自然分形在负荷预测中的应用

传统的负荷预测方法通常只能描述原始数据序列的长期趋势或季节性、周期性趋势等宏观行为.而对于表现微观行为的粗糙数据则难于建模.通过将原始序列分解成光滑部分和粗糙部分,在用传统方法对光滑部分进行建模的基础上以自然分形对粗糙部分进行建模,提高了预测的精度.计算实例表明该模型能很好地与原始序列拟合,且具有良好的预测效果.

APPLICATION OF NATURAL FRACTAL IN LOAD FORECASTING

Traditional load forecasting methods can only describe the macroscopic behavior of original source data series such as the data series of long-term trend, seasonal trend and periodic trend, but it is difficult to build the model of rough data representing the microscopic behavior. Separating the original source data series into the smooth part and rough part respectively and on the basis of building the model for the smooth part by traditional methods, for instance the GM(1,1), the author builds the model for the rough part by natural fractal method to improve the accuracy of load forecasting. The calculation results of actual examples show that the built models can fit with the source data series well and possesses good forecasting performance.