基于RF和ANFIS算法的直驱风电机组故障预警

针对风电机组运行工况复杂和单一状态参数不能较好实现故障早期预警的特点,提出随机森林算法(RF)和自适应模糊神经网络算法(ANFIS)相结合的故障预警方法。该方法充分考虑机组运行数据高维非线性特点,应用随机森林算法,建立有功功率与运行参数的数据驱动模型,计算各运行参数影响有功功率的相关度;构建自适应网络模糊推理系统模型,以训练误差最大值作为故障预警阈值,实时监测发电机运行状态。将该方法应用于某1.5 MW直驱机组发电机故障预警分析,结果表明,该方法能够提前预警发电机健康状态,避免严重事故发生,对风电场开展预防性维护、维修具有重要的指导意义。

Failure Prognostic Prediction of Direct DriveWind Turbines Based on Algorithms of Random Forest and Adaptive Network-based Fuzzy Inferer System

Aiming at the characteristics that the complicated operating conditions and single state parameter of wind turbine is complex and difficult to forceast potential failure，a fault prognostic method based on random forest algorithm (RF) and adaptive fuzzy neural network algorithm (ANFIS) was proposed. The high-dimensional and nonlinear of state features were fully considered，Using random forest algorithm, active power and operation parameters of the data driven model was established，relevancy of active power operation parameters and influence was calculated. ANFIS model was constructed to train error maximum value as the fault early warning threshold, the real-time monitoring running state of generator. This method was applied to 1.5 MW direct drive generator fault early warning analysis, results show that the method can advance warning generator health status，avoid serious accident. The evaluation results can carry out important guiding significance of preventive maintenance on wind farm.