A Predictive Model for Electricity Consumption in University Campuses Using Artificial Neural Networks.
ABSTRACT
Energy efficiency is paramount in the quest to achieve sustainable development in the 21st century. Statistics in recent research have shown that in many sectors in any nation’s economy, which include buildings, industries and transportation, energy consumption in buildings accounts for about 77%, a higher percentage than other sectors in Nigeria; the same is true worldwide.
Energy consumption forecasting is a critical and necessary input to planning and monitoring energy usage, with particular reference to CO2 and other greenhouse gas emissions. According to literature, very little research has been carried out in designing models for energy consumption in institutional buildings.
In this research, the African University of Science and Technology (AUST) is considered as a case study, whereby the data collected is the monthly energy consumption for the period 2012–2014 and 2015–2017.
The data was collected from the monthly electricity utility bills when the school was using a flat rate and when they were using a measured meter rating respectively. The two models were designed for the monthly prediction of electricity consumption of the buildings within the university using an artificial neural network.
Results obtained from the two models were compared and showed that the model designed using the latter dataset could be adopted to forecast the electricity consumption of the school with respect to its population.
This will further assist the university in monitoring the trends of energy consumption, classify factors and components that impact energy consumption within the university community and hence building policies on its usage and consumption. Moreover the possibility of using renewable energy in the university could also be integrated as a future work.
TABLE OF CONTENTS
CERTIFICATION ii
ABSTRACT v
DEDICATION vi
ACKNOWLEDGEMENT vii
LIST OF FIGURES xi
LIST OF TABLES xii
LIST OF ABBREVIATIONS xiii
CHAPTER ONE INTRODUCTION 1
Background of Study 1
Statement of the Problem 3
Aim and Objectives 3
Expected Contributions 3
Scope of the Work 4
Thesis Structure 4
CHAPTER TWO LITERATURE REVIEW
Introduction 5
Overview of Electricity Consumption in Nigeria 5
Electricity Consumption in Institutional Buildings 8
Machine Learning 9
Supervised Learning 10
Unsupervised Learning 10
Reinforcement Learning 11
Machine Learning Techniques in Electricity Consumption Prediction 11
Grey Models and their Applications 11
Statistical Models and their Application 12
Artificial Intelligence Models 13
Review of Related Works on Electricity Consumption Prediction 14
Summary of Literature Review 17
CHAPTER THREE RESEARCH METHODOLOGY
Introduction 18
AUST Campus Information and Data 18
Weather Conditions at Abuja 19
Electricity Consumption Data for AUST 20
Preliminary Data Analysis 21
Demystification of Artificial Neural Networks 22
Forecasting with Artificial Neural Networks 22
Data Collection 24
Input Variables 24
Output Variables 25
Data Preprocessing 25
Model Description of the Network Model for AUST 27
Training the Network 28
Network Model Parameter Investigation 29
Performance Evaluation Analysis 29
Implementation of ANN using MATLAB 30
Neural Fitting Tool 30
Data Selection from the Workspace Area 31
Data Validation and Testing Pane 31
Network Architecture Pane 32
Network Training Pane 33
Network Evaluation Pane 34
Application Deployment Pane 34
Results Pane 35
CHAPTER FOUR RESULTS AND DISCUSSION
Introduction 36
Performance and Comparisons of the Models 36
Validation and Testing Results 38
Prediction of Electricity Consumption with the Built Models 39
CHAPTER FIVE CONCLUSION AND FUTURE WORK
Conclusion 41
Future Work 42
REFERENCES 43
APPENDIX 48
INTRODUCTION
1.1 Background of Study
For any nation to be identified as being extremely industrialized, social, economic and industrial development must exist. Energy Consumption has become a prime focus in global discussions towards ensuring sustainable development.
Recent studies have shown that in many parts of the world, energy consumption of buildings exceeds that of other sectors, including transportation and industries. For example, in the Nigeria, residential buildings consume as much as 77.8%, while transportation, industries and others account for the rest.
In Nigeria, electricity is one of the oldest forms of energy available for daily activities. It is also, unfortunately, in too short supply to meet the demand of an ever-increasing population. This is largely due to inadequate planning (Kofoworola, 2003).
Arimah (1993) gave an overview of the current situation of the Nigerian electricity industry where he mentioned that it is beset with several serious technical, managerial, personnel, financial and logistical problems.
REFERENCES
Abu-El-Magd, M. A., & Sinha, N. K. (1982). Short-Term Load Demand Modeling and Forecasting : A Review. IEEE Transactions on Systems, Man, and Cybernetics, 12(3), 370–382.
Arimah, B. (1993). Electricity consumption in Nigeria: A spatial analysis. OPEC Review, (May 2008). https://doi.org/10.1111/j.1468-0076.1993.tb00465.x
Beale, M. H., Hagan, M. T., & Demuth, H. B. (2017). Neural Network Toolbox TM User â€TM s Guide. The Maths Work.
Braun, M. R., Altan, H., & Beck, S. B. M. (2014). Using regression analysis to predict the future energy consumption of a supermarket in the UK. Applied Energy, 130, 305–313. https://doi.org/10.1016/j.apenergy.2014.05.062
Deb, C., Eang, L. S., Yang, J., & Santamouris, M. (2016). Forecasting diurnal cooling energy load for institutional buildings using Artificial Neural Networks. Energy and Buildings, 121.