To better predict river water quality, this study utilized daily scale water quality data from 16 monitoring stations and daily meteorological data within Heilongjiang Province as the foundation. By employing Kernel Principal Component Analysis (KPCA) for feature extraction and dimensionality reduction of water quality data, Spearman's correlation analysis was used to screen for strongly correlated meteorological factors. These factors, along with the Water Quality Index (WQI), were taken as input data for the models. Three meteorological-water quality coupling prediction models were constructed using Random Forest (RF), Multilayer Perceptron (MLP), and Generalized Regression Neural Network (GRNN), and were compared and validated. The results indicated that the GRNN model had the best predictive performance, with an R-squared (R2) of 98.68% and an accuracy of 87.50% for the water quality conditions of the 16 stations, providing a basis for government decision-making in water resource management.