A data-driven approach towards fast economic dispatch in integrated electricity and natural gas system

Effective economic dispatch for the integrated energy system (IES) can improve energy efficiency and promote renewable energy accommodation. Tradition IES economic dispatch are based on model-based methods that rely on accurate system parameters and uncertainty prediction. This paper proposes a data-driven fast economic dispatch method based on deep reinforcement learning (DRL) in an integrated electricity and natural gas system (IEGS). Unlike other DRL-based studies that spend a lot of computation time to calculate power flow, we employ DNNs to learn the complex nonlinear relationship existing in the IEGS, namely IEGSNet. Then, the economic dispatch problem is formulated as a Markov decision process, and solved by the soft actor–critic​ (SAC) algorithm. Simulation results illustrate that the proposed method achieves the similar operation cost to the tradition DRL-based dispatch method, but takes almost one tenth of the training time. Additionally, the computation time of the proposed method for 10-day dataset is at least two orders of magnitudes shorter that the model-based optimization algorithm.