| The medium and long-term runoff forecast has a positive effect and important significance for water conservancy projects.The unity,limitation and instability of traditional forecasting models make it difficult for forecasting to accurately predict runoff changes.With the rapid development of machine learning technology,Kernel extreme learning machines have been widely used in the field of runoff forecasting,and its prediction performance largely depends on the selection of regularization coefficients and Kernel parameters.However,setting parameters based on traditional experience or methods often leads to poor prediction performance.Using particle swarm optimization to optimize the parameters of the model can improve the accuracy and generalization ability of the model,but the algorithm itself also has some problems,which affects the actual optimization effect.This paper selects the runoff data of the Ganjiang River Basin,first preprocesses the data,and optimizes the parameters of the Kernel Extreme Learning Machine(KELM)model with the improved particle swarm algorithm.The main research contents are as follows:(1)Aiming at the problems that the particle swarm optimization algorithm is easy to fall into the local optimum,the convergence accuracy is not high and the convergence speed is slow,this paper proposes a hybrid hierarchical self-learning quantum particle swarm optimization algorithm HHQPSO.First,the particle population dynamics are divided into three levels according to the particle fitness value and the number of iterations:the upper and lower levels have fewer particle numbers,and the local learning model and the global learning model are used respectively to increase the particle diversity;the middle layer has a higher particle number distribution Many,using a hybrid adaptive quantum learning model.Secondly,in the hybrid quantum model,an improvement of the difference strategy is proposed to update the random position of the particles,and the Levy flight strategy is added to improve the convergence speed and accuracy of the algorithm.Finally,through comparative simulation experiments of 5 improved particle swarm algorithms on 9 typical test functions,the experimental results show that the HHQPSO algorithm has obvious advantages in convergence accuracy,speed and stability,and is especially suitable for multi-peak function search.excellent.(2)In order to improve the search accuracy and stability of the particle swarm,a weighted mutation particle swarm algorithm WVPSO is proposed.First,the adaptive inertia weight and adaptive learning factor are proposed to balance the global search and local search capabilities;then the replacement strategy based on arithmetic crossover mutation and natural selection mechanism is proposed,which increases the diversity ofparticles and improves the algorithm’s performance Convergence accuracy;Finally,Gaussian disturbance is added to make the particles oscillate,making it easier to jump out of the local optimum.Simulation experiments show that,compared with many representative swarm intelligent evolutionary algorithms,WVPSO has better results in solving accuracy and convergence speed,and has better accuracy and stability in high-dimensional function optimization problems.(3)Analyze the monthly and annual runoff of the Ganjiang River Basin.Combine the two improved particle swarm algorithms in this paper with the Kernel extreme learning machine model,and then train two combined models,HHQPSO-KELM and WVPSO-KELM.In order to test the performance of the two combined models of HHQPSO-KELM and WVPSO-KELM,the traditional Kernel extreme learning(KELM)machine prediction model,PSO-KELM model and GA-KELM model of genetic algorithm were selected,and these 5 models were compared to monthly runoff Data and annual runoff data are separately trained and forecasted.The experimental results show that the HHQPSO-KELM and WVPSO-KELM models proposed in this paper have achieved good forecast results and provide an effective theoretical basis for runoff forecasting. |
PDF Full Text Request