Towards improved automation for desalination processes, Part II: Intelligent control

Abstract

The frontiers of automatic control are expanding to keep abreast of the state-of-the-art technology. Innovative concepts in control systems have relegated what was considered a satisfactory solution a decade ago to obsolescence today. One novel feature is artificial intelligence (AI) which has ability to cope with uncertainties that we encounter in today's complex process control. Intelligent control systems are becoming almost commonplace worldwide incorporating special features such as expert systems, self-tuning, self-diagnosis, life management, equipment health monitoring, modelling and simulation. Thus, a trend towards larger desalination plants necessitates more sophisticated automation technology. In view of the benefits that can be realized from the application of artificial intelligence in the control field of desalination processes, intelligent control is the application of AI in control, employing methods like expert systems, neural networks, fuzzy logic and pattern recognition. Thereby, learning and decision making are the most important features of intelligent control. The application of intelligent control is necessary due to computational complexity, nonlinear behaviour with many degrees of freedom, and the presence of uncertainty in control environment. Intelligent control can be used at different levels of control systems. On the highest level, monitoring of data, selection of algorithms, selection of objective functions, and assigning of values for the setpoints are the main aims. On the level of advanced regulatory control, process identification can be supported by intelligent control, e.g. selecting of structures by expert systems, implementing neural networks for parameter estimation etc. Expert systems, assisted by pattern recognition, can be used at this level for tuning of adaptive controllers. Adaptation of controller parameters according to an intelligent strategy can also be provided by neural networks. Also, expert systems can be used to assist the design of advanced control systems. In the large field of design of conventional control systems, artificial intelligence methods can be used to provide intelligent user interface, to select the best algorithms, to consider design heuristics for optimization of feedback control loops etc. The varying and uncertain conditions in a multistage flash (MSF) desalination plant during its operation in the annual cycle of seasons under different situations of loads and disturbances, point out that no single fixed strategy, no matter how advanced it may be, is likely to be either valid or possible at all times under all conditions in all the sub-systems related to the plant. This paper is addressed to some aspects of artificial intelligence (AI) which aims at equipping the process with the capability for continual analysis and assessment of the situations that arise during the course of operation of the plant leading to the choice and implementation of an appropriate course of action in process control. At the outset MSF plant conditions calling for consideration of advanced control strategies with AI support are highlighted. Then some aspects of AI applicable to plant control are reviewed, finally, suggestions towards intelligent control of MSF plants are made pointing also to the possibility of an integrated system for Process Control and Care.