Fuzzy model reference learning control of multi-stage flash desalination plants

• Published in: Desalination 1998-09, Vol.116 (2), p.157-164
• Main Author: Ismail, Abdulla
• Format: Article
• Language: English
• Subjects: Applied sciences; Drinking water and swimming-pool water. Desalination; Exact sciences and technology; Fuzzy control; Learning systems; MSF desalination; Pollution; Water treatment and pollution
• ISSN: 0011-9164; 1873-4464
• DOI: 10.1016/S0011-9164(98)00192-1

The large scale process control of highly complex desalination plants has been largely dealt with conventional proportional-integral-derivative (PID) controllers. Although these conventional techniques may provide a minimum performance requirement, they fall short of the increasing control performance demand of robustness, optimality and adaptation to external disturbances. However, in the last few years, new emerging intelligent control techniques have been gaining acceptance for their attractive design and implementation advantages. These new control methods provides solutions for problems where no mathematical model of the system may exit and where uncertainties in the operating environment are significant. Among the widely spread desalination plants in need of efficient and reliable control mechanism are the distillation-based multi-stage flash (MSF). These complex non-linear systems with inter-coupled control loops have not been studied satisfactorily for their efficient performance during different operating conditions and under changing loads. While fuzzy control can provide effective practical solutions to complex industrial problems, as an alternative to conventional control methods, there are several drawbacks that may limit its use for some problems. One main drawback being the inability of the fuzzy controller, designed for the nominal plant, to perform adequately if significant and unpredictable plant variations may occur. Introducing the capability of a progressively learning mechanism into the system, along with basic ideas of fuzzy sets and control theory, will improve the performance of the overall controlled system when interacting with the environment. The fuzzy reference learning controller (FMRLC) utilises a learning mechanism which observes the plant outputs and adjusts accordingly the rules in a direct fuzzy controller such that the overall system performs satisfactorily. In this paper we would discuss the advantages of using FMRLC in controlling the top brine temperature (TBT) of the brine heater in an 18th stage MSF desalination plant. Comparisons with classical as well as direct fuzzy control of the same plant are investigated. Furthermore, some practical implementation issues of the proposed controller are discussed.