The demand for drinking water continues to increase daily in parallel with population growth. Unplanned urbanization, industrialization, and migration further intensify the need for a reliable water supply. Accurate forecasting of water consumption is crucial for efficient planning of production, transmission, and distribution, as overestimation may lead to unnecessary infrastructure costs, whereas underestimation can result in supply shortages. In this study, future drinking water consumption in the city of Kocaeli was forecasted using three time series methods: Long Short-Term Memory (LSTM), Seasonal Autoregressive Integrated Moving Average (SARIMA), and fully connected Artificial Neural Networks (ANN). Monthly historical consumption data spanning 11 years (2010–2020) were utilized for model training and evaluation. Experimental results indicate that the LSTM-based model effectively captures the dynamic patterns in the consumption data and outperforms SARIMA and ANN models in predictive accuracy. The findings highlight the potential of deep learning approaches, particularly LSTM networks, for reliable short- and long-term water consumption forecasting.
İsmail Gülsoy (Mon,) studied this question.