Research On Power Grid Infrastructure Investment Model Based On Operational Data And Its Application

With the continuous development of the economy and power grid,the increases of social electricity consumption and the investment of power grid,it has been a important problem that how to distribute the investment properly.However,the method of grid investment is still in the traditional way which lacks of scientific and accuracy at present.This paper focuses on the annual investment of the power grid and its operation data.From the point of view of data analysis and the method of machine learning,a set of investment model of power grid infrastructure based on operation data has been proposed.This method has good accuracy and it is adaptive to the existing problems and the characteristics of large scale and small amount of data in power grid operation.The main work of this paper is as follows:1.Aiming at the characteristics of large dimension of grid operation data,this paper proposes a feature selection strategy based on grid operation data by using the grey correlation analysis method.Through this method,the true affective operation data can be found.2.In light of the subjective nature of traditional methods and the insufficiency of the existing power grid investment forecasting model,which is the gray model to deal with the ability to jump data,this paper applies the statistical learning model to the grid investment field and proposes a grid tree-based power grid investment forecasting algorithm.By mining the relationship between operational data and investment amount,the algorithm constructs a regression tree prediction model,realizes the forecasting analysis of the overall investment trend of one province power grid,and provides specific models that affect the allocation of investment amount.By comparing the prediction results of the grey model,the average prediction error of the prediction model is reduced by 49.8%.3.Aiming at the current situation of precise grid investment demand and the lack of sample size of grid operation data for specific cities,this paper proposes a sample data expansion method based on a similarity clustering algorithm that involves the participation of similar city's operational data in model training.Based on this method,a precise forecasting strategy for specific cities was proposed.Comparing the experimental results,the average prediction error of the precision prediction strategy is 66.1% lower than the average prediction error of the overall prediction strategy.At the same time,the training data obtained by the similar city data expansion method based on the clustering algorithm is compared with the average error of the prediction.Randomly selecting the same number of city data reduces the average error of the training data participation prediction by 66.9%.