Rainfall forecasting plays a crucial role in disaster prevention, water resource management, agricultural support, transportation, and industry. It aids in early warnings, production planning, safety assurance, damage mitigation, and resource optimization across various fields. This paper introduces a rainfall forecasting model for the next hour (Nowcasting) using CNN (Convolution Neural network) combined with GRU (Gated Recurrent Unit) neural networks and integrated with an improved genetic algorithm (GA) to achieve better performance. The theoretical contribution of this paper is the creation of a new model that combines CNN layers with GRU and suggests improvements to the GA for optimizing the model’s weights. Practically, the model is tested with radar data from the Phadin station collected from 6:00 AM to 4:00 PM on July 10, 2020. Results indicate that this model with the improved GA outperforms other models.