Forecasting-based energy-efficient design and implementation of containerized mobile processing system for remote power stations using STDO-BiGRU

Remote power stations in isolated or off-grid areas face challenges of limited infrastructure, high operating costs, and inefficient energy utilization. This study proposes a novel Scalable Tasmanian Devil Optimized Bidirectional Gated Recurrent Unit (STDO-BiGRU) framework that combines an adaptive metaheuristic optimizer with a deep recurrent predictor to enhance energy efficiency, load forecasting, and container migration in mobile processing systems for remote power stations. The innovation lies in integrating the STDO algorithm with the BiGRU architecture to dynamically predict workload variations and intelligently switch server states between active and idle modes, reducing unnecessary power consumption. The system employs containerized edge-based architecture, enabling localized decision-making without reliance on centralized infrastructure. To provide both accuracy and consistency for model training, the Remote Power Station Operational Dataset with 10,560 data records are obtained and preprocessed by data cleaning, and z-score normalization methods. Having 94.4% accuracy, and 93.7% F1-score, the model outperformed the baseline techniques while tested with 10-fold cross-validation. The results demonstrate that, in comparison to traditional methods, the created STDO-BiGRU model greatly improves energy effectiveness, functional performance, and prediction dependability. The system provides effective energy use and dependable operation in remote places by offering a versatile and sustainable solution to power storage devices, remote generators, and portable energy infrastructures. Smart forecasting reduces power use and improves reliability in remote energy stations. The novel STDO-BiGRU model boosts accuracy and lowers system delay under changing workloads. Container-based design enables faster, local decisions without needing central infrastructure.