Electric Energy Substitution Potential Prediction Based on Logistic Curve Fitting and Improved BP Neural Network Algorithm

Abstract

In order to predict the potential of electric energy substitution in the next decade in China, this paper proposes a prediction method based on Logistic curve fitting and improved BP neural network algorithm. The amount of electric energy substitution is defined to quantify the potential of electric energy substitution. Then the important influencing factors of electric energy substitution based on the Impact by Population, Affluence and Technology (IPAT) model are established and quantified. For different influencing factors, logistic curve fitting and polynomial function fitting method were used to estimate the data fitting. A two-node output layer model of BP neural network is established and improved with additional momentum factors and adaptive learning rate to learn and train the data related to electric energy substitution from 2003 to 2017, and calculate the amount of electric energy substitution which are substitution potential from 2018 to 2020, 2025 and 2030. The calculation results show that the method has higher computational accuracy and fewer iterations. The prediction results are reasonable and effective, which can be the reference of the research of energy substitution.