Gas turbine (GT) trip is one of the most disrupting events that affect GT operation, since its occurrence causes a reduction of equipment remaining useful life as well as revenue loss because of business interruption. Thus, early detection of incipient symptoms of GT trip is crucial to ensure efficient operation and lower operation and maintenance costs. This paper applies a data-driven methodology that employs a Long Short-Term Memory (LSTM) neural network and a clustering technique to identify the time point at which trip symptoms are triggered. The same methodology also partitions trips into homogeneous clusters according to their most likely trigger position. The methodology is applied to two real-world case studies composed of a collection of trips, of which the causes are different, taken from various fleets of industrial GTs. Data collected from twenty sensors during three days of operation before trip occurrence are analyzed. For each trigger scenario, this paper investigates different lengths of the training and testing time window (namely “trigger time window”), by considering up to 24, 18, 12 or 6 hours before and after the considered trigger position. The results demonstrate that longer time windows allow an improvement of the predictive capability.

Influence of the trigger time window on the detection of gas turbine trip

Losi Enzo
Primo
;
Venturini Mauro
Secondo
;
Manservigi Lucrezia
Penultimo
;
2022

Abstract

Gas turbine (GT) trip is one of the most disrupting events that affect GT operation, since its occurrence causes a reduction of equipment remaining useful life as well as revenue loss because of business interruption. Thus, early detection of incipient symptoms of GT trip is crucial to ensure efficient operation and lower operation and maintenance costs. This paper applies a data-driven methodology that employs a Long Short-Term Memory (LSTM) neural network and a clustering technique to identify the time point at which trip symptoms are triggered. The same methodology also partitions trips into homogeneous clusters according to their most likely trigger position. The methodology is applied to two real-world case studies composed of a collection of trips, of which the causes are different, taken from various fleets of industrial GTs. Data collected from twenty sensors during three days of operation before trip occurrence are analyzed. For each trigger scenario, this paper investigates different lengths of the training and testing time window (namely “trigger time window”), by considering up to 24, 18, 12 or 6 hours before and after the considered trigger position. The results demonstrate that longer time windows allow an improvement of the predictive capability.
2022
% reductions; Data driven; Maintenance cost; Neural-networks; Operations and maintenance; Operations cost; Remaining useful lives; Time windows; Turbine operation; Turbine trip
File in questo prodotto:
File Dimensione Formato  
paper_ATI_trip_conferenza.pdf

accesso aperto

Descrizione: Full text editoriale
Tipologia: Full text (versione editoriale)
Licenza: Creative commons
Dimensione 1.45 MB
Formato Adobe PDF
1.45 MB Adobe PDF Visualizza/Apri

I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11392/2501700
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact