风电机组运行状态识别对风电机组发电性能评估和风电场精细化管理具有重要意义，然而不同风电机组的数据采集与监视控制系统(supervisory control and data acquisition，SCADA)数据分布差异明显，如果将已训练好的单台风电机组正常行为模型直接应用于多风电机组运行状态辨识，辨识精度较低。为了提高辨识精度，需要针对每台风电机组正常行为模型进行重复性训练，工作量大。为此，提出了一种基于迁移成分分析(transfer component analysis，TCA)的多风电机组运行状态划分模型。首先，采用基于最大互信息系数和反向传播(back propagation, BP)双隐层神经网络的变量优选方法挖掘风电机组运行状态关键影响变量；然后，以正常运行状态下的优选变量为输入，功率为输出，构建了基于BP双隐层神经网络的风电机组正常行为模型；最后，基于迁移成分分析，构建多风电机组运行状态划分模型。算例结果表明，所提模型可解决不同风电机组数据分布差异的问题，提高运行状态划分模型的精度和效率。

The identification of the operating state of wind turbines is of great significance to the performance evaluation of wind turbines and the refined management of wind farms. However, the supervisory control and data acquisition (SCADA) data distribution of different wind turbines varies significantly. If the trained normal behavior model of single typhoon turbine is directly applied to the operating state identification of multiple wind turbines, the identification accuracy is low. In order to improve the recognition accuracy, it is necessary to conduct repetitive training for each wind turbine normal behavior model, which is a large workload.Therefore, a multi-wind turbines operating state identification model based on transfer component analysis (TCA) is proposed. Firstly, a variable optimization method based on the maximal information coefficient and back propagation (BP) double hidden layer neural network is used to mine the key influencing variables of wind turbine operating status. Then, the normal behavior model of wind turbine based on BP double-hidden layer neural network is constructed with the optimal variables as input and the power as output. Finally, based on transfer component analysis, a multi-wind turbines operation state classification model is constructed. The results show that the proposed model can solve the problem of data distribution differences between different wind turbines and improve the accuracy and efficiency of the operating state partition model.