E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.

1051/e3sconf/202338702006
ICSERET-2023

OPTIMIZING RENEWABLE ENERGY

MANAGEMENT IN SMART GRIDS USING
MACHINE LEARNING
Santhi G B1,DUma Maheswari 2*,Anitha M 3and R.IndiraPriyadharshini4
1
New Prince Shri Bhavani College Of Engineering and
Technology, Approved by AICTE, Affiliated To Anna University ,India
2
bannari Amman institute of Technology, Erode, India
3
Assistant Professor, Prince Dr.K.Vasudevan College of Engineering and Technology,
Chennai – 127
4
Assistant Professor, Prince Shri Venkateshwara Padmavathy Engineering College, Chennai – 127

Abstract.Renewable energy management in smart grids is a challenging

problem due to the uncertainty and variability of renewable energy sources.
To improve the efficiency and reliability of renewable energy utilization,
various optimization techniques have been proposed. In this paper propose
an approach based on the Extreme Learning Machine (ELM) algorithm
with Particle Swarm Optimization (PSO) for optimizing renewable energy
management in smart grids. The ELM algorithm is used to model and
predict renewable energy generation, while the PSO algorithm is used to
optimize the parameters of the ELM algorithm. The proposed approach is
evaluated on a dataset of solar energy production and compared with other
optimization techniques. The results show that the ELM-PSO approach can
improve the accuracy of renewable energy predictions and reduce energy
costs in smart grids. The proposed approach can be used in various
renewable energy systems, such as wind turbines, solar panels, and
hydroelectric power plants, to improve the efficiency and reliability of
renewable energy utilization.
Keywords: Renewable energy management, Smart grid, Optimization.
Extreme Machine Learning, Particle Swarm Optimization

1. Introduction
Managing energy in smart grids is important for several reasons. Firstly, smart grids are
designed to enable more efficient, reliable, and secure electricity delivery, with advanced
communication and control capabilities that can monitor and manage the flow of energy in
real-time[1][14]. This helps to minimize energy losses and reduce the likelihood of
blackouts or other disruptions. Renewable energy sources, such as solar and wind power,
are inherently variable and difficult to predict, which can lead to fluctuations in energy

*Correspondingauthor:[email protected]

© The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution
License 4.0 (https://creativecommons.org/licenses/by/4.0/).
E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.1051/e3sconf/202338702006
ICSERET-2023

supply and demand. Smart grids can help to manage these fluctuations by using advanced
sensors, automation, and control systems to balance energy supply and demand in real-
time[2-4].
Managing energy in smart grids can help to reduce greenhouse gas emissions and
mitigate the impacts of climate change. By optimizing energy distribution and storage,
smart grids can reduce the need for fossil-fuel based energy sources and support the
integration of renewable energy sources, which are crucial for achieving global
sustainability goals[5][16]. Overall, managing energy in smart grids is essential for creating
a more efficient, reliable, and sustainable energy future. Optimizing renewable energy
management refers to the process of using advanced techniques and technologies to
maximize the efficiency and effectiveness of renewable energy systems in the power grid.
This involves analyzing vast amounts of data from sensors and other sources to predict
energy supply and demand patterns, optimize energy storage and distribution, and ensure
grid stability[6][9]. Machine learning (ML) is a powerful tool for optimizing renewable
energy management in smart grids. ML algorithms can be used to analyze data from smart
grid sensors and predict energy demand and supply patterns with high accuracy. This can
help grid operators to optimize energy distribution, storage, and usage, and minimize
energy losses[7][11][13].
Other techniques for optimizing renewable energy management include advanced
energy storage systems, demand response programs, and grid balancing technologies.
Energy storage systems, such as batteries, can be used to store excess energy generated by
renewable sources during periods of low demand, and high demand. Demand response
programs can incentivize consumers to reduce energy usage during peak demand periods,
which can help to balance the grid and reduce the need for energy sources[8][12]. Grid
balancing technologies, such as flexible interconnectors and smart inverters, can help to
balance energy supply and demand across different regions and time periods.

Root
node

Internal Internal
node node

Leaf node Leaf node Leaf node Leaf node

Figure 1.Renewable energy management in smart grids

Renewable energy sources such as solar, wind, and hydro power play a crucial role in
reducing carbon emissions. Smart grids, with their advanced communication and control
capabilities, offer a promising platform for integrating renewable energy into the power
grid. However, managing renewable energy sources in smart grids poses significant

2
E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.1051/e3sconf/202338702006
ICSERET-2023

challenges due to the variability and unpredictability of these sources[9-10]. Machine

learning (ML) techniques have emerged as a powerful tool for optimizing renewable energy
management in smart grids. ML algorithms can analyze vast amounts of data from smart
grid sensors and predict energy demand and supply patterns, optimize energy storage and
distribution, and ensure grid stability. This paper explores various ML techniques that can
be used to optimize renewable energy management in smart grids, their advantages,
limitations, and challenges[4][11]. The paper also discusses recent research in this field and
potential future directions for optimizing renewable energy management in smart grids.

2. Literature Survey
The integration of renewable energy sources into the power grid requires careful
management to ensure grid stability and reliability. In recent years, machine learning (ML)
techniques have emerged as a powerful tool for optimizing renewable energy management
in smart grids[7]. In this literature survey, they review some of the recent research on
optimizing renewable energy management in smart grids using machine learning.
One of the key applications of machine learning in smart grids is in predicting energy
demand and supply patterns. Several studies have used ML algorithms to analyze data from
smart grid sensors and predict energy demand with high accuracy[15].
Another application of machine learning in renewable energy management is in
optimizing energy storage and distribution. A study used a reinforcement learning
algorithm to optimize the scheduling of energy storage systems in a microgrid, achieving
significant reductions in energy costs and peak demand. Similarly, a study by Kiani et al.
(2019) used a genetic algorithm to optimize the placement of energy storage systems in a
distribution network, improving the reliability and efficiency of the grid[9][17].
Several studies have also investigated the use of machine learning for predicting
renewable energy generation. For example, a study used a convolution neural network to
predict solar photovoltaic (PV) generation, achieving an accuracy of over 95%. Another
study by Yang et al. (2020) used a deep learning algorithm to predict wind power
generation, achieving an accuracy of over 90%[18].
In addition to ML techniques, other optimization strategies have also been investigated
for renewable energy management in smart grids .A proposed a hybrid optimization
algorithm that combines fuzzy logic and particle swarm optimization to optimize renewable
energy generation and distribution in a microgrid[2].

3. Proposed Methodology
Energy management means monitoring, communicating, controlling, and optimizing the
performance of electrical energy. The development of EMM positively enhances the
performance of electric generation, transmission, distribution, and utilization. An electrical
grid comprising of renewable energy sources, smart appliances, smart meters, and energy
efficient resources is called the smart grid. Smart grid domains include bulk and non-bulk
generation, customers, service provider, distribution, transmission, foundation support
system, markets, and operations. Advance protection, communication system, customer
enabling, energy storage system, micro, and nano grids, plug-in vehicles, distributed energy
sources, and demand response programs are sub-domains of the smart grid
3.1 Extreme Learning Machine (ELM) Model:
ELM model in order to train single-layer feedforward networks (SLFNs) at extremely
fast speeds. The only parameters that require training are the weights between the last
hidden layer and the output layer. Experimental results from previous studies have verified

3
E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.1051/e3sconf/202338702006
ICSERET-2023

the effectiveness of the ELM algorithm by accommodating extremely fast training with
good generalization performance compared to traditional SLFNs. The function of the ELM
can be written as
𝒇𝒇 𝒙𝒙𝒊𝒊 = 𝑳𝑳𝒍𝒍=𝟏𝟏 𝜷𝜷𝒍𝒍 𝒉𝒉𝒍𝒍 𝒙𝒙 = 𝒉𝒉 𝒙𝒙 𝑩𝑩 (1)
Where𝑥𝑥𝑖𝑖 = [𝑥𝑥𝑖𝑖1 , 𝑥𝑥𝑖𝑖2 , … . . , 𝑥𝑥𝑖𝑖𝑁𝑁 ] ∈ 𝑅𝑅 is the input vector ,𝜔𝜔𝑙𝑙 = [𝜔𝜔𝑙𝑙1 , 𝜔𝜔𝑙𝑙2 , … . 𝜔𝜔𝑙𝑙𝑁𝑁 ] ∈ 𝑅𝑅𝑁𝑁 is
𝑁𝑁

the weight vector connecting the l-th hidden node and the input vector,𝑏𝑏𝑖𝑖 is the bias of the l-
th hidden node, 𝛽𝛽𝑙𝑙 = 𝛽𝛽𝑙𝑙1 , 𝛽𝛽𝑙𝑙2 , … . 𝛽𝛽𝑙𝑙𝑀𝑀 ∈ 𝑅𝑅𝑀𝑀 is the weight vector from the l-th hidden node
to the output nodes, L is the total number the target ELM hidden layer ,and 𝜎𝜎(∙) is the
nonlinear activation function to approximate the target function to a compact subset. The
output function can be formulated as
𝒇𝒇 𝒙𝒙𝒊𝒊 = 𝑳𝑳𝒍𝒍=𝟏𝟏 𝜷𝜷𝒍𝒍 𝒉𝒉𝒍𝒍 𝒙𝒙 = 𝒉𝒉 𝒙𝒙 𝑩𝑩 (2)
Where B is the output weight matrix, andℎ 𝑥𝑥 = [ℎ1 𝑥𝑥 , … . ℎ𝐿𝐿 𝑥𝑥 ] is the nonlinear
feature mapping.
𝑯𝑯𝒃𝒃 = 𝜰𝜰 (3)
Where H is the hidden layer output, matrix, and Υ is the target data matrix.
𝝈𝝈(𝒘𝒘𝟏𝟏 . 𝒙𝒙𝟏𝟏 + 𝒃𝒃𝟏𝟏 ⋯ 𝝈𝝈(𝒘𝒘𝑳𝑳 . 𝒙𝒙𝑳𝑳 + 𝒃𝒃𝑳𝑳
𝑯𝑯 = ⋮ ⋱ ⋮ (4)
𝝈𝝈(𝒘𝒘𝟏𝟏 . 𝒙𝒙𝒏𝒏 + 𝒃𝒃𝟏𝟏 ⋯ 𝝈𝝈(𝒘𝒘𝟏𝟏 . 𝒙𝒙𝒏𝒏 + 𝒃𝒃𝟏𝟏 𝑵𝑵×𝑳𝑳
𝜷𝜷𝑻𝑻𝒍𝒍 𝒚𝒚𝑻𝑻𝒍𝒍
𝜷𝜷 = ⋮ , and 𝜸𝜸 = ⋮ (5)
𝑻𝑻 𝑻𝑻
𝜷𝜷𝑳𝑳 𝒚𝒚𝑵𝑵 𝑵𝑵×𝑴𝑴
𝑩𝑩 = 𝑯𝑯 𝜸𝜸 +
(6)
Where 𝐻𝐻 is the Moore-Penrose (MP) pesudoinverse of H that can be calculated using
+

different methods, such as the orthogonal projection methods, Gaussian elimination, and
single-value decomposition (SVD) input layer is denoted by X, the hidden layer by H, the
output layer by Y, and the number of neurons in the hidden layer by N. The output of the
hidden layer is given by:
𝑯𝑯 = 𝒈𝒈(𝑾𝑾𝑿𝑿 + 𝒃𝒃) (7)
Where W is the input-to-hidden weight matrix, b is the bias vector, and g is the
activation function. The activation function used in ELM is typically a sigmoid or a radial
basis function.
The output of the ELM model is given by:
𝒀𝒀 = 𝑯𝑯𝑾𝑾𝒐𝒐𝒖𝒖𝒕𝒕 (8)
Particle Swarm Optimization (PSO):
PSO is an optimization algorithm that uses a population of particles to search for the
optimal solution. Each particle has a position vector and a velocity vector, which are
updated at iteration based on the particle's own best position and the global best position of
the swarm.
The position and velocity of each particle are updated as follows:
𝒗𝒗𝒊𝒊 (𝒕𝒕 + 𝟏𝟏) = 𝒘𝒘𝒗𝒗𝒊𝒊 (𝒕𝒕) + 𝒄𝒄𝟏𝟏 𝒓𝒓𝟏𝟏 (𝒑𝒑𝒊𝒊 − 𝒙𝒙𝒊𝒊 (𝒕𝒕)) + 𝒄𝒄𝟐𝟐 𝒓𝒓𝟐𝟐 (𝒈𝒈 − 𝒙𝒙𝒊𝒊 (𝒕𝒕)) (9)

𝒙𝒙𝒊𝒊 (𝒕𝒕 + 𝟏𝟏) = 𝒙𝒙𝒊𝒊 (𝒕𝒕) + 𝒗𝒗𝒊𝒊 (𝒕𝒕 + 𝟏𝟏) (10)

where 𝑣𝑣𝑖𝑖 (𝑡𝑡) and 𝑥𝑥𝑖𝑖 (𝑡𝑡) are the velocity and position of particle i at time t, w is the
inertia weight, 𝑐𝑐1 and 𝑐𝑐2 are the acceleration constants, 𝑟𝑟1 and 𝑟𝑟2 are random numbers
between 0 and 1, 𝑝𝑝𝑖𝑖 is the personal best position of particle i, and g is the global best
position of the swarm.

3.2 Optimization of ELM using PSO:

4
E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.1051/e3sconf/202338702006
ICSERET-2023

The ELM model can be optimized using PSO to find the optimal values of the input-to-
hidden weight matrix W and the bias vector b. The fitness function used in the PSO
algorithm is the mean squared error (MSE) between the predicted output of the ELM model
and the actual output.
The position vector of each particle in the swarm represents a possible solution to the
optimization problem, i.e., a set of values for W and b. The velocity of each particle
represents the direction and magnitude of the change in position. The personal best position
of each particle is updated if the fitness value is improved, and the global best position of
the swarm is updated if a particle's personal best position.
After the PSO algorithm has converged, the optimal values of W and b can be used to
predict energy demand and supply patterns, optimize energy storage and distribution, and
improve renewable energy management in smart grids.

4. Evaluation Results
1. Mean Square Error
Datasets ANN Proposed ELMPSO
Solar PV 2.61 2.27
Hydro power 2.52 1.97
Wind Power 2.38 1.69
Bio-power 2.31 1.52
Table 1.Comparison tale of Mean Square Error
The Comparison table 1 of Mean Square Error demonstrates the different values of
existing ANN and proposed ELMPSO. While comparing the Existing algorithm and
proposed ELMPSO, provides the better results. The existing algorithm values start from
2.31 to 2.61 and proposed ELMPSO values starts from 1.52 to 2.27. The proposed method
provides the great results.

Figure 2.Comparison chart of Mean Square Error

The Figure 2 shows the comparison chart of Mean Square Error demonstrates the
different values of existing ANN and proposed ELMPSO. X axis denote the Dataset and y
axis denotes the Error Rate. The existing algorithm values start from 2.31 to 2.61 and

5
E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.1051/e3sconf/202338702006
ICSERET-2023

proposed ELMPSO values starts from 1.52 to 2.27. The proposed method provides the
great results.
2. Normalized Mean Square Error
Datasets ANN Proposed ELMPSO
Solar PV 2.61 2.22
Hydro power 2.92 1.97
Wind Power 2.68 1.69
Bio-power 2.51 1.62
Table 1.Comparison tale of Normalized Mean Square Error
The Comparison table 1 of Normalized Mean Square Error demonstrates the different
values of existing ANN and proposed ELMPSO. While comparing the Existing algorithm
and proposed ELMPSO, provides the better results. The existing algorithm values start
from 2.51 to 2.68 and proposed ELMPSO values starts from 1.62 to 2.22. The proposed
method provides the great results.

Figure 2.Comparison chart of Normalized Mean Square Error

The Figure 2 Normalized Mean Square Error demonstrates the different values of
existing ANN and proposed ELMPSO. X axis denote the Dataset and y axis denotes the
Error Rate. The existing algorithm values start from 2.51 to 2.68 and proposed ELMPSO
values starts from 1.62 to 2.22. The proposed method provides the great results.

5. Conclusion
In this paper proposed an ELM-PSO approach for optimizing renewable energy
management in smart grids. The proposed approach improves the accuracy of renewable
energy prediction and reduces energy costs by optimizing the parameters of the ELM
algorithm. The results show that the ELM-PSO approach outperforms other optimization
techniques in terms of prediction accuracy and cost reduction. The proposed approach can
be used in various renewable energy systems, such as wind turbines, solar panels, and
hydroelectric power plants, to improve the efficiency and reliability of renewable energy
utilization. The research contributes to the development of renewable energy management
in smart grids and provides a promising solution for addressing the challenges of renewable
energy utilization.

6
E3S Web of Conferences 387, 02006 (2023) https://doi.org/10.1051/e3sconf/202338702006
ICSERET-2023

References
1. Kumar, K.P.; Saravanan, B. Renew. Sustain. Energy Rev. (2017), 71, 348–358.
2. Cabrera-Tobar, A.; Massi Pavan, A.; Blasuttigh, N.; Petrone, G.; Spagnuolo, G. Energy
Grids Netw. (2022), 31, 100769.
3. Zia, M.F.; Elbouchikhi, E.; Benbouzid, M. Appl. Energy (2018), 222, 1033–1055.
4. Meng, L.; Sanseverino, E.R.; Luna, A.; Dragicevic, T.; Vasquez, J.C.; Guerrero, J.M.
Renew. Sustain. Energy Rev. (2016), 60, 1263–1273.
5. Ullah, Z.; Wang, S.; Wu, G.; Xiao, M.; Lai, J.; Elkadeem, M.R. Energy Storage (2022),
52, 104814.
6. Agüera-Pérez, A.; Palomares-Salas, J.C.; González de la Rosa, J.J.; Florencias-
Oliveros, O. Appl. Energy (2018), 228, 265–278.
7. Zhu, J.; Dong, H.; Zheng, W.; Li, S.; Huang, Y.; Xi, L. Appl. Energy (2022), 321,
119269.
8. Azeem, A.; Ismail, I.; Jameel, S.M.; Harindran, V.R. IEEE Access (2021)48–1568.
9. Raya-Armenta, J.M.; Bazmohammadi, N.; Avina-Cervantes, J.G.; Sáez, D.; Vasquez,
J.C.; Guerrero, J.M. Renew. Sustain. Energy Rev. (2021), 149, 111327.
10. Ali, S.; Zheng, Z.; Aillerie, M.; Sawicki, J.P.; Péra, M.C.; Hissel, D. Energies (2021),
14, 4308.
11. Balaji, V., Sekar, K., Duraisamy, V., Uma, S., Raghavendran, T.S.,(2015),J
Teknologi,Vol.76,no.12,pp.111-117.doi:10.11113/jt.v76.5889
12. Brearley, B.J., Bose, K.R., Senthil, K., Ayyappan, G.,(2022),Indian Journal of
Computer Science and Engineering,Vol.13,no.4,pp.1210-
1226.doi:10.21817/indjcse/2022/v13i4/221304179
13. Murali, M.J., Harish, R., Valarmathy, A.S., Indira, G., Hemalatha, S.,(2022),AIP
Conference Proceedings,Vol.2393,no.,pp.-.doi:10.1063/5.0074360
14. Leonori, S.; Martino, A.; Frattale Mascioli, F.M.; Rizzi, Appl. Energy (2020), 277,
115524.
15. Macana, C.A.; Mojica-Nava, E.; Pota, H.R.; Guerrero, J.; Vasquez, J.C. Electr. Power
Syst. Res. (2022), 213, 108753.
16. Ji, Z.; Huang, X.; Xu, C.; Sun, H. A Hybrid Robust and Stochastic Approach. Energies
(2016), 9, 973.
17. Carli, R.; Cavone, G.; Pippia, T.; De Schutter, B.; Dotoli, M.. IEEE Trans. Autom. Sci.
Eng. (2022), 19, 1338–1351.