Water-to-water heat pumps are a key technology for efficient heating, but their performance depends heavily on precise control of operational parameters like flow and return water temperatures. In this study, we used neural networks to predict these temperatures during heating mode operation. Real-world data was collected over a month, with measurements taken every minute, including flow water temperature, return water temperature, and outdoor air temperature. The neural network effectively captured the complex relationships between these variables, delivering highly accurate predictions. By using this approach, we show how machine learning can help optimize heat pump performance, making them more energy-efficient and reliable. This work demonstrates the potential for smarter, data-driven heat pump control strategies that can benefit both energy savings and sustainability. .