According to the characteristics that the conventional model of dam safety early-warning is sensitive to the missing data and the forecast accuracy is prone to be affected by other factors, a model combing the improved PSO algorithm and SVM theory was proposed, which seeks the optimal parameters of SVM model through PSO algorithm. Meanwhile, the convergence de-gree is introduced and the inertia weight factor and study factor are optimized to avoid the premature convergence of PSO, so the global searching ability is improved. The establishing process of the model was introduced and the forecast accuracy was verified by actual monitoring data. With the analysis of the example, it is proved that the improved model is superior to the conventional model, and the application range of PSO algorithm is extended.%针对传统大坝安全变形预警监控模型对缺失数据敏感、精度易受其它因素影响的特点,提出了一种利用粒子群算法与支持向量机相结合的建模方法. 即通过粒子群算法对支持向量机模型的参数进行寻优,同时改进了惯性权重因子与学习因子,并引入参数收敛程度,有效地解决了粒子群算法存在的早熟收敛问题,提高了全局收敛能力. 阐述了模型建立的算法步骤,并利用某水电站观测数据进行了验证. 结果表明,相对于传统优化算法,改进的PSO-SVM模型在大坝安全变形监控上具有很大的优越性,而且也扩展了粒子群算法的应用范围.
展开▼