The Cloud Model Theory of Intelligent Control Method for Non-Minimum-Phase and Non-Self-Balancing System in Nuclear Power

Abstract

Steam generator (SG) is one of the key equipment of nuclear power units. Because of the large range of its loads changing, the water level control of SG effectively is an essential secure guarantee of nuclear power plants. SG is a complex system, besides imbalance and non-minimum phase characteristic, it also has the properties of nonlinearity, time-varying and with small stability margin. There are many difficulties in water level control of SG. Of which false water level and varying parameters are the most severe problems. In this paper, first the water level features and the water level control principle of U-tube steam generator (UTSG) are introduced. Then mathematical model mechanism and both the static and dynamic characteristic of it water level are discussed. Finally various control methods are used for comparing the control effect. Intelligent control is a type of control strategy which imitates human intelligence behavior. It is mainly aimed at the controlled plant with complicate model parameters, or which model structure hard to describe accurately by mathematical method. Cloud Model theory is proposed by Academician Li Deyi based on the idea of artificial intelligence with uncertainty. This theory focus on analyzing the uncertainty of control plant, realizes the uncertain conversion between qualitative concept and quantitative numerical by combining ambiguity and randomness. In the field of control technology, ambiguity and randomness make it difficult to establishing precise mathematical model of control plant, and become a bottleneck during the research of improving stability, accuracy and quickness of control system. In this context, Cloud Model can be a good conversion between qualitative concept and quantitative numerical due to its ability of showing the uncertainty of qualitative concept which described by natural language. Under the action of external input, system control can be realized by inferencing according to the qualitative concept and uncertainty rules of Cloud Model. In this paper, the researched Cloud Model control system is based on normal distribution, because a large number of random events in nature and society obey or approximately obey normal distribution. The rate of convergence of Cloud Model control is evidently faster than PID. Moreover, the capability of Cloud Model control in tracking, adapting, anti-interference and overcoming large time lag are apparently superior when comparing with the control effect of PID.