Centrifugal compressors are widely used for the pressurization of fluids and they have a fundamental role in many industrial processes, especially in the chemical, oil and gas industries. A significant effort in study how to control these machines has been done from the first years of the twentieth century, stimulated from the fact they have a limited operational range and they are very sensitive to changes in the operating conditions. With the advent of the Variable-Speed Drivers (VSD) it was possible to develop new advanced controllers that use the electric motor angular speed as control variable. However this new possibility leads to a multivariable control problem that, if treated with standard methods based on the use of PID controllers, may bring to inefficient operation, system instability and difficulties in parameter tuning, especially in complex plants. In order to overcome these issues, an approach based on Model Predictive Control (MPC) can be chosen. Indeed MPC fits better with multivariable constrained problems and it allows to simultaneous control of the process variables accordingly to a cost function. On the other hand, this strategy requires a significant computing power and has to deal with the stringent real-time requirement due to compressors dynamics. This has led up to nowadays using MPC mostly as a supervisor of standard controllers, as PIDs. Recent years progresses in modern electronics have contributed to a development of faster and cheaper Programmable Logic Controllers (PLC) increasing the interest in MPC as embedded systems. The aim of this thesis is to develop a Real Time Model Predictive Control (RT- MPC) strategy for the stabilization of the outlet pressure of a compression system, implement it on High Performance ABB AC500 PLC and test it on the Compressor Rig Test Stand situated at AGH University in Krakow, Poland. The first part of the work regards the control of a single compressor. Initially the controller design is performed in a simulation environment and later the MPC is tested on the experimental rig. Excellent results are obtained despite measurement disturbances and model discrepancies, also when the compressor operates far away from the nominal condition. Later, two centrifugal compressors working in parallel are analysed, taking into account a load sharing strategy. Although satisfying simulation results are achieved, the increasing of problem complexity leads to face with PLC computing limitations.
I compressori centrifughi sono ampiamente utilizzati per la pressurizzazione dei fluidi e svolgono un ruolo fondamentale in molti processi industriali, in particolare nelle industrie chimiche, petrolifere e del gas. Uno sforzo significativo nello studio di come controllare queste macchine è stato fatto fin dai primi anni del ventesimo secolo, ricercando soluzioni alla loro sensibilità alle condizioni operative e al loro range operativo limitato. Con l'avvento dei Variable Speed Drive (VSD) è stato possibile sviluppare nuove strategie di controllo avanzate che utilizzassero la velocità angolare del motore elettrico come variabile di controllo. Tuttavia questa nuova possibilità porta a un problema di controllo multivariabile che, se trattato con metodi standard basati sull'uso di controller PID, può portare a difficoltà nella sintonizzazione dei parametri ed ad inefficienza ed instabilità del sistema, specialmente in impianti complessi. Per superare questi problemi, è possibile scegliere un approccio basato sul Model Predictive Control (MPC). Infatti tale strategia di controllo si adatta meglio ai problemi multivariabili vincolati e consente il controllo simultaneo delle variabili di processo in base a una funzione di costo. Tuttavia questa strategia richiede una potenza di calcolo significativa e deve far fronte a stringenti requisiti in termini di tempo di calcolo a causa della veloce dinamica dei compressori. Tutto ciò ha portato fino ad oggi ad usare il MPC principalmente come supervisore di controllori standard, come PID. Negli ultimi anni i progressi nell'elettronica hanno portato allo sviluppo di Programmable Logic Controller (PLC) più veloci ed economici che hanno aumentato l'interesse per lo sviluppo del MPC in sistemi integrati. Lo scopo di questa tesi è quello di sviluppare una strategia di Real Time Model Predictive Control (RT-MPC) per la stabilizzazione della pressione di un sistema di compressione, implementarla sul PLC ABB AC500 e testarla sul sistema di compressione costruito dall’ABB e situato presso l'Università AGH di Cracovia, in Polonia. La prima parte del lavoro riguarda il controllo di un singolo compressore, spaziando dalla progettazione iniziale del MPC in ambiente di simulazione fino ai test sull'impianto di compressione. I risultati ottenuti sono eccellenti anche quando il compressore funziona lontano dalle condizioni nominali, nonostante disturbi di misurazione e imprecisioni del modello. Successivamente, vengono analizzati due compressori centrifughi che lavorano in parallelo, tenendo conto di una strategia di condivisione del carico. Sebbene siano stati raggiunti risultati di simulazione soddisfacenti, l'aumento della complessità del problema porta a scontrarsi con i limiti del PLC in termini ti potenza di calcolo e tempo di esecuzione.
Control of centrifugal compressors via embedded real time MPC
FALASCHI, COSIMO
2016/2017
Abstract
Centrifugal compressors are widely used for the pressurization of fluids and they have a fundamental role in many industrial processes, especially in the chemical, oil and gas industries. A significant effort in study how to control these machines has been done from the first years of the twentieth century, stimulated from the fact they have a limited operational range and they are very sensitive to changes in the operating conditions. With the advent of the Variable-Speed Drivers (VSD) it was possible to develop new advanced controllers that use the electric motor angular speed as control variable. However this new possibility leads to a multivariable control problem that, if treated with standard methods based on the use of PID controllers, may bring to inefficient operation, system instability and difficulties in parameter tuning, especially in complex plants. In order to overcome these issues, an approach based on Model Predictive Control (MPC) can be chosen. Indeed MPC fits better with multivariable constrained problems and it allows to simultaneous control of the process variables accordingly to a cost function. On the other hand, this strategy requires a significant computing power and has to deal with the stringent real-time requirement due to compressors dynamics. This has led up to nowadays using MPC mostly as a supervisor of standard controllers, as PIDs. Recent years progresses in modern electronics have contributed to a development of faster and cheaper Programmable Logic Controllers (PLC) increasing the interest in MPC as embedded systems. The aim of this thesis is to develop a Real Time Model Predictive Control (RT- MPC) strategy for the stabilization of the outlet pressure of a compression system, implement it on High Performance ABB AC500 PLC and test it on the Compressor Rig Test Stand situated at AGH University in Krakow, Poland. The first part of the work regards the control of a single compressor. Initially the controller design is performed in a simulation environment and later the MPC is tested on the experimental rig. Excellent results are obtained despite measurement disturbances and model discrepancies, also when the compressor operates far away from the nominal condition. Later, two centrifugal compressors working in parallel are analysed, taking into account a load sharing strategy. Although satisfying simulation results are achieved, the increasing of problem complexity leads to face with PLC computing limitations.File | Dimensione | Formato | |
---|---|---|---|
2018_4_Falaschi.pdf
non accessibile
Descrizione: Thesis Text
Dimensione
95.35 MB
Formato
Adobe PDF
|
95.35 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/10589/141182