燃气短期负荷预测的小波分析综合模型  

A SYNTHESIS WAVELET ANALYSIS METHOD FOR SHORT-TERM GAS LOAD PREDICTION

作　　者：李持佳[1] 焦文玲[1] 赵林波[2] 

机构地区：[1]哈尔滨工业大学市政环境工程学院 [2]东北林业大学

出　　处：《天然气工业》2007年第8期103-105,108,共4页Natural Gas Industry

基　　金：国家自然科学基金(编号:50376014);哈尔滨工业大学校基金(编号:HIT.2002.65);北京市"供热;供燃气;通风及空调工程"重点实验室;哈尔滨市青年科学研究基金(编号:2003AFXXJO22)资助

摘　　要：城市燃气负荷预测对于保证燃气企业的供气安全、优化调度等具有重要意义。燃气负荷受天气、温度、节假日及一些随机因素等影响,很难建立准确的预测模型。为此,根据H市燃气短期日负荷变化特点,提出了用于燃气短期负荷预测的小波分析综合方法。首先用信息熵函数最小选择最优小波基,然后用其对燃气负荷进行二层分解得到负荷的低频信号和高频信号。低频信号受各种主要因素的影响,反映燃气负荷整体的变化趋势,对低频信号利用神经网络进行建模与预测;高频信号则主要受随机因素的影响,可将其看作"白噪声",对其建立时间序列自回归预测模型。低频信号和高频信号的预测值合成得到预测结果。实例验证表明,燃气短期负荷预测小波分析综合模型有效地提高了负荷预测精度。The urban gas load prediction has great significance for safe gas supply and optimized dispatching for gas enterprises.As affected by weather,temperature,holidays as well as some other random factors,it is very difficult to establish prediction models for gas load.A synthesis wavelet analysis method for short-term gas load prediction was put forward on the basis of characteristics of daily gas load changes of a certain city in a short period.Firstly,the best wavelet basis selection was conducted via message entropy function;then the gas load was decomposed to low frequency signal and high frequency signal by two times through it.As influenced by various kinds of major factors,low frequency signal reflects the overall trend of the gas load.The prediction model for the low frequency signal was established by means of the nerve network.Mainly influenced by random factors,high frequency signal was known as 'white noise'.The time series auto regression prediction model was established for it.The prediction result was finally acquired by synthesizing the forecast values of the low frequency signal and high frequency signal.The example results indicated that the synthesis method based on the wavelet analysis on short-term gas load prediction has effectively improved the prediction precision.

关 键 词：城市燃气 负荷预测 小波分析 神经网络 时间序列 数学模型 

分 类 号：TU996[建筑科学—供热、供燃气、通风及空调工程]