基于改进的偏最小二乘回归的酸雨pH值预测

The Prediction of pH Values in Acid Precipitation Based on Modified Partial Least Squares Regression Model

酸雨pH值受到酸性离子[SO42-]、[NO3-]和碱性离子[Ca2+]、[NH4+]等的影响。这些影响因素之间存在多重相关性。用一般最小二乘回归分析预测pH值,参数估计存在很大的误差且物理意义明显不足。应用偏最小二乘回归技术建立pH值预测模型,克服了自变量之间多重相关性的问题,因而更具有先进性,计算结果更为可靠,而改进的偏最小二乘回归则从预测角度对偏最小二乘回归模型进行了改进。以我国17个城市pH值预测为例,说明了改进的偏最小二乘回归法比普通偏最小二乘回归法效果好。

pH values of acid precipitation are affected by not only acid ions [ SO4^2- ] and [ NO3^-] but also alkaline ions [Ca^2＋ ], [NH4^＋] and so on. There are some multiple correlations between these factors, As the multiple correlations existed, the estimated regressive parameters with the least squares method include a good deal of errors and in the multiple regression equation it cannot reflect its physical meaning, The partial least squares method can easily solve the multiple correlated problems. The method is simple and quick to calculate, The estimated regressive parameters from it are robust, The modified partial least squares regression improves the partial least squares regression model in the prediction. A case study, the pH values prediction of 17 cities in China, has been researched, The results show that the modified partial least squares regression is better than the former.