Research On Short-term Natural Gas Procurement Strategy Based On Load Forecasting

As a clean energy source related to the national economy and the people's livelihood,natural gas has an important strategic position in China's primary energy structure.The existing research may discuss the natural gas load forecasting method from a microscopic point of view,or analyze the natural gas industry and the natural gas market from a macro perspective,but relatively lack the relevant research based on the enterprise's own perspective.In recent years,as the natural gas consumer market continues to expand,the resource load of natural gas companies is also changing.How to accurately predict the user's gas demand and rationally purchase natural gas has become a problem that enterprises must solve to ensure natural gas supply.This is also related to the formulation of government-related policies.In view of the above problems,this paper is based on the enterprise perspective,drawing on the research results of domestic and foreign scholars' load forecasting.Based on the research on the cause of natural gas load,a new natural gas load forecasting combined model was constructed,which consists of BP neural network model,empirical mode decomposition(EMD)model,fuzzy entropy algorithm,and the LSTM neural network model.Based on this,a short-term natural gas dynamic purchasing strategy was designed to enable enterprises to achieve the goal of ensuring natural gas supply at a lower procurement cost.Specific work includes:Firstly,the mature natural gas trading market structure and its trading model are discussed.From the perspective of enterprises,the natural gas procurement methods suitable for different characteristics of natural gas load are analyzed.Secondly,the daily load characteristics of natural gas are analyzed,and the main influencing factors are extracted from it.The main characteristics of daily load are captured by BP neural network combined with the main influencing factors.The residuals generated by BP neural network prediction are reconstructed and reconstructed by EMD model and fuzzy entropy algorithm,and the LSTM model is established to predict the reconstructed items.Then combining the two-part model together to get the final forecasting results.At last,using the actual load data to verify the combined model,it is found that the BPNN-EMD-LSTM combined model has better prediction accuracy for short-term natural gas load than other models.Finally,a dynamic natural gas procurement strategy based on combined forecasting model is designed.The BPNN-EMD-LSTM model is used to predict the natural gas loads with high precision.Take the enterprise's gas source attribute and gas supply mode into consider,set the company's procurement strategy and gas supply order for different objective conditions.The actual natural gas load data was selected for the experiment.By calculating the procurement cost of the strategy at different load levels,the effectiveness of the strategy is fully verified.