Multi-agent bargaining learning for distributed energy hub economic dispatch

Abstract

This paper proposes a novel multi-agent bargaining learning (MABL) for the distributed energy hub economic dispatch (EHED) of multiple energy carrier systems (MECS). Distributed EHED is developed by extending the conventional economic dispatch (ED) into MECS in a distributed manner, in which each energy hub is regarded as a learning agent for self-scheduling. The classical Q-learning with associative memory is employed for knowledge learning of each agent, while the non-uniform mutation operator is adopted for handling the continuous control variables. To maximize the total payoff of all the energy hubs, the bargaining game is presented for achieving an effective coordination between the buyer agents and a seller agent, where the slack energy hub is designed as the seller agent and the others are the buyer agents. MABL has been thoroughly evaluated for the distributed EHED on a high-complex 39-hub MECS with 29 energy hub structures and 76 energy production units. Case studies verify the superior performance of MABL for the distributed EHED compared with six centralized heuristic optimization algorithms.