STEAM PREDICTION MODEL

Propylene oxide (PO) is essential for many everyday products, such as furniture foams, car seats, and building insulation. Producing propylene oxide on an industrial scale requires pressurized steam to heat the reactors in the process. But how can we determine the appropriate amount of steam needed?

PO sales are planned one month in advance, allowing for the prediction of the steam needed to meet those sales targets. To accomplish this, Seeq was utilized to extract all the variables that might impact steam usage. Following that, a statistical analysis was conducted to identify the 5 most significant variables. With this information, an ordinary least squares (OLS) model was developed to accurately predict steam consumption.

The results were promising: the new steam prediction model reduced the error by 55%. This improvement gives a more accurate estimate of the steam needed each month, which ultimately improves decision making regarding when to activate additional steam generators.

Excel | JMP | Chemical Processes | Data Analysis