基于IQPSO优化SVM在径流预报中的应用

Application of IQPSO optimized SVM in runoff forecasting

提高径流预报精度的关键因素是选取合适的预报模型和预报因子。选择支持向量机作为径流预报模型,针对支持向量机模型参数在应用中存在选取困难的缺点,在标准量子粒子群算法中加入早熟判定准则、高斯扰动和自适应权重,提出改进量子粒子群算法(IQPSO),并使用该算法实现支持向量机参数的自动优选。为了验证效果,分别采用PSO-SVM、QPSO-SVM和径向基神经网络模型预报作对比,并使用多种评价指标进行对比分析。结果表明,使用改进量子粒子群算法优化支持向量机(IQPSO-SVM)模型能够有效提高月径流预报精度。

The key factor to improve the accuracy of runoff forecasting is to select the appropriate forecast model and forecast factor. Selecting support vector machine（ SVM） as the runoff forecasting model to overcome the shortcomings of the selection of support vector machine（ SVM） model parameters in the application,we introduce the precocious decision criterion,the Gaussian perturbation and the adaptive weight to the standard quantum particle swarm optimization algorithm.（ QPSO）,and use this algorithm to realize the automatic optimization of SVM parameters. In order to validate the results,PSO-SVM,QPSOSVM and RBF neural network model were used respectively to compare and forecast,and a variety of evaluation indexes were used for comparative analysis. The results show that using improved quantum particle swarm optimization to optimize support vector machine（ IQPSO-SVM） model can effectively improve the accuracy of monthly runoff forecasting.