摘要
电力负荷受众多因素的共同作用表现为复杂不规则的混沌规律,须采取合适的方法才能获得准确的短期负荷预测值。考虑因训练样本数目的不同而产生迥异的预测效果,先以经典混沌时间序列为例,比较训练样本数目从10变化到2 000时的各预测方法性能。仿真结果表明,经典混沌方法对小数目训练样本效果明显,随着样本数目的增多,智能混沌方法的优势渐显,其中最小二乘支持向量机有优异的预测精度和运算速度,且较神经网络对样本数目的依赖性小。欧洲智能技术网络(EUNITE)预测结果表明,最小二乘支持向量机能灵敏捕获小样本混沌电力负荷的变化规律,有效提高了短期电力负荷的预测精度。
Affected by many factors such as population, weather, social activities, and so on, the power load was performed as complex irregular chaotic character. The appropriate method was taken in order to obtain an accurate short-term load forecasting. The predicted effects were different with different number of training samples. Firstly, the classical chaotic time series were proposed for example with the number of training samples varying from 10 to 2 000. The simulation results show that the few training samples are well performed under the classic chaotic prediction methods. The advantages of chaotic prediction methods were visible with the increasing of the samples. The significant forecast accuracy and operation speed were supported by the least-square vector machine. The NN was less relying on the numbers of samples. The EUNITE instance indicates that the variation rule of small quantity chaotic electdc load is captured sensitively by the least-square support vector, and the short-term prediction accuracy is effectively improved.
出处
《电源技术》
CAS
CSCD
北大核心
2014年第3期528-531,共4页
Chinese Journal of Power Sources
基金
国家自然科学基金(61203110)
关键词
混沌时序
电力负荷
短期预测
预测方法
样本数目
chaotic time series
power load
short-term forecast
prediction methods
samples number
