Multi-Carrier Energy Home Energy Management System Using Genetic Algorithms and Random Forest Predictions

Abstract

The energy transition encourages using heat pumps at the residential level, which results in a multi-carrier energy system when combined with PV and battery storage. Optimally controlling such systems has proven challenging. The numerous constraints required, different response times per energy carrier, and the need for forecasting methods also increase the complexity and computational cost. We propose an adaptable energy management system strategy for any system architecture with a reduced number of constraints using genetic algorithms with a discrete-continuous approach for the power setpoints. Using random forest regression, we also created short-term estimation models for the PV generation and electric and thermal demand, with error distributions centred near 0 %. Our results demonstrate that the strategy can solve the power allocation problem in the order of 1 s, including forecasting 60 minutes, minimizing electric costs, and ensuring thermal comfort.