Analysis Method of Electric Energy Substitution Potential Based on Time Series and BP Neural Network

Abstract

In recent years, as science and technology continue to advance, energy consumption has been constantly growing. As a result, environmental problems have been highlighted and become a hot topic. In contrast to coal, petroleum, natural gas, and other energy sources, electric energy has significant advantages in quickness, safety, and cleanliness. Thus, the implementation of electric energy substitution is capable of easing energy shortage and reducing environmental pollution to some extent. Study on electric energy potential provides a reference basis for electric energy programs. Therefore, in view of the quantization of electric energy substitution potential with electric energy substitution amount, this paper proposes a time series and back propagation (BP) neural networks-based combined prediction method. This means electric energy substitution amount can be predicted by the cubic exponential smoothing method-based time series model and the prediction results can be modified by BP neural network. In this study, the energy consumption from National Bureau of Statistics is used as data for forecasting and comparative analysis. The calculation results indicate that, compared to a single-method prediction, the prediction accuracy of electric energy substitution can be greatly enhanced via the combined prediction based on time series and BP neural networks. These findings have implications for analyzing electric energy substitution potential.