基于SMA-LSSVM的径流中长期预测

Medium-and Long-term Runoff Prediction Based on SMA-LSSVM

径流中长期预测对防洪减灾和提高水资源利用效率极为重要。为解决预测模型参数对预测精度的影响,提出一种基于黏菌算法(SMA)优化LSSVM的径流中长期预测模型。首先,选取5个标准测试函数,对比在不同维度条件下SMA和PSO算法的仿真结果;其次,利用SMA优化LSSVM的惩罚参数和核参数,并构建LSSVM、PSO-LSSVM对比模型;最后,通过漫湾水电站水库入库月径流和莺落峡水文站月径流预测实例对各模型进行验证。结果表明,SMA-LSSVM模型相比LSSVM、POS-LSSVM模型,对漫湾站月径流预测的均方误差分别降低了29.26%、7.42%,对莺落峡站月径流预测的均方误差分别降低了32.61%、6.61%,预测精度更高。提出的SMA-LSSVM模型综合预测性能更好,也为中长期径流预测提供了一种新方法。

Medium-and long-term runoff prediction is extremely important for flood control,disaster reduction and the utilization efficiency improvement of water resources.To avoid the influence of prediction model parameters on prediction accuracy,this paper proposes a medium-and long-term runoff prediction model based on least squares support vector machine(LSSVM)optimized by the slime mold algorithm(SMA).Firstly,five standard test functions are selected to compare the simulation results of SMA and particle swarm optimization(PSO)algorithms in different dimensions.Secondly,SMA is used to optimize the penalty parameters and kernel parameters of LSSVM,and the comparison models of LSSVM and PSO-LSSVM are constructed.Finally,the models are verified with the monthly runoff of Manwan Hydropower Station Reservoir and Yingluoxia Hydrological Station as prediction examples.The results show that the mean square error of the SMA-LSSVM model is 29.26%and 7.42%lower than those of the LSSVM and PSO-LSSVM models,respectively,in the monthly runoff prediction of the Manwan station,and 32.61%and 6.61%lower,respectively,in the monthly runoff prediction of the Yingluoxia station.The proposed SMA-LSSVM model has better comprehensive prediction performance and also provides a new method for medium-and long-term runoff prediction.