摘要
该文根据分形拼贴定理,由分形插值方法求取一个吸引子与电力负荷历史数据相近的迭代函数系统(IFS),建立分形预测模型,实现电力负荷预测。检验点的预测结果表明,最大相对误差为-2.92%,平均相对误差为-0.40%。该方法不存在收敛问题,数据收集简便,具有较好的实用价值。另外,该文应用N阶迭代累加和构造分形,应用分维定义对用电量构成的发展趋势进行分形预测,3项经济指标,6个检验点的预测结果表明,最大相对误差为2.5%,平均相对误差为0.46%。该方法同样不存在收敛问题,计算速度较快,也具有较好的实用价值。
According to the fractal collage theorem andwith the fractal interpolation, we establish fractal forecastingmodel to forecast the medium long term electric consumption. The predicted results show that the maximal relative error is2.92% and the average relative error is 0.40%. The method has no convergence property, and collects the data convenientlyTherefore, it holds good value in practice.In addition, we use N steps iterative accumulated value to form fractal and with thedefinition of fractal dimension, the whole society electricconsumption constitution is predicted in the article. 3 economic indicators and 6 checkpoints predicted results show that themaximal relative error is 2.5% and the average error is 0.46%. This method has no convergence property either, and itcalculates fast, so it also holds good value in practice.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2004年第11期91-95,共5页
Proceedings of the CSEE
关键词
用电量
最大
预测
检验点
经济指标
问题
平均
收敛
累加
吸引子
Electric machinery and electrotechnology
Electric consumption
Electric consumption constitution
Fractal collage
Fractal dimension
Scale independent range
Forecasting
