Data-driven-based dynamic pricing method for sharing rooftop photovoltaic energy in a single apartment building

In this study, a novel data-driven based dynamic pricing framework is proposed for sharing rooftop photovoltaic (PV) energy in a single apartment building. In this framework, the input includes the load data, electricity price data, and PV power data and the output includes the pricing strategy for local PV generations. Specifically, the building energy management system operator is responsible for setting internal uniform prices of their own rooftop PV productions to facilitate the local PV energy sharing with apartment building users, aiming to maximise the economic profits. To protect the privacy of apartment building users and meanwhile improve the computational efficiency, a neural network is designed for simulating their demand response (DR) behaviours. Besides, the uncertain rooftop PV generations can be duly addressed by using a well-trained long short-term memory network, which can capture the future trends of rooftop PV generations in a rolling-horizon manner. With the information on PV predictions and DR results, a model-free reinforcement learning method is developed for finding the near-optimal dynamic pricing strategy. The simulation results verify the effectiveness of the proposed framework in making dynamic pricing strategies with partial or uncertain information.