基于OGA-KSVR的电能计量设备测量误差预测

Measurement error prediction of power metering equipment based on OGA-KSVR

针对多应力条件下电能计量设备测量误差难以预测的问题,提出一种基于核支持向量回归(KSVR)的电能计量设备测量误差预测方法,并提出一种优化遗传算法(OGA)对核参数进行优化。首先,提出一种线性加权多核函数融合多个应力特征,利用核权值系数刻画不同应力对电能计量设备的影响。然后,在核函数参数选择阶段,为了避免对参数人工调整的局限性,提出一种交叉概率与变异概率自适应调整的优化遗传算法,并将其应用到核参数优化选择问题中。国网新疆高干热试验基地智能电表运行数据分析表明,本文所提模型具有较高的准确性,预测结果的平均均方误差为0.00018,最高拟合优度可达0.989,可以为电能计量设备在多环境应力下的健康管理提供针对策略。

Aiming at the problem that the measurement error of power metering equipment is difficult to predict under multi-stress conditions,a measurement error prediction method of power metering equipment is proposed based on kernel support vector regression(KSVR),and an Optimized Genetic Algorithm(OGA)is proposed to optimize the kernel parameters.Firstly,a linear weighted multi-kernel function is proposed to fuse multiple stress features,and the kernel weight coefficient is used to describe the influence of different stresses on the power metering equipment.Then,in the parameter selection stage of the kernel function,to avoid the limitation of manual parameter adjustment,an OGA with crossover probability and mutation probability adaptive adjustment is proposed and applied to the optimization selection issue of kernel parameters.The operation data analysis of the smart electricity meters in Xinjiang High Dry Heat Test Base of State Grid shows that the proposed model has high accuracy,the average mean square error of the prediction results is 0.00018,and the highest goodness of fit can reach 0.989,which can provide a targeted strategy for the health management of power metering equipment under multiple environmental stresses.