PID Tuning peltier

[JasperBFH]

Good afternoon,

I came across your github and I see that you are very experienced in realizing PID controllers. I am working on creating a PID controller for controlling the water temperature using a peltier, H-bridge and PID. The PID for heating has been tuned using your document PID_Controller_Calculus. I have also tried this with cooling, but this does not work at all. Can you help me to tune the PID for cooling.

[Emile666]

You closed this, so you probably figured it out already. A normal STC1000 works as a thermostat, meaning that heating (and cooling) is switched off if the temperature is reached. I added a PID controller for heating to this, since a heating element is typically controlled by a SSR, which is controlled by a PWM signal. Which is a perfect for a PID controller.

Cooling (compressor in a fridge) on the other hand cannot be controlled by a PID controller. So there you have the standard thermostat algorithm.

Hope this helps understanding the function of the PID controller I implemented.

[lalo-uy]

For the cooling side, if you use chiller an recirculating pump or electro valves, you ca use a very slow pwm, 1 minute period or even more.

El El lun, 20 nov. 2023 a la(s) 13:11, Emile @.***> escribió:

You closed this, so you probably figured it out already. A normal STC1000 works as a thermostat, meaning that heating (and cooling) is switched off if the temperature is reached. I added a PID controller for heating to this, since a heating element is typically controlled by a SSR, which is controlled by a PWM signal. Which is a perfect for a PID controller.

Cooling (compressor in a fridge) on the other hand cannot be controlled by a PID controller. So there you have the standard thermostat algorithm.

Hope this helps understanding the function of the PID controller I implemented.

— Reply to this email directly, view it on GitHub https://github.com/Emile666/stc1000_stm8/issues/18#issuecomment-1819362372, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACXBW4P54MMUNB3FUFV6K4TYFN6JNAVCNFSM6AAAAAA7OEH3YOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMJZGM3DEMZXGI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[JasperBFH]

I have 2 sides. On one side a water pump with fan. And on the other side a water pump with a container of water where the temperature must be regulated (cooling and heating). I control the H-bridge (Heating & cooling) with a PWM of 20KHz. The sample time for the PID algorithm is 3 seconds. With the help of Ziegler Nichlos and pidtuner.com i determined my parameters for heating & cooling.

Cooling Kp= 175, Ki = 1.45, Kd = 0, Heating Kp = 20, Ki = 0.0889, Kd = 0,

I tried your type C Takashi PID controller. Unfortunately I'm getting crazy results with your code. I used a different PID code and it does produce the desired results.

What am I doing wrong? Surely a better PID controller (Type C Takashi) should work and not change the parameters?

float PID_Update (PIDController *PID, float setpoint, float measurement
{
// Calculate Error
float error = setpoint - measurement;

//Calculate the integral error
pid->integrator = pid->integrator + pid->Ki  * error;

//calculate the measurement derivative
float dpv = (measurement - pid->prevMeasurement);

//calculate the PID output
pid->P  = pid->Kp * error;
pid->I  = pid->integrator;
pid->D  = -pid->Kd * dpv;
pid->out = pid->P + pid->I + pid->D;

// implement anti-reset windup
if (pid->out > pid->limMax)
{
    pid->I =  pid->I - pid->Ki  * error;
    pid->integrator = pid->I;
    pid->out = pid->limMax;
}
else if (pid->out < pid->limMin)
{
    pid->I =  pid->I - pid->Ki  * error;
    pid->integrator = pid->I;
    pid->out = pid->limMin;
}

pid->prevMeasurement = measurement;
return pid->out;}

[JasperBFH]

I guess the problem is in the PID algorithm. I will solve this myself to realize it for heating and cooling.

I do have one more question. If I start heating with the PID for heating, do I have to use the PID for cooling to make the calculations for cooling in the background so that the integrator adjusts when the temperature rises again

[Emile666]

There's a define GMA_LLIM that is currently set to 0, but if you set this to -1000, it will do cooling as well. You only have to add some code that a negative output is sent to the Peltier element (as a positive percentage), and that a positive output is sent to the heating element as usual. In this way you can do a step-response as described in the document.

[JasperBFH]

Okay, that's clear. The problem I am experiencing is that the P value is negative and therefore the system/PID controller does nothing when heating with the formula pid->P = pid->Kp * (pid->xk_1 - xk) I get negative numbers.

If xk_1 is 25 degrees and xk = 26 degrees then you get a value of 20 * (25 - 26) = -20

This is strange if the set point is 30 degrees and the system needs to heat up. Then it seems to me that the P value must be positive?

[Emile666]

My PID function is a type C velocity algorithm, it calculates a delta, so it is perfectly normal to get negative numbers, meaning that the PID-output should be decreased. The PID-output can still be positive, but it is getting a lower value. And since the setpoint is removed from the P and D action, it is less obvious to see. If xk_1 - xk < 0 in your case, then xk_1 < xk, meaning that ek < ek_1, which indicates that the error-value is decreasing. And therefore it is logical that the PID delta is less than 0, so that the PID output will be decreased as well.

[JasperBFH]

Okay. Hmm okay, that's clear. But the problem remains that the output with PID never starts up. Like your example in the brewing process where the output goes to 100% when starting.

My tuned/calculated parameters (with the other PID algorithm) Kp, Ti, Td, Ts values ​​are 20, 239, 0, 3.

If I run the PID algorithm for the first time with room temperature then it is Kp = 20 * (0-22.09) Kp = -441.875

Ki = 0.25 * (30 - 22.09) Ki = 1.98

Kd = 0

Out += P + I + D Out = 0

the PID returns 0 and the water will never heat. What am I doing wrong?

[Emile666]

So you run only 1 iteration of the PID-controller and based on that you conclude that the 'water will never heat'? That's what I call premature. If xk and xk_1 was properly initialized to the environmental temperature prior to starting the controller, you wouldn't have the large negative transient (that is called ' bumpless transfer'). If you run the PID-controller for at-least 6 seconds prior to enabling it, the controller starts without negative transients. That's why I always use an enable signal for the controller. It always runs every Ts seconds, but only starts calculating an output when the controller is enabled.

I am not going to give a lecture on digital control systems here, but hopefully you can convince yourself by running multiple iterations (every 3 seconds in your case) and see what it does. Because the I-action keeps on integrating, while the P-action becomes less negative (because xk_1 is approximately equal to xk), so you have a controller that starts in a nice and relaxed way. If your Ki is 1.98, then after 10 * 3 seconds, you already have something like 40 % total output. But if you only look at the very first iteration, then you are not getting the picture. Again, you have to understand what exactly a velocity type algorithm is (controller calculates a delta and you should add that to the cumulative output) and what a type-C controller does.