基于IGWO-SVR的输电线路覆冰预测模型研究

Research on Transmission Line Icing Prediction Model Based on IGWO-SVR

为降低输电线路覆冰事故对电网安全生产的影响,电网运维部门通过预估线路覆冰厚度来指导电网的除冰融冰工作。本文提出采用支持向量机回归(support vector regression,SVR)方法预测覆冰厚度,研究了1种基于改进的灰狼优化算法(improved grey wolf optimization algorithm,IGWO)优化SVR的覆冰厚度预测模型。首先,采用非线性收敛因子a和精英反向学习(elite opposition-based learning,EOBL)提高了灰狼优化算法(grey wolf optimizer,GWO)的收敛速度并解决了易陷入局部最优的问题。其次,构建以环境温度、相对湿度和环境风速为输入量,覆冰厚度为输出量的IGWO-SVR覆冰预测模型,将实验数据代入模型计算其适应度值为0.019,远优于GWO-SVR模型。最后,将IGWO-SVR模型与PSO-SVR和GWO-SVR模型相比较,通过预测数据和实测数据的对比验证了IGWO-SVR模型的均方误差(mean square error,MSE)值为1.16,低于PSO-SVR和GWO-SVR模型,对输电线路覆冰预测研究具有一定的参考价值。

In order to reduce the impact of the transmission line icing accident on the safe production of the power grid,the power grid operation and maintenance department guides the deicing and ice melting of the power grid by estimating the line icing thickness.In this paper,support vector regression(SVR)method is proposed to predict the ice thickness,and an ice thickness prediction model based on improved grey wolf optimization algorithm(IGWO)is studied.Firstly,the nonlinear convergence factor a and elite opposition-based learning(EOBL)are adopted to improve the convergence speed of grey wolf optimizer(GWO)and solve the problem of falling into local optimization.Secondly,the IGWO-SVR icing prediction model with ambient temperature,air relative humidity and wind speed as input and icing thickness as output is constructed.The fitness function is 0.019 by substituting the experimental data into the model,which is far better than the GWO-SVR model.Finally,the IGWO-SVR model is compared with PSO-SVR and GWO-SVR models,and the mean square error(MSE)value of IGWO-SVR model is 1.16,which is also lower than that of PSO-SVR and GWO-SVR models,which has certain reference value for the research of transmission line icing prediction.