PID implementation using code

I have implemented a PID controller using simulink pid block for a system. I was trying to implement the same thing using a PID code (formula) in a matlab function block and compare the performances. However, I am getting differences in both the cases. Unable to figure out where exactly is the problem.

For simulink block I have implemented antiwindup, similarly I have done it in the code.

PID code as in the matlab function is as below:

function y = fcn(error,kp,ki)

dt = sampletime;

persistent integral newintegral

if isempty(integral)

integral = 0;

end

if isempty(newintegral)

newintegral = 0;

end

newintegral = integral + (error*dt);

output = kp*error + ki * integral;

if output>48000

output=48000;

elseif output<3000

output =3000;

else

integral = newintegral;

end

y = output;

Is there anything that I am implementing wrong in the code? Requesting some help/direction.

Thank you

Answer 1

[Sam Chak]

Hi @Sanjeet Kulkarni

Wouldn't it be simpler if you obtain the signal directly from the Integrator block, and then feed it into the MATLAB function block? Something like this

function y = fcn(err, errint, Kp, Ki)

where err is the error signal and errint is the intergral of error signal.

[Sanjeet Kulkarni]

Feeding it to the block can be done, but, I want to implement complete PID with anti-windup in a matlab function. Hence not using the blocks.

[Sam Chak]

Two questions.

Does your Anti-Windup involve some sort of integration but not the Integral action of the PID controller?

Is "integral" as a special variable or the built-in MATLAB integral function?

[Sanjeet Kulkarni]

I have used the clamp method of antiwindup, where in the integral action should be clamped to the maximum and mininum limits of the actuator.

Have declared integral as a persistent variable (not the built in function).