<p>Mobile Wireless Rechargeable Sensor Networks (MWRSNs) have emerged as a pivotal technology for enabling sustainable operations in Wireless Sensor Networks (WSNs) by mitigating energy depletion issues. However, optimizing the deployment and scheduling of Mobile Chargers (MCs) in such networks remains an NP-hard problem, especially in large-scale and dynamic environments. To address the challenges of efficient charger deployment and dynamic energy demands, machine learning algorithms are employed to enable accurate prediction and adaptive decision-making. Initially, three predictive models Neural Networks (NN), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM)- were employed to accurately predict the minimum number of MCs required to maintain continuous sensor node operations. To further optimize post-deployment operations, we introduce TRIDENT (Trajectory-based Iterative Estimation for Node Tracking), a novel algorithm that performs cluster-wise angular partitioning followed by Prim’s algorithm within each sector to generate locally optimized Minimum Spanning Trees (MSTs). This sector-based modular optimization significantly reduces path overhead, computational complexity, while ensuring efficient dynamic energy replenishment. Extensive experiments validate the efficacy of the proposed approach, achieving MC prediction accuracies of 99.84%, 99.72%, and 99.88% for NN, LSTM, and XGBoost models respectively, while substantially minimizing traversal distances and enhancing the overall network lifetime. The framework demonstrates strong potential for real-world applications including smart cities, environmental monitoring, and industrial automation.</p>