论文部分内容阅读
针对电力系统运行过程中负荷及故障的不确定性,在经济调度中引入风险评估原理,并提出了一种全新的基于知识迁移的细菌觅食强化学习优化算法。该算法将细菌觅食算法的寻优模式与Q学习算法的试错迭代机制结合,利用多主体协同合作来更新共有的知识矩阵,并以基于知识延伸的维度缩减方式避免了“维数灾难”。在预学习获得最优知识矩阵后,利用知识迁移加速在线学习进程。IEEE RTS-79测试系统的仿真结果表明:所提算法在保证获得高质量最优解的同时,寻优速度可达经典智能算法的9~20倍,适合求解大规模复杂电网的风险调度快速优化。
Aiming at the uncertainty of loads and faults in the process of power system operation, risk assessment principle is introduced in economic dispatch, and a new optimization algorithm of bacteria foraging based on knowledge transfer is proposed. This algorithm combines the optimization model of bacterial foraging algorithm with the trial-and-error iteration mechanism of Q learning algorithm to update the common knowledge matrix by multi-agent collaborative cooperation and avoids the dimensionality reduction of ". After pre-learning obtains the optimal knowledge matrix, knowledge transfer is used to speed up the online learning process. The simulation results of IEEE RTS-79 test system show that the proposed algorithm can obtain 9-20 times of classical intelligent algorithm while guaranteeing high quality optimal solution, which is suitable for rapid optimization of risk scheduling of large-scale and complex grids .