A Study Of Rural Coal-to-gas Load Prediction Based On The Cumulative Effect Of Temperature

As a clean,environmentally friendly and highly efficient energy source,the share of natural gas in the energy consumption sector has been growing year on year in recent years,and it has penetrated into every aspect of residents’ lives.However,there are still problems with the gas network in terms of uneven gas supply,which can have an unnecessary impact on the normal life of residential customers;especially as the scale of gas users in the country continues to expand,the impact of these problems is more obvious in rural areas where gas facilities are not adequate.In order to ensure the normal and stable consumption of gas by residential customers,especially those in rural areas,this paper develops a technology for predicting the daily gas load of rural "coal-to-gas" customers in Shandong,taking into account the cumulative effect of temperature,in the hope of providing a reference for gas scheduling in rural areas.It is expected to provide a reference for gas scheduling in rural areas.Firstly,this paper analyses the application area and research content of load forecasting through the clustering analysis mapping of load forecasting keywords,and then introduces the concept and characteristics of gas load forecasting.The actual load characteristics of this type of customer were obtained through a combination of field research and analysis of historical gas load data;the steps of the load forecasting study were also developed according to the actual needs.By analysing the load characteristics of the coal-to-gas customers and the results of the field research,the factors affecting the load of coal-to-gas customers were analysed and identified,namely: average daily temperatures,holidays,weather types and seasons."The concept of cumulative temperature effect was proposed to improve the prediction accuracy of the model,and the causes and effects of the cumulative temperature effect were analysed through the actual load data,as well as the characterisation of the cumulative temperature effect.Then,by pre-processing the collected raw data,including the interpolation of missing values in the raw load data,identification,labelling,analysis and replacement of anomalous values,etc..In addition,by using the correlation principle,the correlation between each influencing factor and the daily gas load and the magnitude of the correlation were clarified;with the help of the maximum-minimum standardisation method,the input characteristic parameters were normalised to eliminate the prediction errors caused by using characteristic parameters of different magnitudes;three commonly used evaluation indicators were selected to evaluate the prediction performance of different models.On this basis,a temperature correction model based on the cumulative effect of temperature was established to correct the influencing factor-temperature(in this paper,it refers to the daily average temperature);the daily average temperature before and after the correction was taken as the input parameters respectively,and the BP model was used for validation.The example shows that the modified daily mean temperature can improve the prediction accuracy of the model,and the MAPE of the prediction result is reduced from 6.78% to 3.74%,i.e.taking into account the influence of the cumulative effect of temperature will help to improve the prediction accuracy of the model.Finally,the modified daily average temperature and other selected influencing factors were used to forecast the daily gas load of a rural "coal-to-gas" customer in January of the 2020/2021 heating season in Shandong,for which WNN neural network models,BP neural network models,grey prediction models,ARIMA models and Multiple regression analysis forecasting models were developed.The results show that the WNN model performs best with a MAPE of2.62%;the BP model comes second with a MAPE of 3.49%;the ARIMA and grey models have similar prediction accuracy with MAPEs of 14.34% and 16.00% respectively;and the multiple regression analysis model has the worst prediction performance with a MAPE of 17.08%.The case study demonstrates that the selection of appropriate factors and prediction models,with adequate consideration of the cumulative effect of temperature,will result in high prediction accuracy,which will provide reference value for the operation and scheduling of gas pipeline networks.