A Two-Stage Cooperative Game-Theoretic Based Approach for Distributed & Model-Free Control of Gas Transmission Networks

Abstract

Model dependency in the majority of control approaches makes them costly "if not infeasible" to get implemented for large-scale and cyclic gas network control. This work, hence, introduces a distributed, model-free, and game-theoretic approach comprising a simultaneous game followed by a sequential game. The two-stage cooperative approach is adopted in the control structure of gas transmission networks. The humans supervising the compressor stations in the network are considered as game players (agents), the rationality (utility function) of which is emulated via a designed fuzzy inference system such that decisions made by the agents closely match those made by the human operators. At the first stage of the game, players prevent the network from collapsing due to pressure drop in deliveries. In the subsequent stage, all players strive to enhance their utilities until the desired condition is met. Contrary to previous studies on gas network control, controllers learn appropriate action in the proposed strategy rather than calculating it using specific mathematical models. The performance and robustness of the proposed algorithm are assessed by its utilization to control a cyclic and interdependent gas transmission network. The results evidence that supervisory controllers operate the gas network in the permissible range in the presence of various loads; thus, such a comprehensive model-free control approach is a pragmatic solution in the systems whose models are not applicable for controller design purposes or exact utility evaluation in their game-based model is almost impossible.