基于相关向量机的发电机进相能力建模  

Generator leading phase capability model based on relevance vector machine

作　　者：翟学锋[1] 卫志农[2] 范立新[1] 徐钢[1] 王成亮[1] 刘亚南[1] 

机构地区：[1]江苏方天电力技术有限公司,江苏南京211102 [2]河海大学可再生能源发电技术教育部工程研究中心,江苏南京210098

出　　处：《电力自动化设备》2015年第3期146-151,共6页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(51277052;51107032;61104045)~~

摘　　要：发电机是一个多变量、强耦合的非线性系统,传统的分析方法难以建立精确的发电机进相能力分析模型。提出一种基于相关向量机(RVM)的发电机进相能力模型,以发电机有功功率和无功功率为输入、发电机的功角和电网电压为输出。以典型工况下发电机进相运行试验结果作为训练样本和测试样本,建立某600 MW发电机进相能力RVM模型,并讨论了核函数的选择对RVM模型收敛精度的影响。结果表明所建立的发电机进相RVM模型较之BP神经网络、径向基函数(RBF)神经网络和支持向量机(SVM)模型,精度更高、泛化能力更强,能有效地克服传统方法的局限性,适用于发电机进相运行实时控制。As generator is a multivariable and strongly-coupled nonlinear system,it is difficult to establish an accurate leading phase capability model of generator by traditional analysis method. A generator leading phase capability model based on the RVM（Relevanee Vector Machine） is proposed,which takes the active power and reactive power of generator as its inputs and the generator power-angle and grid voltage as its outputs. With the test results of generator leading phase operation in typical operating conditions as the training samples and test samples,a RVM-based model of generator leading phase capability is built for a 600 MW generator. The influence of the kernel function selection on the convergence accuracy of RVM-based model is discussed. Simulative results show that,the model based on RVM has higher accuracy and better generalization ability than that based on BP neural network,RBF（Radial Basis Function） neural network or SVM（Support Vector Machine）. It overcomes the limitations of traditional methods effectively and is suitable for the real-time control of generator leading phase operation.

关 键 词：发电机 进相 相关向量机 BP神经网络:RBF神经网络 支持向量机 建模 

分 类 号：TM715[电气工程—电力系统及自动化]