Parameter Optimization for Forecasting of Boiler Losses and Efficiency using Statistical Model

Abstract

Estimation of boiler efficiency to the highest degree of acceptance is the need of the hour.
The correct evaluation of boiler efficiency is hypothetical due to several reasons, as
mathematical equations are complex, time consuming and prone to human error. The
statistical model based estimation is more efficient. This paper presents the optimization of
boiler parameters based on statistical models for the prediction of boiler losses and
efficiency. The data is collected from a well established and sophisticated cement plant and is
used for building the models and testing the forecast values of boiler losses and efficiency.
Linear regression is performed for the prediction of boiler losses and boiler efficiency.
Independent variables for regression analysis are selected from a large set of boiler
parameters. Parameter selection for the prediction of both boiler losses and boiler efficiency
is based on the basis of experience. The final independent variables are optimized to get
higher accuracy. Six models are built. Three of these models are for individual boiler loss
predictions and other three models for the prediction of boiler efficiency. Out of three models
for boiler efficiency prediction, each model is built by considering a particular independent
variable. The independent variables are: hydrogen content in coal, moisture in coal, and both
hydrogen and moisture content in coal. The dependent variable in all three cases is boiler
efficiency. The optimized model for prediction of boiler efficiency is built considering
hydrogen and moisture content in coal as an independent variable.