自回归综合移动平均预测模型.doc

《自回归综合移动平均预测模型.doc》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、自回归综合移动平均预测模型数据采集本文选取了2011年某省电力系统从1月1日开始之后80天的电力负荷观测，如表一。第n天负荷量第n天负荷量第n天负荷量第n天负荷量12565957.38212705368.6412429907.99612743833.5622588923..55422476962.26622736933.5232595037.39232667444..4632773791.842621899..34442614097.2642748178.3752605604.4252646095.544

2、52680843.85652737334.2262597404..14462775056.43662720053.6172363386.42272641570.43472728907.25672700061.1582620185.38282584430.88482611172.72682709553.0492615940.83292474001.24492601989.82692681309.47102615480.96302396095.97502668757.4702683185.5611261234

3、8.58312288598...7122610054.23322166399.62522695802.63722644097.64132610964.36332062979.7532689571.21732685694.93142637653.21341997281.18542654423.52742702991.02152633388.14351925136.26552642984.4.375162640311.3361970438..78762680354.45172678530.1137197655

4、7.678572754918.32772682596.37182687189.9382050309.54582758839.28782695560.6192694733..52592817728.94792674342.97202709637.218402384011.84602759327.72802685891.98表1数据处理利用spass绘制时间序列原始数据的散点图根据图1，我们可以看出原始数据的时间序列，是非平稳时间序列。因此我们下面对原始数据进行平稳化处理，首先，我们进行一阶差分，得到表2表2

5、n差分值n差分值n差分值n差分值222965.6522-27404.054247054.2762-6900.0436114.3623-10520.544399293.146336858.28426861.7624-7457.674437841.864-25613.435-16294.7525-13890.84566746.6565-10844.156-8200.27266219.64694212.5866-17280.617-234017.7127-10744.7147-46149.1867-19992.

6、468256798.9628-57139.5548-117734.53689491.899-4244.5529-110429.6449-9182.969-28243.5710-459.8730-77905.275066767.58701876.0911-3132.3831-107497.84518632.6671-21347.8612-2294.3532-122198.515218412.5772-17740.0613910.1333-103419.9253-6231.427341597.29142668

7、8.8534-65698.5254-35147.697417296.0915-4265.0735-72144.9255-11439.5150875-15966.64532166923.163645301.85669158.7750876-6669.924681738218.81376119.618485742775.54772241.92188659.793873751.86152583920.967812964.23197543.1139104178.985958889.6679-21217.63201

8、4904.23.3260-58401.228011549.0121-4268.617824145896.1561-15494.16然后，绘制出一阶差分的时间序列的自相关函数图和折线图，如图2和图3图2根据图2，我们可以看出一阶差分后的电负荷量逐渐趋向0.图3一阶差分值的自相关系数ACF图由图3，我们可以看出差分序列的自相关函数迅速衰减到0，所以我们判断一阶差分时间序列平稳。模型识别由上述经处理好的基本符合要求的数据，用spas