This project aims to develop an innovative model based on neural networks and differential equations (DEs) to analyze, simulate, and predict the dynamics of pollutants in water bodies. The approach focuses on mixing problems, where the concentration of substances changes over time due to the inflow and outflow of pollutants within the system. Through mathematical modeling, the project seeks to accurately represent the physical and chemical processes affecting water quality, while neural networks provide the capability to learn complex patterns and make precise predictions based on observed data. The proposed methodology enables the identification of critical pollution peaks, predicts their evolution over time, and assesses their environmental impact, offering essential information for preventing and mitigating risks associated with water quality.