| Unsaturated soil moisture movement is a major component of the natural water cycle and plays an important role in farmland irrigation and drainage and water resource evaluation.The movement process of soil water can be reflected by the basic equation of soil moisture movement,but the acquisition of soil moisture movement parameters is the premise for solving the equation of unsaturated soil water movement.How to select an effective method to obtain parameters is a hot spot in current research.Based on whale optimization algorithm and grey wolf optimization algorithm,this paper inverts the parameters of soil water motion equation by algebraic methods,numerical methods and machine learning.The main research is as follows:(1)Based on the whale optimization algorithm and the grey wolf optimization algorithm,the objective function of the inversion model was constructed by using the analytical solution of the vertical infiltration problem of unsaturated soil moisture,and the soil hydrodynamic parameters randomly generated by the swarm intelligent optimization algorithm were substituted into the analytical solution to obtain the analytical solution data of cumulative infiltration amount I,infiltration rate i,soil water content 0 and infiltration time T,and compared with the numerical solution obtained by HYDRUS-1D simulation to optimize the optimal parameters with the smallest objective function.The results showed that both methods can better invert the soil moisture movement parameters,and the parameter error obtained by the fixed parameters θ,and θs inversion under the whale optimization algorithm is small.(2)Based on the whale optimization algorithm and the grey wolf optimization algorithm,the numerical simulation model of soil moisture motion equation was constructed by using the finite difference method,and the soil hydrodynamic parameters randomly generated by the group intelligent optimization algorithm were substituted into the numerical model,and the numerical solution data of cumulative infiltration I,infiltration rate i and soil water content θare simulated and compared with the numerical solution obtained by HYDRUS-1D simulation,and the optimal parameters with the smallest objective function were optimized.The results showed that the error of the parameters inverted by the grey wolf optimization algorithm when the parameters θr and θs are fixed,and the numerical method had higher inversion accuracy and lower efficiency than the algebraic method.(3)Based on machine learning,this paper used the Regression Learner toolbox in Matlab for model training,took the cumulative infiltration and wet front depth as the training set,the soil parameters θr,θs,hd,n,and Ks as the response variables,set the cross-validation fold to 10,trained under the linear regression model,regression tree,and Gaussian regression model,and outputted the regression model with the best training effect,and predicted soil parameters by the model with the best training effect.The results showed that machine learning can better invert soil hydrodynamic parameters,which can provide reference for the parameter inversion of soil water movement equations. |
