摘要
在“负荷趋势加混沌”短期负荷预测法中,将原始负荷序列拆分成“负荷趋势”和“混沌”两部分,其中负荷趋势是可以准确预测的多周期行为,混沌子序列可以用相空间重构的线性回归法预测。以多元回归分析和矩阵计算的奇异性理论为基础,优化了混沌子序列预测中的三个参数。首先根据取样序列的“平稳性”及避免矩阵计算的“奇异性”,选取合适的延时时间;然后根据混沌子序列功率谱上的峰值选取嵌入窗长,并由此确定嵌入空间维数;最后选取邻近矢量的数目为嵌入空间维数的3倍以上。
The original power load series can be divided into two components, namely load trend and chaotic component. The load trend component is composed of multi-periods behavior and can be forecasted exactly. The latter can be forecasted by linear regression in phase space reconstruction. Based on the theories of multivariate regressive analysis and singularity in matrix calculation, three parameters in chaotic component forecasting are optimized. Firstly the delay time is optimized by the smoothness and non-singularity in matrix calculation of the sampled series. Secondly the embedded dimensions are selected upon the length of embedded window, which is given by a peak value in the power spectra of the chaotic component. Lastly the numbers of neighboring vectors selected are more than three times of the embedded dimensions.
出处
《电网技术》
EI
CSCD
北大核心
2005年第4期27-30,44,共5页
Power System Technology
关键词
电力系统
短期负荷预测
参数优化
“负荷趋势加混沌”法
Chaos theory
Electric power systems
Linear systems
Optimization
Regression analysis
Vectors