风电功率GARCH预测模型的应用研究周晖.pdfVIP

第 39 卷 第 5 期 电力系统保护与控制 Vol.39 No.5 2011 年 3 月 1 日 Power System Protection and Control Mar.1, 2011 风电功率 GARCH 预测模型的应用研究 周 晖，方江晓，黄 梅 （北京交通大学电气工程学院，北京 100044） 摘要：根据风速变化的特点, 选择了适于描述波动变化特性时间序列的 GARCH 方法。分析风速小时变化曲线的残差项, 发 现其存在着 ARCH 效应, 满足 ARCH 的建模条件。采用美国夏威夷岛Lalamilo的风速数据, 建立了ARCH和GARCH风速变 化时间序列模型, 预测日的日逐点预测误差的平均值为25.1%。经过与ARIMA算法的比较, 预测的精度有所提高。运用风电 机组出力与风速的关系, 转换后得到了所需要的风电机组出力。对集群性不同的时间序列进行了多次数值计算, 发现GARCH 模型对波动性序列具有更好的适应性。 关键词：风速预测；风电功率；时间序列；GARCH；波动集群性 Application research of wind power forecasting model GARCH ZHOU Hui ，FANG Jiang-xiao ，HUANG Mei （School of Electrical Engineering，Beijing Jiaotong University ，Beijing 100044 ，China ） Abstract ：Considering the variation characteristics of wind speed curves，we selected the GRACH approach ，which has excellent advantage in tracing the variation of those fluctuated time series．Based on the analysis of those error entries of hourly wind speed curves，it is found that the ARCH effect does exist，which means that the series meets the requirement of constructing the ARCH model ．With the data from Hawaii Island Lalamilo，America ，the ARCH and GARCH model which reflect the wind speed variation are built and the average point-by-point forecasting error on the anticipated day is 25.1% ．Compared with the result by ARIMA approach，the predication precision has been improved ．With help of the equation between wind turbine output and wind speed ，the needed wind turbine output is easily gotten after transformation ．Finally ，amount of numerical calculation regarding those times series with different volatility clustering show that GARCH has better adaptability to fluctuation sequences ． Key words ：wind speed forecasting ；wind power