This thesis represents the first step of a larger work finalized with the development of an advance multivariable control system for compressors in the oil & gas industry. The project has been developed in the ABB corporate research center of Baden-Dättwil (CH) for the ABB Group. The developed controller should be able to control both the load of the compressor and to avoid a dangerous unstable operating condition, called surge. The technology chosen to develop the control system is Nonlinear Model Predictive Control (NMPC): Model Predictive Control is the most successful advanced control system used in the industry, due to its multivariable nature, to the opportunity of including constraints on variables and the possibility to incorporate process models. The use of a nonlinear control system is related to the need of accurate predictions of a strongly nonlinear process like gas compressions and in order to efficiently forecast the onset of surge. The control of compressors relies of the usage of valves with the purpose of changing the system discharge pressure and flow: however this approach is extremely inefficient and wastes a significant amount of energy. A new trend is using also compressors moved by variable speed drives so as to change the compressor operating point by changing its speed: this approach reduces the need to utilize valves and produces a relevant energy saving and lowers the exhaust emissions that would otherwise be produced by a combustion turbine. The first step of the thesis work has focused on the analysis of compressors and compression systems, by analyzing the literature and the most successful developed models in the last fifty years and by investigating on the most common plant layouts used in the oil and gas industry. Next, a five states model suitable for control purposes and describing all the relevant components of a compression plant is derived: the models inputs are the opening setpoint for the valves (suction valve, discharge valve and a recycle valve) and the speed setpoint for the compressor’s variable speed driver; the model’s outputs are the pressures, the flows and the temperatures typically measured in real plants, plus the compressor current speed. After the derivation of the mathematical model, a simulator has been developed in the Matlab-Simulink software environment: in order to verify the quality of the developed model, it has been validated using data from an experimental setup. The next step of the work has been the study of the current state of art in compressor’s control and compression processes control: load control and anti-surge systems have been investigated and a decentralized PI-based control system has been designed in order to control the five states model: this control system take full advantage of variable speed drives by adjusting the compressor speed setpoint so as to adjust the plant discharge flow and/or pressure. Finally, after an overview on MPC and Nonlinear MPC, a NMPC controller has been designed and implemented using an ABB proprietary toolbox, called NOMOT, also running on the Matlab-Simulink environment. The toolbox allows the design of complex objectives, it can use variable constraints and performs a nonlinear optimization at every control-step. Then the designed controller has being tested to evaluate its disturbances rejection. The designed decentralized and the NMPC controller’s performances have been tested and compared by simulating several scenarios. The simulations have shown potential benefits of the NMPC controller in terms of energy saving, up to about 40%, plant throughput flow, up to about 30%, and response time, up to 75%, when compared with the PI-based controller. These results are mostly related to the use of a multivariable centralized controller and to the effective use of the compressor speed as main control variable. The last part of the thesis discusses the advantages of this control strategy and the downside, the computational time needed to perform the calculations for one control step.
Questa tesi rappresenta il primo passo di un progetto più grande, finalizzato allo sviluppo di un sistema di controllo avanzato e multivariabile per compressori nell’industria dell’oil & gas. Il lavoro di tesi è stato svolto presso il centro di ricerca di ABB a Baden-Dättwil (CH) ed è stato finanziato dal gruppo ABB. Il sistema di controllo deve essere in grado di gestire il carico del compressore ed evitare una pericolosa condizione operativa, chiamata pompaggio o surge. La tecnologia scelta per sviluppare il sistema di controllo è il Nonlinear Model Predictive Control (NMPC): il Model Predictive Control (MPC) è la tipologia di controlli avanzati di maggiore successo nel campo industriale, a causa dell’intrinseca natura multivariabile, dell’opportunità di includere vincoli sulle variabili e della possibilità di incorporare nel controllore un modello del processo controllato. La scelta di un sistema di controllo nonlineare è legata alla necessità di predizioni accurate per un processo molto nonlineare come la compressione di gas e per prevedere efficacemente l’insorgere del pompaggio. Il controllo dei compressori si basa sull’uso di valvole al fine di variare la pressione e il flusso in uscita dall’impianto: tuttavia questo approccio risulta estremamente inefficiente dissipando un ingente quantitativo di energia. Recentemente si sta diffondendo l’uso di compressori mossi da azionamenti a velocità variabile al fine di mutare la condizione operativa del compressore variandone la velocità di rotazione: questo approccio riduce l’uso delle valvole producendo un notevole risparmio energetico e riducendo le emissioni di gas di scarico, altrimenti prodotto da turbine a gas. Il primo passo del lavoro di tesi è stato lo studio dei compressori e dei sistemi di compressione in generale, attraverso l’analisi della letteratura e dei modelli di maggior successo negli ultimi cinquanta anni e investigando le strutture più comuni negli impianti usati nell’industria dell’oil & gas. In seguito, si è ricavato un modello a cinque stati compatibile per un controllore in real-time e rappresentante tutti i principali componenti dei sistemi di compressione: gli ingressi del modello sono i riferimenti per l’apertura delle valvole (valvola di aspirazione, valvola di scarico e una valvola di ricircolo) e il riferimento di velocità per l’azionamento a velocità variabile; le variabili di uscita del modello sono le pressioni, i flussi e le temperature tipicamente misurati negli impianti reali, con in più la velocità del compressore. Quando si è terminato lo sviluppo del modello matematico, si è proceduto sviluppando un simulatore nell’ambiente Matlab-Simulink: per verificare la qualità del modello sviluppato, questo è stato validato usando dati provenienti da un apparato sperimentale. Il passo successivo del lavoro è stato lo studio dell’attuale stato dell’arte nel controllo dei compressori e dei processi di compressione: sono state investigate le strategie di controllo del carico e i sistemi anti-surge, portando allo sviluppo di un controllore decentralizzato basato su PI, finalizzato al controllo del modello a cinque stati: il controllore sviluppato fa pieno uso dell’azionamento a velocità variabile, correggendo la velocità del compressore per variare la pressione e/o il flusso in uscita dall’impianto. In seguito, dopo una panoramica su MPC e MPC nonlineare, si è un controllore NMPC, realizzato usando la piattaforma proprietaria di ABB, chiamata NOMOT, la quale può funzionare anche all’interno dell’ambiente Matlab-Simulink. La piattaforma permette l’inserimento di obiettivi complessi, l’uso di vincoli sulle variabili e svolge un processo di ottimizzazione nonlineare a ogni istante di controllo. In seguito il controllore realizzato è stato testato per valutarne la reiezione ai disturbi. Le prestazioni del controllore decentralizzato e del controllore NMPC sono state testate e comparate simulando alcuni scenari rilevanti. Le simulazioni hanno mostrato le potenzialità del controllore NMPC rispetto al controllore basato su PI in termini di risparmi energetico, fino al 40%, throughput dell’impianto, fino al 30% in più, e tempo risposta, diminuito fino al 75%. L’ultima parte della tesi è stata la discussione dei vantaggi del controllo NMPC e del principale svantaggio, ovvero l’eccessivo tempo richiesto per svolgere le computazioni per un singolo istante di controllo.
Nonlinear model predictive control of compressors with variable speed drives in the oil and gas industry
PARESCHI, DIEGO
2009/2010
Abstract
This thesis represents the first step of a larger work finalized with the development of an advance multivariable control system for compressors in the oil & gas industry. The project has been developed in the ABB corporate research center of Baden-Dättwil (CH) for the ABB Group. The developed controller should be able to control both the load of the compressor and to avoid a dangerous unstable operating condition, called surge. The technology chosen to develop the control system is Nonlinear Model Predictive Control (NMPC): Model Predictive Control is the most successful advanced control system used in the industry, due to its multivariable nature, to the opportunity of including constraints on variables and the possibility to incorporate process models. The use of a nonlinear control system is related to the need of accurate predictions of a strongly nonlinear process like gas compressions and in order to efficiently forecast the onset of surge. The control of compressors relies of the usage of valves with the purpose of changing the system discharge pressure and flow: however this approach is extremely inefficient and wastes a significant amount of energy. A new trend is using also compressors moved by variable speed drives so as to change the compressor operating point by changing its speed: this approach reduces the need to utilize valves and produces a relevant energy saving and lowers the exhaust emissions that would otherwise be produced by a combustion turbine. The first step of the thesis work has focused on the analysis of compressors and compression systems, by analyzing the literature and the most successful developed models in the last fifty years and by investigating on the most common plant layouts used in the oil and gas industry. Next, a five states model suitable for control purposes and describing all the relevant components of a compression plant is derived: the models inputs are the opening setpoint for the valves (suction valve, discharge valve and a recycle valve) and the speed setpoint for the compressor’s variable speed driver; the model’s outputs are the pressures, the flows and the temperatures typically measured in real plants, plus the compressor current speed. After the derivation of the mathematical model, a simulator has been developed in the Matlab-Simulink software environment: in order to verify the quality of the developed model, it has been validated using data from an experimental setup. The next step of the work has been the study of the current state of art in compressor’s control and compression processes control: load control and anti-surge systems have been investigated and a decentralized PI-based control system has been designed in order to control the five states model: this control system take full advantage of variable speed drives by adjusting the compressor speed setpoint so as to adjust the plant discharge flow and/or pressure. Finally, after an overview on MPC and Nonlinear MPC, a NMPC controller has been designed and implemented using an ABB proprietary toolbox, called NOMOT, also running on the Matlab-Simulink environment. The toolbox allows the design of complex objectives, it can use variable constraints and performs a nonlinear optimization at every control-step. Then the designed controller has being tested to evaluate its disturbances rejection. The designed decentralized and the NMPC controller’s performances have been tested and compared by simulating several scenarios. The simulations have shown potential benefits of the NMPC controller in terms of energy saving, up to about 40%, plant throughput flow, up to about 30%, and response time, up to 75%, when compared with the PI-based controller. These results are mostly related to the use of a multivariable centralized controller and to the effective use of the compressor speed as main control variable. The last part of the thesis discusses the advantages of this control strategy and the downside, the computational time needed to perform the calculations for one control step.| File | Dimensione | Formato | |
|---|---|---|---|
|
2011_03_PARESCHI.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
18.46 MB
Formato
Adobe PDF
|
18.46 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/17321