Prediction Model of Metro Traction Power Consumption Based on Grey Neural Network

Abstract

Objective In order to improve train operational efficiency, it is necessary to monitor metro traction power consumption and establish a relevant energy consumption model for prediction analysis of metro traction power consumption. Method The basic principles of grey prediction model and BP (backpropagation) neural network are introduced. Taking the traction daily electricity consumption data for a typical metro station in Tianjin in June 2021 as example, the grey correlation analysis method is used to select the influencing factors with high correlation to the daily traction power consumption of metro. Based on the GM (1,1) grey prediction model, the short-term traction daily power consumption is predicted. The selected influencing factors with high correlation, the short-term traction daily power consumption predicted by GM (1,1) grey model, and the adjacent historical traction daily power consumption data are used as input for training in BP neural network model to establish the GM-BP grey neural network model. The required short-term metro traction daily power consumption prediction data is generated. Result & Conclusion Compared with conventional GM (1,1) grey prediction model and BP neural network model, the prediction error of short-term traction daily power consumption predicted by the GM-BP grey neural network model shows significant improvement, and can be used as effective metro traction power consumption data for short-term prediction data analysis.

Keywords

• metro
• estimated traction daily power consumption prediction
• grey neural network