346. [61] Longanesi L, Pereira AP, Johnston N, Chuck CJ. Oxidative stability of biodiesel:
[37] Andreo-Martinez P, Ortiz-Martinez VM, Salar-Garcia MJ, Veiga-del-Bano JM, recent insights. Biofuels Bioprod Biorefin 2022:265–89. https://doi.org/10.1002/
Chica A, Quesada-Medina J. Waste animal fats as feedstock for biodiesel bbb.2306.
production using non-catalytic supercritical alcohol transesterification: A [62] Setyawardhani DA, Ammar T, Ammar Y. Tertiary butylhydroquinone influence
perspective by the PRISMA methodology. Energy Sustain Dev 2022:150–63. over oxidation stability of biodiesel from waste cooking oil. Jurnal Rekayasa
https://doi.org/10.1016/j.esd.2022.06.004. Proses 2022:1–6. https://doi.org/10.22146/jrekpros.67177.
[38] Kurczyński D, Wcisło G, Le{\acute{s}}niak A, Kozak M, Łagowski P. Production [63] V. Barbosa Dos Santos, A. Moreno Ferreira Dos Santos, J.R. da Silva Cabral de
and testing of butyl and methyl esters as new generation biodiesels from fatty Moraes, I.C. de Oliveira Vieira, G. de Souza Rolim. Machine learning algorithms
wastes of the leather industry. Energies 2022:8744. https://doi.org/10.3390/ for soybean yield forecasting in the Brazilian Cerrado. J Sci Food Agric. 2022;
en15228744. 3665-3672. doi: 10.1002/jsfa.11713.
[39] Monika S, Banga VVP. Biodiesel production from waste cooking oil: A [64] Kumar S, Singhal MK, Sharma MP. Predictability of biodiesel fuel properties from
comprehensive review on the application of heterogenous catalysts. Energy Nexus the fatty acid composition of the feedstock oils. Arab J Sci Eng 2021:1–21.
2023:100209. https://doi.org/10.1016/j.nexus.2023.100209. https://doi.org/10.1007/s13369-021-05564-0.
[40] Catarino M, Ferreira E, Dias APS, Gomes J. Dry washing biodiesel purification [65] Sharma V, Tsai ML, Chen CW, Sun PP, Nargotra P, Dong CD. Advances in machine
using fumed silica sorbent. Chem Eng J 2020:123930. https://doi.org/10.1016/j. learning technology for sustainable biofuel production systems in lignocellulosic
cej.2019.123930. biorefineries. Sci Total Environ 2023:163972. https://doi.org/10.1016/j.
[41] Eldiehy KSH, Bardhan P, Borah D, Gohain M, Ahmad Rather M, Deka D, et al. scitotenv.2023.163972.
A comprehensive review on microalgal biomass production and processing for [66] Mishra H, Mishra D. Artificial intelligence and machine learning in agriculture:
biodiesel production. Fuel 2022:124773. https://doi.org/10.1016/j. transforming farming systems. research trends in agriculture. Science 2023:1–16.
fuel.2022.124773. [67] Liakos KG, Busato P, Moshou D, Pearson S, Bochtis D. Machine learning in
[42] Miyuranga KV, Arachchige US, Jayasinghe RA, Samarakoon G. Purification of agriculture: a review. Sensors (Basel) 2018:2674. https://doi.org/10.3390/
residual glycerol from biodiesel production as a value-added raw material for s18082674.
glycerolysis of free fatty acids in waste cooking oil. Energies 2022:8856. https:// [68] Alabi TR, Abebe AT, Chigeza G, Fowobaje KR. Estimation of soybean grain yield
doi.org/10.1016/j.techfore.2021.120784. from multispectral high-resolution UAV data with machine learning models in
[43] Kılıçarslan MN, Argun ME. Holistic effect of operating conditions and purification West Africa. Remote Sens Appl: Soc Environ 2022:100782. https://doi.org/
methods on the pollution characteristics of washing water and biodiesel quality. 10.1016/j.rsase.2022.100782.
J Water Process Eng 2023:103989. https://doi.org/10.1016/j. [69] Rajković D, Marjanović Jeromela A, Pezo L, Lončar B, Zanetti F, Monti A, et al.
jwpe.2023.103989. Yield and quality prediction of winter rapeseed—Artificial neural network and
[44] Šalić A, Kučan KZ, Gojun M, Rogošić M, Zelić B. Biodiesel purification: real-world random forest models. Agronomy 2021;58. https://doi.org/10.3390/
examples, case studies, and current limitations. Sustainable Biodiesel 2024: agronomy12010058.
185–237. https://doi.org/10.1016/B978-0-12-820361-3.00005-X. [70] Niazian M, Sadat-Noori SA, Abdipour M. Artificial neural network and multiple
[45] Pasae Y, Salla Y, Bulo L. Properties of biodiesel purified by membrane technology. regression analysis models to predict essential oil content of ajowan (Carum
Asian Journal of Applied Chemistry Research 2020:21–7. https://doi.org/ copticum L.). J Appl Res Med Aromat Plants 2018:124–31. https://doi.org/
10.9734/ajacr/2020/v5i430142. 10.1016/j.jarmap.2018.04.001.
[46] Khujamberdiev R, Cho HM. Biodiesel raw material characterization and its [71] O. Akande, J.A. Okolie, R. Kimera, C.C. Ogbaga. A Comprehensive Review on
production process-a review. Nveo-Natural Volatiles & Essential Oils Journal Deep Learning Applications in Advancing Biodiesel Feedstock Selection and
NVEO 2021:12112–20. Production Processes. Available at SSRN 4898486. 2024.doi: 10.2139/
[47] Dhabhai R, Koranian P, Huang Q, Scheibelhoffer DS, Dalai AK. Purification of ssrn.4898486.
glycerol and its conversion to value-added chemicals: A review. Sep Sci Technol
2023:1383–402. https://doi.org/10.1080/01496395.2023.2189054.

17
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

[72] Kumar S, Singhal MK, Sharma MP. Analysis of oil mixing for improvement of Mühendislik Bilimleri Dergisi 2023:1593–605. https://doi.org/10.28948/
biodiesel quality with the application of mixture design method. Renew Energy ngumuh.1342996.
2023:809–21. https://doi.org/10.1016/j.renene.2022.11.097. [96] Agrawal P, Gnanaprakash R, Dhawane SH. Prediction of biodiesel yield
[73] Fadhil AB, Al-Tikrity ET, Albadree MA. Biodiesel production from mixed non- employing machine learning: interpretability analysis via Shapley additive
edible oils, castor seed oil and waste fish oil. Fuel 2017:721–8. https://doi.org/ explanations. Fuel 2024:130516. https://doi.org/10.1016/j.fuel.2023.130516.
10.1016/j.renene.2022.11.097. [97] Huang Y, Li F, Bao G, Xiao Q, Wang H. Modeling the effects of biodiesel chemical
[74] Farghali M, Osman AI, Chen Z, Abdelhaleem A, Ihara I, Mohamed IM, et al. composition on iodine value using novel machine learning algorithm. Fuel 2022:
Social, environmental, and economic consequences of integrating renewable 123348. https://doi.org/10.1016/j.fuel.2022.123348.
energies in the electricity sector: a review. Environ Chem Lett 2023:1381–418. [98] Satya S, Kolakoti A, Rao R. Optimization of palm methyl ester and its effect on
https://doi.org/10.1007/s10311-023-01587-1. fatty acid compositions and cetane number. Mathematical models in engineering
[75] Sun X, Opulencia MJC, Alexandrovich TP, Khan A, Algarni M, Abdelrahman A. 2019:25–34. https://doi.org/10.21595/mme.2019.20469.
Modeling and optimization of vegetable oil biodiesel production with [99] Tan D, Suvarna M, Tan YS, Li J, Wang X. A three-step machine learning
heterogeneous nano catalytic process: Multi-layer perceptron, decision regression framework for energy profiling, activity state prediction and production
tree, and K-Nearest Neighbor methods. Environ Technol Innov 2022:102794. estimation in smart process manufacturing. Appl Energy 2021:116808. https://
https://doi.org/10.1016/j.eti.2022.102794. doi.org/10.1016/j.apenergy.2021.116808.
[76] Karimi M, Jenkins B, Stroeve P. Multi-objective optimization of transesterification [100] Hosseinpour S, Aghbashlo M, Tabatabaei M, Khalife E. Exact estimation of
in biodiesel production catalyzed by immobilized lipase. Biofuels Bioprod biodiesel cetane number (CN) from its fatty acid methyl esters (FAMEs) profile
Biorefin 2016:804–18. https://doi.org/10.1002/bbb.1706. using partial least square (PLS) adapted by artificial neural network (ANN). Energ
[77] Aghbashlo M, Hosseinpour S, Tabatabaei M, Soufiyan MM. Multi-objective Conver Manage 2016:389–98. https://doi.org/10.1016/j.
exergetic and technical optimization of a piezoelectric ultrasonic reactor applied enconman.2016.07.027.
to synthesize biodiesel from waste cooking oil (WCO) using soft computing [101] Giwa SO, Taziwa RT, Sharifpur M. Dependence of composition-based approaches
techniques. Fuel 2019:100–12. https://doi.org/10.1016/j.fuel.2018.07.095. on hybrid biodiesel fuel properties prediction using artificial neural network and
[78] Giwa SO, Aasa SA, Taziwa RT, Sharifpur M. Deploying artificial neural network to random tree algorithms. Renew Energy 2023:119324.
predict hybrid biodiesel fuel properties from their fatty acid compositions. Int J [102] D. Baron. Machine learning in astronomy: A practical overview. arXiv preprint
Ambient Energy 2024:2262466. https://doi.org/10.1080/ arXiv:1904.07248. 2019. doi: 10.1016/j.renene.2023.119324.
01430750.2023.2262466. [103] Pusdiktasari ZF, Fitriani R, Sumarminingsih E. Simulation study using average
[79] Mowbray M, Savage T, Wu C, Song Z, Cho BA, Del Rio-Chanona EA, et al. difference algorithm to analyze the outlierness degree of spatial observations. In:
Machine learning for biochemical engineering: A review. Biochem Eng J 2021: 2020 4th International Conference on Informatics and Computational Sciences
108054. https://doi.org/10.1016/j.bej.2021.108054. (ICICoS); 2020. p. 1–6. https://doi.org/10.1109/icicos51170.2020.9298999.
[80] Mondal PP, Galodha A, Verma VK, Singh V, Show PL, Awasthi MK, et al. Review [104] Ozsahin DU, Mustapha MT, Mubarak AS, Ameen ZS, Uzun B. Impact of feature
on machine learning-based bioprocess optimization, monitoring, and control scaling on machine learning models for the diagnosis of diabetes. In: 2022
systems. Bioresour Technol 2023:128523. https://doi.org/10.1016/j. International Conference on Artificial Intelligence in Everything (AIE); 2022.
biortech.2022.128523. p. 87–94. https://doi.org/10.1109/aie57029.2022.00024.
[81] I. Walsh, M. Myint, T. Nguyen-Khuong, Y.S. Ho, S.K. Ng, M. Lakshmanan, [105] R.S. Freitas, Á.P. Lima, C. Chen, F.A. Rochinha, D. Mira, X. Jiang. Prediction of
Harnessing the potential of machine learning for advancing “quality by design” in liquid fuel properties using machine learning models with Gaussian processes and
biomanufacturing, MAbs, Taylor & Francis, 2022, p. 2013593. doi: 10.1080/ probabilistic conditional generative learning. arXiv preprint arXiv:2110.09360.
19420862.2021.2013593. 2021. doi: 10.48550/arXiv.2110.09360.
[82] Steinwandter V, Borchert D, Herwig C. Data science tools and applications on the [106] Sharma P, Said Z, Kumar A, Nizetic S, Pandey A, Hoang AT, et al. Recent advances
way to Pharma 4.0. Drug Discov Today 2019:1795–805. https://doi.org/ in machine learning research for nanofluid-based heat transfer in renewable
10.1016/j.drudis.2019.06.005. energy system. Energy Fuel 2022:6626–58. https://doi.org/10.1021/acs.
[83] Wali WA, Hassan KH, Cullen JD, Shaw A, Al-Shamma’a AI. Real time monitoring energyfuels.2c01006.
and intelligent control for novel advanced microwave biodiesel reactor. [107] Elango RK, Sathiasivan K, Muthukumaran C, Thangavelu V, Rajesh M,
Measurement 2013:823–39. https://doi.org/10.1016/j. Tamilarasan K. Transesterification of castor oil for biodiesel production: Process
measurement.2012.10.004. optimization and characterization. Microchem J 2019:1162–8. https://doi.org/
[84] Naderloo L. Energy ratio of produced biodiesel in hydrodynamic cavitation 10.1016/j.microc.2018.12.039.
reactor equipped with LabVIEW controller and artificial intelligence. Energy Rep [108] Galvan D, Cremasco H, Mantovani ACG, Bona E, Killner M, Borsato D. Kinetic
2020:1456–67. https://doi.org/10.1016/j.egyr.2020.05.029. study of the transesterification reaction by artificial neural networks and
[85] Wali WA, Al-Shamma’a AI, Hassan KH, Cullen JD. Online genetic-ANFIS parametric particle swarm optimization. Fuel 2020:117221. https://doi.org/
temperature control for advanced microwave biodiesel reactor. J Process Control 10.1016/j.fuel.2020.117221.
2012:1256–72. https://doi.org/10.1016/j.jprocont.2012.05.013. [109] Liao M, Yao Y. Applications of artificial intelligence-based modeling for bioenergy
[86] Galvan D, Cremasco H, Gomes Mantovani AC, Bona E, Killner M, Borsato D. systems: A review. GCB Bioenergy 2021:774–802. https://doi.org/10.1111/
Kinetic study of the transesterification reaction by artificial neural networks and gcbb.12816.
parametric particle swarm optimization. Fuel 2020:117221. https://doi.org/ [110] Ajao J, Akanbi O, Akinwole I, Adedokun A, Ogunsakin O, Ige A. Research
10.1016/j.fuel.2020.117221. advances on machine learning technologies for enhanced biodiesel production: a
[87] Nasim MAG, Khan O, Parvez M, Bhatt BK. Optimizing ultrasonic reactor comprehensive review. Pet Coal 2024.
operating variables using intelligent soft computing models for increased [111] Yusuff AS, Ishola NB, Gbadamosi AO, Azeez TM, Onibonoje MO. An artificial
biodiesel production. Green Technol Sustainability 2023:100033. https://doi. intelligence approach to model and optimize biodiesel production from used
org/10.1016/j.grets.2023.100033. cooking oil using CaO incorporated zeolite catalyst. Energy Convers Manage: X
[88] Alviso D, Artana G, Duriez T. Prediction of biodiesel physico-chemical properties 2023:100452. https://doi.org/10.1016/j.ecmx.2023.100452.
from its fatty acid composition using genetic programming. Fuel 2020:116844. [112] Samuel OD, Okwu MO, Oyejide OJ, Taghinezhad E, Afzal A, Kaveh M. Optimizing
[89] Cunha CL, Torres AR, Luna AS. Multivariate regression models obtained from biodiesel production from abundant waste oils through empirical method and
near-infrared spectroscopy data for prediction of the physical properties of grey wolf optimizer. Fuel 2020:118701. https://doi.org/10.1016/j.
biodiesel and its blends. Fuel 2020:116344. https://doi.org/10.1016/j. fuel.2020.118701.
fuel.2019.116344. [113] Veza I, Afzal A, Mujtaba MA, Tuan Hoang A, Balasubramanian D, Sekar M, et al.
[90] Shahbaz K, Baroutian S, Mjalli FS, Hashim MA, AlNashef IM. Prediction of Review of artificial neural networks for gasoline, diesel and homogeneous charge
glycerol removal from biodiesel using ammonium and phosphunium based deep compression ignition engine. Alex Eng J 2022:8363–91. https://doi.org/10.1016/
eutectic solvents using artificial intelligence techniques. Chemom Intel Lab Syst j.aej.2022.01.072.
2012:193–9. https://doi.org/10.1016/j.chemolab.2012.06.005. [114] Pomeroy B, Grilc M, Likozar B. Artificial neural networks for bio-based chemical
[91] Bukkarapu KR, Krishnasamy A. Biodiesel composition based machine learning production or biorefining: A review. Renew Sustain Energy Rev 2022:111748.
approaches to predict engine fuel properties. Proceedings of the Institution of https://doi.org/10.1016/j.rser.2021.111748.
Mechanical Engineers, Part D: Journal of Automobile Engineering 2024:1844–60. [115] Ahmad T, Madonski R, Zhang D, Huang C, Mujeeb A. Data-driven probabilistic
https://doi.org/10.1016/j.rser.2021.111925. machine learning in sustainable smart energy/smart energy systems: Key
[92] Kumbhar V, Pandey A, Sonawane CR, El-Shafay A, Panchal H, Chamkha AJ. developments, challenges, and future research opportunities in the context of
Statistical analysis on prediction of biodiesel properties from its fatty acid smart grid paradigm. Renew Sustain Energy Rev 2022:112128. https://doi.org/
composition. Case Stud Therm Eng 2022:101775. https://doi.org/10.1016/j. 10.1016/j.rser.2022.112128.
csite.2022.101775. [116] Mosavi A, Bahmani A. Energy consumption prediction using machine learning. a
[93] Suvarna M, Jahirul MI, Aaron-Yeap WH, Augustine CV, Umesh A, Rasul MG, et al. review 2019. https://doi.org/10.20944/PREPRINTS201903.0131.
Predicting biodiesel properties and its optimal fatty acid profile via explainable [117] Esonye C, Augustine SC, Chukwu GE, Ude G. Adaptive neuro-fuzzy inference
machine learning. Renew Energy 2022:245–58. https://doi.org/10.1016/j. system-genetic algorithm versus response surface methodology-desirability
renene.2022.02.124. function algorithm modelling and optimization of biodiesel synthesis from waste
[94] Bukkarapu KR, Krishnasamy A. Biodiesel composition based machine learning chicken fat. J Taiwan Inst Chem Eng 2022:104389. https://doi.org/10.1016/j.
approaches to predict engine fuel properties. Proceedings of the Institution of jtice.2022.104389.
Mechanical Engineers, Part D: Journal of Automobile Engineering 2023:1844–60. [118] Zhou L, Pan S, Wang J, Vasilakos AV. Machine learning on big data: Opportunities
https://doi.org/10.1177/09544070231158240. and challenges. Neurocomputing 2017:350–61.
[95] Gülüm M. Prediction of cetane numbers of pure biodiesels based on various fuel [119] Meena M, Shubham S, Paritosh K, Pareek N, Vivekanand V. Production of biofuels
properties using multiple non-linear correlations. Ömer Halisdemir Üniversitesi from biomass: Predicting the energy employing artificial intelligence modelling.

18
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

Bioresour Technol 2021:125642. https://doi.org/10.1016/j. [143] Lopes MV, Barradas Filho AO, Barros AK, Viegas IMA, Silva LCO, Marques EP,
biortech.2021.125642. et al. Attesting compliance of biodiesel quality using composition data and
[120] Hoang AT, Nižetić S, Ong HC, Tarelko W, Le TH, Chau MQ, et al. A review on classification methods. Neural Comput & Applic 2019:539–51. https://doi.org/
application of artificial neural network (ANN) for performance and emission 10.1007/s00521-017-3087-4.
characteristics of diesel engine fueled with biodiesel-based fuels. Sustainable [144] Komariah LN, Arita S, Rendana M, Ramayanti C, Suriani NL, Erisna D. Microbial
Energy Technol Assess 2021:101416. https://doi.org/10.1016/j. contamination of diesel-biodiesel blends in storage tank; an analysis of colony
seta.2021.101416. morphology. Heliyon 2022:e09264.
[121] Sivaramakrishnan K, Tannous JH, Chandrasekaran V. Prediction of [145] Awogbemi O, Kallon DVV. Application of machine learning technologies in
thermogravimetric data for asphaltenes extracted from deasphalted oil using biodiesel production process—A review. Front Energy Res 2023:1122638.
machine learning techniques. Ind Eng Chem Res 2023. https://doi.org/10.1021/ https://doi.org/10.3389/fenrg.2023.1122638.
acs.iecr.3c01798. [146] Abd Manaf IS, Embong NH, Khazaai SNM, Rahim MHA, Yusoff MM, Lee KT, et al.
[122] Wang K, Khoo KS, Leong HY, Nagarajan D, Chew KW, Ting HY, et al. How does A review for key challenges of the development of biodiesel industry. Energ
the Internet of Things (IoT) help in microalgae biorefinery? Biotechnol Adv 2022: Conver Manage 2019:508–17. https://doi.org/10.1016/j.
107819. https://doi.org/10.1016/j.biotechadv.2021.107819. enconman.2019.02.019.
[123] Fabris M, Abbriano RM, Pernice M, Sutherland DL, Commault AS, Hall CC, et al. [147] Coşgun A, Günay ME, Yildirim R. Machine learning for algal biofuels: A critical
Emerging technologies in algal biotechnology: toward the establishment of a review and perspective for future. Green Chem 2023. https://doi.org/10.1039/
sustainable, algae-based bioeconomy. Front Plant Sci 2020:279. https://doi.org/ D3GC00389B.
10.3389/fpls.2020.00279. [148] Hupp AM, Perron J, Roques N, Crandall J, Ramos S, Rohrback B. Analysis of
[124] Loganathan G, Kannan M. Optimized production of biodiesel using internet of biodiesel-diesel blends using ultrafast gas chromatography (UFGC) and
things sensed temperature with hybrid particle swarm optimization. J Chem chemometric methods: Extending ASTM D7798 to biodiesel. Fuel 2018:264–70.
2022:3793739. https://doi.org/10.1155/2022/3793739. https://doi.org/10.1016/j.fuel.2018.05.102.
[125] D.B. KS, G. Adarsh, K. Karan, S.K. PB, S. Rupesh, R. Jayachandran, A Novel [149] Selvan SS, Pandian PS, Subathira A, Saravanan S. Artificial neural network
scheme for IoT based Real Time Monitoring of Biodiesel Quality, 2023 modeling-coupled genetic algorithm optimization of supercritical methanol
International Conference on Recent Trends in Electronics and Communication transesterification of Aegle marmelos oil to biodiesel. Biofuels 2021. https://doi.
(ICRTEC), IEEE, 2023, pp. 1-4. doi: 10.1109/icrtec56977.2023.10111864. org/10.1080/17597269.2018.1542567.
[126] Clauser NM, Felissia FE, Area MC, Vallejos ME. Integrating the new age of [150] Karmee SK, Patria RD, Lin CS. Techno-economic evaluation of biodiesel
bioeconomy and industry 4.0 into biorefinery process design. BioResources 2022; production from waste cooking oil–a case study of Hong Kong. Int J Mol Sci 2015:
5510–5531. https://doi.org/10.15376/biores.17.3.Clauser. 4362–71. https://doi.org/10.3390/ijms16034362.
[127] Singh SK, Chauhan A, Sarkar B. Resilience of sustainability for a smart production [151] Ishola NB, Epelle EI, Betiku E. Machine learning approaches to modeling and
system to produce biodiesel from waste animal fat. J Clean Prod 2024;142047. optimization of biodiesel production systems: State of art and future outlook.
https://doi.org/10.1016/j.jclepro.2024.142047. Energy Convers Manage: X 2024:100669. https://doi.org/10.1016/j.
[128] Meindl B, Ayala NF, Mendonça J, Frank AG. The four smarts of Industry 4.0: ecmx.2024.100669.
Evolution of ten years of research and future perspectives. Technol Forecast Soc [152] Oke EO, Okolo BI, Adeyi O, Adeyi JA, Ude CJ, Osoh K, et al. Process design,
Chang 2021;120784. https://doi.org/10.1016/j.techfore.2021.120784. techno-economic modelling, and uncertainty analysis of biodiesel production
[129] Wang J, Ma Y, Zhang L, Gao RX, Wu D. Deep learning for smart manufacturing: from palm kernel oil. Bioenergy Res 2022:1355–69. https://doi.org/10.1007/
Methods and applications. J Manuf Syst 2018;:144–56. https://doi.org/10.1016/ s12155-021-10315-y.
j.jmsy.2018.01.003. [153] Ajala OO, Oke EO, Odejobi OJ, Adeoye BK, Oyelade JO. Artificial neuro-fuzzy
[130] Selvan SS, Pandian PS, Subathira A, Saravanan S. Comparison of response surface intelligent prediction of techno-economic parameters of computer-aided scale-up
methodology (RSM) and artificial neural network (ANN) in optimization of aegle for palm kernel oil based biodiesel production. Cleaner Chem Eng 2023:100098.
marmelos oil extraction for biodiesel production. Arab J Sci Eng 2018:6119–31. https://doi.org/10.1016/j.clce.2023.100098.
https://doi.org/10.1007/s13369-018-3272-5. [154] Omidkar A, Alagumalai A, Li Z, Song H. Machine learning assisted techno-
[131] Mohadesi M, Rezaei A. Biodiesel conversion modeling under several conditions economic and life cycle assessment of organic solid waste upgrading under
using computational intelligence methods. Environ Prog Sustain Energy 2018: natural gas. Appl Energy 2024:122321. https://doi.org/10.1016/j.
562–8. https://doi.org/10.1002/ep.12698. apenergy.2023.122321.
[132] Wang X-M, Zeng Y-N, Wang Y-R, Wang F-P, Wang Y-T, Li J-G, et al. A novel [155] Xia Y, Tang Z-C. A novel perspective for techno-economic assessments and effects
strategy for efficient biodiesel production: Optimization, prediction, and of parameters on techno-economic assessments for biodiesel production under
mechanism. Renew Energy 2023:385–97. https://doi.org/10.1016/j. economic and technical uncertainties. RSC Adv 2017:9402–11. https://doi.org/
renene.2023.03.027. 10.1039/c6ra25754b.
[133] Ada E, Köksal MA. Modelling residential end-use electricity consumption using [156] Su C. Literature review on the net present value method of project investment
statistical and artificial intelligence approaches, energy efficiency in domestic decision. advances in economics. Management and Political Sciences 2024:60–6.
appliances and lighting. In: Proceedings of the 10th International Conference https://doi.org/10.54254/2754-1169/60/20231156.
(EEDAL’19); 2022. p. 171–87. https://doi.org/10.1007/978-3-030-79124-7-13. [157] T.S. Milessi, S.C.T. Tabuchi, T.D. Esteves, D.B. Hirata, R.S. Capaz, A.A. Mendes.
[134] Wali W, AI-Shamma’a A, Hassan KH, Cullen J. Artificial intelligent control Biodiesel production in oil biorefinery and by-products utilization. in: A.K.
techniques for nonlinear real time chemical reactor. International Journal of Chandel, F. Segato (Eds.), Production of Top 12 Biochemicals Selected by USDOE
Computer and Information Technology 2012:62–72. from Renewable Resources. Elsevier2022, pp. 109-150. doi: 10.1016/b978-0-12-
[135] Rouchi MB, Khorrami MK, Garmarudi AB, de la Guardia M. Application of 823531-7.00010-x.
infrared spectroscopy as Process Analytics Technology (PAT) approach in [158] Ebadian M, van Dyk S, McMillan JD, Saddler J. Biofuels policies that have
biodiesel production process utilizing Multivariate Curve Resolution Alternative encouraged their production and use: An international perspective. Energy Policy
Least Square (MCR-ALS). Spectrochim Acta A Mol Biomol Spectrosc 2019: 2020:111906. https://doi.org/10.1016/j.enpol.2020.111906.
347–53. https://doi.org/10.1016/j.saa.2019.01.046. [159] Julio AAV, Milessi TS, Batlle EAO, Lora EES, Maya DMY, Palacio JCE. Techno-
[136] López-Zapata BY, Adam-Medina M, Álvarez-Gutiérrez PE, Castillo-González JP, economic and environmental potential of Renewable Diesel as complementation
Hernández-de León HR, Vela-Valdés LG. Virtual sensors for biodiesel production for diesel and biodiesel in Brazil: A comprehensive review and perspectives.
in a batch reactor. Sustainability 2017:455. https://doi.org/10.3390/su9030455. J Clean Prod 2022:133431. https://doi.org/10.1016/j.jclepro.2022.133431.
[137] Sebayang AH, Kusumo F, Milano J, Shamsuddin AH, Silitonga AS, Ideris F, et al. [160] Wirawan SS, Solikhah MD, Setiapraja H, Sugiyono A. Biodiesel implementation in
Optimization of biodiesel production from rice bran oil by ultrasound and Indonesia: Experiences and future perspectives. Renew Sustain Energy Rev 2024:
infrared radiation using ANN-GWO. Fuel 2023:128404. https://doi.org/10.1016/ 113911. https://doi.org/10.1016/j.rser.2023.113911.
j.fuel.2023.128404. [161] B. Akkaya, N. Çolakoğlu. Comparison of multi-class classification algorithms on
[138] Abdelbasset WK, Alrawaili SM, Elsayed SH, Diana T, Ghazali S, Felemban BF, early diagnosis of heart diseases. 2019.
et al. Optimization of heterogeneous Catalyst-assisted fatty acid methyl esters [162] Hamolia V, Melnyk V, Zhezhnych P, Shilinh A. Intrusion detection in computer
biodiesel production from Soybean oil with different Machine learning methods. networks using latent space representation and machine learning. International
Arab J Chem 2022:103915. https://doi.org/10.1016/j.arabjc.2022.103915. Journal of Computing 2020:442–8. https://doi.org/10.1016/j.
[139] Rocabruno-Valdés C, González-Rodriguez J, Díaz-Blanco Y, Juantorena A, Muñoz- enconman.2020.113345.
Ledo J, El-Hamzaoui Y, et al. Corrosion rate prediction for metals in biodiesel [163] Ahmad Sobri MZ, Redhwan A, Ameen F, Lim JW, Liew CS, Mong GR, et al.
using artificial neural networks. Renew Energy 2019:592–601. https://doi.org/ A review unveiling various machine learning algorithms adopted for biohydrogen
10.1016/j.renene.2019.03.065. productions from microalgae. Fermentation 2023:243. https://doi.org/10.3390/
[140] Marichal GN, Avila D, Hernandez A, Padron I, Castejon C. Feature extraction from fermentation9030243.
indirect monitoring in marine oil separation systems. Sensors (Basel) 2018:3159. [164] Alsaleem F, Tesfay MK, Rafaie M, Sinkar K, Besarla D, Arunasalam P. An IoT
https://doi.org/10.3390/s18093159. framework for modeling and controlling thermal comfort in buildings. Front Built
[141] Ismail FB, Singh D, Nasif MS. Adoption of Intelligent computational techniques Environ 2020:87. https://doi.org/10.3389/fbuil.2020.00087.
for steam boilers tube leak trip. Malays J Comput Sci 2020:133–51. https://doi. [165] Sakunthala S, Kiranmayi R, Mandadi PN. A review on artificial intelligence
org/10.22452/mjcs.vol33no2.4. techniques in electrical drives: Neural networks, fuzzy logic, and genetic
[142] Ning Y, Niu S, Wang Y, Zhao J, Lu C. Sono-modified halloysite nanotube with algorithm. In: 2017 International Conference on Smart Technologies for Smart
NaAlO2 as novel heterogeneous catalyst for biodiesel production: optimization Nation (SmartTechCon); 2017. p. 11–6. https://doi.org/10.1109/
via GA_BP neural network. Renew Energy 2021:391–404. https://doi.org/ SmartTechCon.2017.8358335.
10.1016/j.renene.2021.04.135. [166] Ashouri R, Jafari D, Esfandyari M, Vatankhah G, Mahdavi M. Valorization of
slaughterhouse wastes through transesterification for sustainable biodiesel

19
M. Arif et al. Sustainable Energy Technologies and Assessments 73 (2025) 104097

production using potassium hydroxide as a heterogeneous catalyst. J Clean Prod Journal of Industrial Chemistry 2017:253–62. https://doi.org/10.1007/s40090-
2024:141596. https://doi.org/10.1016/j.jclepro.2024.141596. 017-0122-3.
[167] Suhel A, Abdul Rahim N, Abdul Rahman MR, Bin Ahmad KA, Teoh YH, Zainal [181] Silitonga AS, Mahlia TMI, Kusumo F, Dharma S, Sebayang AH, Sembiring RW,
Abidin N. An experimental investigation on the effect of ferrous ferric oxide nano- et al. Intensification of Reutealis trisperma biodiesel production using infrared
additive and chicken fat methyl ester on performance and emission characteristics radiation: Simulation, optimisation and validation. Renew Energy 2019:520–7.
of compression ignition engine. Symmetry 2021:265. https://doi.org/10.3390/ https://doi.org/10.1016/j.renene.2018.10.023.
sym13020265. [182] Ardabili SF, Najafi B, Alizamir M, Mosavi A, Shamshirband S, Rabczuk T. Using
[168] Takase M, Essandoh PK. Two-step biodiesel production using high free fatty acid SVM-RSM and ELM-RSM approaches for optimizing the production process of
containing pig fat. Int J Green Energy 2021:381–9. https://doi.org/10.1016/j. methyl and ethyl esters. Energies 2018. https://doi.org/10.3390/en11112889.
sajce.2023.01.004. [183] Kumar S, Jain S, Kumar H. Prediction of jatropha-algae biodiesel blend oil yield
[169] Kanthasamy P, Arul Mozhi Selvan V. FTIR and GCMS analysis on useful methyl with the application of artificial neural networks technique. Energy Sources Part
ester compound from industrial waste animal fleshing oil (WAFO). Materials A 2019:1285–95. https://doi.org/10.1080/15567036.2018.1548507.
Today. Proceedings 2021:10072–8. https://doi.org/10.1016/j. [184] Liu Z, Baghban A. Application of LSSVM for biodiesel production using
matpr.2021.06.255. supercritical ethanol solvent. Energy Sources Part A 2017:1869–74. https://doi.
[170] Parida S, Pali HS, Chaturvedi A, Sharma A, Balasubramanian D, Ramegouda R, org/10.1080/15567036.2017.1380732.
et al. Production of biodiesel from waste fish fat through ultrasound-assisted [185] Guo J, Baghban A. Application of ANFIS strategy for prediction of biodiesel
transesterification using petro-diesel as cosolvent and optimization of process production using supercritical methanol. Energy Sources Part A 2017:1862–8.
parameters using response surface methodology. Environ Sci Pollut Res 2024: https://doi.org/10.1080/15567036.2017.1380731.
25524–37. https://doi.org/10.1007/s11356-024-32702-6. [186] Sivakumar R, Gokulsankar S, Aravazhi T, Baskar G, Maheswaran S. Investigation
[171] Shafiq F, Mumtaz MW, Mukhtar H, Touqeer T, Raza SA, Rashid U, et al. Response on optimization of biodiesel production using machine learning techniques. In:
surface methodology approach for optimized biodiesel production from waste 2024 2nd International Conference on Artificial Intelligence and Machine
chicken fat oil. Catalysts 2020:633. https://doi.org/10.3390/catal10060633. Learning Applications Theme: Healthcare and Internet of Things (AIMLA); 2024.
[172] Ejeromedoghene O. Acid-catalyzed transesterification of Palm Kernel Oil (PKO) to p. 1–7. https://doi.org/10.1109/AIMLA59606.2024.10531481.
biodiesel. Mater Today Proc 2021:1580–3. https://doi.org/10.1016/j. [187] Agu CM, Orakwue CC, Menkiti MC, Agulanna AC, Akaeme FC. RSM/ANN based
matpr.2021.04.042. modeling of methyl esters yield from Anacardium occidentale kernel oil by
[173] Wang X, Dai M, Xie Y, Han J, Ma Y, Chen C. Experimental investigation of transesterification, for possible application as transformer fluid. Curr Res Green
evaporation characteristics of biodiesel-diesel blend droplets with carbon Sustainable Chem 2022:100255. https://doi.org/10.1016/j.crgsc.2021.100255.
nanotubes and nanoceria as nanoadditives. Appl Surf Sci 2020:144186. https:// [188] Paul AK, Borugadda VB, Bhalerao MS, Goud VV. In situ epoxidation of waste
doi.org/10.1016/j.renene.2023.03.027. soybean cooking oil for synthesis of biolubricant basestock: a process parameter
[174] Anwar F, Tariq M, Nisar J, Ali G, Kanwal H. Optimization of biodiesel yield from optimization and comparison with RSM, ANN, and GA. Can J Chem Eng 2018:
non-food karanja seed oil: Characterization and assessment of fuel properties. 1451–61. https://doi.org/10.1002/cjce.23091.
Sustainable Chemistry for the Environment 2023:100035. https://doi.org/ [189] Mostafaei M. Prediction of biodiesel fuel properties from its fatty acids
10.1016/j.scenv.2023.100035. composition using ANFIS approach. Fuel 2018:227–34. https://doi.org/
[175] Guo J, Sun S, Liu J. Conversion of waste frying palm oil into biodiesel using free 10.20944/preprints201903.0131.v1.
lipase A from Candida antarctica as a novel catalyst. Fuel 2020:117323. https:// [190] Al-Shanableh F, Evcil A, Savaş MA. Prediction of cold flow properties of biodiesel
doi.org/10.1016/j.fuel.2020.117323. fuel using artificial neural network. Procedia Comput Sci 2016:273–80. https://
[176] Lawer-Yolar G, Dawson-Andoh B, Atta-Obeng E. Synthesis of biodiesel from tall doi.org/10.1016/j.procs.2016.09.401.
oil fatty acids by homogeneous and heterogeneous catalysis. sustainable. [191] Özgür C, Tosun E. Prediction of density and kinematic viscosity of biodiesel by
Chemistry 2021:206–21. https://doi.org/10.3390/suschem2010012. artificial neural networks. Energy Sources Part A 2017:985–91. https://doi.org/
[177] Chakraborty R, Sahu H. Intensification of biodiesel production from waste goat 10.1080/15567036.2017.1280563.
tallow using infrared radiation: Process evaluation through response surface [192] Mairizal AQ, Awad S, Priadi CR, Hartono DM, Moersidik SS, Tazerout M, et al.
methodology and artificial neural network. Appl Energy 2014:827–36. https:// Experimental study on the effects of feedstock on the properties of biodiesel using
doi.org/10.1016/j.apenergy.2013.04.025. multiple linear regressions. Renew Energy 2020:375–81. https://doi.org/
[178] Adewale P, Dumont M-Je, Ngadi M. Enzyme-catalyzed synthesis and kinetics of 10.1016/j.renene.2019.06.067.
ultrasonic-assisted methanolysis of waste choice white grease for fatty acid [193] Cao Y, Doustgani A, Salehi A, Nemati M, Ghasemi A, Koohshekan O. The
methyl ester production. Energy Fuel 2015:6412–21. https://doi.org/10.1016/j. economic evaluation of establishing a plant for producing biodiesel from edible
cej.2015.08.053. oil wastes in oil-rich countries: Case study Iran. Energy 2020:118760. https://doi.
[179] Amenaghawon AN, Omede MO, Ogbebor GO, Eshiemogie SA, Igemhokhai S, org/10.1016/j.energy.2020.118760.
Evbarunegbe NI, et al. Optimized biodiesel synthesis from an optimally [194] Feng L, Liu J, Lu H, Liu B, Chen Y. Techno-economic and profitability analysis of
formulated ternary feedstock blend via machine learning-informed methanolysis plant for producing biodiesel from fresh vegetable oil and waste frying oil on
using a composite biobased catalyst. Bioresour Technol Rep 2024:101805. large-scale. Fuel 2022:124304. https://doi.org/10.1016/j.fuel.2022.124304.
https://doi.org/10.1016/j.biteb.2024.101805. [195] El Shimi H, Fawzy AS, Attia NK, El Diwani GI, El-Sheltawy ST. Evaluation of
[180] Banerjee A, Varshney D, Kumar S, Chaudhary P, Gupta VK. Biodiesel production biodiesel production from spent cooking oils, a techno-economic case study of
from castor oil: ANN modeling and kinetic parameter estimation. International Egypt. ARPN Journal of Engineering and Applied Sciences 2016:10280–90.

20