Transfer learning for operational planning of batteries in commercial buildings

摘要: Recently, building owners are investing in rooftop photovoltaic (PV) installations and batteries order to meet the (facility) load their buildings. As a consequence, several commercial research solutions have emerged for battery energy management such Most of these rely on sufficiently accurate system models tailor-made those systems. This work proposes use transfer learning model-free reinforcement (RL) control operation enables knowledge from one be used by RL algorithm another with similar characteristics. In this paper, K-shape clustering is group buildings characteristics - based consumption patterns. To plan batteries, we fitted Q-iteration, algorithm. Simulation results using real-world data show that including forecast information PV generation feature space algorithm, competes mixed integer linear programming which assumes perfect system. We also investigate through simulation, effect transferring policy learned all belonging same cluster. faster convergence achieved fewer training samples required near optimal policy.