采用GA-BPNN与TLS模型的风电机组异常辨识方法  

Anomaly Identification Method of Wind Turbine Based on Genetic Algorithm-Back Propagation Neural Network and t-location Scale Model

作　　者：李泽宇[1] 郭创新[1] 朱承治[1] LI Zeyu;GUO Chuangxin;ZHU Chengzhi(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]浙江大学电气工程学院,浙江省杭州市310027

出　　处：《电力系统自动化》2020年第9期95-102,共8页Automation of Electric Power Systems

基　　金：国家重点研发计划资助项目(2017YFB0902600);国家自然科学基金资助项目(51877190);国家电网公司科技项目(52110418000T)。

摘　　要：基于反向传播神经网络(BPNN)建立了风电机组状态参数预测模型,并采用遗传算法(GA)对BPNN模型的初始权重与阈值进行优化,有效消除环境因素对风电机组状态参数的影响;采用TLS(t-location scale)分布模型刻画不同风速区间下预测残差的分布特性,基于矩估计方法实现TLS分布参数估计,并在此基础上提出了计及风速影响的状态残差异常程度量化指标。以某风电场的1.5 MW双馈风电机组为例进行了异常分析,结果验证了模型的有效性和准确性。Based on back propagation neural network(BPNN), a wind turbine(WT) state parameter prediction model is established, and the initial weight and threshold of BPNN model are optimized by genetic algorithm(GA), which effectively eliminates the influence of environmental factors on WT state parameters. The t-location scale(TLS) distribution is used to characterize the distribution characteristics of state parameter prediction errors with different wind speed intervals. Moment estimation is used to calculate the TLS distribution parameters and error anomaly index(EAI) is defined to quantify the anomaly level of prediction errors, which is verified to be an indicator of the WT anomalies. The proposed method is used for 1.5 MW WT with doubly fed induction generators and the results show that the proposed method is effective and accurate.

关 键 词：风电机组 数据采集与监控系统 预测模型 TLS分布模型 异常辨识 

分 类 号：TM315[电气工程—电机]