Truncated Importance-Sampling Emphatic TD3 at Deregulated Electricity Market

This paper introduces a novel reinforcement learning method, called Truncated Importance-Sampling Emphatic TD3 (TISE-TD3), which is tested in a Home Energy Management Systems (HEMS) framework. Operating in a free-energy market, the model uses real-world household consumption data, weather signals (irradiance and temperature), and price series at 5-minute granularity. By integrating emphatic weightings with importance-sampling truncation into the TD3 architecture, the method addresses off-policy distribution mismatch while ensuring stability. Results demonstrate that TISE-TD3 offers higher performance and lower computational complexity than the standard TD3 and TISE-SAC baselines, making it ideal for real-time energy optimization in energy storage systems.