Statistical Early Warning Model for Dam Based on Improved Particle Swarm Coupled Method

Because the dam safety influencing indexes are complex,sometimes the results are bad when the statistical model is applied to early warning evaluation for dam. Suppose regression coefficients is transformed to the linear programming question. The coefficients of multi-statistic regression can be determined by the particle swarm optimization algorithm. But when the PSO algorithm was applied to high dimension space optimization question,the convergence rate would be slow and the calculations could easily fall into local extreme points. In order to overcome these shortcomings,the PSO is improved,and a new self-adapting strategy is proposed. The coupled method combines a new self-adapting strategy and simulated annealing algorithm,and the statistical early warning model for dam is based on it (SA-APSOR). Applications in practical engineering show that this method can improve the convergence ability of PSO,and can avoid the algorithm falling into the local extreme points. Hence,the convergence rate of this method is quick. Furthermore,compared with the traditional least square regression,the forecast precision of this model based on SA-APSOR is high and its early warning evaluation results almost correspond with the practice operating condition.