A Deep-Learning-Aided Optimal Operation Model for Integrated Electrical and Heat Systems with Data Privacy Protection

In order to protect the privacy of different entities, this paper proposes a deep-learning-aided optimal operation model for integrated electrical and heat systems (IEHSs), where the electric power system (EPS) and district heating system (DHS) are connected through combined heat and power (CHP) units. First, a deep neural network (DNN) is trained to predict the feasible region of CHP units using real data or high-fidelity simulation data. Then, based on the Big-M method, the trained DNN with ReLU active function is reformulated as a set of mixed-integer linear constraints. The reformulated DNN is incorporated into the optimal operation model of IEHS only with inputs and outputs of DNN. With the aid of powerful neural networks, effective privacy protection is achieved between the EPS and DHS without compromising accuracy. Finally, the proposed method is validated on a modified IEHS with 39-node EPS and 7-node DHS. Numerical results show that effective privacy protection is achieved between IES and DHS with data privacy considerations, and the final result accuracy reaches 98%.