基于典型相关和人工粒子群的大坝监测分析模型

Dam Monitoring Analysis Model Based on Typical Correlation and PSO

针对大坝安全监测多效应量统计分析模型存在相关性而影响模型精度问题,提出了优化方案,预先对效应量和影响因子进行典型相关性分析,实现变量降维并提取与效应量相关性大的自变量作为模型输入因子构建多元线性回归模型,并基于人工粒子群算法求解最优偏回归系数。经与全回归、逐步回归、偏最小二乘回归模型比较验证,实例表明所建模型有较优的拟合效果和预测精度、鲁棒性强。

In this paper, an optimization program is proposed for accuracy problem caused by the correlation of multi- factors on dam safety monitoring analysis model. Firstly, typical correlation analysis is carried out with effect variables and influencing factor to obtain dimension reduction and principal components, on which a multiple linear regression mod el would be built as input factors. And then, partial regression coefficients are solved based on artificial particle swarm optimization algorithm. Compared with full regression, stepwise regression, and partial least squares regression, the re- suits show that the proposed model has good fitting effect and forecast accuracy, as well as strong robustness.