An Improved CFPOS Combined With Neural Networks And Their Applications In Control System

This paper mainly introduces a method combining the neural network with improved particle swarm algorithm. Using this method to design the CSTR PID control and liquid level control system PID control to be compared with the traditional method.At first, aiming at the problem that slow convergence speed and thus fall into the local minima limitations when BP neural network calculating weights, Someone bring particle swarm algorithm in neural network to optimize the network’s weight. But in view of the basic particle swarm optimization has a lot of uncertainty and randomness, it also has many defects,such as easy to premature at the iteration, show convergence speed at the later iteration. Then I think the algorithm of particle swarm should be improved. The particle swarm optimization in these text is based on the algorithm of CFPSO. By introducing two parameters v(velocity factor) and d(location factor) shows the current position and the optimal position of the distance. When speed of the particles close to the optimal value is less than v, set it stagnation.When stagnant situation appearing, initializes particle, enhances the vitality of the particle and the late convergence speed. This method can disperse the particles and make the population diversity. This method gives proof of its convergence. By six examples simulation, this text illustrates the feasibility of the algorithm and its high efficiency effectively.Then, PID control scheme of CSTR based on the improved CFPSO neural network has been proposed. The control system of CSTR is always a nonlinear, time-varying system with lag behind, noise disturbance and the other impacts at the same time. So the classical PID control theory could not have a good effect on CSTR control. The text got the essential reasons of poor precision of control system by analyzing the control properties of CSTR systemically. Based on this, this article proposed CSTR control system design schemes based on the improved CFPSO neural network. Using the improved CFPSO to instead of the traditional gradient descent method can optimize the weights in throughout the control process. The properties of approximation and adaptive of BP neural network can greatly improve the control effect compared with traditional PID. Through the example simulation to contrast control effect. The result shows that, under the transient response of the same time,control effect improved CFPSO neural network controller comparing with the BP neural network controller, the overshoot of the former quantity is smaller, and has a smaller output.At last, using improved CFPSO optimization neural network PID control to design the string loop liquid control system. The string loop liquid control system is a typical system in industry process control. This paper has built systemic mathematical model based on analysis of the structure and control principle of string loop liquid control system. It also provides the block diagram of the improved particle swarm neural network PID string loop liquid control system. It shows that the overall processes and calculation steps. Through the example simulation, it has proved that this improved particle swarm neural network PID control hasrobustness and instantaneity, and its characteristics are very suitable for nonlinear level cascade control system.