超过指数增长速度的年度用电量曲线拟合预测

Forecasting Annual Electricity Consumption of Ultra-Exponent Increase Trend by Curve Fitting

为有效预测超过指数增长速度的年度用电量,选用超阶乘、二重指数等可线性化的函数,对年度用电量进行曲线直接拟合外推预测.采用这些新的数学函数预测北京年度用电量,拟合平均相对误差绝对值小于4.2%,2006年校验误差小于1.9%.超阶乘、二重指数函数不仅可用于超过指数增长的负荷预测,还可用于纠正线性回归、指数平滑和移动平均等方法对增长负荷预测值偏小的系统误差.

To forecast the annual electricity consumption of ultra-exponent increase trend, linearizing mathematical functions, such as ultra-factorial and duplicate-exponent, were adopted to directly fit the historical loads and forecast the future annual loads. Using these new functions to fit the historical data, the mean relative percentage error is less than 4.2%, and forecast error of the year of 2006 is less than 1.9% for the annual electricity consumption of Beijing. Besides forecasting the annual electricity consumption of ultra-exponent increase trend accurately, the functions of ultra-factorial and duplicateexponent can be used to rectify the systematic errors in increasing load forecasting by linear regression, exponential smoothing and moving average.