This study presents a water quality prediction model for urban waterways using wavelet packet denoising (WPD) and long short-term memory (LSTM) neural networks. The model was developed using real-time water quality data from four monitoring stations in the Fenjiang River Basin of Foshan City, collected between 2016 and 2021. The data included eight parameters, including pH, conductivity, water temperature, flow rate, turbidity, NH3-N, COD, and DO. The main factors affecting water quality were identified as COD and NH3-N. To improve data accuracy, the original data were denoised using WPD. An LSTM-based prediction model was then developed to forecast NH3-N and COD levels. The model demonstrated improved predictive accuracy compared to traditional LSTM models, with significant reductions in RMSE values for both parameters. In the 12-hour prediction, the RMSE values of NH3-N predictions decreased by 55% to 67%, with an average decrease of 61%, while the RMSE values of COD predictions decreased by 18% to 51%, with an average decrease of 29%. In the 3-day prediction, the RMSE values of NH3-N predictions decreased by 40% to 83%, with an average decrease of 65%, while the RMSE values of COD predictions decreased by 50% to 69%, with an average decrease of 60%. The study concludes that the WPD-LSTM model provides more accurate predictions for water quality in urban waterways, supporting effective water environment management. The model can also be applied to predict water quality in other sections of the Fenjiang River and other rivers in the future.This study presents a water quality prediction model for urban waterways using wavelet packet denoising (WPD) and long short-term memory (LSTM) neural networks. The model was developed using real-time water quality data from four monitoring stations in the Fenjiang River Basin of Foshan City, collected between 2016 and 2021. The data included eight parameters, including pH, conductivity, water temperature, flow rate, turbidity, NH3-N, COD, and DO. The main factors affecting water quality were identified as COD and NH3-N. To improve data accuracy, the original data were denoised using WPD. An LSTM-based prediction model was then developed to forecast NH3-N and COD levels. The model demonstrated improved predictive accuracy compared to traditional LSTM models, with significant reductions in RMSE values for both parameters. In the 12-hour prediction, the RMSE values of NH3-N predictions decreased by 55% to 67%, with an average decrease of 61%, while the RMSE values of COD predictions decreased by 18% to 51%, with an average decrease of 29%. In the 3-day prediction, the RMSE values of NH3-N predictions decreased by 40% to 83%, with an average decrease of 65%, while the RMSE values of COD predictions decreased by 50% to 69%, with an average decrease of 60%. The study concludes that the WPD-LSTM model provides more accurate predictions for water quality in urban waterways, supporting effective water environment management. The model can also be applied to predict water quality in other sections of the Fenjiang River and other rivers in the future.