Underfloor heating with PID controller, NodeRed

Updated 2024-11-14 !

In a former post I did some hacking/checking how our underfloor heating control system handled the valves, actuators, regulating the heat. 

Now it was time to do some tests with NodeRed(NR), with a PID controller, and a Shelly 1(S1) connected via WiFi and driving the actuator.

The actuators are in my case controlled by 24 VAC and the S1 accepts 110 - 240 VAC, 24 - 60 VDC and 12 VDC. So in this case it was simple to just add a rectifier and a capacitor, to the 24 VAC, to get approx 36 VDC to source the S1.

Hardware setup

Raspberry Pi

I have a Raspberry Pi which is running 24/7, with other functions, so added the below flow to the NR installation.

Relay

So connect according to the schematics, seen at the right, and then setup the S1 for your WiFi.

Thermometer

A very nice sensor is the Ruuvi tag but due to it is Bluetooth the range could be limited. Another solution could be using cheep 433 MHz temperature sensors.

Software

Prerequisites

As said I have used NodeRed with the additional nodes

• node-red-contrib-pid, PID node. Check the documentation with links to a sample flow
• node-red-dashboard, which visualizes the data
• node-red-contrib-timeprop, gives a time proportioned output from a linear input.

The flow

Every minute the measured temperature is retrieved, via the "link in" Node, and processed by the PID node. The PID output, range from 0 - 1, is then turned into a time proportioned output. This output is then sent to the S1 with a turn on, http://192.168.1.145/relay/0/?turn=on, or turn off URL.

As an extra function the valve is exercised to max open during 10 minutes, at 06:45, so it will not get stuck. Especially useful during summertime when it's probably closed most of the time.

Tuning 

There are lot of documentation for tuning PID controllers but the guy who wrote the PID node have done an article and Eurotherm, a machine automating company, also have a good compact writing. 

Start up with configure the PID node to use ON/OFF control

• "Proportional band” set to zero. 
• The integral effect, "Integral time", will be disabled with a large number. One day in seconds  => 24x60x60 => 86400.  
• Disable the derivate effect, "Derivative time" and set to zero.
• "Initial integral" set to zero.
• "Derivate smoothing factor" set to zero.  

Start up the process and check the temperature progress in the graph. Get the values for "Xosc" and "tosc". As a first approach change the values in the PID to; 

P = 2.0 * Xosc, Proportional band
I = 1.5 * tosc, Integral time
D = I / 5, Derivative time

and maybe wait with the D part untill you have gained some knowledge of the process.

As said by Eurotherm "These values result in a slightly over damped response and consequently all values can be reduced slightly to give a more responsive loop."

Wrap up

As seen in the picture, set point 20 °C, the measured temperature values are in 0,1 °C steps which makes the PID output react with a quick raise in the output. This could be changed with a smaller "Derivative time", or a low pass filter on the input, but the best way is to have a measurement which is more stable and have a better resolution. 

Please note that the data with the valve opening status have another time base.

For the moment, 2024-11, the setup is running with the mentioned Ruuvi, temperature with 2 decimals, which really improves performance.

If you want to learn more about the PID controller and it's parameters just search on Youtube and, as said, use the sample flow.