摘要
针对发电机组励磁与汽门的综合控制,研究了一种多模型自学习控制(MMSC).首先,建立机组不同工况下的样本数据并归纳模糊控制器(FLC)规则,随后采用模糊聚类算法将样本约简为典型工况,并得到对应于典型工况的模型库与控制器库.MMSC的控制量为多个FLC输出的加权集成,而加权系数由模型匹配程度决定.采用学习能力强的支持向量机来实现FLC的自学习和在线优化.仿真实验验证了MMSC的控制性能和效果.
As excitation and turbine control of generators confront with challenges of strong nonlinear characteristics and varying operation points, this paper proposed a multiple models self-learning control(MMSC). Firstly, fuzzy control rules for generators at various operation points were derived from operation samples. Then fuzzy clustering algorithm was employed to reduce the models at various operation points to a multi-model bank with corresponding fuzzy logic controller (FLC). Here the control signal of MMSC was simply the weighted sum of FLC, which were decided by their matching degree of multiple models based on fuzzy logic. Support vector machines (SVM), a power machine learning algorithm, were applied to the self-learning of FLC. Simulation results showed the desirable performance and control capability of the proposed MMSC.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2008年第1期47-52,56,共7页
Control Theory & Applications
基金
国家自然科学基金资助项目(60775047)
"863"计划资助项目(2007AA04Z244)
关键词
模糊控制
学习系统
支持向量机
多模型
发电机组
fuzzy control
learning systems
support vector machines(SVM)
multiple models
generator