Adaptation of Swarm Intelligence Algorithm in Support Vector Machine Modeling for Monthly Runoff Prediction
Accurate prediction of runoff in changing environments is becoming more and more difficult.The integra-tion of robust optimization algorithms has been the driving force behind the development of machine learning in recent years and is an effective way to improve the accuracy of runoff prediction.In order to prevent the inadequacy of the selec-tion of the fitness function in the SVM optimization modeling process,the SVM parameter optimization paradigm based on swarm intelligence algorithm is proposed.Five typical algorithms,particle swarm algorithm(PSO),difference algo-rithm(DE),gray wolf algorithm(GWO),whale algorithm(WOA)and sparrow algorithm(SSA),are simulated and validated by taking the monthly average flow prediction of Xinjiang Kizil reservoir as an example.The results show that the qualified rate(QR)of the SVM established by PSO is less than 60%,and the prediction accuracy is not up to stand-ard.The mean absolute relative error(MAPE)and Nash coefficient(NSE)of the SVM models established by the remai-ning four algorithms are between 10%-20%and 0.75-1,respectively,with good prediction results.Among the results of 20 independent operations,there are poor prediction results for PSO,WOA,SSA and DE,among which,PSO has the worst stability.In general,the support vector machine model(SVMGWO)optimized by the gray wolf algorithm has the best accuracy,stability and reliability in the monthly runoff prediction.
NETL
NSTL
万方数据
