<![CDATA[The Simpang Lima PUPR Pump Station on Jalan Radial, Palembang, serves as a critical drainage point for the largest water discharge in the downstream area, making the surrounding region highly vulnerable to surface runoff and flooding, especially during short-duration high-intensity rainfall events. This study aims to develop a 24-hour ahead multi-step river water level prediction model using the Gated Recurrent Unit (GRU) algorithm, powered by real-time data from Internet of Things (IoT) sensors installed at the pump station. The collected dataset spans from June to July and includes water level, rainfall, temperature, humidity, and barometric pressure. The data was preprocessed through normalization before being used as input to the GRU model. The GRU-based prediction model demonstrated strong performance with a Mean Squared Error (MSE) of 0.394, Root Mean Squared Error (RMSE) of 0.628, coefficient of determination (R²) of 0.99, and Nash-Sutcliffe Efficiency (NSE) of 0.9853. These results indicate high predictive accuracy and model reliability. The proposed model has strong potential for integration into early warning dashboards to support flood mitigation strategies and improve the operational efficiency of pump stations in high-risk urban zones. Additionally, this research offers a data-driven framework for the Ministry of Public Works and Housing (PUPR) to design real-time, predictive flood control systems. The approach can optimize pump operations, enhance emergency response planning, and guide drainage infrastructure improvements. Furthermore, it promotes climate-resilient flood adaptation policies and serves as a model for smart technology deployment in other Indonesian cities.]]>
