Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : ASEAN Journal for Science and Engineering in Materials

Analysis of Bed Temperature on Circulated Fluidized Bed Boiler Using Simple Multivariable Regression Kurniawan, Asep; Kurniawan, Teguh; Alwan, Hafid; Bindar, Yazid; Irawan, Anton
ASEAN Journal for Science and Engineering in Materials Vol 3, No 2 (2024): AJSEM: Volume 3, Issue 2, September 2024
Publisher : Bumi Publikasi Nusantara

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

A Circulated Fluidized Bed (CFB) boiler is a type of steam boiler with more complex phenomena of fluidization and combustion occurring in the furnace. One of the operating problems is the temperature bed which is difficult to predict. Bed temperature prediction is important as a reference to know the combustion process and heat transfer along the furnace. The purpose of this study is multivariable data analysis to predict bed temperature based on historical data. The amount of historical data is then prepared for the dataset and passes through the stages of data cleansing, visualization, exploration, and engineering judgment. The parameters selected as control variables after going through the first principal analysis are 5 parameters, namely gross power, coal feed (X1), primary air (PA) flow (X2), secondary air (SA) flow (X3), and average bed temperature (y). The dataset is then divided based on the load into 2 groups a low load of 20.03-30.00 MW and a high load of 30.01-54.41 MW. Each parameter is converted to the natural logarithm (ln) then multivariable regression is performed. The result is a low load model equation  with Root Mean Square Error (RMSE) = 23.2813 and a high load model equation  with RMSE = 4.8416. This model can be used to predict the average bed temperature at certain input conditions of coal feed, PA flow, and SA flow according to operating load. Prospects for bed temperature prediction with this multivariable can be developed using data-based machine learning so that the operating patterns obtained are more accurate and real-time forecast prediction.