Rasearch On Daily Gas Load Combination Forecast Model Of Towns

As a clean and efficient energy source,natural gas plays an increasingly important role in building modern energy systems and meeting the challenges of global climate change.China's urban gas industry has made remarkable achievements in the development of nearly 30 years.The investment of various gas companies in science and technology innovation has been continuously increased,and safety and service levels have been improved through continuous introduction and integration of some new technologies.However,compared with the overall urban gas industry level in developed countries,there is still a certain gap,especially the balance of supply and demand of urban gas and the problem of peak regulation,in the context of “coal to gas” and large-scale construction of urban pipe network in recent years.All cities have experienced different levels of seasonal supply tension and weak supply security.Therefore,achieving accurate urban gas daily load forecasting has very important practical significance for solving the problem of natural gas supply status.In this paper,regression analysis,neural network,support vector machine,combined forecasting and other forecasting methods are used to conduct in-depth research on urban gas daily load forecasting.The main contents are as follows:The daily load of town gas is affected by various factors.Accurately grasping the change law of historical daily load data,the influencing factors and the mechanism of its action are important prerequisites for establishing a suitable forecasting model.This paper first analyzes the change law of urban gas historical daily load data in the heating season and non-heating season,and uses the correlation analysis method to study the main influencing factors of daily load.Finally,the LOF algorithm based on density detection theory is used to analyze the outlier data.The positioning is performed,and the causes of the outliers detected one by one are further judged,and the interpolation method is used for correction.The urban gas daily load forecasting is a time-varying system with many uncertain factors such as random and fuzzy.The single-load forecasting model established at a specific moment has certain limitations in practical long-term application.According to the daily load influencing factors and the daily load sequence,a variety of single load forecasting models including regression analysis,time series method,neural network,support vector machine and extreme gradient lifting tree are established.Errors,root mean square error,correlation coefficient,grey correlation degree,Theil inequality coefficient,etc.The comprehensive evaluation index values calculated by each single evaluation criterion eliminate redundant models,and based on information entropy theory and ant colony algorithm respectively,based on weights The distributed combination forecasting model proves that the generalization ability and stability of the combined forecasting model are better than any single model.The application and development of gas load forecasting system with more accurate prediction and higher intelligent level will be the research hotspot of future load forecasting.Based on the Python language and MySQL database,this paper develops a town gas daily load forecasting software,which implements login verification,data management,load forecasting and other related functions.