Power quality management as an operational capability: A dynamic capabilities perspective

Industrial energy management requires distinguishing between electricity demand and energy efficiency. This study examines associations between electrical operating conditions and electricity consumption using 35,040 fifteen-minute observations from one steel facility. Within an interpretive dynamic capabilities framework, Ordinary Least Squares (OLS), Random Forest (RF), Gradient Boosting Machine (GBM), and SHAP are applied, and an Operational Energy Review Score (OERS) is developed to prioritize monitoring. Lagging reactive energy is the dominant predictor. OLS explains 91.7% of full-sample variation; random-holdout R² values are 0.9995 for RF and 0.9986 for GBM. Chronological RF validation yields R² = 0.9985. Physical reconstruction using reactive energy and power factor achieves R² = 0.9999 within the applicable subset, suggesting that the predictive accuracy largely reflects electrical coupling rather than a newly identified managerial mechanism. OERS tracks consumption, but internal comparisons do not establish technical efficiency, avoidable waste, or realized savings. In the absence of production and process variables, high scores indicate intervals requiring contextual review. The contribution is a capability-oriented interpretation of operational monitoring; dynamic capabilities remain a theoretical lens rather than an empirically tested organizational construct. Cross-facility and cross-sector validations are needed.The study employed Ordinary Least Squares (OLS), Random Forest (RF), Gradient Boosting Machine (GBM), and SHapley Additive exPlanations (SHAP) within the framework of Dynamic Capabilities Theory and developed an Operational Energy Risk Score (OERS) to identify operating conditions associated with elevated electricity-utilization risk. The findings identified lagging reactive power as the strongest predictor of electricity consumption across all analytical approaches. While the OLS model explained 91.7% of the variance, the machine-learning models achieved substantially higher predictive performance (RF R² = 0.9995; GBM R² = 0.9986). SHAP analysis confirmed the dominant influence of power-quality variables, whereas temporal factors contributed comparatively little explanatory power. The proposed OERS was validated as a practical monitoring tool for identifying high-risk operating conditions. The study extended Dynamic Capabilities Theory by conceptualizing power-quality management as an operational capability and demonstrated that industrial energy performance could be enhanced through data-driven operational capabilities alongside technological investments