Energy Demand Forecasting in an Automotive Manufacturing Plant

Energy analysis is an essential topic within a sustainable manufacturing strategy. To better understand the energy demand in a manufacturing plant, consideration of trends and patterns of energy consumption, and making predictions based on historical data is a promising approach. Time series analysis is a favorable method to be used; because of the rapid development in metering/sensor technology and computational systems, time series analysis can now be deployed on larger-scale systems. However, the application of time series models to manufacturing plant energy modeling is rare due to complexity. This paper augments traditional time series forecasting for manufacturing energy study, with the consideration of data trend and patterns, exogenous influential inputs, and potential overfitting issues. Automotive manufacturing plant electricity demand was used as a study case for the proposed modeling approach validation. In this research, time series analysis is shown to effectively capture the increasing trend and seasonal patterns in the energy demand of a vehicle manufacturing plant. Models with exogenous inputs show a better accuracy as measured by Mean Square Error, and are more robust to sudden deviations. 1 INTRDUCTION Consideration of energy prediction is an essential topic in sustainable manufacturing. Energy consumption in the production plant is not only expensive, but also environmentally alarming. Manufacturers are eager to act on energy conservation because of the pressure from public image, financial budgets, regulations, and internal strategies. Understanding the energy demand in the manufacturing plant is the first step for further smart energy management. To better understand the energy demand in a manufacturing plant, a promising approach is studying the trends and patterns of the energy consumption and making predictions based on historical data. With accurate demand forecasting, manufacturers can better operate an on-site energy supply system [1], further realize the situational intelligence (integrated historical and real-time data to implement near-future situational awareness), and guarantee the supply stability. It can also assist to intelligently schedule production and manage the working conditions to avoid peak load, and to create deeper knowledge on how the manufacturing plants affect the local energy distribution. In addition, the results of forecasting can support budget and investment decisions, and aid negotiation with utility suppliers; therefore, cutting down the energy cost.