小波分析和考虑外生变量的广义自回归条件异方差模型在电价预测中的应用  

Application of Wavelet Analysis and Generalized Autoregressive Conditional Heteroscedastic Model Considering Exogenous Variables in Electricity Price Forecast

作　　者：刘达[1] 王尔康[2] 牛东晓[1] 

机构地区：[1]华北电力大学工商管理学院,北京市昌平区102206 [2]武汉大学经济与管理学院,湖北省武汉市430072

出　　处：《电网技术》2009年第18期99-104,共6页Power System Technology

基　　金：国家自然科学基金资助项目(70671039);华北电力大学校内基金资助(200822048)~~

摘　　要：电力市场中的电价序列存在很大的随机波动和价格尖峰。文章提出根据电价序列的变化特点,通过小波变换将其分解为概貌序列和细节序列,从而在不同尺度上反映电价的变化规律。通过概貌分量找出电价的主要波动规律,并由此对电价进行预测,剔除细节分量所反映的电价的随机波动影响。建立考虑异方差的广义自回归条件异方差模型(generalized autoregressive conditional heteroscedasticity,GARCH)对概貌序列建模,并在GARCH模型中加入外生变量形成GARCHX模型,以弥补传统时间序列模型忽略外界影响的缺陷。对美国PJM电力市场的实例研究表明,所建立的W-GARCHX模型比传统时间序列模型的预测精度有明显提高。There are evident random fluctuations and peak price in the electricity price sequence of electricity market. According to its variation features, the price sequence is decomposed into approximate sequence and detailed sequence by wavelet transform to reflect the variation law of electricity price in different scales. By use of approximate components the principal fluctuation law of electricity price can be found and from this the electricity price is forecasted while the influence of random fluctuation due to detailed components is rejected. During the modeling of approximate sequence, a generalized autoregressive conditional heteroscedasticity （GARCH） model, to which the exogenous variables are added, is built to form GARCHX model to avoid the defect of traditional time sequence model that the external impacting factors are neglected. Case study results of PJM electricity market show that forecasted electricity prices by the built W-GARCHX model are much more accurate than those by traditional time sequence model.

关 键 词：电力市场 电价预测 小波分析 广义自回归条件 异方差(GARCH) 自回归移动平均(ARMA) 

分 类 号：F123.9[经济管理—世界经济]