The improved method of least squares support vector machine modeling and its application

Least squares support vector machines (LS-SVM) method is used for modeling, and its penalty factors and kernel parameters with different values will affect the accuracy of the soft sensor model. This paper presents a particle swarm optimization (PSO) algorithm with mutation to automatically search the parameters for LS-SVM, and is applied to real-time measurement problem of saturated vapor dryness in gas driving oil extraction. The proposed algorithm is based on statistical learning theory to map the complex nonlinear relationship between dryness and its influence factors by learning from empirical data, therefore, saturated vapor dryness can be forecasted. The experimental results show that soft sensor modeling based on particle swarm optimization with mutation has high precision, adaptability, and ease of practical application.