Combined Use of Support Vector Machine and Extreme Gradient Boosting System for Cost Prediction of Ultra High Voltage Transmission Projects

The cost prediction of an ultra-high voltage (UHV) power transmission project represents an important part of management and control of the project. In order to improve the prediction accuracy of an UHV transmission project, a combinational forecasting method is applied. An UHV power transmission project could be decomposed into several subprojects depending on the characteristics of the project, and the cost prediction could be made first for each subproject and then integrated. The principal component analysis and expertise of domain experts are first combined to get the key factors of the subprojects. Then the genetic algorithm based support vector machine and extreme gradient boosting system are used for cost prediction respectively. The Shapley value theory in game theory is next employed to determine the weights of combinatorial forecasting so as to attain an appropriate combinatorial forecasting model. Finally, an actual UHV project is employed to demonstrate the prediction accuracy of the presented combinatorial forecasting model.

I. Introduction

In recent years, rapid development in UHV power transmission has been achieved in China. It is highly demanded to establish an accurate cost prediction method for an UHV transmission project to ensure investment efficiency. The traditional method of determining the project cost according to the project quota and estimated budget cannot make full use of historical data of similar projects and cannot well meet the project investment needs. Especially in the middle and late stages of a project, it is very difficult for the cost management department to judge the advancement and rationality of the project design scheme and construction technology by the project quota and estimated budget.