Study On The Parameters Optimized PID Control Based On Neural Network

The traditional PID control which based on mathematic model was steady and reliable, and was applied in many fields. However, with the development of technology, the object becomes more and more complex. So, the conditional PID controller will not get ideal control effect for some object which can not be described by accurate mathematic model. And neural network afford new method to solve the problem.Neural network can learn and adopt dynamics of uncertain system, so it has strongly robustness, and can deal with the questions that can not be described by model and rule, and has been successful in many control of uncertain system. The work of this paper mainly consists of two parts. At first, for the shortcomings of BP algorithm, LM algorithm optimization is used to training neural network, at the same time, in allusion to the two problem of the learning rate and the choice of the inverse matrix solving of LM algorithm that seriously influence the training time and the accuracy of convergence, the paper proposed three ways to improve the LM algorithm, and present Programming by MATLAB. In the next place, the paper uses the improved BP network algorithm into parameter optimization of PID controller, and show the programming by MATLAB.At last, based on the temperature control model of heating furnace, the paper uses MATLAB 7.0 to carry out PID control emulation experiment, and does contrast research of the algorithms' control performance. The results indicates obviously that the BP neural network PID controller optimized by improved LM algorithm is better than other algorithms in training velocity and constringency precision, which confirm the availability and practicability of new algorithm, and achieve the expectant intention of the paper.