Deep Learning-based Multi-Scenario Forecasting of Beijing's Energy Footprints

Abstract

In the context of ‘dual carbon’, China's economic and social progress is gradually transitioning from a high-speed growth mode to a high-quality growth mode, and it is particularly important to promote the green and low-carbon energy transition and the high-quality development of energy. As the political and cultural centre of China, Beijing needs to play an exemplary leading role, and predicting the development trend of Beijing's energy footprint is of good research value. Taking Beijing as the research object, this paper firstly accounts for the energy footprint in the past years, and then quantitatively analyses the impact of influencing factors on the change of energy footprint based on the LMDI model. Finally, a forecast model based on GRU deep learning neural network was constructed, three scenarios were set up, and the time of peak energy footprint was predicted for each scenario. The results show that the four factors of energy consumption structure, energy consumption intensity, economic development level, and population size all have influence on the energy footprint of Beijing, and that the energy footprint of Beijing will peak in 2026 under the baseline scenario, with the two scenarios of low-footprint and high-footprint advancing or delaying the time.