Abstract: Using the daily and hourly data of gas load and simultaneous meteorological observational data in Hangzhou from 2008 to 2013,the variation of gas load in Hangzhou and its relationship with meteorological factors were statistically analyzed.A gas load forecasting model was established based on a SVM (Support Vector Machines) method.The results show that the gas load in Hangzhou from 2008 to 2013 exhibits obvious seasonal variation,with the highest daily mean meteorological load rate in winter and the lowest in summer.Diurnal variation of gas load,with a single peak,is similar among different seasons.The daily meteorological load rate has a negative correlation with daily air temperature in all months except for June and September,with the largest correlation coefficient in December.The daily meteorological load rate has positive values under the conditions of daily average temperature ≤ 13 ℃,and it reaches a maximum value with daily average temperature of about 3 ℃.A positive correlation between daily mean air pressure and daily meteorological load rate is observed from January to April and from October to December.The hourly meteorological load rate correlates negatively to daily air temperature in all seasons except for summer,and the best correlation occurs in autumn.Considering the main meteorological influencing factors,a gas load daily/hourly forecasting model is established based on a SVM regression method.This model has good performance,with a mean error of daily gas load forecasting of 4.36% and a mean error of hourly gas load forecasting of 4.18%.

Key words: Gas load, Meteorological factor, Support Vector Machines (SVM), Correlation analysis, Forecasting model