Sliding Window–Neural Network and Round-Robin Framework: A Cloud-Based Data-Driven Energy Load Scheduling for Household Renewable Energy Systems

This study proposed a data-driven framework for real-time battery load scheduling that uses a sliding window and an artificial neural network (ANN) integrated with a round-robin switching mechanism. The system utilizes voltage-only measurements from multiple lead-acid batteries and processes real-time data using a 36-lag sliding window to capture nonlinear temporal discharge behavior. The ANN model, trained on historical discharge datasets and deployed in a pretrained format, generates a similarity score that informs switching decisions based on a predefined threshold. Experimental results showed that the proposed method accurately predicts low State of Charge (SoC) conditions, with decision outputs consistently operating near the 10 V threshold, thereby preventing deep discharge. In addition, the ANN model achieved a computation time of less than 1s, demonstrating its suitability for real-time implementation. The integration of round-robin scheduling ensures balanced battery utilization and system reliability. Overall, the proposed approach provides an efficient and practical solution for intelligent battery management in real-world energy systems.