信息物理系统下区域发电Q学习控制方法

Q-learning control method for regional power generation based on cyber-physical system

为解决传统区域电网发电控制依赖计算控制响应函数,导致控制算法效率较低的问题,在信息物理系统构架下提出了一种区域发电Q学习控制方法.Q学习算法通过分析历史数据,生成控制智能体,避免了复杂的控制响应函数计算环节.基于IEEE-30节点系统的算例表明,该方法具有更高的响应效率,能够有效避免由于控制延迟导致的断面越限问题,断面越限时间相比于传统方法降幅30%以上,对提高区域电网运行控制能力具有显著作用.

In order to solve the problem that the traditional and regional automatic generation control(AGC)relies on the calculation of control response function,which leads to the low efficiency of control algorithm,a Q-learning control method for regional power generation was proposed under a cyber-physical system(CPS)architecture.This Q-learning algorithm generated control agents by analyzing historical data,to avoid the complex calculation of control response function.The calculation example based on IEEE-30 node system proves that the as-proposed method has higher response efficiency and can effectively avoid the problem of operational section over-limit due to control delay.The over-limit time can be reduced by more than 30%compared with that of traditional method,showing a significant effect on improving the operation control ability of regional power grid.