基于功率预测的风力机健康状态监测

随着风力发电的广泛应用,对风力机健康状态进行准确监测的重要性日益凸显,为此提出了一种基于风力机功率预测的健康状态监测方法,即结合多项式模型和自回归模型特点,考虑风速与风力机输出功率之间的相关性和滞后性,利用改进非线性自回归模型对某风场风力机输出功率进行预测,并将预测结果与传统灰色模型、BP神经网络模型预测结果进行对比,计算与实测数据之间的误差。最后,选取功率预测系数中变化较为稳定的系数项作为观测系数,通过标准残差法确定异常观测系数反推风力机健康状态。分析结果表明,改进非线性自回归模型预测值与实测数据较为接近,趋势较为吻合。相比于传统灰色模型、BP神经网络模型,改进非线性自回归模型预测误差较小,精度较高。可见通过分析功率预测系数变化能够及时发现风力机健康状态变化,为故障发现提供参考。

Health Status Monitoring of Wind Turbine Based on Power Forecasting

With the wide application of wind power generation, health status monitoring of wind turbine is becoming more and more important. A status monitoring method based on power forecasting was proposed. By seeking out the time delay from the coherences between wind speed and output power, an improved nonlinear autoregressive model, which combined with polynomial model and autoregressive model, was applied to predict wind power output. The prediction result was compared with the traditional grey model and BP neural network model as well as calculating their forecast error respectively. Finally, the most stable prediction coefficient was chosen as the observation coefficient, standardized residual method was used to determine abnormal observation coefficient and infer the health status of wind turbine. The results show that the predicted value based improved nonlinear autoregressive model is consistent with the practical data in terms of trend. Compared with the traditional grey model and BP neural network model, the improved nonlinear autoregressive model has smaller error and higher accuracy. Therefore, analyzing the change of observation coefficient is an effective method for health monitoring and fault diagnosis of wind turbine.