Eventos Anais de eventos
ENCIT 2018
Brazilian Congress of Thermal Sciences and Engineering
EVALUATION OF AN ARTIFICIAL NEURAL NETWORK MODEL COMPARED WITH A REGRESSION MODEL TO ESTIMATE THE STEAM FLOW GENERATION OF A REAL THERMOELECTRIC POWER PLANT DATA
Submission Author:
Helena Haas Reichert , RS
Co-Authors:
Helena Haas Reichert, Joao Fonseca, Paulo Smith Schneider
Presenter: Helena Haas Reichert
doi://10.26678/ABCM.ENCIT2018.CIT18-0207
Abstract
Due to its ability to model complex phenomena, applications of artificial neural networks (ANN) have been growing in heat transfer problems. The objective of this study is to develop and evaluate an ANN to estimate the steam flow generation of a real thermoelectric power plant. The power plant behavior is assessed based on past records of its operation. ANN configuration is obtained after a series of tests aiming to reduce the prediction error, and predictions are compared to a reference model based on multiple linear regression analysis. ANN prediction quality is assessed through the calculated mean absolute percentage error MAPE of 3 validation sets, which display results as 3.30%, 2.22% and 2.52%. Regression procedures applied to the same set of data have MAPE values of 6.64%, 2.24% and 3.94%, and show the ANN ability to estimate steam production from actual equipment.
Keywords
Artificial neural networks, Steam Generator, Thermoelectric power plant operation modelling
