Metaheuristic-optimized forecasting in a smart Edge—Fog—Cloud energy management framework: An industrial mining case study

One of the most energy intensive industry is the mining operation, and with an abnormal and non efficient energy usage there is a high chances to increase the operational cost of the operation, and still many mining sites lack a unified real time monitoring frame work that make a proactive decisions. This article will present a case study that focuses on the Open Pit Mine of Benguerir, with different power meters technologies deployed to monitor the 3 main stations the site, Destoning, Screening and train loading stations. It proposer a hybrid IIoT AI SCADA architecture, and a comparative study between machine learning models including the hyper parameters optimization using meta heuristics for better results. This study proposes a modular architecture for Data acquisition, Data storing, and monitoring through OPC UA, PostgreSQL and Python. After acquiring and cleaning the data, the initial test show that 3 models are typically good, then increasing their efficiency by tuning the hyper parameters using. For forecasting, a comparative analysis of multiple machine learning models—including linear regression, ANN, tree-based ensembles, and support vector regression was conducted. XGBoost emerged as the best-performing model, and its hyperparameters were further optimized using metaheuristic algorithms (GA, PSO, GWO, WOA, SHO, SFO), leading to significant improvements in accuracy and robustness across three critical stations: destoning, screening, and train loading. The results demonstrate that metaheuristic-optimized XGBoost achieves up to 90% reduction in prediction errors compared to baselines, enabling proactive load management and reducing operational uncertainties. Beyond technical performance, the proposed framework offers a scalable blueprint for energy forecasting in energy-intensive industries, supporting cost reduction and digital transformation in mining operations. • Novel Smart Edge–Fog–Cloud architecture for energy monitoring and forecasting. • Validated with smart power meters, SCADA, and industrial data acquisition. • Metaheuristic optimization improves ML forecasting performance in different mining operations. • Real case study: destoning, screening, and train loading stations in Morocco Experimental Open Pit Mine of Benguerir. • Scalable IIoT-based EMS combining AI, optimization, and digital twin principles.