Flexible deep reinforcement learning building load demand response method considering participation of energy storage

The invention discloses a flexible deep reinforcement learning building load demand response method considering energy storage participation, and the method mainly comprises the following steps: firstly, collecting historical load data and energy storage system data of multiple types of buildings, building a load model, and extracting an action space and an observation space; secondly, designing a reward function, and establishing a Markov process model for the demand response process of the building; thirdly, establishing an action value network, a target value network and a strategy network; and finally, historical load data and energy storage system data are used to train the network model, and the trained network can output a load action sequence and a load adjustable potential according to the load state of the current building. According to the method, the situation that the dimensionality of a demand response action space can be increased due to participation of an energy storage system, discretization