The influence of machine learning technologies and deep learning algorithms is rapidly increasing and penetrating almost in every field, and water quality prediction is not being excluded from those fields. This paper propose deep learning models such as gated recurrent unit (GRU), long short-term memory (LSTM) and hybrid convolutional neural networks-long short-term memory (CNN-LSTM) to predict water quality in the smart cities. We start by collecting and prepocessing data in order to analyse correlation between different water parmeters. Then, we deploy our deep learning models, we train and test them using the dataset of water. Finally, we evaluate the three proposed models by evaluting their prediction results and examine the prediction accuracy using mean squared error (MSE), root mean square error (RMSE) and the coefficient of determination ( \(R^2\) ) between the predicted values and the real values.