Artificial Intelligence for Renewable Energy Systems

10Forecasting of Electricity Consumption for G20 Members Using Various Machine Learning Techniques

Jaymin Suhagiya¹, Deep Raval¹*, Siddhi Vinayak Pandey², Jeet Patel², Ayushi Gupta³ and Akshay Srivastava³

¹Department of Information and Communication Technology, Adani Institute of Infrastructure Engineering, Ahmedabad, Gujarat, India

²Department of Electrical Engineering, Adani Institute of Infrastructure Engineering, Ahmedabad, Gujarat, India

³Department of Electronics and Communication Engineering, Pranveer Singh Institute of Technology, Kanpur, Uttar Pradesh, India

Abstract

Forecasting the actual amount of electricity consumption with respect to demand of the load hasa always been a challenging task for each electricity generating station. In this manuscript, electricity consumption forecasting has been performed for G20 members. Recurrent Neural Networks, Linear Regression, Support Vector Regression, and Bayesian Ridge Regression have been used for forecasting, while sliding window approach has been used for the generation of the dataset. During experimentation, we have achieved Mean Absolute Error of 16.0714 TWh, R² score of 0.9995, and Root Mean Squared Error of 31.3758 TWh with LSTM-based model trained on dataset created with window size of 6. Furthermore, predictions of electricity consumption have also been included till 2025.

Keywords: Electricity consumption, forecasting, machine learning, LSTM, GRU

10.1 Introduction

Electric consumption is the form of energy consumption ...