Data-Driven Insights into HVAC Plant Performance: Regression Versus Random Forest on Real-World Manufacturing Data

This study aims to explore the application of data analytics by assessing and improving the energy performance of the heating, ventilation, and air-conditioning (HVAC) system in a packaging manufacturing facility. Using the real-world operational data of the last 21 months of summer, a multiple linear regression model was developed with energy consumption of HVAC plant (kWh) as the outcome/dependent variable and four key predicting variables: production volumes (tonnage), weather temperature (°C), building volume (ft3), and plant operating duration (hours). The model achieved high explanatory power, with an R2 (coefficient of determination) of 0.973, indicating that 97.3% of the variance in energy consumption can be attributed to these factors. To validate and strengthen the reliability of the findings, a Random Forest regression model was employed as a benchmark for machine learning. The Random Forest model validated the significance of the selected predictors and demonstrated comparable predictive performance. The study highlights how combining traditional statistical techniques with machine learning can provide actionable insights into energy management, supporting data-informed decision-making and operational efficiency in industrial environments. Moreover, this study identifies the key variables that influence the energy performance of an HVAC plant, particularly in a packaging manufacturing context.