基于SCADA数据和改进BP神经网络的塔筒应力预测

针对风电机组载荷监测中应变片寿命短的缺陷,基于风电场海量状态监测数据,利用遗传算法和粒子群算法对BP神经网络进行改进,建立塔筒应力预测模型,并通过综合相关系数实现输入参量的有效选择。仿真结果表明,改进后的GA-BP神经网络预测模型和PSO-BP神经网络模型,预测结果的最大、最小相对误差等指标均比BP神经网络预测模型好;GA-BP神经网络预测模型的塔筒应力预测平均误差为7.04%,相对BP神经网络预测结果误差减少了4.38%,预测精度满足工程需求。所提出的方法建立风电场海量监测数据和塔筒应力数据之间的有效关系模型,可为风电场长期有效的载荷监测提供新的手段。

Stress Prediction ofWind Turbine Tower Drums Based on SCADA Data and Improved BP Neural Network

In view of the short lifespan of strain gauges in load monitoring of wind turbines,based on the massive state monitoring data of wind power plants,genetic algorithm and particle swarm optimization are used to improve BP neural network.The wind turbine tower drum stress prediction model is established.Through the integrated correlation coefficient,the effective choice of input parameters is realized.Simulation results show that the improved GA-BP neural network prediction model and PSO-BP neural network model are better than the BP neural network prediction model.The average error of the prediction model of GA-BP neural network is 7.04%,which is 4.38%less than that of the BP neural network.The prediction accuracy meets the engineering requirements.Therefore,the proposed method has established an available relationship between the massive monitoring data and the stress data of the tower barrels in wind power plants,and provided a new method for the long-term effective load monitoring of wind power plants.