短期负荷可预报天数的初步研究

只依据历史数据的短期负荷预报，预报准确率会随预报时间的增大而明显降低。这种现象来源于负荷记录的混沌特性。通过傅里叶级数展开，以及吸引子的相关维数、最大李雅普诺夫指数等分析，发现某地的负荷是双周期行为和混沌行为的合成。双周期成分是可以精确预报的，负荷预报的误差主要来自混沌成分。混沌成分的可预报天数约为15 d。按混沌成分占负荷的比重折算后，得到的负荷可预报天数约为3 d。

A PRELIMINARY STUDY ON PREDICTABLE SHORT-TERM LOAD FORECASTING TIME

When short terms loads are forecasted only by the historical load records, the forecasting precision will decrease with the forecasting time increases. This phenomenon is caused by the chaotic characteristic of the load records. By Fourier series decompositions, relative dimension testing of attractor, and largest Lyapunov exponent calculating, it is proved that the load records of an area in china are composed of the double period and chaotic components. Since the double period component can be forecasted accurately, the load forecasting errors are chiefly caused by the chaotic component. The predictable time of chaotic component is about 1.5 days, so the load predictable time is about 3 days, which is gotten by converting that of the chaotic component according to its proportion in the load records.