Fault detection and diagnosis in fiscal metering stations using … · 2019-05-27 · Fiscal metering • The longer the fault takes to be detected, the higher the probabilities of

Fault detection and diagnosis infiscalmetering stationsusing canonicalvariate analysis

AfrânioMelo1,Nayher A.Clavijo Vallejo1,MaurícioM.Câmara1,ThiagoFeital1,ThiagoK.Anzai2,PedroH.Thompson2,FábioC.Diehl2 and JoséC.Pinto1

1- ProgramadeEngenhariaQuímica/COPPE– UFRJ- UniversidadeFederaldoRiodeJaneiro.2- CentrodePesquisasLeopoldoAméricoMiguez deMello- CENPES,Petrobras- PetróleoBrasileiroS.A.

CidadeUniversitária,RiodeJaneiro,[email protected]

Outline

• The context: fiscal metering;

• The system:AMP, a sytem for process monitoring;

• The application: the metering station;

• Results;

• Conclusions.

ProgramadeEngenhariaQuímica,COPPE/UFRJCENPES,Petrobras

ThecontextFiscalmetering


Fiscalmetering

• Fiscal meters are one of the most crucial equipments of a process plant.

• Values related to royalties, allocation and custody transfer depends on them.

• Small errors in measurement can lead to large losses in revenue.

• Thus, accuracy plays a very important role in their operation!


Fiscalmetering

• Flow measurements are constantly supervised by governmental regulatoryagencies.

• The agencies establish protocols that must be followed in order to notifyabnormalities.


Fiscalmetering

• Below we have the protocol imposed by ANP (Agência Nacional do Petróleo,Gás Natural e Biocombustíveis):

• once a fault is identified, an official notification must be sent to the regulatory agencywithin 72 hours;

• the problemmust be fixed in a 240 hours limit;

• after this limit, the fiscal meter could be put out of operation - which would lead to a greatimpact in terms of production.


Fiscalmetering


Measure estimation methodology applied by ANP to the measurements of a faulty sensor.

Fiscalmetering

• The longer the fault takes to be detected, the higher the probabilities ofroyalties costs being overestimated!

• Moreover, there are other economic impacts of faults for the company, suchas fines.

• This justifies the development of a system for process monitoring of fiscalmeters, as presented in this work.


ThesystemAMP,asystemforprocess monitoring


AMP:PredictiveMaintenance Application

• The AMP system is a web application operating on the AMP server andconsists of 5 components:

• client system,• data acquisition unit,• communication unit,• engine system,• database.


AMP:PredictiveMaintenance Application

• We applied Canonical Variate Analysis (CVA), a technique based on latentvariable modeling and dimensionality reduction, to model the statistics ofprocess data:


TheapplicationThemetering station


Themetering station

• The metering station (EMED) is located at an onshore field of Petrobras.

• The EMED receives gas and oil from four satellite stations distributed alongthe field.

• The meters monitored are two orifice plate devices as gas meters and onecoriolis device as oil meter.


Oil and gas metering station PFD


Results


Results


SPE values and gas fiscal temperature contribution during 8months.

Results


SPE values and gas fiscal temperature contribution during 1month.

Results


Three scenarios for the calculation of the royalty-bearings. The black point is the maximum daily mean of production in each hatched period.

Conclusions


Conclusions

• Test simulations of the AMP system on real data were able to detect an earlyfailure at a metering station days before the operation crew.

• It could also successfully identify, among hundreds of variables, the rootcause of the abnormal situation with fast response.

• Thus, we conclude that the application of AMP monitoring system can leadto large long-term economic gains.


Thank you!Obrigado!


Fault detection and diagnosis in fiscal metering stations using … · 2019-05-27 · Fiscal metering • The longer the fault takes to be detected, the higher the probabilities of

Documents