相似日短期负荷预测的非线性理论基础与改进

Nonlinear Theoretical Foundation and Improvement of Similar Days Method for Short Term Load Forecasting

从非线性动力系统理论角度看,相似日预测的实质是对负荷序列中平衡点和准周期行为的预测。特定相似日预测方法所使用的负荷序列频谱越宽,预测结果就越准确,其实质是加强了准周期和混沌成分的预测;另外,采用负荷水平则增强了对平衡点的预测能力。采用“连续多天负荷”作为改进的相似日预测的基本单元、采用原始负荷的“负荷趋势+相似日”分解预测、对原始负荷序列按照素数间隔二次抽样等,都是提高预测准确率的具体手段。

From the viewpoint of nonlinear dynamic system theory, the substance of load forecasting by the loads in similar days is the forecasting for the behaviors of equilibrium points and quasi-periods. The wider the frequency spectrum of the load sequence used in specified similar days method is, the more accurate the forecasting result will be, because the forecasting for quasi-periodic and chaotic components are enhanced; moreover, the forecasting ability of equilibrium points can be improved by use of mean of loads. Following measures can be used to improve forecasting accuracy： using load data of continuous multi-days as the basic unit of improved similar days method; decomposing original load data into load trend and similar days load series, then applying the similar days method to the latter; and sub-sampling the original load series according to prime number intervals and so on.