基于组合函数和遗传算法最优化离散灰色模型的电力负荷预测

Load forecasting based on discrete grey model optimized by composite function and genetic algorithm

鉴于传统DGM(1,1)模型建模过程中假定原始数据序列服从近似指数增长规律,且以数据序列的第1个数据保持不变得出预测结果的缺陷,利用组合函数"对数-幂函数"对原始数据进行处理,使其符合灰色预测模型的建模规律,引入遗传算法寻求离散灰色模型初始迭代值的最优解,建立了基于组合函数和遗传算法改进的离散灰色模型。负荷预测案例得出所建模型的平均相对误差(MAPE)为0.892%,而GM(1,1)预测的MAPE为1.580%,DGM(1,1)预测的MAPE为1.343%,证明该改进模型有效提高了预测精度。

It is assumed during the modeling process of traditional DGM（1,1） that the original data sequence conforms to the approximate exponential growth rule and the first datum keeps unchanging,for which,the logarithm-power composite function is applied to preprocess the original data. Genetic algorithm is then introduced to find the optimal initial iteration value of the discrete gray model. Improved gray model based on composite function and genetic algorithm is established. Case study of load forecasting shows that the MAPE （Mean Absolute Percentage Error） of the proposed model is 0.892 % while the MAPE of GM（1,1） is 1.580 % and the MAPE of DGM（1,1） is 1.343%,which proves the proposed model improves the forecasting accuracy.