Utilization of MLP and Linear Regression Methods to Build a Reliable Energy Baseline for Self-benchmarking Evaluation

Abstract

This paper presents a reliable energy baseline model for self-benchmarking evaluation of energy saving potential by using multilayer perceptron (MLP) method. The measured energy data and product quantities of the sample plant in daily period dating back since 2011 to 2016 are used as variables and then normalized to represent the energy baseline (EnB) of the manufacturing plant. A comparison of MLP and linear regression (LR) methods for creating the baseline model is investigated during the factory expansion capacity. For LR method, we use the ASHRAE Guideline 14-2002 as a reference in recommended values for modeling uncertainty. As the uncertainty problem, the LR method is more sensitivity to the outliners, because the nature of plant variables has more complexity and nonlinearity. So we introduce the MLP method to solve or reduce the effect of nonlinearity by supervised learning in the short-term and long-term period of the production. For simulation results, in short-term period the LR method demonstrates some better results of uncertainty parameters. However, the proposed MLP with LR method can build a reliable baseline showing in better R-square values than LR method. This is useful for energy evaluation when the plant is expanding capacity to protect misleading interpretation occurring during the year. For long-term period, the MLP method can overcome the LR method in all uncertainty parameters. Therefore, the MLP method may be able to the alternative choice for creating the EnB in nonlinearity circumstances of the plants for short-term and long-term period.