灰色模型结合微粒群算法的城市用水量预测

为了提高城市用水量的预测精度,基于灰色GM(2,1)模型,采用参数ρ进行数乘变换,利用参数λ修正其背景值,引入微粒群算法(PSO)寻求参数λ、ρ的最优解,构建PSO-GM(2,1,λ,ρ)模型,对某市1990—2001年用水量进行预测,并与灰色神经网络(GNNM)算法预测结果进行对比。结果表明:引入PSO算法,利用其全局搜索、局部搜索相结合的搜索模式确定λ、ρ,可以提高灰色模型的预测精度;参数λ、ρ的随机性、灵活性加上PSO算法的搜索性、寻优高效性使PSO-GM(2,1,λ,ρ)模型比GNNM模型预测精度更高。

Urban Water Consumption Prediction by Gray Model Combined With PSO

In order to improve the forecast accuracy of urban water consumption,based on the GM(2,1)model,by using parameter ρ to do the transformation,using parameter λ to amend the background value,and using particle swarm optimization(PSO) to obtain λ and ρ,this paper proposed PSO-GM(2,1,λ,ρ) model to predict the actual annual water consumption of 1990 to 2001,and compared with the results of gray neural network(GNNM) algorithm.The results show that a) by using PSO’s search mode which combined global search and local search to obtain λ and ρ,the prediction accuracy of the gray model is improved;b) because parameter λ and ρ have randomness and flexibility,and PSO has searching method and high efficiency,the PSO-GM(2,1,λ,ρ) model has better prediction accuracy than GNNM model.