Optimizing the Energy Consumption Costs of a Production Plant

This project presents a comprehensive study to optimize the energy consumption costs of one of FATER’s manufacturing plants. The global trend in energy consciousness and the desire to reconcile cost-efficiency with environmental responsibility are critical pillars on which this project at FATER is based. Reducing these costs is a major motive for energy optimization as energy expenses account for a significant amount of operational costs for the plant, emphasizing the direct financial impact of wasteful energy consumption on the company. Therefore, this thesis aims to understand and analyze the historical records of energy consumption within the plant and forecast this later using machine learning models as well as calculating the associated costs and presenting a set of actionable strategies adapted to the plant’s specific needs. The machine learning models that predicted energy consumption were compared based on metrics such as R2, MAE, RMSE, and MAPE. The final model achieved a 4.47% MAPE across all cross-validation folds and less than 10% for each predicted label. After that, the associated costs for the predicted energy consumption were calculated based on various constraints for when the engine is on and when it is off. These costs include the manufacturing costs, the methane costs, the energy purchased costs, the revenues from the energy sold, and the Wartsila costs. Finally, the best strategy was assigned to each day of the week and represented either “ON”, or “OFF, indicating whether to use the engine and generate power or buy it directly from Enel. This project is a step towards improving sustainability by prioritizing long-term cost-effectiveness and environmental responsibility within FATER.