摘要
提出了采用遗传算法求解支持向量机最优参数的算法,并将其应用于Chickasaw河的日流量预测。为评估由不同时刻的流量、降水量及蒸发量组成的输入向量对模型预测精度的影响,设计了四种模型输入方案,以方差和确定性系数为标准对其进行评价。结果表明,采用前3 d流量、前1 d降水量及蒸发量的方案3预测精度最高。与BP神经网络模型预测结果对比显示,支持向量机模型预测精度较优,可用于水库或水电站的日流量预测。
This paper applied a genetic algorithm （GA） to optimize the parameters of Support vector machine （SVM） for daily flow forecasting in Chickasaw creek at Mobile County. To investigate the impact of variable enabling/disabling of streamflow, rainfall and evapotranspiration on model prediction accuracy, four modelling structures with different input vectors were developed and their performance was evaluated in terms of the mean square error and the coefficient of deter-mination. The results show that the third modelling structure consisting of the past 3 days＇s treamflow, the past rainfall and evapotranspiration as the inputs is superior to other model structure in performance. Compared with the BP neural network, experimental results show that accuracy of the proposed SVM model is slightly better than the latter in this article and can be used for forecasting the daily flow of reservoir or hydropower plant.
出处
《水电能源科学》
2008年第4期14-17,共4页
Water Resources and Power
基金
国家自然科学基金重点资助项目(50239030)
