Research On Ant Optimization-Based Neural Network Intelligent Pid Control Strategy

PID controller is widely used in various industrial control circumstance for its simple structure,easy to implement,good control effect and strong robustness features.However, the traditional PID controller, because its parameters once set will not be changed,especially when it comes to the system with characteristics of nonlinear complex,parameters time changing,uncertainties ,etc, it is often difficult to obtain satisfied control performance. Luckily, with the development of intelligent control theory, conventional PID controller combined with intelligent control technology forming variety of intelligent PID controllers , strengthening their own merits and complementing each other , provides a new way to solve the complex,dynamic,uncertain systems. It's true of its characteristics: not relying on accurate mathematical model of the system and better adaptability to changes , making intelligent PID controllers become effective,practicable and posses very broad prospects.As the most popular training algorithm for forward neural network learning, BP learning algorithm modify BP neural network'additional coefficient according to steepest descent method based on non-linear programming methods, making learning error reached the minimum with the least-squares sense. But falling into local minimum,the sensitivity of the initial value of additional coefficient and slow convergence are its serious drawbacks, that is the accuracy and good real-time charactertistic of algorithm itself reminning to solve. The key of BP neural network PID parameter tuning is that while the traditional PID based on BP neural network , it is inevitable that there are some common shortcomings of BP neural networks. Thus, when the network structure, learning rate factor, the activation function are defined reasonablely , the research of the randomness selection of initial weights and non-precision of the object'Jacobian information ( the sensitivity between system output and control input) has a very important significance. Therefore, a new intelligent PID control strategy based on Ant Colony Optimization(ACO) off-line global optimizating initial weights of BP neural network,BP neural network self-learning on-line and RBF neural network dentificating Jacobian information in real-time is proposed in this paper, and solved the systemic defects immersed in a local minimum and accelerate the network's convergence during the training of BP neural network effectively.This method can automatically adjust PID parameters to obtain satisfied performance, and was applied in the position server electro-hydraulic material testing system. Simulation analysis based on MATLAB and real-time experimental platform built based on LabVIEW,The simulation and experimental result show that the Intelligent PID controller possessed better dynamic response characteristics and robust performance than traditional ones.