19(2): S2: 400-405, 2024 www.thebioscan.

com

Implementation of novel Machine Learning Technique using several Meta with

Naive Bayes Models to Analyse the Performance of Wave Energy Converters

Dr. K. Ravikumar1 Dr. G. Simi Margarat2 Dr. A. Karthikayen3 Dr. G. Gomathy4 Dr. N. Kalyana
Sundaram5 Dr. M. Rajalakshmi6
1
Professor, Department of IT, Dhanalakshmi Srinivasan College of Engineering and Technology, Chennai, Tamilnadu, India.
[email protected]
2
Professor, Department of CSE (Cyber Security) ,New Prince Shri Bhavani College of Engineering and Technology, Chennai,
Tamil Nadu, India. [email protected]
3
Professor, Department of ECE, P.T. Lee Chengalvaraya Naicker College of Engineering and Technology, Oovery,
Kanchipuram, Tamilnadu, India. [email protected]
4
Associate Professor, Department of Electrical and Electronics Engineering , Jaya Engineering College, Thiruninravur,
Chennai, Tamil Nadu, India. [email protected].
5
Associate Professor, Department of Information Technology, New Prince Shri Bhavani College of Engineering and
Technology, Chennai, Tamilnadu, India. [email protected]
6
Associate Professor,Department of IT, Tagore Engineering College, Chennai, Tamilnadu, India. [email protected]

DOI: https://doi.org/10.63001/tbs.2024.v19.i02.S2.pp400-405

KEYWORDS ABSTRACT

Wave Energy Converters, Optimization of the layouts of arrays of wave energy converters (WECs) is a challenging problem. The hydrodynamic
Ada Boost, analysis and performance estimation of such systems are performed using semi-analytical and numerical models such as the
F-Measure, boundary element method. However, the analysis of an array of such converters becomes computationally expensive, and the
Stochastic Gradient computational time increases rapidly with the number of devices in the system. As such determination of optimal layouts of
Boost, WECs in arrays becomes extremely difficult. This paper explores the Ada Boost with Naïve Bayes perform well as well it
Received on: showing an efficient outcome. It has the greatest accuracy result of 85.75%. The Ada Boost with Naïve Bayes produces the
greatest precision result of 0.86. The Ada Boost with Naïve Bayes and Stochastic Gradient Boost with Naïve Bayes produce
01-08-2024
the maximum recall of 0.86. The Ada Boost with Naïve Bayes has the greatest F-Measure result of 0.86. The Ada Boost
with Naïve Bayes model has the highest MCC value of 0.65. The Ada Boost with Naïve Bayes model has the greatest kappa
Accepted on:
value of 0.66. The Ada Boost with Naïve Bayes model has an optimal results compare with other models.

18-11-2024

INTRODUCTION renewable energy sources. Most wave energy technologies

require extensive R&D. Parallel technology and power
This work introduces wave energy converters and power
electronics standardization have improved prospects since the
systems. [1]Energy demand will rise globally in the coming
1970s (new converters can be cheaper and more reliable). Wave
decades. Meanwhile, sustainable development and
energy technology has reduced kWh production costs by one
environmental protection must limit the energy industry's
order of magnitude in the last 20 years. Many demonstration
growth.[2] Oil shortages are inevitable. Renewable energy could
wave power plants have been developed (a few were full-size).
help reduce energy industry carbon dioxide emissions. Such
Recently, some prototypes have proven their technology and
technologies are booming.[3,4] Europe leads wind energy with
may be suitable for pre-commercialization.
thousands of MW added in the last few years. Governments and
Ocean wave energy is a renewable energy source that could help
industry are focused on offshore wind energy. This rapidly
meet global electricity demand. Concentrated solar energy has
increasing field interacts favourably with some marine
high short- and long-term variability. Well-built devices can
renewables, and combined offshore power plants to utilise more
transform sea wave energy into electricity by following
resources have been suggested. Wave energy, the most
numerous ideas and concepts. Systems must survive tremendous
concentrated renewable energy source, may develop in the next
loads in storms. So, any cost-effective wave energy tapping
years. It might produce high pressure for reverse osmosis, which
strategy must balance efficiency and reliability, which presents
could desalinate water on islands. Tidal streams, oceanic
technical hurdles.
temperature and salinity gradients, and waves are other marine

400
What follows is the rest of the paper's outline: The associated demand by Cheng et al. [25]. LSTM improves forecasting by
work is described in Section 2. The proposed technique is 21.80% and 28.57%. Lin et al. [26] improved LSTM error-based
introduced in Section 3, followed by a brief overview of the power prediction. LSTM yielded the best results. Deep learning
results and discussion in Section 4. In Section 5, we wrap up the projected wave energy converter power for Ni et al. High-
paper and its findings. frequency waves strongly affect modelling efficiency when
II Literature Survey comparing deep learning systems.
Fossil fuels provide most energy [5]. Climate change and III Materials and Methods
pollution are caused by non-renewable energy [6]. Low-carbon The Dataset gathered from UCI's open data repository. Positions
energy is essential. Non-petroleum energy sources have grown and absorbed power outputs from four actual wave scenarios off
significantly. Solar, wind, tidal, and geothermal energy are most the southern coast of Australia make up this data set (Sydney,
popular [7]. Wave energy is the second-most potential ocean Adelaide, Perth and Tasmania). The CETO [1] model of totally
renewable energy [8]. WECs generate power from ocean waves. submerged three-tether converter is used in this application. In
WECs differ [9,10]. Investment and power management depend a space-restricted setting, 16 WECs are strategically located.
on reliable WEC forecasts. Investors worldwide need WEC system The issue is classified as a costly optimisation problem because
power generation potential predictions [11]. Numerical the examination of each farm takes several minutes.
calculations and experiments using WEC system output power Features Information:
[12,13]. "Searaser," a breakthrough wave energy converter, was 1. WECs position {X1, X2… X16; Y1, Y2… Y16}
used in this study [14]. Ocean waves are affordable, safe, continuous from 0 to 566 (m).
predictable, and clean. Large-scale integrated maritime energy 2. WECs absorbed power: {P1, P2... P16}
systems are vulnerable to uncertainty [15,16]. Hence, accurate 3. Total power output of the farm: Powerall
ocean wave energy estimates reduce power generation building 4. Location: Perth, Adelaide, Sydney, and Tasmania
costs and pilot programmes. Wave energy is abundant and Methods:
predictable [7,18]. Engineers cannot predict ocean wave power The following method are applied in this research work
from random data. Researchers wish to replace numerical • Borrowed dataset
solutions because solving equations with intricate boundary • Data preprocessing
conditions is time-consuming and expensive. AI estimates energy • Apply for Ensemble machine learning
system production capacity fast and affordably. Hence, algorithms:
engineering AI researchers have developed algorithms to predict • Gradient Boosting Machine with Naïve Bayes
ocean wave energy systems' electrical power from effective (GBM with NB)
parameters [19,20]. Zhenqing et al. [21] predicted ocean waves • Stochastic Gradient Boosting with Naive
using machine learning and genetic algorithms. Converters are Bayes(SGD with NB)
shown using wave periods, wave height, and ocean depth. • AB with NB(Ada with NB)
Tuning converters solved industrial technological concerns. Li et • Extreme Gradient Boost with Naive
al. [22] examined wave power parameters. Machine learning and Bayes(XGB with NB)
an artificial neural network predicted the wave's free surface
• Light Gradient Boosting Machine with Naive
height and force. Mistakes showed a power capture efficiency-
Bayes(LGBM with NB)
parameter relationship. Gomez et al. [23] created a new
• To get Optimal results
software tool with a user-friendly guiding interface to predict
output from two meteorological data sources using the latest • Find a best Model
machine learning methods. Butt et al. [24] introduced AI system To produce an efficient result, these strategies were applied in
forecasting. 24 h load prediction. These technologies improve python API. This study uses only 10% of the total dataset and
maintenance by assessing error kinds. LSTM projected electricity uses tenfold cross validation for all categories.

Analysis on wave energy converters

Data Collection

Data Preprocessing

Hyper Parameter Tuning

Classifiers

Yes No
GBM SGB Ada Boost XG Boost Light GBM

Base Classifier: Naïve Bayes

Result?

An Optimal model
Figure 1: Proposed System

401
 Table 2: Performance of selected classifiers

S.No Classifiers Accuracy Precision Recall F-Measure MCC Kappa

1 AB with NB 85.75% 0.86 0.86 0.86 0.65 0.66

2 XGB with NB 84.10% 0.84 0.84 0.83 0.56 0.55

3 LGB with NB 83.85% 0.83 0.83 0.84 0.56 0.58

4 GDM with NB 79.02% 0.81 0.8 0.79 0.54 0.54

5 SGB with NB 85.00% 0.85 0.86 0.84 0.58 0.59

The above table shows that the various selected ensemble The LGBM with NB produces accuracy level 83.85%, a precision
classifiers. value 0.83, recall value 0.83, an F-Measure value 0.84,an MCC
The AB with NB has an accuracy level of 85.75%, a precision value 0.56 and a kappa statistic value 0.58.
value of 0.86, a recall value of 0.86, an F-Measure value of 0.86, The GBM with NB produces a yield of 79.02% an accuracy, a
an MCC value of 0.65 and a kappa statistic value of 0.66. precision value of 0.81, a recall of 0.80, an F-Measure of 0.79,
The XGB with NB results in an accuracy level of 84.10%, a an MCC of 0.54 and a kappa statistic of 0.54.
precision value of 0.84, a recall value of 0.84, an F-Measure The SGB with NB results in an accuracy level of 85%, a precision
value of 0.83, an MCC value of 0.56 and a kappa statistic value value of 0.86, a recall value of 0.86, an F-Measure value of
of 0.55. 0.84,an MCC value of 0.58 and a kappa statistic value of 0.59.

Model Vs Accuracy
88.00%
86.00%
84.00%
Accuracy

82.00%
80.00%
78.00%
76.00%
74.00%
GBM with
AB with NB XGB with NB LGB with NB SGB with NB
NB
Accuracy 85.75% 84.10% 83.85% 79.02% 85.00%
Figure 2: Performance of Ensemble classifiers with their accuracies
The above diagram shows that the accuracy performances of result of 79.02%. The accuracy of the LGBM with NB, XGB with
selected models. The AB with NB has the greatest accuracy NB, and SGB with NB is 83.85%, 84.10%, and 85%, respectively.
result of 85.75%. The GBM with NB produces the lowest accuracy

Model Vs Precision
0.87
0.86
0.85
0.84
Precision

0.83
0.82
0.81
0.8
0.79
0.78
XGB with LGB with GBM with SGB with
AB with NB
NB NB NB NB
Precision 0.86 0.84 0.83 0.81 0.85

402
 Figure 3: Performance of Ensemble Classifiers with their Precision values
The precision performances of selected models are depicted in accuracy result of 0.81. The accuracy levels of the LGBM with
the diagram above. The AB with NB produces the greatest NB, XGB with NB, and SGB with NB are 0.83, 0.84, and 0.85,
precision result of 0.86. GBM with NB produces the lowest respectively.

Model Vs Recall
0.87
0.86
0.85
0.84
0.83
Recall

0.82
0.81
0.8
0.79
0.78
0.77
GBM with
AB with NB XGB with NB LGB with NB SGB with NB
NB
Recall 0.86 0.84 0.83 0.8 0.86
Figure 4: Performance of Ensemble Classifiers with their Recall values
The graph above depicts the recall performances of selected 0.80. The recall levels for the LGBM with NB and the Extreme
models. The AB with NB and SGB with NB produce the maximum GBM with NB are 0.83 and 0.84, respectively.
recall of 0.86. GBM with NB produces the lowest recall result of

Model Vs F-Measure
0.88

0.86

0.84
F-Measure

0.82

0.8

0.78

0.76

0.74
XGB with LGB with GBM with SGB with
AB with NB
NB NB NB NB
F-Measure 0.86 0.83 0.84 0.79 0.84
Figure 5: Performance of Ensemble Classifiers with their F-Measure values
The graph above depicts the F-Measure performances of selected 0.79. The XGB with NB has an F-Measure of 0.83, whereas the
models. The AB with NB has the greatest F-Measure result of LGB with NB and SGB with NB have the same value of 0.84.
0.86. The GBM with NB produces the lowest F-Measure result of

403
 Model Vs MCC
0.7

0.6

0.5

0.4
MCC
0.3

0.2

0.1

0
GBM with
AB with NB XGB with NB LGB with NB SGB with NB
NB
MCC 0.65 0.56 0.56 0.54 0.58
Figure 6: Performance of Ensemble Classifiers with their MCC values
The graphic above depicts the MCC performance of selected the Light Gradient Boosting Machine with NB Decision Trees
models. The AB with NB model has the highest MCC value of model, have the same MCC value of 0.56. The MCC value for SGB
0.65. GBM with NB produces the lowest MCC result (0.54). The with NB 0.58.
remainder of the models, such as the XGB with NB model and

Model Vs Kappa
0.7

0.6

0.5

0.4
Kappa

0.3

0.2

0.1

0
GBM with
AB with NB XGB with NB LGB with NB SGB with NB
NB
Kappa 0.66 0.55 0.58 0.54 0.59
Figure 7: Performance of Ensemble classifiers with their Kappa statistic values
The graph above depicts the kappa value performances of of 0.86, a recall of 0.86, an F-Measure result of 0.86, an MCC
selected models. The AB with NB model has the greatest kappa value of 0.65 and a kappa value of 0.66.This model recommends
value of 0.66. The GBM with NB produces the lowest kappa the AB with NB compare with other models.
result of 0.54. Other models with kappa values between 0.55 REFERENCES
and 0.59 are Extreme Gradient Boost with NB, Light Gradient
Boosting Machine with NB, and Stochastic Gradient Boosting with • L. D. Mann, A. R. Burns, , and M. E. Ottaviano. 2007. CETO,
NB. a carbon free wave power energy provider of the future. In
CONCLUSION the 7th European Wave and Tidal Energy Conference
This work finds that the AB with NB has an F-Measure of 0.86, an (EWTEC).
MCC of 0.65, a kappa statistic of 0.66, a recall of 0.86, a • Neshat, M., Alexander, B., Wagner, M., & Xia, Y. (2018,
precision of 0.86, and a precision of 0.86. The combined XGB July). A detailed comparison of meta-heuristic methods for
and NB achieves an impressive 84.10% accuracy, 0.84 precision, optimising wave energy converter placements. In
0.84 recall, 0.83 F-Measure, 0.56 MCC, and 0.55 kappa statistic. Proceedings of the Genetic and Evolutionary Computation
Accuracy of 83.85%, precision of 0.83%, recall of 0.83%, F- Conference (pp. 1318-1325). ACM.
Measure of 0.84%, mean correlation coefficient of 0.56%, and • Neshat, M., Alexander, B., Sergiienko, N., & Wagner, M.
kappa statistic of 0.58% are all generated by the LGBM with NB. (2019). A new insight into the Position Optimization of
Using NB, the GBM achieves a return of 79.02% accurate Wave Energy Converters by a Hybrid Local Search. arXiv
predictions, 0.81 precision, 0.80 recall, 0.79 F-Measure, 0.54 preprint.
MCC, and 0.54 kappa. Accuracy is at 85%, precision is 0.86, • Nataliia Sergiienko (2023). Wave Energy Converter (WEC)
recall is 0.86, the F-Measure is 0.84, the MCC is 0.58, and the Array
kappa statistic is 0.59 when using the SGB with NB. The AB with Simulator (https://www.mathworks.com/matlabcentral/fil
NB has the greatest accuracy result of 85.75%. a precision result eexchange/71840-wave-energy-converter-wec-array-

404
 simulator), MATLAB Central File Exchange. Retrieved April force for arrays of wave energy devices. IEEE Trans.
4, 2023. Sustain. Energy 2018, 9, 1672–1680.
• Raj, S.., Shankar, G., Murugesan, S., Raju, M. N. ., Mohan, • Sang, Y.; Karayaka, H.B.; Yan, Y.; Yilmaz, N.; Souders, D.
E., & Rani, P. J. I. (2023). Exploratory Data Analysis on 1.18 Ocean (Marine) Energy. In Comprehensive Energy
Blueberry yield through Bayes and Function Systems; Dincer, I., Ed.; Elsevier: Amsterdam, The
Models. International Journal on Recent and Innovation Netherlands, 2018; Volume 1, pp. 733–769.
Trends in Computing and Communication, 11(11s), 634–641. • Reikard, G.; Robertson, B.; Bidlot, J.-R. Combining wave
https://doi.org/10.17762/ijritcc.v11i11s.8299 energy with wind and solar: Short-term forecasting. Renew.
• Thivakaran, T. K. ., Priyanka, N. ., Antony, J. C. ., Energy 2015, 81, 442–456.
Surendran, S. ., Mohan, E. ., & Innisai Rani, P. J. . (2023). • Bento, P.; Pombo, J.; Mendes, R.; Calado, M.; Mariano, S.
Exploratory Data Analysis for Textile Defect Ocean wave energy forecasting using optimised deep
Detection. International Journal on Recent and Innovation learning neural networks. Ocean Eng. 2021, 219, 108372.
Trends in Computing and Communication, 11(9s), 121–128. • Ni, C.; Ma, X. Prediction of wave power generation using a
https://doi.org/10.17762/ijritcc.v11i9s.7403 convolutional neural network with multiple inputs. Energies
• Rani, P. J. I. ., Venkatachalam, K. ., Sasikumar, D. ., 2018, 11, 2097.
Madhankumar, M. ., A., T. ., Senthilkumar, P. ., & Mohan, • Liu, Z.; Wang, Y.; Hua, X. Prediction and optimization of
E. . (2024). An Optimal Approach on Electric Vehicle by oscillating wave surge converter using machine learning
using Functional Learning . International Journal of techniques. Energy Convers. Manag. 2020, 210, 112677.
Intelligent Systems and Applications in • Li, L.; Gao, Y.; Ning, D.; Yuan, Z.M. Development of a
Engineering, 12(13s), 197–206. constraint non-causal wave energy control algorithm based
• Sinsel, S.R.; Riemke, R.L.; Hoffmann, V.H. Challenges and on artificial intelligence. Renew. Sustain. Energy Rev. 2020,
solution technologies for the integration of variable 138, 110519.
renewable energy sources—A Review. Renew. Energy 2020, • Gómez-Orellana, A.M.; Fernández, J.C.; Dorado-Moreno,
145, 2271–2285. M.; Gutiérrez, P.A.; Hervás-Martínez, C. Building suitable
• Aderinto, T.; Li, H. Ocean wave energy converters: Status datasets for soft computing and machine learning
and challenges. Energies 2018, 11, 1250. techniques from meteorological data integration: A case
• Jiang, B.; Li, X.; Chen, S.; Xiong, Q.; Chetn, B.-F.; Parker, study for predicting significant wave height and energy flux.
R.G.; Zuo, L. Performance analysis and tank test validation Energies 2021, 14, 468.
of a hybrid ocean wave-current energy converter with a • Butt, F.M.; Hussain, L.; Mahmood, A.; Lone, K.J. Artificial
single power takeoff. Energy Convers. Manag. 2020, 224, Intelligence based accurately load forecasting system to
113268. forecast short and medium-term load demands. Math.
• Ahamed, R.; McKee, K.; Howard, I. Advancements of wave Biosci. Eng. 2020, 18, 400–425.
energy converters based on power take off (PTO) systems: A • Cheng, Y.; Xu, C.; Mashima, D.; Thing, V.L.; Wu, Y.
review. Ocean Eng. 2020, 204, 107248. Ocean Energy PowerLSTM: Power demand forecasting using long short-
Systems. Annual Report Ocean Energy Systems 2016. Ocean term memory neural network. In Proceedings of the
Energy Systems Website. 2017. Available online: International Conference on Advanced Data Mining and
https://report2016.ocean-energy-systems.org Applications, Singapore, 5–6 November 2017; Springer:
• Ruehl, K.; Forbush, D.D.; Yu, Y.-H.; Tom, N. Experimental Cham, Switzerland, 2017.
and numerical comparisons of a dual-flap floating • Lin, Z.; Cheng, L.; Huang, G. Electricity consumption
oscillating surge wave energy converter in regular waves. prediction based on LSTM with attention mechanism. IEEJ
Ocean Eng. 2020, 196, 106575. Trans. Electr. Electron. Eng. 2020, 15, 556–562.
• Giorgi, G.; Gomes, R.P.; Henriques, J.C.; Gato, L.M.; • Ni, C.; Ma, X.; Wang, J. Integrated deep learning model for
Bracco, G.; Mattiazzo, G. Detecting parametric resonance predicting electrical power generation from wave energy
in a floating oscillating water column device for wave converter. In Proceedings of the 2019 25th International
energy conversion: Numerical simulations and validation Conference on Automation and Computing, Lancaster, UK,
with physical model tests. Appl. Energy 2020, 276, 115421. 5–7 September 2019. He, J. Coherence and cross-spectral
• Shahriar, T.; Habib, M.A.; Hasanuzzaman, M.; Shahrear-Bin- density matrix analysis of random wind and wave in deep
Zaman, M. Modelling and optimization of Searaser wave water. Ocean Eng. 2020, 197, 106930.
energy converter based hydroelectric power generation for • Dripta Sarkar, Emile Contal, Nicolas Vayatis, Frederic Dias,
Saint Martin’s Island in Bangladesh. Ocean Eng. 2019, 192, Prediction and optimization of wave energy converter
106289. arrays using a machine learning approach, Renewable
• Reikard, G.; Robertson, B.; Bidlot, J.-R. Wave energy Energy, Volume 97, 2016, Pages 504-517, ISSN 0960-1481,
worldwide: Simulating wave farms, forecasting, and https://doi.org/10.1016/j.renene.2016.05.083.(https://ww
calculating reserves. Int. J. Mar. Energy 2017, 17, 156–185. w.sciencedirect.com/science/article/pii/S096014811630493
• Pena-Sanchez, Y.; Garcia-Abril, M.; Paparella, F.; 1)
Ringwood, J.V. Estimation and forecasting of excitation

405