基于GM(1,1)模型和模拟退火算法的中长期负荷预测

Long-term load forecasting based on GM (1,1) model and the simulated annealing algorithm

分析了灰色模型(GM)和模拟退火模型(SA),GM(1,1)学习参数的计算采用最小二乘法,而最小二乘法是基于残差平方和最小寻优,容易陷入局部最小,对于非线性较强的负荷,会产生很大的偏差。提出了一种GM(1,1)与SA相结合的方法,根据模拟退火原理,结合概率突跳特性在解空间中随机寻找目标函数的全局最优解,自动优化GM(1,1)的参数,在负荷预测的实例中取得良好效果。

The gray model （GM） and simulated annealing model （SA） were analyzed.Learning parameters of GM （1,1） were calculated by the least squares method, while the least squares method was based on the minimum residual sum of squares optimization.This method was easy to fall into local minimum and would have a huge bias for the strong non-linear load. A method based on GM （1,1） and SA was proposed,combined with the probability of sudden jump in the solution space characteristics of the objective function of a random search for global optimal solution, automatic optimization of GM（1,1） of the parameters.This proposed method can efficiently select the parameters of LS-SVM methed ,and the accuracy of load forecasting is effectively improved.