Research On Power Grid Infrastructure Investment Model Based On AdaBoost Regression Tree

With the rapid development of the economy,the power consumption of the whole society continues to increase,which makes the scale of investment in the power grid continue to increase.The grid investment includes many special projects,of which infrastructure investment accounts for more than 85 percent of the total investment,and there are many factors affecting the grid infrastructure investment.It has always been a difficult problem for grid companies to solve the infrastructure investment of their subsidiaries.Therefore,it is necessary to study a scientific,reasonable,effective and feasible grid infrastructure investment budgeting method,which provides a reference for grid companies to allocate capital investment to their subsidiaries.Therefore,this paper has done the following work:1.In view of the current situation of the absence of grid operation data,several common data filling methods are studied and analyzed,and a data filling method based on the combination of fuzzy c-means clustering and Lagrange interpolation is proposed.This method fully considers the integrity and locality of the sample set.Finally,we construct multiple missing rate data sets on single attribute and multi-attribute respectively,and compare with a variety of padding algorithms.Using the method of this paper to fill compared to the traditional grid using the averaging method to fill,the average filling accuracy on single attributes and multiple attributes increased by 17.07% and 16.18%,respectively.The experimental results show that the proposed method has a good filling effect on both properties.2.In view of the fact that there are many factors affecting the investment of power grid infrastructure and the number of total samples is small,it is impossible to select the relevant indicators reasonably as the characteristics of the model.This paper calculates the slope correlation degree between each operation index and the amount of capital investment,analyses the physical meaning of the leading index of the correlation degree,and synthesizes the expert experience.At the same time,considering the demand of power grid,the economic benefits of power grid,the social benefits of power grid and other factors,the reasonable selection of indicators that have a greater impact on the investment of infrastructure as the characteristics of the investment model of power grid infrastructure.3.In view of the current situation that the traditional methods and existing methods do not take into account the overall data and that the impact of each feature in the model on infrastructure investment is not clear,machine learning is applied to the field of power grid infrastructure investment.A power grid infrastructure investment model based on AdaBoost regression tree is proposed.This method can clearly see the impact of each feature on infrastructure investment and is a "white box" test.Finally,the three methods of AdaBoost regression tree,regression tree and SVR are compared experimentally.The average prediction errors of the three methods are 17.27%,37.00% and 136.87%.The experimental results show that this method is effective in forecasting the investment of power grid infrastructure.