基于PCA-NAR神经网络的风电机组健康趋势评估

Evaluation of Wind Turbine Health Trend Based on the PCA-NAR Neural Network

为了提高风电机组的实时可靠性,避免维修不足及维修过剩问题,降低风场运维成本,基于监控与数据采集(supervisory control and data acquisition,SCADA)系统得出的监测数据,应用主元评价和非线性自回归(non-linear auto-regressive,NAR)时间序列神经网络,通过建立评价及预测模型。通过实际数据验证模型有效性。结果表明:采用PCA-NAR的方法对风电机组的健康状况进行评估及预测有较好的效果,有助于帮助风场运营人员提前识别故障趋势并作出具体决策,以免机组遭受更为严重的故障。

Aiming at improvement of the real-time reliability of wind turbines,avoidance of insufficient maintenance and excessive maintenance,and reduction of wind farm operation and maintenance cost,based on the monitoring data acquired from the supervisory control and data acquisition(SCADA)system,principal component analysis(PCA)and non-linear auto-regressive(NAR)time series neural network were applied to establish an evaluation and prediction model.The effectiveness of the model was verified through actual data.The results showed that the PCA-NAR approach could achieve quite good effects in the evaluation and prediction of the health status of wind turbines,and could help wind farm operators to identify fault trend in advance and make specific decisions so as to avoid more serious turbine failures.