Industrial lighting is a major contributor to global electricity consumption, driving the need for optimized energy management strategies. Traditional energy consumption models, constrained by static assumptions, often fail to capture the temporal and operational complexities of industrial lighting systems. This research aims to address this gap by leveraging statistical and machine learning techniques to analyze high-resolution lighting energy consumption data from South Korean industrial facilities. The proposed work utilizes Recurrent Neural Networks (RNN) and Bidirectional Long Short-Term Memory (BILSTM) for predictive modeling. Our methodology includes advanced imputation techniques (GAIN) for addressing missing data, Z-Score normalization for consistency, and K-Fold Cross-Validation to enhance model robustness. Among the predictive models evaluated, BILSTM achieved superior performance, with a Mean Absolute Percentage Error (MAPE) of 1.83%, demonstrating its efficacy in sequence modeling and anomaly detection. The proposed approach not only uncovers operational patterns and consumption anomalies but also provides actionable insights for optimizing industrial lighting systems. This research offers a scalable framework for improving energy efficiency, aligning with global sustainability goals and fostering both economic and environmental benefits in industrial practices.