发电机组汽门系统的多模型自学习控制

针对发电机组的非线性、大范围运行等实际问题，研究了用于汽门系统的多模型自学习控制（MMSC），首先根据各种工况下的样本数据归纳出模糊控制规则；然后由模糊聚类算法将多种工况约简为典型工况，得到相应的子模型模糊控制器（FLC）．以子模型FLC输出的加权集成作为MMSC的控制输出，而加权系数取决干子模型匹配度．在子模型FLC学习优化中，由支持向量机离线逼近模糊规则曲面，再由梯度下降算法在线自学习．仿真实验验证了所设计控制器的优良性能．

Multi-model self-learning control for turbine valving control

For the problem that turbine valving control of synchronous generator faces practical challenges as nonlinear characteristics and changing operation points, a multi-model self-learning control (MMSC) system is proposed. Fuzzy control rules for turbine valving control at variable operation points are derived from operation samples. Then a fuzzy clustering algorithm is employed to reduce variable operation points to typical points, and sub-model fuzzy logic controller (FLC) is obtained. The control output of MMSC is the sum of sub-model FLC multiplying respective weights which are decided by the matching degree of typical points using fuzzy logic. For the self-learning of sub-model FLC, support vector machines are used to approximate fuzzy rules curve offline firstly, then a gradient descend algorithm is used for online learning. Simulations show the capability of the proposed controller.