提出一种基于ＭＬＰ模型实现适应式ＰＩＤ控制的策略．该策略不需要被控对象的精确模型，以经典的ＰＩＤ为基础，用神经网络实现，在给定评价函数下进行自学习．把它应用于单机运行的水轮机组频率扰动和负荷扰动仿真试验中，结果表明该策略对于被控系统的严重非线性有很好的适应性和鲁棒性 A strategy of adaptive PID control based on MLP model is proposed. This strategy does not need the exact model of the controlled object. Based on the classical PID, it is implemented by neural network and self-learning under a given evaluation function. It is applied to the test of frequency disturbance and load disturbance in single-unit turbine. The results show that this strategy is very adaptable and robust to the severe nonlinearity of the controlled system.