单神经元PID控制算法matlab文件

clear; x = [0,0,0]'; %初始化输入向量 kP = 0.40; kI = 0.35; kD = 0.4; %初始化PID的三个参数的学习速率 last_w_x1 = 0.1; last_w_x2 = 0.1; last_w_x3 = 0.1; %初始化输入网络的三个权重系数 error_1 = 0; %记录上一次误差以及上上次误差 y_1 = 0; y_2 = 0; y_3 = 0; %被控量的离散序列，用于计算被控对象方程以及计算误差 u_1 = 0; u_2 = 0; u_3 = 0; %控制输出的离散序列，用于计算被控对象方程以及更新神经元权重系数。 ts = 0.001; %设置迭代周期为1ms learning_mode = 4; %设置学习规则 error_integral = 0; %积分记录 for k = 1:1:1000 time(k)=k*ts; %横轴变化 yd(k) = sin(50*2*pi*k*ts);%0.5*sign(sin(2*2*pi*k*ts)); %输入期望的信号函数-阶跃函数 y(k)=0.368*sin(y_1) + 0.26*y_2 + 4*sin(u_1) + 0.632*u_2; %被控对象的离散方程。 x' = Ax + Bu error(k) = yd(k) - y(k); %计算当前误差 %神经元的输入 if learning_mode == 1 %无监督的HEBBE学习规则 tmp_x1 = last_w_x1 + kP*u_1*x(1); tmp_x2 = last_w_x2 + kI*u_1*x(2); tmp_x3 = last_w_x3 + kD*u_1*x(3); w_x1(k) = exp(tmp_x1)/(exp(tmp_x1)+(1/exp(tmp_x1))); w_x2(k) = exp(tmp_x2)/(exp(tmp_x2)+(1/exp(tmp_x2))); w_x3(k) = exp(tmp_x3)/(exp(tmp_x3)+(1/exp(tmp_x3))); K = 0.06; %无监督的HEBBE学习规则 elseif learning_mode == 2 %有监督的delta学习规则 tmp_x1 = last_w_x1 + kP*error(k)*x(1); tmp_x2 = last_w_x2 + kI*error(k)*x(2); tmp_x3 = last_w_x3 + kD*error(k)*x(3); w_x1(k) = exp(tmp_x1)/(exp(tmp_x1)+(1/exp(tmp_x1))); w_x2(k) = exp(tmp_x2)/(exp(tmp_x2)+(1/exp(tmp_x2))); w_x3(k) = exp(tmp_x3)/(exp(tmp_x3)+(1/exp(tmp_x3))); K = 0.06; %有监督的delta学习规则 elseif learning_mode == 3 %有监督的hebbe学习规则 tmp_x1 = last_w_x1 + kP*error(k)*u_1*x(1); tmp_x2 = last_w_x2 + kI*error(k)*u_2*x(2); tmp_x3 = last_w_x3 + kD*error(k)*u_3*x(3); w_x1(k) = exp(tmp_x1)/(exp(tmp_x1)+(1/exp(tmp_x1))); w_x2(k) = exp(tmp_x2)/(exp(tmp_x2)+(1/exp(tmp_x2))); w_x3(k) = exp(tmp_x3)/(exp(tmp_x3)+(1/exp(tmp_x3))); K = 0.12; %有监督的hebbe学习规则 elseif learning_mode == 4 %改进后的有监督的hebbe学习规则 tmp_x1 = last_w_x1 + kP*error(k)*u_1*(error(k) - error_1); tmp_x2 = last_w_x2 + kI*error(k)*u_2*(error(k) - error_1); tmp_x3 = last_w_x3 + kD*error(k)*u_3*(error(k) - error_1); w_x1(k) = exp(tmp_x1)/(exp(tmp_x1)+(1/exp(tmp_x1))); w_x2(k) = exp(tmp_x2)/(exp(tmp_x2)+(1/exp(tmp_x2))); w_x3(k) = exp(tmp_x3)/(exp(tmp_x3)+(1/exp(tmp_x3))); K = 0.12; %改进后的有监督的hebbe学习规则 end wadd(k)=abs(w_x1(k))+abs(w_x2(k))+abs(w_x3(k)); w11(k) = w_x1(k)/wadd(k); w22(k) = w_x2(k)/wadd(k); w33(k) = w_x3(k)/wadd(k); w=[w_x1(k),w_x2(k),w_x3(k)]; %权重系数归一化操作 x(1) = error(k); %比例 x(2) = error_integral; %积分 x(3) = error(k)-error_1; %微分 u(k)=K*w*x; %控制律 error_1 = error(k); %更新误差序列 error_integral = error_integral + error(k); %更新误差积分 u_3 = u_2; u_2 = u_1; u_1 = u(k); %更新神经元输出控制律序列 y_3=y_2; y_2=y_1; y_1=y(k); %更新被控对象输出序列 last_w_x1=w_x1(k); last_w_x2=w_x2(k); last_w_x3=w_x3(k); %更新权重系数 end figure(1); plot(time,yd,'r',time,y,'k:','linewidth',2); xlabel('time');ylabel('yd;y'); title('track') %打印被控对象跟踪结果 figure(2); plot(time,u,'r','linewidth',2); xlabel('time');ylabel('controlinput U'); title('control law') %打印控制律曲线 figure(3); plot(time,w11,'r',time,w_x1,'b','linewidth',2); xlabel('time');ylabel('kp'); title('kp line') %打印控制律曲线 figure(4); plot(time,w22,'r',time,w_x2,'b','linewidth',2); xlabel('time');ylabel('ki'); title('ki line') %打印控制律曲线 figure(5); plot(time,w33,'r',time,w_x3,'b','linewidth',2); xlabel('time');ylabel('kd'); title('kd line') %打印控制律曲线