Energy Savings Prediction: a Case Study

Abstract

This case study illustrates the prediction of energy savings at a chemical plant after energy saving improvementswere implemented at the plant. The study shows the danger in fitting prediction equations by automatic procedures,and how over fitting an equation can actually increase the error of prediction. Errors of prediction estimatedfrom the data used to find the equation are shown to be optimistic. Data splitting is demonstrated as amethod to validate a prediction equation, and a more reliable prediction equation than the one found by theautomatic procedure was developed along with a more realistic estimate of prediction error.