Data-Driven Prioritization of Grid Interruptions: A Lean Six Sigma Study Using Public Outage Logs

Abstract

The supply of power that is reliable and uninterrupted is one of the pillars of the contemporary energy infrastructure. Nonetheless, the issue of grid interruptions which are frequent and long-lasting is still compromising the resilience of the system and customer satisfaction. Traditional approaches to outage-management and prioritization tend to be reactive and based on manual evaluation and missing records of events. To overcome such shortcomings, this paper suggests a hybrid meta-learning model that would combine Random Forest (RF), Extreme Gradient Boosting (XGBoost), and a Deep Artificial Neural Network (Deep ANN) to come up with the data-driven prioritization of grid interruptions. The data were processed extensively with the Missing-value treatment, derivation of impact-score, feature scaling using robust scale and SMOTE and controlled Gaussian perturbation using publicly available data 15 Years of Power Outages (2000 - 2014) dataset. The proposed hybrid model integrates RF and XGBOOST in an ensemble layer while capturing the variance decomposition and residual optimization for variance decomposable nonlinearities and the Deep ANN meta-learner captures complex non-linear dependencies. Instead of maintenance, proactive and timely, the system prioritizes the outage events on the basis of the predicted severity of impact in the framework of Lean Six Sigma DMAIC. Comparative analysis with four baseline models which included RF, XGBoost, Deep ANN, and RF-XGB models showed that the hybrid model was better as it had 99.43% accuracy, 98.92% precision, 99.15% recall, and 99.03% F1-score. The obtained results support that the suggested meta-learning technique offers a significant enhancement in the robustness, scalability, and interpretability of prediction of outage impact. This study creates a solid information-driven approach to the implementation of the Six Sigma principles into the contemporary energy-analytics processes that opens the way to increased intelligence and agile power-grid operations.