Open Access System for Information Sharing

Login Library


Cited 0 time in webofscience Cited 0 time in scopus
Metadata Downloads

New Methods of Nonparametric Identification and PID Controller Design and their Applications to a Molten Carbonate Fuel Cell System

New Methods of Nonparametric Identification and PID Controller Design and their Applications to a Molten Carbonate Fuel Cell System
Date Issued
In this research, improved methods in Proportional-Integral-Derivative (PID) auto-tuning are proposed: frequency response identification, model reduction and PID tuning. In addition, the proposed auto-tuning method is applied to a Molten Carbonate Fuel Cell (MCFC) system to test the methods. The Chapter 2, new nonparametric process identification methods for continuous-time processes and discrete-time are proposed. The two methods provide the frequency response model from given process input and output data of continuous-time and discrete-time processes, respectively. The proposed algorithms can estimate exact models for all desired frequencies. They are applicable to various process conditions (both initial/final steady state, initial steady state/final cyclic steady state, initial cyclic steady state/final steady state and both initial/final cyclic steady state) and require a smaller amount of memory than previous methods. Also, they provide exact models even in the presence of a static disturbance and show an acceptable robustness to measurement noises. In Chapter 3, a new model reduction method and an analytic PID tuning rule for PID auto-tuning are proposed. They are based on a fractional order plus time delay model to handle an extensive class of linear self-regulating processes. The model reduction method fits frequency response data to the fractional order model by solving a simple single-variable optimization problem. In addition, the PID tuning rule for the fractional order plus time delay model is developed by fitting the optimal tuning parameters. The optimal tuning parameters are obtained by the Integral of the Time weighted Absolute Error (ITAE) minimization and then the proposed PID tuning rule is developed. It provides almost the same performance as the optimal tuning parameters. Simulation study confirms that the auto-tuning strategy based on the proposed model reduction method and the PID tuning rule is superior to the existing approaches based on the integer order models. In Chapter 4, the proposed methods of Chapter 2 and 3 are applied to a MCFC system. The investigated MCFC system contains complicated configuration and interactive subsystems therefore it has nonlinearity and highly complex dynamics. The proposed control strategy is the PID auto-tuning methodology in Chapter 2 and 3 combined with an operational optimization and gain scheduling technique. The operational optimization problem is formulated by a Radial Basis Functional (RBF) neural network model. The neural network model is produced by a database of the integrated MCFC system model. The optimization results are utilized in the auto-tuning procedure as well as a static feedforward look-up table controller. For the gain scheduling, the frequency response identification is implemented repeatedly for several operating points and then appropriate PID controllers are obtained for each operating region. The proposed strategy is aimed at designing a control system to follow power demand changes. The results in the simulation study confirm that the proposed control strategy provides good control performance for the successive power demand changes and successfully compensates the nonlinearity of the system.
Article Type
Files in This Item:
There are no files associated with this item.


  • mendeley

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Views & Downloads