| As an efficient,high quality and clean energy source,natural gas can effectively reduce air pollution caused by burning coal.In the context of Chinese efforts to achieve the " Carbon peak" and " Carbon neutral ",natural gas as a low-carbon energy source will play an important role in the energy transition.With the implementation and further promotion of the coal-to-gas policy in recent years,the use of natural gas has increased in scope and usage.The year-on-year increase in gas demand and the volatile international energy situation have posed challenges to the use and dispatch of natural gas,with many regions experiencing varying degrees of seasonal gas supply constraints and dispatch difficulties.Therefore,achieving accurate and effective natural gas load forecasting can better provide a basis for production at the gas source end,optimise the scheduling and operation of the pipeline network,protect people’s livelihoods with gas,and has important research significance in addressing the current situation of natural gas supply.Based on thorough research,this paper analyses the daily natural gas load characteristics in depth,screens the main influencing factors,establishes a natural gas load forecasting model using various advanced algorithms,and develops load forecasting software for engineering applications.The main research elements are as follows:An accurate grasp and analysis of gas consumption patterns is an important basis for forecasting,and the selection of influencing factors and the quality of the underlying data also largely affect the accuracy of forecasts.In this paper,a comprehensive analysis of the daily gas load characteristics is carried out to discover the gas consumption patterns of customers.The main influencing factors of gas load are determined by integrating the actual gas consumption and correlation analysis results,and historical load data are processed for missing values and outliers using the mean value method and Local Outliers Factor(LOF)to ensure data integrity.Natural gas loads are influenced by many factors,which have been under-considered or over-introduced in previous studies,resulting in poor model predictions.To address this problem,this paper uses a BP neural network to fit the daily load and the main influencing factors,capturing the primary feature information and outputting a residual sequence containing the secondary feature information.The signal processing method Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)is introduced to decompose the residual series,reconstruct the components according to the t-test results,and build a Gated Recurrent Unit(GRU)timing prediction model for each of the reconstructed components,and combine the prediction results to obtain the final prediction results.The average absolute percentage error and root mean square error of the combined prediction model was 3.87%,both better than BP’s 4.68% and GRU’s 5.85%,as verified by the example,indicating that the combined model has better prediction results.To facilitate engineering applications,a natural gas load forecasting software was developed based on the APP Designer module in MATLAB,which enables forecasting work for different regions and user types,while having the advantages of fast forecasting and high accuracy.A mobile APP for use on the Android platform was developed using Android Studio,and a server was established using a My SQL database and intranet penetration software to upload the forecasting results from the forecasting software on the computer side to the public network,enabling data reading on the mobile side and the ability to view the forecasting data on the mobile phone at anytime and anywhere,providing a more convenient way for practical engineering applications. |
PDF Full Text Request