风险调度中引入知识迁移的细菌觅食强化学习优化算法

发布时间：2018-01-21 22:21

本文关键词： 知识迁移细菌觅食强化学习风险调度　出处：《电力系统自动化》2017年08期 　论文类型：期刊论文

【摘要】：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。
[Abstract]:In view of the uncertainty of load and fault during the operation of power system, the risk assessment principle is introduced in the economic dispatch. A new bacterial foraging reinforcement learning optimization algorithm based on knowledge migration is proposed, which combines the optimization pattern of bacterial foraging algorithm with the iterative mechanism of Q-learning algorithm. Multi-agent collaborative cooperation is used to update the common knowledge matrix, and the dimensionality disaster is avoided by dimensionality reduction based on knowledge extension. The optimal knowledge matrix is obtained after pre-learning. The simulation results of the IEEE-based RTS-79 test system using knowledge migration to accelerate the online learning process show that the proposed algorithm can obtain the best solution with high quality. The speed of optimization is 9 ~ 20 times that of the classical intelligent algorithm, which is suitable for the rapid optimization of large-scale complex power grid risk scheduling.
【作者单位】：华南理工大学电力学院;
【基金】：国家重点基础研究发展计划(973计划)资助项目(2013CB228205) 国家自然科学基金资助项目(51477055)~~
【分类号】：TM73;TP18
【正文快照】： 上网日期:2017-01-10。0引言为更好地权衡系统安全性与经济效益,增强调度操作抵御运行风险的水平,国内外学者已经在发电优化中引入电力系统的风险理论,对风险调度进行了大量研究。目前,实际电力系统中多采用确定性“N-1”原则指导安全约束经济调度[1],然而确定性准则下的结果

【参考文献】