基于改进SVM算法的输变电工程异常数据检测方法设计

Design of abnormal data detection method for power transmission and transformation engineering based on improved SVM algorithm

输变电工程的异常造价数据检测是电力工程数据分析的重要组成部分,但要对其进行深入的分析处理并投入工程应用,仍存在较多障碍。为解决输变电工程数据难以量化的问题且实现对异常数据的检测,文中基于建筑通用模型BIM技术,将建筑模板与实际数据相融合,设计完成了一套建筑成本估算系统。其针对多维度异常数据难以分类的问题,在改进SVM算法的基础上开发了多维逻辑回归分类器,进而实现了分布式的主动数据分类学习方法。通过实验验证可知,相比于常规算法,所提改进SVM算法的准确率更高,且可对不同工程方案、不同工程建设阶段的造价数据进行快速、准确识别,因此具有良好的应用价值。

The abnormal cost data detection of power transmission and transformation project is an important part of power engineering data analysis,but there are still some obstacles to its in⁃depth analysis and processing and put into engineering application.In order to solve the problem that it is difficult to quantify the data of power transmission and transformation projects and realize anomaly detection,this paper designs and implements a set of construction cost estimation system based on the general building model BIM Technology and the integration of building template and actual data.Aiming at the problem that multidimensional abnormal data is difficult to classify,a multidimensional logistic regression classifier is designed based on the improved SVM algorithm,and a distributed active data classification learning method is realized.Through experimental verification,compared with the conventional algorithm,the improved SVM algorithm proposed in this paper has higher accuracy,can quickly and accurately identify the cost data of different engineering schemes and different engineering construction stages,and has good application value.