基于虚拟模型重建与智能校核的电力工程数据处理技术研究

Research on power engineering data processing technology based on virtual model reconstruction and intelligent verification

针对海量电力工程数据自动校核过程中存在对异常数据识别精度较低的问题,文中提出了基于Grubbs-PCA-PSO-SVR的电力工程造价预测算法。其利用Grubbs法剔除异常数据,完成对电力工程数据的智能校核。同时采用PCA算法对高纬度数据进行压缩降维,以降低非关键因素对预测结果的干扰。通过PSO算法对SVR模型的惩罚系数及核函数参数加以优化,进而获得具有最优性能的SVR模型。并使用优化的SVR模型对电力工程数据进行挖掘分析,实现了电力工程造价的精准预测。算例分析结果表明,所提Grubbs方法能够准确识别异常数据,而PCA-PSO-SVR组合算法则相较于单一算法具有更高的准确性。

Aiming at the problem that the accuracy of identifying abnormal data is not high in the process of automatic verification of massive power engineering data,a power engineering cost prediction algorithm based on Grubbs⁃PCA⁃PSO⁃SVR is proposed in this paper.The algorithm uses Grubbs method to eliminate abnormal data and realize intelligent verification of power engineering data.PCA algorithm is used to compress and reduce the dimension of high latitude data and reduce the interference of non key factors on the prediction results.Through PSO algorithm,the penalty coefficient and kernel function parameters of SVR model are optimized to obtain the SVR model with the best performance.The optimized SVR model is used to mine and analyze the data of power engineering to realize the accurate prediction of power engineering cost.The results of example analysis show that the proposed Grubbs method can accurately identify abnormal data,and the designed PCA⁃PSO⁃SVR combined algorithm has higher accuracy than a single algorithm.