基于模型融合的多维条件下现场校验仪计量误差建模研究

Study on Modeling of Electric Parameter Field Calibrator Measurement Error under Multidimensional Conditions Based on Model Ensemble

针对复杂的现场条件对关口电能表现场校验仪计量性能造成影响的问题,文中基于一种融合模型研究了多维条件下现场校验仪计量误差的建模。首先搭建了多维条件试验平台,试验平台可控条件包括:电流、电压、功率因数、温度、湿度。其次采用随机森林算法分别建立原始变量和交叉变量作为输入下的基础作用模型和耦合作用模型,并基于一种改进Stacking方法对上述模型进行融合得到一种反映多维条件对现场校验仪计量误差影响的融合模型。最后进一步试验验证了该模型的准确性和泛化性,结果表明该融合模型的预测精度优于单一模型。

According to the fact that complicated conditions influence the the metering performance of electric parameter field calibrator,this paper study modeling of electric parameter field calibrator measurement error under multidimensional conditions based on model ensemble.Firstly,constructed multidimensional test platform that can set current,voltage,power factor,temperature and humidity.Secondly random forest adopted to establish the basic effect model with primitive variable and coupling effect model with coupled variables respectively.The improved stacking fusion of the proposed models is conducted to get a comprehensive model to response the effect of multidimensional conditions on measurement error.Lastly further experiment verify the accuracy and generalization of the model,and show that the performance of fusion model is better than that of single model with prediction accuracy.