Machine Learning Based Energy Forecasting for Residential Communities

Energy forecasting has become a well-known theme due to the intensifying use of Renewable Energy Sources (RES) in today's power system. Energy forecasting is a method for predicting future energy requirements in order to achieve demand and supply balance. This paper addressed the estimation of household communities' overall energy usage considering different prediction horizons. Solar generation and electricity consumption data are collected from households and government databases. Gaussian Process Classifier (GPC) and Quadratic Discriminant Analysis (QDA) are used to process the data and it will be stored in a database. The user interface is created using the Python Open-Source tool. GPC and QDA are used for the classification and prediction of electricity consumption. After compilation, future energy demand will be predicted based on parameters such as solar power generation, and behavioral aspects. Energy Forecasting helps the household communities to plan and control their electricity consumption. Gaussian classifier obtained a prediction accuracy of 87% compared to the Quadratic discriminant classifier, Support Vector Machine (SVM) and K Nearest Neighbours (KNN).