See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/358905399

Prediction of Solar Power Using Machine Learning Algorithm

Chapter · February 2022

DOI: 10.1007/978-981-16-7610-9_39

CITATION READS

1 842

5 authors, including:

Rupesh Mailugundla
BVRIT Hyderabad
17 PUBLICATIONS 62 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rupesh Mailugundla on 07 December 2022.

The user has requested enhancement of the downloaded file.

Prediction of Solar Power Using Machine
Learning Algorithm

M. Rupesh, J. Swathi Chandana, A. Aishwarya, C. Anusha, and B. Meghana

Abstract The stability and reliability of power in an integrated renewable energy

system vary according to changes in environmental conditions, as the radiation,
temperature, and humidity change continuously; the power generation through PV
system also changes, and hence, the power scheduling and operation mainly depend
on the estimation of power through renewable energy sources. The power generation
in PV systems initially depends on the radiation and temperature, so prediction of
weather conditions helps in predicting the power generation through PV solar plant.
The stability of a power system can be improved by predicting solar energy which
can tell approximately how much solar power can be generated in the future at a
particular location. Solar power forecasting has several methods, one of all methods
is using machine learning/neural networks. In this paper, the power generation with
a solar plant is forecasted by predicting the future weather generation using machine
learning algorithms. The accuracy of forecasting will be checked directly with the
practical data which is generated and simulated data using MATLAB/Simulink by
applying various machine learning algorithms.

Keywords Weather forecasting · Solar power forecasting · Artificial neural

network (ANN) · Machine learning · Feed-forward back-propagation algorithm

M. Rupesh (B) · J. Swathi Chandana · A. Aishwarya · C. Anusha · B. Meghana

Electrical & Electronics Engineering Department, BVRIT Hyderabad College of Engineering for
Women, Hyderabad, India
e-mail: [email protected]
J. Swathi Chandana
e-mail: [email protected]
A. Aishwarya
e-mail: [email protected]
C. Anusha
e-mail: [email protected]
B. Meghana
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 529
D. J. Hemanth et al. (eds.), Intelligent Data Communication Technologies and Internet
of Things, Lecture Notes on Data Engineering and Communications Technologies 101,
https://doi.org/10.1007/978-981-16-7610-9_39
530 M. Rupesh et al.

1 Introduction

In present days, the economic growth is directly proportional to electrical demand,

so in developed and developing countries like India, fossil fuels are exhaustive in
nature and their by-products are resulting in pollution and greenhouse gases so power
generation mainly depends on renewable energies to satisfy the customers demand.
The PV system [1] plays a major role in renewable energy systems because it is
abundant in nature and is renewable [2], as the PV power is varying due to the
changes in weather conditions it will affect the operation of the grid adversely. To
maintain the system stable in variable power generation conditions, the fossil fuels
should act as spinning reserve and proper planning and estimation of operating time,
capacity is very important to save the environment and reduce fossil fuel usage.
Prior knowledge of power produced through PV systems helps in estimating the
fossil fuel power requirements and reduces the cost for power generation. In solar
power generation, the prediction mainly depends on weather prediction and past
data. Hence, the collection of past data is very important in solar power forecasting.
Various researchers have proposed many forecasting mechanisms with good results,
but still, there is room to improve the results. Artificial neural network (ANN) in
machine learning gives a very high degree of accuracy in predicting the weather and
power in solar PV systems [3].
The paper is organized like the proposed method in Sect. 2, implementation of
ANN forecasting method in Sect. 3, simulation and results in Sect. 4, and Sect. 5
deals with conclusion.

2 Proposed Model

Reliable data availability and choosing the right attributes from the collected data
is very important to predict accurately especially in solar power generation [4]. In
this work, the data is collected from the BV Raju Institute of Technology, Narsapur,
Medak Dist., India.
We choose the 3 years of weather information dataset from the above-said location.
The dataset available is minutes-based values of weather parameters like irradiance,
temperature, panel temperature, wind direction, and speed. Different weather values
are collected from the year 2012 to 2014, i.e., dataset of 201,235 to analyze the
relationship between the weather parameters and the power generation for accurate
prediction (Figs. 1 and 2).
Prediction of Solar Power Using Machine Learning Algorithm 531

Fig. 1 Correlation of weather parameters

3 Machine Learning Algorithm

State of the art of solar power generation will only be established if forecast algorithm
predicts that how much power will be generated at any location and time [5–8]. The
machine learning algorithm will be developed with training, testing, and validating
the collected data, the workflow [9] is as followed in the flowchart as shown in Fig. 3.

3.1 Feed-Forward Back-Propagation Algorithm

for Prediction

We have tested the selected data of the given work in the commonly used machine
learning algorithm feed-forward neural network with back-propagation to evaluate
the performance with weather data. In this weather, parameters are given as an input
to the ANN [10] and it gives the output as predicted solar power. The neuron functions
532 M. Rupesh et al.

Fig. 2 Relationship between the temperature and irradiance

used in this model are used to train the dataset with the learning rate as 1000 epochs.
It is analyzed and observed that the RMS error value decreases as the increased
learning rate.
The given dataset is split into three categorized as testing, training, and validating
datasets.
A multilayer feed-forward neural network in our proposed method consists of
input layer, two hidden layers and output layer [11]. The input layer consists of the
weather parameters as attributes, and output layer consists of solar power and voltage
as attributes.
Figure 4 [12] is representing multilayer feed-forward back-propagation neural
network.
Cost function of gradient is defined as [13].

1 
h w,b (x) − y 2
J (W, b; x, y) = (1)
2
From the above, the squared-error cost function is defined as
Prediction of Solar Power Using Machine Learning Algorithm 533

Fig. 3 Flowchart for process of ANN

534 M. Rupesh et al.

Fig. 4 Multilayer feed-forward back-propagation neural network model [13]

1  
m 
 λ nl−1   (l) 2
Sl Sl+1
J (W, b) = [ J W, b; x (i) , y (i) + W ji (2)
m i=1 2 l=1 i=1 j=1

1  1
m
   
nl−1   (l) 2
Sl Sl+1
J (W, b) = h w,b x (i) − y (i) 2 + λ W ji (3)
m i=1 2 2 l=1 i=1 j=1

One iteration of gradient decent updates the parameters W, b as follows.

∂
Wilj = Wilj − α J (W, b) (4)
∂ Wilj
∂
bi(l) = bi(l) − α J (W, b) (5)
∂bil

Partial derivation of cost function

1   
m
∂
J (W, b) = [ J W, b; x (i) , y (i) + λWi(l)
j (6)
∂ Wi j
l m i=1

1   
m
∂
J (W, b) = [ J W, b; x (i) , y (i) (7)
∂bil m i=1
Prediction of Solar Power Using Machine Learning Algorithm 535

4 Simulation and Results for Solar Prediction Using ANN

Model

The feed-forward back-propagation model is used as neural network model to study

the prediction of solar generation using the collected dataset by following the various
steps in developing neural network model. The steps might be included like dividing
the collected data as 70% as training data, 15% data as testing, and 15% data as
validating data, after dividing the data develop the neural network model [14] using
nntool in matlab to predict the solar generation [14, 15]. The flowchart for developing
the neural network model using feed-forward back-propagation algorithm is shown
in Fig. 5.
The proposed neural network model using feed-forward back-propagation algo-
rithm is shown in Fig. 6.
Figure 7 shows the training of NN for 100 epochs.
Figure 8 shows the performance of the NN model using FFBP algorithm for
training, testing, and validation processes.
Figure 9 shows the regression values of the NN model using the FFBP algorithm
for the prediction of solar generation at any location.

Fig. 5 Flowchart for

preparing the neural network
Start
model using FFBP algorithm

Read Training Data

Read TesƟng Data

Use Neural Network for Training and

Get PredicƟon result

Plot PredicƟon Result

Plot Regression PredicƟon Result

End
536 M. Rupesh et al.

Fig. 6 NN model using FFBP algorithm

Fig. 7 Training the NN model

The R value for the given model is about 0.994, which tells about the accuracy of
the predicted model.

5 Conclusion

In this paper, the Sun irradiance, temperature, wind velocity, humidity, as input
variables, solar-generated voltage, power as output variables have been connected
from the BVRIT Narsapur, Medak Dist. Telangana, and generalized artificial
Prediction of Solar Power Using Machine Learning Algorithm 537

Fig. 8 Performance curves of NN model

neural network model using machine learning algorithm, i.e., feed-forward back-
propagation algorithm have been developed for weather and solar power forecasting
using MATLAB/Simulink application. Finally, it can be concluded that the solar
forecasting is achieved with the accuracy of 99.4%; hence, our model can be used to
estimate the power generation from any solar plant at any location.
538 M. Rupesh et al.

Fig. 9 Regression plot for the NN model with FFBP algorithm

Acknowledgements The authors would like to thank BVRIT, Narsapur Solar Plant in charge Mr.
N. Ramchandar, Associate Professor, EEE, BVRIT, Narsapur, and Mr. M. Sudheer Kumar, Assistant
Professor, BVRIT HYDERABAD College of Engineering for Women, Hyderabad.

References

1. Rupesh M, Shivalingappa TV (2019) Evaluation of optimum MPPT technique for PV system

using MATLAB/simulink fig 2. Equivalent circuit of PV cell. 5:1403–1408
2. Revana G, Kota VR (2020) Simulation and implementation of resonant controller based PV
fed cascaded boost-converter three phase five-level inverter system. J King Saud Univ—Eng
Sci 32(7):411–424
3. Trivedi S (2021) Evaluation of the use of artificial neural networks to predict the photovoltaic
power generation factors by using feed forward back propagation (FFBP) technique. Int J Curr
Sci Res Rev 04(02):113–119
4. Jawaid F, Nazirjunejo K (2017) Predicting daily mean solar power using machine learning
regression techniques. 2016 6th Int Conf Innov Comput Technol INTECH 2016:355–360
 Prediction of Solar Power Using Machine Learning Algorithm 539

5. Wu YK, Chen CR, Abdul Rahman H (2014) A novel hybrid model for short-term forecasting
in PV power generation. Int J Photoenergy 2014
6. Coelho JP, Boaventura-Cunha J (2014) Long term solar radiation forecast using computational
intelligence methods. Appl Comput Intell Soft Comput 2014(December):1–14
7. Gupta A, Kumar P, Pachauri RK, Chauhan YK (2014) Performance analysis of neural network
and fuzzy logic based MPPT techniques for solar PV systems. 2014 6th IEEE Power India Int
Conf 1–6
8. Khan I, Zhu H, Khan D, Panjwani MK (2018) Photovoltaic power prediction by cascade
forward artificial neural network. 2017 Int Conf Inf Commun Technol ICICT 2017
2017(December):145–149
9. Ahmed R, Sreeram V, Mishra Y, Arif MD (2020) A review and evaluation of the state-of-the-
art in PV solar power forecasting: techniques and optimization. Renew Sustain Energy Rev
124(June 2019):109792
10. Aljanad A, Tan NML, Agelidis VG, Shareef H (2021) Neural network approach for global
solar irradiance prediction at extremely short-time-intervals using particle swarm optimization
algorithm. Energies 14(4)
11. Shekher A, Khanna V (2016) Modelling and prediction of 150KW PV array system in Northern
India using artificial neural network. 5(5):18–25
12. Kabilan R, et al. (2021) Short-term power prediction of building integrated photovoltaic (BIPV)
system based on machine learning algorithms. Int J Photoenergy 2021
13. (2015) Multi-layer neural network neural network model. http://deeplearning.stanford.edu/tut
orial/supervised/MultiLayerNeuralNetworks/. pp 1–6
14. Shaik NB, Pedapati SR, Ammar Taqvi SA, Othman AR, Abd Dzubir FA (2020) A feed-forward
back propagation neural network approach to predict the life condition of crude oil pipeline.
Processes 8(6)
15. Choudhary A, Pandey D, Bhardwaj S (2020) Artificial neural networks based solar radiation
estimation using backpropagation algorithm. Int J Renew Energy Res 10(4):1566–1575

View publication stats