Attention-Based Deep Learning Methods for Predicting Gas Turbine Emissions

Rebecca Lauren Potts; Georgios Leontidis

Attention-Based Deep Learning Methods for Predicting Gas Turbine Emissions

Rebecca Lauren Potts^*
, Georgios Leontidis

^*Corresponding author for this work

Research output: Contribution to conference › Poster › peer-review

11 Downloads (Pure)

Abstract

Predictive emissions monitoring systems (PEMS) for gas turbines are critical for monitoring harmful pollutants being released into the atmosphere, while reducing the use of expensive continuous emissions monitoring systems (CEMS) which require daily maintenance to achieve accurate readings. We consider two attention-based deep learning models, FT-Transformer and SAINT, and compare with classical tree-based XGBoost to predict emissions from gas turbines. We find that the attention-based models outperform XGBoost for both prediction tasks, i.e. carbon monoxide (CO) and nitrogen oxides (NOx).

Original language	English
Number of pages	3
Publication status	Published - 23 Jan 2023
Event	Northern Lights Deep Learning Conference 2023 (Extended Abstracts) - Tromso, Tromso, Norway Duration: 9 Jan 2023 → 13 Jan 2023

Conference

Conference	Northern Lights Deep Learning Conference 2023 (Extended Abstracts)
Country/Territory	Norway
City	Tromso
Period	9/01/23 → 13/01/23

Bibliographical note

This work was supported by the Engineering
and Physical Sciences Research Council
[EP/W522089/1].

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 7 Affordable and Clean Energy
SDG 9 Industry, Innovation, and Infrastructure

Keywords

Artificial intelligence
Deep Learning
Gas Turbines
predicting emissions

Access to Document

Potts_NLDL extended abstractFinal published version, 185 KBLicence: Unspecified

Cite this

@conference{b19619093c8a4562ad56caf67e7e4cdb,

title = "Attention-Based Deep Learning Methods for Predicting Gas Turbine Emissions",

abstract = "Predictive emissions monitoring systems (PEMS) for gas turbines are critical for monitoring harmful pollutants being released into the atmosphere, while reducing the use of expensive continuous emissions monitoring systems (CEMS) which require daily maintenance to achieve accurate readings. We consider two attention-based deep learning models, FT-Transformer and SAINT, and compare with classical tree-based XGBoost to predict emissions from gas turbines. We find that the attention-based models outperform XGBoost for both prediction tasks, i.e. carbon monoxide (CO) and nitrogen oxides (NOx).",

keywords = "Artificial intelligence, Deep Learning, Gas Turbines, predicting emissions",

author = "Potts, \{Rebecca Lauren\} and Georgios Leontidis",

note = "This work was supported by the Engineering and Physical Sciences Research Council [EP/W522089/1].; Northern Lights Deep Learning Conference 2023 (Extended Abstracts) ; Conference date: 09-01-2023 Through 13-01-2023",

year = "2023",

month = jan,

day = "23",

language = "English",

}

TY - CONF

T1 - Attention-Based Deep Learning Methods for Predicting Gas Turbine Emissions

AU - Potts, Rebecca Lauren

AU - Leontidis, Georgios

N1 - This work was supported by the Engineering and Physical Sciences Research Council [EP/W522089/1].

PY - 2023/1/23

Y1 - 2023/1/23

N2 - Predictive emissions monitoring systems (PEMS) for gas turbines are critical for monitoring harmful pollutants being released into the atmosphere, while reducing the use of expensive continuous emissions monitoring systems (CEMS) which require daily maintenance to achieve accurate readings. We consider two attention-based deep learning models, FT-Transformer and SAINT, and compare with classical tree-based XGBoost to predict emissions from gas turbines. We find that the attention-based models outperform XGBoost for both prediction tasks, i.e. carbon monoxide (CO) and nitrogen oxides (NOx).

AB - Predictive emissions monitoring systems (PEMS) for gas turbines are critical for monitoring harmful pollutants being released into the atmosphere, while reducing the use of expensive continuous emissions monitoring systems (CEMS) which require daily maintenance to achieve accurate readings. We consider two attention-based deep learning models, FT-Transformer and SAINT, and compare with classical tree-based XGBoost to predict emissions from gas turbines. We find that the attention-based models outperform XGBoost for both prediction tasks, i.e. carbon monoxide (CO) and nitrogen oxides (NOx).

KW - Artificial intelligence

KW - Deep Learning

KW - Gas Turbines

KW - predicting emissions

M3 - Poster

T2 - Northern Lights Deep Learning Conference 2023 (Extended Abstracts)

Y2 - 9 January 2023 through 13 January 2023

ER -

Attention-Based Deep Learning Methods for Predicting Gas Turbine Emissions

Abstract

Conference

Bibliographical note

UN SDGs

Keywords

Access to Document

Fingerprint

Cite this