短期负荷预报的“双周期加混沌”法中的子模型优选理论探讨

SHORT TERM LOAD FORECASTING BY SYNTHESIZING DOUBLE PERIODS AND CHAOTIC COMPONENT: OPTIMIZATION BY SUB-MODELS

对于只依据历史记录的短期负荷预报,可以采用多个子模型来进行预报。拆分子模型的一种优化原则是定性和定量结合的原则。定性原则即子模型类型的选取顺序:首先选择可以精确预报的周期行为,其次是可以部分预报的混沌行为,最后采用随机行为。定量原则就是根据功率谱上的密度分布来选取子模型。对于某负荷记录时间序列,根据其功率谱等,可以发现该负荷记录是以双周期行为为主的。双周期以外的剩余成分是以混沌为主的。因此可以采用“双周期加混沌”这2个子模型来进行负荷预报。由于双周期行为不直接引起预报误差,因此“双周期+混沌”方法会有更高的预报精度和可靠性。

Multi sub-models are useful for the short-term load forecasting with historical data of loads. An optimalprinciple to split sub-models is to unify the qualitative andquantitative principles. The qualitative principle is the order for selecting sub-models: firstly choosing the predictable periodsub-models, then the partial predictable chaotic one, last theunpredictable random one. The quantitative principle is toselect sub-models according to the density distribution ofpower spectra of time series of load records. As an example, for the given load records it is found that their dominatingcomponents is the double periodic and the residual component subtracted from the load records is mainly chaotic, so thesynthesis of the two sub-models, i.e., the double periodicsub-model and the chaotic sub-model, can be used to forecast load. This method can bring higher forecasting precision and reliability, because the double periodic sub-model does notdirectly cause forecasting errors.