Deep Reinforcement Learning for Home Energy Optimization

Efficient energy management in prosumer households is key to alleviating grid stress in an energy transition marked by electric vehicles (EV), renewable energies and battery storage. However, it is unclear how households optimize prosumer EV charging. Here we study real-world data from 90 households in German-speaking countries to investigate the potential of Deep Reinforcement Learning (DRL) and other control approaches (Rule-Based, Model Predictive Control) to address the dynamic and uncertain environment of Home Energy Management (HEM) and optimize household charging patterns. We find that half of households align EV charging with solar energy generation while the other half shows a more irregular pattern and overnight charging. A detailed analysis of nine households (1 hour resolution, 1 year) demonstrates that high battery capacity facilitates self optimization; in this case further algorithmic control shows little value. In cases with relatively little battery capacity, algorithmic control with DRL improves energy management and cost savings by a relevant margin. This result is further corraborated by our simulation of synthetic household with varying photovoltaic (PV) systems and battery capacities. Benchmarking reveals that Model Predictive Control with full information can fully optimize HEMs, negating the need for DRL. We conclude that about half of prosumer households with EVs would profit from control approaches, thus benefiting also the full electricity system and its decarbonization.

Dr Alona Zharova is a Postdoctoral Researcher at the Chair of Information Systems of the Humboldt-Universität zu Berlin. Her research focuses on AI forecasting and optimization, recommendation systems, reinforcement learning, multi-agent systems, and complex data visualization. She applies these methods to domains such as data-driven decision support, demand-side climate change mitigation, load management in households, and smart home and smart city.