Development of Gas Turbine Predictive Emissions Monitoring System Model based on GRNN-GA Hybrid Method

Winarto, Rudy (2024) Development of Gas Turbine Predictive Emissions Monitoring System Model based on GRNN-GA Hybrid Method. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text THESIS_Rudy Winarto_6047222026_FINAL APPROVAL.pdf - Accepted Version Restricted to Repository staff only Download (27MB) | Request a copy

Abstract

Gas turbines are crucial for the process industries. They are used in several critical operational areas and are an important contributor of atmospheric pollution by emitting particulate matter, Carbon Monoxide (CO), Nitrogen Oxide (NO), and Sulphur Dioxide (SO2).
For several process sectors, Predictive Emissions Monitoring Systems (PEMS) have shown to be a practical substitute for CEMS. By integrating PEMS, the emission prediction technique could be streamlined and a more cost-effective solution that complies to regulations.
The development of the PEMS system in this research is based on the hybrid method of GRNN-GA. The primary objective is to enhance the predictive capabilities by applying a genetic algorithm to optimize its variables.
Turbine and CEMS emissions-related process data collected at intervals from 00:00 on January 1, 2021, to 23:59 on December 31, 2023, from Process Historian Data. An individual of 50,370 data of each process variable and a total of 1,006,740 data sets is employed for PEMS model development.
Using a setup with 17 input parameters, a 32-neuron hidden layer with ReLU activation, a single output neuron with linear activation, and 100 epochs of Adam's optimization, the GRNN-GA hybrid method is applied as a predictive model. For hyperparameter optimization, a genetic algorithm with a population size of 2000, a crossover probability of 0.7, a mutation probability of 0.2 and 50 generations was applied.
The GRNN-GA model resulted a significant improvement over the GRNN model, with MAPE values range far below 1%. Additionally, the R² values for the GRNN-GA model are near-perfect, ranging from 0.994 to 1.0, indicating a high level of predictive accuracy compared to the GRNN model's R² values which range from 0.373 to 0.985. It demonstrates better model fit, greater predictive power, improved accuracy, reduced error margins, and more precise predictions, making it a superior choice for accurate and reliable gas emissions prediction. Continuous evaluation will ensure its reliability and stability for long-term applications.

Keywords: emission, GRNN, GRNN-GA, model, prediction, PEMS, ReLU

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Emission, GRNN, GRNN-GA, model, prediction, PEMS, ReLU Emisi, GRNN, GRNN-GA, pemodelan, prediksi, PEMS, ReLU
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 78201-System And Technology Innovation
Depositing User:	Rudy Winarto
Date Deposited:	04 Aug 2024 12:38
Last Modified:	04 Aug 2024 12:38
URI:	http://repository.its.ac.id/id/eprint/110313

Actions (login required)

View Item