CAJ | 학술논문

本文以研究最基本的神经网络PID控制即单神经元PID控制为主，辅助研究BP神经网络。PID控制要取得好的控制效果，就必须调节好比例、积分和微分参数在控制系统中的比例，他们的数值存在着一种相互制约的关系，然而在工业实际生产过程中，完全的线性时不变系统是不存在的，一般都是非线性时变系统，并且难以建立数学模型，参数整定困难。该文通过MATLAB建立仿真，在其中建立控制对象以及控制器，即需要对比的两种神经网络PID控制，在变化被控对象为一阶、二阶系统进行分别的仿真并分析对比其控制效果；最后发现BP神经网络PID控制算鲁棒性最好。
본문이연구최기본적신경망락PID공제즉단신경원PID공제위주，보조연구BP신경망락。PID공제요취득호적공제효과，취필수조절호비례、적분화미분삼수재공제계통중적비례，타문적수치존재착일충상호제약적관계，연이재공업실제생산과정중，완전적선성시불변계통시불존재적，일반도시비선성시변계통，병차난이건립수학모형，삼수정정곤난。해문통과MATLAB건립방진，재기중건립공제대상이급공제기，즉수요대비적량충신경망락PID공제，재변화피공대상위일계、이계계통진행분별적방진병분석대비기공제효과；최후발현BP신경망락PID공제산로봉성최호。
This paper studies the basic neural network PID control that single neuron PID control-oriented , assisted on BP neural network ., and added to the control condition BP neural network PID control for comparison . PID control To achieve good control effect, it must regulate the proportional, integral and differential ratio control system parameters, and their values there is a rela?tionship of mutual restraint, but in actual industrial production process, completely linear time invariant system does not exist, are generally nonlinear time-varying systems , and it is difficult to establish the mathematical model, parameter tuning difficult. This paper established by MATLAB simulation, in which the establishment of the control object and a controller that need to compare two types of neural network PID control , changes in the controlled object for first, second and simulation systems were analyzed and compared its control effect;and finally found that BP neural network PID control algorithm robustness of the best .