Research on Enterprise Power Consumption Monitoring Based on LSTM Model

With the continuous development and update of social energy-using technologies, energy consumption is also increasing in a stepwise manner. Given the limited and tense energy resources, the country has proposed relevant policies for controlling energy consumption, mainly targeting high-energy-consuming and high-polluting industries, to save energy and reduce consumption. The supervision of energy consumption of enterprises has become a major issue. Traditional energy consumption monitoring mainly relies on a large number of data collection units and manuals to collect various types of energy data for aggregation, which requires a lot of manpower and material resources in terms of equipment investment and operation and maintenance costs. This article selects the monthly energy consumption data of enterprises with a capacity of over 5000tce in Chongqing from 2019 to the first quarter of 2022, as well as monthly electricity consumption data, and uses an LSTM (Long Short-Term) model to predict enterprise energy consumption data. The traditional ARIMA time series model is introduced for comparison and verification. Experimental results show that the LSTM model is always superior to the ARIMA model, not only in terms of the fitting of prediction results but also in performance on multiple evaluation indicators. This also points out a meaningful research direction for short-term enterprise energy consumption prediction and enables faster and more efficient implementation of enterprise-level energy consumption monitoring.