基于改进PSO-LSTM算法的风电机组状态监测方法研究

Research on Wind Turbine Status Monitoring Methods Based on Improved PSO-LSTM Algorithm

通过改进粒子群算法(particle swarm optimization,PSO)优化长短期记忆神经网络算法(long short-term memory,LSTM)的参数,提出了一种基于改进PSO-LSTM算法的直驱式风电机组运行状态监测方法。首先将数据采集与监控系统(supervisory control and data acquisition,SCADA)采集到的数据利用随机森林的方法进行特征筛选,得到模型的输入参数;其次采用改进PSO-LSTM网络建立有功功率的预测模型,计算出预测值与实际值的残差,根据残差的分布来确实直驱式风电机组的状态;最后利用某风电机组SCADA数据对所提预测模型进行验证分析,结果表明,PSO-LSTM预测模型相比其他三种预测模型,具有较高的预测精度,并在状态异常后最短时间内发出故障警报,保证电场的健康稳定运行。

Using improved particle swarm optimization(PSO)to optimize the parameters of long short-term memory(LSTM),a direct drive wind turbine operation status monitoring method based on improved PSO-LSTM algorithm was proposed.Firstly,random forest method was applied to select feature from the collected supervisory control and data acquisition(SACDA)data,and the input parameters of the model were obtained.Secondly,the modified PSO-LSTM network was used to establish the active power prediction model,the residual between the predicted value and the actual value was calculated,and then the state of the direct-driven wind turbine was obtained according to the residual distribution.Finally,a wind turbine SCADA data was used to verify and analyze the proposed prediction model.The results show that compared with the other three prediction models,the PSOLSTM prediction model can send out fault alarm in the shortest time when abnormal situation appears,achiving a higher precision and thus ensuring the healthy and stable operation of the wind farm.