Natural Gas Load Forecasting For Residential Users Based On LSTM

In recent years,energy consumption has been a hot topic of national and people's attention.At present,energy shortage and environmental pollution make the optimization of energy structure imminent.In recent years,the introduction of some relevant policies has guided the development of natural gas industry.As a clean energy,the demand for natural gas has increased significantly.However,natural gas itself has its own particularity,and a large number of natural gas storage will cause certain security risks.Therefore,how to predict the consumption of natural gas and ensure the smooth transportation of natural gas is of great significance for economic development and the protection of residents' gas safety It is of great practical significance.This paper mainly analyzes and studies the natural gas load of residents in Zhengzhou.Based on the monthly data of 34 periods of natural gas users from May2016 to February 2019,it forecasts the monthly data from March 2019 to may 2019.Considering that the random fluctuation of single user's natural gas load series is strong,when the number of residents reaches a certain scale,the fluctuation tends to be stable,and the natural gas load is stable The company usually faces thousands of users,so when forecasting natural gas,it needs to forecast the average gas consumption of users in the residential user group,which can reduce the prediction error to a certain extent and facilitate the gas distribution of natural gas companies to downstream subsidiaries of different sizes.By setting different scales of user groups,considering the nonlinear relationship between natural gas load data,ARIMA model,prophet model and LSTM model are selected to predict the average natural gas load of different scale user groups.It is found that ARIMA model and LSTM model have better performance,and LSTM model has the best performance,and its stability is better than ARIMA model It is found that when the residential user set is higher than500,the prediction deviation rate of LSTM for each month is less than 5%,while the performance of Prophet model is poor because the data set is monthly data.However,with the improvement and development of the model,it is expected to achieve good prediction results.Finally,some suggestions are put forward for the development of natural gas and other clean energy.