Analyzing the Feasibility of Different Machine Learning Techniques for Energy Imbalance Classification in Smart Grid

Abstract

Smart grid involves distributed energy resources (DERs) such as Solar PV, Wind, Battery generations, etc to meet the increasing demand of energy. However, due to the increasing demand of energy, there is need of generating more energy from the available resources to avoid the energy imbalance. Energy imbalance is the core reason behind poor power quality and energy outage problems, and therefore it is required to predict the energy demand and supply on a daily basis for better preparedness to avoid energy crisis. This leads to the energy imbalance classification problem. In literature, many machine learning techniques have been used to address this problem. However, it is required to perform the feasibility analysis of such techniques. In this paper, we have investigated well-known machine learning techniques for energy imbalance classification in smart grid. For the feasibility analysis of different machine techniques, we have collected the energy generation and consumption data from Maharashtra region, India and preprocessed these data using the principal component analysis (PCA) which can help in improving the classification accuracy. For energy imbalance classification, we have considered different machine learning techniques such as Naive bayesian, neural network, support vector machine, decision tree, random forest, bagging and boosting. We have evaluated the performance results in terms of accuracy, root mean square error and mean absolute error.