Prediction and Analysis of Household Energy Consumption Integrated with Renewable Energy Sources using Machine Learning Algorithms in Energy Management

Abstract

- The project aims to develop an Energy management system (EMS) that can adapt to changes in the building’s environment and optimize energy consumption without the need for manual intervention. In an IoT-based smart home context, this work intends to provide predictive models that are driven by data gathered from various sensors to simulate the appliances energy usage. This paper specifically consider two approaches to the prediction problem. First is to determine the important features in forecasting the appliance energy consumption and a general energy consumption model is created utilising machine learning (ML) techniques. A publicly accessible dataset made consisting of historical readings from various humidity and temperature sensors as well as information on the overall amount of energy consumed by household appliances in an smart home that is used to evaluate the performance of the suggested models. By exhibiting significant variation with both the training as well as test data for certain characteristics, among the proposed ML models interms of accuracy score logistic regression beats other models in the prediction outcomes comparison. In order to accurately predict future energy use based on historical energy usage data, we secondly create a time-series models using MA, ARIMA, SARIMAX and LSTM (univariate and multivariate) approach. The suggested predictive models will, in general, allow customers to reduce the power consumption of household appliances and also for utility to more accurately plan and estimate future energy demand to support green urban development.