Optimal scheduling of integrated electricity and heat system with explicit data-driven model of thermal dynamics

• A high-resolution model is formulated to describe thermal dynamics of pipeline networks. • A neural-network surrogate is proposed to encode the nonlinear thermal dynamics. • The nonlinear activation function is converted into explicit mixed-integer linear constraints. • A computationally-efficient optimization operation model is developed for IEHSs. District heating system can provide substantial flexibility for electric power systems by utilizing the huge thermal storage potential in pipeline networks. However, the thermal dynamics of pipeline networks is complicated, which is challenging to model with high computational efficiency and accuracy. Therefore, this paper proposes a novel optimal operation model for integrated electricity and heat systems with explicit data-driven model of thermal dynamics, which improves both solution accuracy and computational efficiency. First, a neural-network surrogate with rectified linear unit activation functions is proposed to model the complicated thermal dynamics of pipelines. Then, the nonlinear neural network is converted into a set of explicit mixed-integer linear constraints through the big-M method, which can then be embedded into the optimal operation of integrated electricity and heat systems directly. Compared to the widely-used partial differential equation models, the proposed model based on neural network greatly reduces the decision variables while guaranteeing accuracy. Finally, an optimal operation model with the formulated explicit data-driven thermal dynamics is developed for the integrated electricity and heat system to exploit the flexibility from thermal pipelines. Numerical results demonstrate that compared with existing partial differential equation models, the computational efficiency of the proposed neural network model for thermal dynamics is improved by 95% with satisfying accuracy. The effectiveness of proposed thermal dynamics model in virtual energy storage to promote renewable energy consumption is also verified.