COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

SOLAR POWER FORECASTING USING DEEP LEARNING

Article Published by -
Meftah Elsaraiti and Adel Merabet

Submitted to - Harish Sharma Sir

PRESENTATION BY -
Happy Saini (22/290)
Munish Kumar (22/318)
Pankaj Kumar (22/326)
Harshita Narban (23/724)
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

OVERVIEW -

Abstruct
Introduction
Literature Review
Methodology
Result and Discussions
Conclusion
Reference
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Absrtuct

This article discusses a method for predicting the solar

enegry using a deep learning technique based on the Long
Short Term Memory (LSTM) algorithm.
An evaluation of the performance of the LSTM network has
been compared it with the Multi- layer Perceptron (MLP).
Network using: Mean Absolute Error (MAE), Mean Absolute
Percentage Error (MAPE), Root Mean Squared Error (RMSE)
and Coefficient of Determination (R2).
The prediction result shows that the LSTM network gives
the best results for each category of days.
Thus, it provides reliable information that enables more
efficient operation of photovoltaic power plants in the
future.
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Introduction
Speaking of solar energy in particular, its unexpected
behavior brings with it a series of problems when
generating energy, such as voltage variations, power
factor details, and stability

ARIMA - Auto-Regressive Integrated Moving Averages models have proven to

be extremely useful for the short-term prediction of high- frequency
time series.

ANNs - powerful,
In contrast to ARIMA models, artificial neural networks are more
especially in representing complex relationships that
exhibit nonlinear behaviors

In recent years, as a result of the research on artificial intelligence,

deep learning based on ANN has come to light to become very
popular due to its capability to accelerate the solution of some
difficult computer problems
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

MLP
Multi-layer perceptron (MLP)-type ANNs can be used to complex
relationships, they are incapable of capturing the long- and
short-term dependencies present in historical data.
ANN to identify and remember behavior patterns from the
distant past and the near past, respectively.
An ANN is needed to make predictions of sequential data
behavior

RNN
the characteristic of these networks to have internal feed- back
loops
these networks presented a great disadvantage, which is known
the vanishing gradient problem.
This problem refers to the difficulty of training these networks
with methods based on gradients or backpropagation algorithms
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

LSTM
Long Short Term Memory (LSTM)-type RNNs emerged
in 1997 as a solution due to their memory units in the
network cells

In this study, the task of predicting the photovoltaic

power for the coming days in short time intervals (30
minutes) from data previously recorded during one
year was considered using Long Short Term Memory
(LSTM)
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Literature Review
In the presence of noisy data, the LSTM model provides a more
accurate prediction compared to the ARIMA model
The use of deep learning techniques has allowed excellent
results in classification and regression problems to be obtained.
Research has focused on describing vari- ous artificial
intelligence (AI) techniques used in renewable energy for their
prediction and optimization
This study examines the importance and benefits of artificial
neural networks in predicting environmental variables and
estimating some unconventional energy systems.
LSTM has become one of the most widely used artificial neural
networks for short-term wind speed predictions
Long short term memory (LSTM) neural networks were used to
predict solar power panels and wind power in the medium and
long term.
The LSTM has been proposed to make forecasts of wind speed on
24-hour wind farms. This work compares the results obtained by
LSTM with MLP, deep versions of MLP, and classical methods.
LSTM showed better efficiency prediction results than MLP and
classical methods
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Methodology
In this section, the sources of information and data
processing methods are described, and the building of deep
learning- based neural networks to predict solar radiation is
presented.

Source of information
Prediction techniques for solar power forecasting
Multilayer perceptrons (MLP)
Long short term memory (LSTM)
performance evaluation
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Source Of Information
The original photo- electric data were collected at 5-
minute intervals and included 365 × 120 = 43800
measurements. Each missing value is processed by
averaging the previous 60 min values. Data is
summarized at 30-minute intervals as the task is to
make a forecast for each half-hour for the next day.
Thus, we have 20 values in one day and 20 × 365 × 1 =
7300 values in one year.
The fact that the observation recording period is 5 minutes.
Thus, within one hour, you will have 12 observations. In turn,
188 (12 × 24) values.

In this study, the MATLAB software

(R2019b) was used for the training
process of the LSTM
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

PREDICTION TECHNIQUES FOR

SOLAR POWER FORECASTING

Artificial Neural Networks (ANN) are part of the area of

knowledge of Artificial Intelligence (AI) and Deep
Learning, The ability to learn nonlinear relationships and
their ability to model complex systems have made them a
useful tool in different scientific fields

Deep learning techniques can be used in forecasting and

identifying characteristics of each one; among them:
stages implemented for the approximation, the accuracy
with which the power generation is approximated, the
convergence time and the uncertainty associated with
the forecast.
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Multilayer Perceptrons (MLP)

They are a type of ANN capable of modeling complex
functions. They are apt to ignore noise and irrelevant
inputs, and they can easily adapt their weights.

The learning process of an MLP can be divided into two main phases:
the input of data through the inputs of the MLP and the correction of
the errors,
The errors are calculated by comparing the real data against the
answer that it delivers to the model through a technique known as
Backpropagation.
This iteration is repeated multiple times to reduce the error, using an
algorithm to obtain a result with better aptitude.

The neuron performs two types of operations: first, the

propagation rule; and later, the activation function
where, Zj is the result of the propagation rule applied to neuron j, xi is
The propagation rule (Z) the input vector i, wij is the weight that joins input i with neuron j, and
bj is the threshold associated with neuron j.

The activation function (A) where, Aj is the activation of neuron j and f

is the activation function.
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

LONG SHORT TERM MEMORY (LSTM)

LSTM networks are currently one of the most popular models in deep
learning applied to time series prediction.
LSTM structures are distinguished by their ability to model and
recognize temporal patterns from data sequences. In this way, they
are able to process the sequences and records of the available
operational data,

The recurrent neural network has a structure where information from the
previous step is passed to the next step and used for estimation.
Accordingly, the recurrent neural network has achieved great results in
estimating sequential information, that is, time series data. However, as the
length of the past data, required for data estimation, increases, the
vanishing gradient problem occurs in the existing recurrent neural network.
This problem can be solved by using the gates of the LSTM algorithm

The memory unit consists of three gates: input gate, forget

gate, and output gate (Input gate (it), Forget gate (ft), and
Output gate (ot))
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

In general, the LSTM structure consists of three different layers:

forget, input and output. In the LSTM architecture, first, xt and
st−1 information are used as inputs, and it is decided which
information to delete. These operations are done in the Forget
layer ft determined -

where, g is the activation function, which is the sigmoid in this work.

where, g is the activation function, which is the

sigmoid in this work.

Then, the candidate information, that will form the new

information, is expressed The drive has one input xt and two drive previous state feedbacks which
are the previous state output st−1 and the state variable ct−1. The gates
use a sigmoid activation function g, while states use a tanh function. The
memory unit of the LSTM can be defined by a set of equations, where w
are the parameters and b are the biases
Finally, the output data is obtained by using the
following expressions in the output layer.
The process described above continues iteratively. Weight
parameters (w) and bias parameters (b) are learned by the model in a
way that minimizes the difference between actual training values and
LSTM output values
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

PERFORMANCE EVALUATION
1. Mean absolute error (MAE) 4. Coefficient of determination (R2)
It expresses the mean absolute deviation of the difference between It represents the strength of the linear relationship between
the predicted values and the actual values and is calculated by the predicted values of the models and the actual values and is
calculated by

2. Root mean square error (RMSE)

It represents the standard deviation of the estimation
errors and is calculated by

where, N is the number of samples used for statistical evaluation

criteria, oi is the actual value of the observation, and pi is the
forecasted value of the observation and o ̃ is the average of the
actual observation values.
3. Mean absolute percentage error (MAPE)
It measures the prediction accuracy of the models as a
percentage and is calculated by
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Result & Discussion

In order to implement the proposed model, 43,800 records were
taken during daylight hours with a 5-minute solar interval in Halifax,
corresponding to the days from January 1, 2017 to December 31,
2017 available at the Nova Scotia Community College (NSCC) and to
find the missing values. A prediction was made every 30 minutes for
a winter day (December 31), and a summer day (June 30) for the
total of 7300 values, which were divided into 7280 training data, and
the last 20 for the prediction test.

A case study was carried out using original photovoltaic data for
one year during daylight hours, from 8 am to 5 pm at 5-minute
intervals. In this work, data is summarized at 30-minute intervals to
make a half-hour forecast for the next day

The forecasted values of the proposed LSTM model used to predict

the solar power and the time graph showing the
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

The results show the importance of the time series used

to train the models. The quality of the data has a
significant impact on the performance obtained by the
forecasting model, especially the outliers. A specific
time series may perform better with one model, while
the same model may perform poorly with another time
series model. Consequently, it is convenient to consider
multiple models to search for suitable predictions.
However, there are several techniques for time series
forecasting that can be implemented according to the
methodology proposed in this study.
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

Conclusion
This data was split into two categories: training and test sets. While only the
training data was used in the learning process of the model, the test data was not
used in the learning process.
The task of predicting photovoltaic power for the coming days at 30-minute
intervals was considered. The results of the suggested model were compared
using the Multi-Layer Perceptron (MLP) algorithm, which is the most extensively
used technique in the literature, in order to evaluate and examine their
correctness and performance.
When compared to the MLP algorithm, the prediction performance of the
suggested LSTM model offered more effective values in all performance
parameters MAE, MAPE, RMSE, and R2 for each category of days, according to the
findings of simulation tests and the results presented in Tables 1 and 2. The
proposed model gives trustworthy data that will allow photovoltaic power plants
to operate more efficiently in the future.

The quality of the data, particularly the outliers, has a substantial impact on the
prediction model’s performance. Several time series prediction approaches can
be implemented using the methodology provided in this work.
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

References
D. Anderson and M. Leach, ‘‘Harvesting and redistributing renewable energy:
T.Rojat,R.Puget,D.Filliat,J.DelSer,R.Gelin,andN.Díaz-Rodríguez, ‘‘Explainable artificial intelligence (XAI) on TimeSeries data: A
survey,
J.BediandD.Toshniwal,‘‘Deeplearningframeworktoforecastelectricity demand,’
S.Hochreiter,‘‘Thevanishinggradientproblemduringlearningrecurrent neural nets and problem solutions,’’
M. Elsaraiti and A. Merabet, ‘‘A comparative analysis of the ARIMA and LSTM predictive models and their effectiveness for
predicting wind speed,’’
M. Elsaraiti and A. Merabet, ‘‘Application of long-short-term-memory recurrent neural networks to forecast wind speed,’’
J. Bottieau, F. Vallee, Z. De Greve, and J.-F. Toubeau, ‘‘Leveraging provi- sion of frequency regulation services from wind
generation by improving day-ahead predictions using LSTM neural networks,’’
E.Pasero,G.Raimondo,andS.Ruffa,‘‘MULP:Amulti-layerperceptron application to long-term, out-of-sample time series prediction,’’
M. Abdel-Nasser and K. Mahmoud, ‘‘Accurate photovoltaic power fore- casting models using deep LSTM-RNN,’’
COMPUTER SCIENCE DEPARTMENT OF RAJASTHAN TECHNICAL UNIVERSITY

THANKYOU
 RECENT TOPICS
Total Question – 15
Very Short Answer – 5
Long Answer – 5
Very Long Answer – 5

Here is a question paper based on the article "Solar Power Forecasting Using Deep
Learning," along with answers.

1.What is the primary challenge in solar photovoltaic (PV) technology?

Answer: The primary challenge is the accurate prediction of solar PV generation due to
its dependence on climatic and meteorological conditions.

2.What are the two types of deep learning models compared in the study?
Answer: The study compares Long Short-Term Memory (LSTM) networks and Multilayer
Perceptron (MLP) networks.

3.What is the function of the 'forget gate' in an LSTM network?

Answer: The forget gate decides which information should be discarded from the cell
state, allowing the network to focus on relevant data for predictions.

4.What kind of data is used for training the LSTM model in this study?
Answer: Solar power measurements taken every 5 minutes during daylight hours over a
year in Halifax, Nova Scotia.

5.What are the four performance evaluation criteria used in this study?
Answer: The criteria are Mean Absolute Error (MAE), Mean Absolute Percentage Error
(MAPE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R²).

1.Explain the vanishing gradient problem in RNNs and how LSTM networks address
it.
Answer: The vanishing gradient problem occurs in RNNs when gradients become too
small during backpropagation, preventing the network from learning effectively over
long sequences. LSTM networks solve this issue with memory cells and gates (input,
forget, and output) that regulate the flow of information, retaining important data over
extended time periods, thereby mitigating the vanishing gradient problem.
2.Discuss the methodology used for solar power forecasting in the study.
Answer: The study collects solar power data at 5-minute intervals over a year and
processes it by summarizing into 30-minute intervals. Missing data is handled by
averaging the previous 60 minutes' values, and the LSTM model is trained using
MATLAB. The performance is compared using statistical evaluation metrics such as
MAE, MAPE, RMSE, and R². The LSTM model outperforms MLP in these criteria.

3.Compare the performance of LSTM and MLP models based on the results.
Answer: The LSTM model consistently outperforms the MLP model in predicting solar
power generation. LSTM has superior accuracy across MAE, MAPE, RMSE, and R² due to
its ability to recognize temporal dependencies, while the MLP model, which does not
consider time-series data, struggles with complex patterns【5†source】.

4.Describe the role of Artificial Neural Networks (ANNs) in renewable energy

forecasting.
Answer: ANNs, particularly deep learning models like LSTM, play a crucial role in
renewable energy forecasting by modelling complex, nonlinear relationships in data.
They surpass traditional methods like ARIMA, which are limited to simpler, linear
relationships, making ANNs ideal for scenarios such as solar power forecasting, where
data is often noisy and influenced by external factors.

5.Analyse the impact of data preprocessing techniques on the LSTM model’s

performance.
Answer: Data preprocessing, such as noise reduction and outlier elimination,
enhances the accuracy of the LSTM model. By ensuring cleaner and more consistent
data, preprocessing allows the model to focus on identifying meaningful patterns
without being misled by anomalies, ultimately improving prediction reliability.

1.Explain in detail how the LSTM architecture works and why it is suitable for time-
series forecasting.
Answer: LSTM networks consist of memory cells and three gates: input, forget, and
output. The input gate controls the flow of new information into the cell, the forget gate
decides what information to discard, and the output gate determines the information to
be passed to the next cell. This architecture makes LSTMs effective at modelling long-
term dependencies in sequential data, making them ideal for time-series forecasting
where future predictions depend on past values. Unlike MLPs, LSTMs can remember
relevant information over extended periods, avoiding the issue of vanishing gradients
typically faced by standard RNNs.

2.Discuss the importance of solar power forecasting and how deep learning
models contribute to improving its accuracy.
Answer: Solar power forecasting is essential for effective energy management and grid
reliability due to the intermittent nature of solar energy production. Accurate forecasts
allow grid operators to balance supply and demand, prevent power outages, and
optimize energy storage. Deep learning models, particularly LSTMs, enhance the
accuracy of these forecasts by modelling complex, nonlinear relationships in weather
and solar power data. LSTMs are capable of learning from historical patterns, capturing
temporal dependencies that are crucial in predicting solar generation trends.

3.Compare traditional statistical methods and deep learning approaches in

renewable energy forecasting. What are the strengths and limitations of each?
Answer: Traditional statistical methods, such as ARIMA, are based on linear
relationships and work well for simple, predictable patterns. However, they struggle
with nonlinearities and complex relationships, which are common in renewable energy
data. Deep learning models, such as LSTMs, overcome these limitations by
automatically learning complex, nonlinear relationships from large datasets, making
them more accurate for forecasting. However, deep learning models require more
computational power, larger datasets, and can be prone to overfitting if not properly
regularized.

4. Describe the data collection, processing, and normalization techniques used in

the study. Why are these steps critical for model performance?
Answer: The data for solar power forecasting is collected every 5 minutes and
summarized into 30-minute intervals for the study. Missing data is addressed by
averaging previous values, and normalization is applied to scale the data, making it
suitable for training deep learning models. These preprocessing steps are critical
because they ensure that the data is consistent and clean, reducing the impact of noise
or outliers that could distort the model's learning process. Properly processed data
allows the LSTM model to identify meaningful patterns, leading to more accurate
predictions.

5. Evaluate the benefits and challenges of using LSTM networks for solar power
forecasting compared to other machine learning techniques.
Answer: LSTM networks offer significant benefits in solar power forecasting due to their
ability to model temporal dependencies, making them well-suited for time-series data.
They outperform traditional machine learning models like MLPs in accuracy and
robustness. However, the challenges include the need for extensive computational
resources, longer training times, and the potential complexity of tuning
hyperparameters like learning rate and epochs. Additionally, LSTM networks are
sensitive to the quality of the input data, making preprocessing a critical factor in their
success.