Hybrid physical and data driven transient modeling for natural gas networks

Abstract With the development of gas-fired generators, the increasing gas-electricity systems bring a new challenge to accurately analyze pressure or flow rate fluctuation in gas pipeline networks caused by gas-fired generators ramping. This paper proposes a hybrid physical and data driven modeling approach, which can take full advantage of physical laws and measurement data to characterize the transient behavior of gas flow in pipeline networks. Based on the linear graph theory, the gas flow dynamics and network topology are uniformly represented in the directed graph, so that the state-space model can be extracted automatically. Considering that the increasing number of model parameters with networks scale, a decentralized parameters identification algorithm is designed to enhance computational efficiency. The overall network is partitioned into a certain number of non-overlapping subnetworks, and each one formulizes the parameter identification as a nonlinear optimization problem solved by an iterative Levenberg-Marquardt algorithm. To demonstrate the effectiveness of the proposed method, a real-world gas network from a steel industrial park in China is studied, and simulation results are compared with the commercial simulation software Apros™. The results indicate that the proposed method improves the accuracy in gas network transient modeling markedly.