Improving Demand-Side Energy Management With Energy Advisor Using Machine Learning

Abstract

The prediction of consumers’ electricity consumption is an important demand-side energy management technique for adaptable dynamic tariff pricing. In this regard, historical consumption patterns play an important role in accurately predicting the future energy consumption. In this paper, we propose a novel system based on machine learning (ML) algorithms to accurately predict not only the future energy consumption based on historical data and various environmental factors but also energy-saving recommendations for the consumers. The proposed system is evaluated on real-world energy consumption data, and the results show that random forest (RF) more accurately predicts energy consumption with 98% R-squared value and 0.51 root mean square error, whereas decision tree (DT) is best fit for personalized energy recommendations resulting in 85% R-squared and 0.14 root mean square error. The implementation of this system will greatly assist utility companies in making precise future energy consumption forecasts to formulate demand-side energy management policies which are cost-effective for both consumers and utility companies.