Abstract:In order to deal with automatic generation control (AGC) under strong random grid en-vironment considering emerging renewable energy sources, a comprehensive new multi-agent simulation platform for AGC based on java agent development Environment (JADE) was proposed. This platform consists of three main parts: detailed and overall process dynamic load frequency response model, real time data transporting module and control algorithm module. The mechanism of agent communication, behavior and action of this platform are analyzed in details. Besides, PI control algorithm with fixed gain, single-agent Q(λ) learning algorithm, decentralized correlated equilibrium Q(λ) (DCEQ(λ)) learning algo-rithm are analyzed in details. Different control algorithms can be embedded into control algorithm module and its control performance can be evaluated, so three algorithms were integrated in control algorithm module, through comparing and analyzing the performance index of three algorithms, it has been proved that this platform embedded different algorithms respectively and applied to China Southern Power Grid could achieve coordination control between AGC areas, when one of the areas could not drive its area control error (ACE) to zero. It demonstrates the effectiveness, scalability, flexibility and stability of this platform.

Research on Multi-agent simulation Platform for AGC based on JADE