FORECASTING OF FUTURE HOUSEHOLD

ELECTRICAL LOAD USING MACHINE LEARNING

A

Project Report

Submitted to

JHARKHAND UNIVERSITY OF TECHNOLOGY, RANCHI

In Partial Fulfillment of the Requirement for the Award of Degree of

BACHELOR OF TECHNOLOGY
IN
ELECTRICAL ENGINEERING

Submitted by

Kunal Ranjan Singh 18030450044

Piyush Kumar 18030450058
Roopam Kr. Raj 18030450072
Sanskar Ram 18030450081
Soham Vivek 18030455015

Under the guidance of

Dr. Amit Kumar Choudhary

Assistant Professor

DEPARTMENT OF ELECTRICAL ENGINEERING

BIT SINDRI, DHANBAD-828123 (JHARKHAND)
June, 2022
 BIT SINDRI, DHANBAD-828123
Department of Higher & Technical Education, Government of Jharkhand

CANDIDATE’S DECLARATION
We hereby certify that the project entitled “FORECASTING OF FUTURE
HOUSEHOLD ELECTRICAL LOAD USING MACHINE LEARNING” which is
being submitted in the partial fulfillment of the requirement for the award of Bachelor of
Technology in the Department of Electrical Engineering, is a record of our own work
carried out under the supervision and guidance of Dr. Amit Kumar Choudhary, Assistant
Professor, Department of Electrical Engineering, BIT Sindri.
All information in this document has been obtained and presented as per academic
rules and ethical conduct.
To the best of our knowledge, the contents presented in this project report have
not been submitted elsewhere for the award of any other degree/diploma.

Date: Signature:
Place: Kunal Ranjan Singh (18030450044)

Signature:
Piyush Kumar (18030450058)

Signature:
Roopam Kr. Raj (18030450072)

Signature:
Sanskar Ram (18030450081)

Signature:
Soham Vivek (18030455015)
 BIT SINDRI, DHANBAD-828123
Department of Higher & Technical Education, Government of Jharkhand

CERTIFICATE OF DECLARATION FROM THE

SUPERVISOR

This is to certify that the work presented in the project report entitled
“FORECASTING OF FUTURE HOUSEHOLD ELECTRICAL LOAD USING
MACHINE LEARNING” in partial fulfillment of the requirement for the award of
Degree of Bachelor of Technology in the Department of Electrical Engineering at BIT
Sindri is an authentic work carried out under my supervision and guidance.
To the best of my knowledge, the content of this report does not form a basis for
the award of any previous Degree to anyone else.

Date: Signature of supervisor

Place: Name: Dr. Amit Kumar Choudhary
Designation: Assistant Professor
Department Name: Electrical Engineering

2
 BIT SINDRI, DHANBAD-828123
Department of Higher & Technical Education, Government of Jharkhand

CERTIFICATE OF APPROVAL FROM THE EXAMINERS

The foregoing project report entitled “FORECASTING OF FUTURE

HOUSEHOLD ELECTRICAL LOAD USING MACHINE LEARNING”, is hereby
approved as a creditable study of the research topic and has been presented satisfactorily
to warrant its acceptance as a prerequisite to the degree for Bachelor of Technology in
the Department of Electrical Engineering at BIT Sindri.
It is understood that by this approval, the undersigned do not necessarily endorse
any conclusion drawn or opinion expressed therein, but approve the report for the purpose
for which it is submitted.

Signature of Internal Examiner Signature of External Examiner

Name: Name :
Designation: Designation:
Name of Department: Name of Department:
Name of the Institute: Name of the Institute:

Signature of HOD
Name: Dr. Md Abul Kalam
Head, Department of Electrical Engineering, BIT Sindri

3
 ACKNOWLEDGMENT
We would like to articulate our deep gratitude to our project guide Dr. Amit
Kumar Choudhary, Assistant Professor, Department of Electrical Engineering who have
always been source of motivation and firm support for carrying out the project. We also
express our gratitude to him for his invaluable suggestion and constant encouragement all
through the project work.

We would also like to convey our sincerest gratitude and indebtedness to the Head
of the Department and all other faculty members and staff of Department of Electrical
Engineering, BIT Sindri, who bestowed their great effort and guidance at appropriate
times without it, would have been very difficult on our work.

An assemblage of this nature could have been attempted with our reference to and
inspiration from the works of others whose details are mentioned in references section.
We acknowledge our indebtedness to all of them. Further, we would like to express our
feeling towards our parents and god who directly or indirectly encouraged and motivated
us during this dissertate.

Date: Signature:
Place: Kunal Ranjan Singh (18030450044)

Signature:
Piyush Kumar (18030450058)

Signature:
Roopam Kr. Raj (18030450072)

Signature:
Sanskar Ram (18030450081)

Signature:
Soham Vivek (18030455015)

4
 INDEX

TITLE PAGE NO.

Declaration 1
Certificate of Declaration from the Supervisor 2
Certificate of Approval from the Examiners 3
Acknowledgement 4
Index 5
Abstract 7
List of Figures 8
List of Tables 8
1. INTRODUCTION 9
1.1 Overview of Machine Learning 9
2. OBJECTIVE 10
2.1 Objective of the work 10
3. METHODOLOGY OF THE WORK 11
3.1 Flowchart of the work 11
3.2 Gathering of data 11
3.3 Data analysis and Pre-processing 13
3.3.1 Data Cleaning 14
3.3.2 Data Integration 14
3.3.3 Data Reduction 15
3.3.4 Data Transformation 15
3.4 Data Visualization 15
3.4.1 Line Charts 16
3.4.2 Bar Graphs 17
3.4.3 Histograms 17
3.4.4 Scatter Plots 18
3.4.5 Violin Plots 18
3.4.6 Box Plots 19
3.5 Training, Testing and Evaluating Machine Learning Models 19
3.5.1 Splitting the dataset 19

5
 3.5.2 Model training and testing 21
3.5.3 Hyper parameter tuning 24
3.5.4 Model Evaluation 25
4. RESULTS AND DISCUSSION 27
5. CONCLUSIONS AND FUTURE SCOPES 29
5.1 Conclusions 29
5.2 Future Scopes 29
REFRENCES 30
LIST OF PUBLICATION 31

6
 ABSTRACT
The objective of the project is to identify & predict future large-scale electrical
energy consumption, a more flexible power consumption so as to reduce energy costs and
the impact on the environment as well as utilize periods with large environmental power
generation and minimum load on the grid.
Machine learning (ML) methods has recently contributed very well in the
advancement of the prediction models used for energy consumption. Such models highly
improve the accuracy, robustness, and precision and the generalization ability of the
conventional time series forecasting tools. This project reviews the state of the art of
machine learning models used in the general application of energy consumption. Through
a novel search and taxonomy, the most relevant literature in the field is classified
according to the ML modelling technique, energy type, perdition type, and the application
area. A comprehensive review of the literature identifies the major ML methods, their
application and a discussion on the evaluation of their effectiveness in energy
consumption prediction. This proposed work further makes a conclusion on the trend and
the effectiveness of the ML models. As the result, this work reports an outstanding rise in
the accuracy and an ever-increasing performance of the prediction technologies using the
novel hybrid and ensemble prediction models.

7
 LIST OF FIGURES
Fig No. Title Page No.
3.1 Flowchart of the work 11
3.2 Code Snippet showing the location of dataset in the drive. 12
3.3 Code Snippet for loading of the dataset in the dataframe. 12
3.4 Data Pre-processing of raw data. 13
3.5 Line Chart 16
3.6 Bar Graph 17
3.7 Histogram 17
3.8 Scatter Plot 18
3.9 Violin Plot 18
3.10 Box Plot 19
3.11 Distribution of dataset into different segment 19
3.12 Code snippet for splitting of the dataset 20
3.13 Training dataset after splitting of the dataset 20
3.14 Testing dataset after splitting of the dataset 21
3.15 Code Snippet for training & testing of the linear regression model 22
3.16 Code Snippet for training & testing of the svm regression (linear kernel) 22
model
3.17 Code Snippet for training & testing of the svm regression (poly kernel) 22
model
3.18 Code Snippet for training & testing of the decision tree regression model 23
3.19 Code Snippet for training & testing of the random forest regression model 23
3.20 Code Snippet for training & testing of the xgboost regression model 24
3.21 Hyperparameter tuning 24
3.22 Hyperparameter tuning technique 24
3.23 Models and their respective mse and r2_score values 25
3.24 Plot of r2_score of models 26
3.25 Plot of mse of models 26
4.1 Plot of r2_score of models 28

4.2 Plot of mse of models 28

LIST OF TABLES
Table No. Title Page No.
4.1 Comparison Between Mean Squared Error & R2_Score 27

8
 CHAPTER 1
INTRODUCTION

1.1 Overview of machine learning

Machine learning (ML) methods has recently contributed very well in the
advancement of the prediction models used for energy consumption. Such models highly
improve the accuracy, robustness, and precision and the generalization ability of the
conventional time series forecasting tools [1].
Types of Machine Learning Algorithms:
• Supervised Machine Learning: In supervised learning, both the input and desired
output are provided and the machine must learn how to map the former to the
latter. To accomplish this, the machine is trained on a statistically representative
set of example inputs and corresponding outputs. Supervised Learning deals with
two main tasks Regression and Classification.
• Unsupervised Machine Learning: In unsupervised learning, the machine is not
provided labelled examples or previous patterns on which to base the analysis of
the data inputs. The machine must uncover patterns and draw inferences by itself,
without having the correct answers. It will classify or cluster data by discovering
the similarity of features on its own. Unsupervised Learning deals with clustering
and associative rule mining problems.
• Reinforcement Learning: Reinforcement learning continuously improves its model
based on feedback from experiences. It learns through trial and error – from the
consequences of its action and by new choices. As an action is taken, the success
of the outcome is graded and receives either a positive or negative score.
Reinforcement Learning deals with exploitation or exploration, Markov’s decision
processes, Policy Learning, Deep Learning and value learning [2].

9
 CHAPTER 2
OBJECTIVE
2.1 Objective of the Work

This project reviews the state of the art of machine learning models used in the
general application of energy consumption. Through a novel search the most relevant
literature in the field is classified according to the ML modelling technique, energy type,
perdition type, and the application area.
A comprehensive review of the literature identifies the major ML methods, their
application and a discussion on the evaluation of their effectiveness in energy
consumption prediction. This literature reviews the training of the different machine
learning models and their performance score on the standard comparison metrics.
This project further makes a conclusion on the trend and the effectiveness of the
data analysis & pre-processing, data visualization, outlier removal and data splitting prior
to start building ML models. The different state of the art machine learning will be trained
with the help of the available dataset with hyper parameter tuning to improve their
accuracy and appropriate model will be selected among all the trained models. As the
result, this project reports an outstanding rise in the accuracy and an ever-increasing
performance of the prediction technologies using the novel hybrid and ensemble
prediction models [3] [4].
With the use of machine learning the below optimization are produced in the power
system:
1. The decrease in the maintenance cost of the power system.
2. Sufficient supply of electrical energy to household as per the requirement
3. To predict upcoming future load.
4. Improved energy efficiency, reliability, and security
5. Reducing power generation costs [5] [6] [7].

10
 CHAPTER 3
METHODOLOGY OF THE WORK
3.1 Flowchart of the work

The below flowchart that is represented by Fig. 3.1 shows the various steps
involved in the project.

Gathering of the data / Data Ingestion

Data Analysis

Data Pre-Processing

Data Visualization

Training of the machine learning

models

Testing of the machine learning

models

Hyperparameter tuning of the

machine learning models

Evaluation and comparison of the

machine learning models with the
help of performance metrics

Fig. 3.1 Flowchart of the work.

3.2 Gathering of the data

The process of gathering data depends on the type of project one desires to make,
if anyone want to make an ML project that uses real-time data. The data set can be
collected from various sources such as a file, database, sensor and many other such

11
sources but the collected data cannot be used directly for performing the analysis process
as there might be a lot of missing data, extremely large values, unorganized text data or
noisy data. Data pre-processing is one of the most important steps in machine learning [1]
[3] [4]. It is the most important step that helps in building machine learning models more
accurately. In machine learning, there is an 80/20 rule.
In this project the data for training the machine learning model is collected from
the UC Irvine Machine learning repository, which is an open-source machine learning
dataset platform for performing machine learning. The dataset on which machine learning
will be performed is stored in the hierarchical directory and it is further loaded in pandas
data frame with the help of python library for performing various kind of exploratory data
analysis and data visualization tasks [3]. Fig 3.2 and fig 3.3 depicts the location of dataset
in drive and loading of dataset in data frame respectively

Fig 3.2 Code Snippet showing the location of dataset in the drive.

Fig 3.3 Code Snippet for loading of the dataset in the data frame.

12
3.3 Data Analysis & Pre-processing

Data pre-processing is the process of transforming raw data into an

understandable format. It is also an important step in data mining while working with raw
data. The quality of the data should be checked before applying machine learning or data
mining algorithms [1].
Importance of Data pre-processing
Pre-processing of data is mainly to check the data quality and consistency. The quality
can be checked by the following:
1. Accuracy: To check whether the data entered is correct or not.
2. Completeness: To check whether the data is available or not recorded.
3. Consistency: To check whether the same data is kept in all the places that do or
do not match.
4. Timeliness: The data should be updated correctly.
5. Believability: The data should be trustable.
6. Interpretability: The understandability of the data.

Major Tasks in Data Pre-processing is shown in fig. 3.4.

1. Data cleaning
2. Data integration
3. Data reduction
4. Data transformation

Fig 3.4 Data Pre-processing of raw data.

13
3.3.1 Data Cleaning
It is the process to remove incorrect data, incomplete data and inaccurate data
from the datasets, and it also replaces the missing values. There are some techniques in
data cleaning [3] [4] [5] [6].
Handling missing values:
1. Standard values like “Not Available” or “NA” can be used to replace the missing
values.
2. Missing values can also be filled manually but it is not recommended when that
dataset is big.
3. The attribute’s mean value can be used to replace the missing value when the data
is normally distributed wherein in the case of non-normal distribution median
value of the attribute can be used.
4. While using regression or decision tree algorithms the missing value can be
replaced by the most probable value.
Noisy:
Noisy generally means random error or containing unnecessary data points. Here are
some of the methods to handle noisy data.
1. Regression: This is used to smooth the data and will help to handle data when
unnecessary data is present. For the analysis, purpose regression helps to decide
the variable which is suitable for our analysis.
2. Clustering: This is used for finding the outliers and also in grouping the data.
Clustering is generally used in unsupervised learning [1].
3.3.2 Data Integration
The process of combining multiple sources into a single dataset. The Data
integration process is one of the main components in data management. There are some
problems to be considered during data integration [1] [4]:
• Schema integration: Integrates metadata (a set of data that describes other data)
from different sources.
• Entity identification problem: Identifying entities from multiple databases. For
example, the system or the use should know student _id of one database and student
name of another database belongs to the same entity.
• Detecting and resolving data value concepts: The data taken from different
databases while merging may differ. Like the attribute values from one database
may differ from another database [1].
14
3.3.3 Data Reduction

This process helps in the reduction of the volume of the data which makes the
analysis easier yet produces the same or almost the same result. This reduction also helps
to reduce storage space. There are some of the techniques in data reduction are
Dimensionality reduction, Numerosity reduction, Data compression [1] [4].
• Dimensionality reduction: This process is necessary for real-world applications as
the data size is big. In this process, the reduction of random variables or attributes is
done so that the dimensionality of the data set can be reduced. Combining and
merging the attributes of the data without losing its original characteristics. This
also helps in the reduction of storage space and computation time is reduced. When
the data is highly dimensional the problem called “Curse of Dimensionality” occurs.
• Numerosity Reduction: In this method, the representation of the data is made
smaller by reducing the volume. There will not be any loss of data in this reduction.
• Data compression: The compressed form of data is called data compression. This
compression can be lossless or lossy. When there is no loss of information during
compression it is called lossless compression. Whereas lossy compression reduces
information but it removes only the unnecessary information.

3.3.4 Data Transformation

The change made in the format or the structure of the data is called data
transformation. This step can be simple or complex based on the requirements. There are
some methods in data transformation [4].

• Smoothing: With the help of algorithms, noises are removed from the dataset and
this helps in knowing the important features of the dataset. By smoothing we can
find even a simple change.
• Aggregation: In this method, the data is stored and presented in the form of a
summary. The data set which is from multiple sources is integrated into with data
analysis description.
• Discretization: The continuous data here is split into intervals. Discretization
reduces the data size.
• Normalization: It is the method of scaling the data so that it can be represented in a
smaller range. Example ranging from -1.0 to 1.0.

3.4 Data Visualization

15
 The process of finding trends and correlations in the data by representing it
pictorially is called Data Visualization. To perform data visualization in python, we will
use the various python data visualization modules such as Matplotlib, Seaborn and Plotly
are used. Data visualization is a field in data analysis that deals with visual representation
of data. It graphically plots data and is an effective way to communicate inferences from
data [1] [3] [4] [5].
Using data visualization, a visual summary of the dataset can be obtained. With
pictures, maps and graphs, it becomes very easy to get the insight about the data. Data
visualization plays a significant role in the representation of both small and large data
sets, but it is especially useful when the dataset is large, in which it is impossible to see all
of the available data. Python offers several plotting libraries, namely Matplotlib, Seaborn
and many other such data visualization packages with different features for creating
informative, customized, and appealing plots to present data in the most simple and
effective way [3] [5].
Matplotlib and Seaborn are two python libraries used for data visualization. They
have inbuilt modules for plotting different graphs. While Matplotlib is used to embed
graphs into applications, Seaborn is primarily used for statistical graphs [5].
Other different data visualization techniques are Line Charts, Bar Graphs,
Histograms, Scatter Plots, Violin-plot and Box-plot. These techniques are discussed
herewith for brief understanding.

3.4.1 Line Charts

A line chart as shown in fig. 3.5 is a graph that represents information as a series
of data points connected by a straight line. In line charts, each data point or marker is
plotted and connected with a line or curve.

Fig 3.5 Line Chart.

16
3.4.2 Bar Graphs
When the data is categorical data, it can be represented with a bar graph. A bar
graph plots data with the help of bars, which represent value on the y-axis and category
on the x-axis. Bar graphs use bars with varying heights to show the data which belongs to
a specific category. The fig. 3.6 represents information as bar graph of datapoints with
help of vertical bars.

Fig 3.6 Bar Graphs.

3.4.3 Histogram

A Histogram is a bar representation of data that varies over a range. It plots the
height of the data belonging to a range along the y-axis and the range along the x-axis.
Histograms are used to plot data over a range of values. It uses a bar representation to
show the data belonging to each range. A histogram is shown in fig. 3.7.

Fig 3.7 Histogram.

17
3.4.4 Scatter Plots
Scatter plots shown in fig. 3.8 are used to plot two or more variables present at
different coordinates. The data is scattered all over the graph and is not confined to a
range. Two or more variables are plotted in a Scatter Plot, with each variable being
represented by a different colour.

Fig 3.8 Scatter Plot

3.4.5 Violin-Plot
A violin plot shown in fig.3.9 is a method of plotting numeric data. It is similar to
a box plot, with the addition of a rotated kernel density plot on each side. Violin plots are
similar to box plots, except that it also shows the probability density of the data at
different values, usually smoothed by a kernel density estimator.

Fig 3.9 Violin Plot.

18
3.4.6 Box-Plot
A box plot shows the distribution of quantitative data in a way that facilitates
comparisons between variables or across levels of a categorical variable. The box shows
the quartiles of the dataset while the whiskers extend to show the rest of the distribution,
except for points that are determined to be “outliers” using a method that is a function of
the inter-quartile range. The box plot is represented in fig 3.10.

Fig 3.10 Box Plot

3.5 Training, Testing and Evaluating Machine Learning Models

Training a machine learning model is a process in which a machine learning
algorithm is fed with training data from which it can learn. ML models can be trained to
benefit businesses in numerous ways, by quickly processing huge volumes of data,
identifying patterns, finding anomalies or testing correlations that would be difficult for a
human to do unaided [1] [3] [4] [7].

3.5.1 Splitting the dataset

The data collected for training needs to be split into three different sets: training,
validation and test as shown in fig 3.11. The code snippet for splitting the dataset is
shown in fig 3.12.

Fig 3.11 Distribution of dataset into different segment

19
 Training Dataset: Up to 75 percent of the total dataset is used for training. The
model learns on the training set; in other words, the set is used to assign the weights and
biases that go into the model. Training dataset is depicted in fig 3.13.
Validation Dataset: Between 15 and 20 percent of the data is used while the
model is being trained, for evaluating initial accuracy, seeing how the model learns and
fine-tuning hyperparameters. The model sees validation data but does not use it to learn
weights and biases.
Testing Dataset: Between five and 10 percent of the data is used for final
evaluation. Having never seen this dataset, the model is free of any of its bias. Testing
dataset is depicted in fig 3.14.

Fig 3.12 Code snippet for splitting of the dataset.

Fig 3.13 Training dataset after splitting of the dataset.

20
 Fig 3.14 Testing dataset after splitting of the dataset.
3.5.2 Model training and testing
Model training for machine learning includes splitting the dataset, training
different machine learning models, tuning hyperparameters and performing batch
normalization. In this work with the available dataset the below mentioned machine
learning models were trained [1] [2]:
• Linear Regression Model (Fig 3.15)
• Support Vector Machine Regression Model (linear kernel) (Fig 3.16)
• Support Vector Machine Regression Model (poly kernel) (Fig 3.17)
• Decision Tree Regression Model (Fig 3.18)
• Random Forest Regression Model (Fig 3.19)
• XgBoost Regression Model (Fig 3.20)

21
 Fig 3.15 Code Snippet for training & testing of the linear regression model.

Fig 3.16 Code Snippet for training & testing of the svm regression (linear kernel) model

Fig 3.17 Code Snippet for training & testing of the svm regression (poly kernel) model

22
Fig 3.18 Code Snippet for training & testing of the decision tree regression model

Fig 3.19 Code Snippet for training & testing of the random forest regression model

23
 Fig 3.20 Code Snippet for training & testing of the xgboost regression model

3.5.3 Hyperparameter tuning

Hyperparameters can be imagined as settings for controlling the behaviour of a
training algorithm, as shown below, in fig 3.21 [9] [10] [11].

Fig 3.21 Hyperparamater tuning

There are two common approaches to tuning hyper parameters, as depicted in the fig
3.22.

Fig 3.22 Hyperparamater tuning technique.

24
 The first, standard grid search optimization, is a brute-force approach through a
predetermined list of combinations of hyperparameters. All possible values of
hyperparameters are listed and looped through in an iterative fashion to obtain the best
values. Grid search optimization takes relatively little time to program and works well if
the dimensionality in the feature-vector is low. But as the number of dimensions
increases, tuning takes longer and longer [1] [7].
The other common approach, random search optimization, involves randomly
sampling values instead of exhaustively searching through every combination of
hyperparameters. Generally, it produces better results in less time than grid search
optimization [7].

3.5.4 Model Evaluation

Evaluation of the machine learning models on a performance metrics is an

essential step after the training of the models. It is very useful step for the selection of the
appropriate model for the current need of application. There are several evaluation
metrics, like r2_score, mean squared error, confusion matrix, cross-validation, AUC-ROC
curve etc [7].

Fig 3.23 Models and their respective mse and r2_score values.

Different evaluation metrics are used for different kinds of problems. In this work
r2_score and mean squared error metrics of the scikit-learn library will be used to
compare the performance of different models as shown in fig 3.23, fig 3.24 and fig 3.25.

25
Fig 3.24 Plot of r2_score of models. Fig 3.25 Plot of mse of models.

The next chapter will discuss about the results obtained in this proposed work.

26
 CHAPTER 4
RESULTS AND DISCUSSION
The various machine learning models were trained along with hyperparameter
tuning with the help of the training dataset obtained after the splitting of the dataset. After
the training of the various machine learning models with the help of testing dataset,
r2_score metric of scikit-learn and mean squared error metric of scikit-learn their
performance were checked.

The models which were trained are linear regression model, support vector
machine model with linear kernel, support vector machine model with poly kernel,
decision tree model, random forest model, Xgboost model as tabulated in table 4.1

From the table, it can be clearly observed that, the Xgboost model has the highest
r2_score. The obtained score is satisfactorily acceptable for the final inference purpose.

TABLE 4.1: COMPARISON BETWEEN MEAN SQUARED ERROR & R2_SCORE

Model r2_score mean-squared_error
Linear regression model 0.8156 15.1421

Support vector machine model 0.8130 15.3605

with linear kernel
Support vector machine model 0.6076 32.2294
with poly kernel

Decision tree model 0.7961 16.7435

Random forest model 0.8093 15.6625

Xgboost model 0.8455 12.6865

The fig 4.1 and fig 4.2 obtained by the training in the earlier chapter depict the acceptable
result from the Xgboost model for mean squared error and r2_score of all the models.

27
Fig 4.1 Plot of r2_score of models. Fig 4.2 Plot of mse of models.

28
 CHAPTER 5
CONCLUSIONS AND FUTURE SCOPES

5.1 Conclusions

Despite the increasing requirement of electrical energy, there must be a proper

technology and system that can predict the upcoming requirement of the household
electrical load. Based on that prediction the energy can be fulfilled as per the requirement.
To address this problem, there is a need of infrastructure that can make the traditional
power system more efficient and optimized.
This project reviews the machine learning technology that can be used to forecast
future electrical load of household. The implementation of the project is done by the use
electrical household consumption dataset. Based on the available data, it is observed that
the available data has 1.25% missing values. To address the data inconsistency problem
various statistical techniques are used. Different graphs and plots are plotted with the help
of the available training data. After plotting it is observed that data has lot of outliers
which makes data inconsistent, so proper outliers removal method need to be used inorder
to address the outlier problem.
After the pre-processing of the available dataset, the different machine learning
models are trained. The different machine learning models are trained successfully by
using the training dataset. It is observed that the highest r2_score is obtained for
XG_boost model among all the trained models. The r2_scores of various models are
shown in the table 4.1. The performance of the trained models can be improved by
performing hyperparameter tuning. By this work with the help of machine learning
algorithms the future household electrical load can be predicted precisely.

5.2 Future Scopes

This system provides precise prediction of future household electrical load with
the help of parameters of a power system. This system can make the traditional power
system more intelligent. This system can form the basis for the implementation of smart
grid in future. This system can become more adaptable than the current system, allowing
it to quickly predict energy demand of an area.

29
 REFERENCES

[1] Machine Learning Course,Andrew Ng,Deeplearning.ai, Stand-ford University.

[2] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning.MIT Press, 2016.

[3] Python, https://docs.python.org/3/

[4] Pandas, https://pandas.pydata.org/docs/

[5] Matplotlib, https://matplotlib.org/stable/

[6] Numpy, https://numpy.org/doc/

[7] Scikit, https://scikit-learn.org/stable/

[8] M. Hayati and Y. Shirvany, “Artificial neural network approach for short term load
forecasting for illam region,” World Academy of Science, Engineering and Technology,
vol. 28, pp. 280–284, 2007.

[9] N. Kandil, R. Wamkeue, M. Saad, and S. Georges, “An effi-cient approach for short
term load forecasting using artificial neural networks,” International Journal of Electrical
Power & Energy Systems, vol. 28, no. 8, pp. 525–530, 2006.

[10] D. C. Park, M. El-Sharkawi, R. Marks, L. Atlas, and M. Damborg, “Electric load

forecasting using an artificial neural network,” IEEE transactions on Power Systems, vol.
6, no. 2, pp. 442–449, 1991.

[11] G. Zhang, B. E. Patuwo, and M. Y. Hu, “Forecasting with artificial neural networks::
The state of the art,” International journal of forecasting, vol. 14, no. 1, pp. 35–62, 1998.

[12] H. S. Hippert, C. E. Pedreira, and R. C. Souza, “Neural networks for short-term load
forecasting: A review and evaluation,” IEEE Transactions on power systems, vol. 16, no.
1, pp. 44–55, 2001.

30
 LIST OF PUBLICATION

01. Sanskar R., Kunal R. Singh, Piyush K., Roopam Kr. Raj, Soham V., Amit K.
Choudhary, “Forecasting of future household electrical load using machine
learning algorithm” IEEE transactions on Power Systems. (Communicated)

31