By the way, this should be the algorithm implemented, and not an on/off cycle. This is not an electric boiler, and I have a feeling that someone took that kind of algorithm and put it in KE100... or there is a bug, where whatever the valve measures, it ends up with 0% or 100% as the desired state.

When the temperature outside is constant, e.g., +2, and there are no changes in the room (new heat source, open window, etc.), I do not see how more than 2-3 corrections would be necessary during the night. Meanwhile, KE100 goes through a "heat -> cooldown -> heat" cycle endlessly, lowering comfort and annoying occupants.

Algorithm with Stability Factor

1. Initialize Variables

   • Set the target temperature (e.g., 20°C).
   • Define a hysteresis range (e.g., ±0.5°C).
   • Initialize a proportional control factor (P) for initial adjustments.
   • Introduce a stability factor (S) to maintain the valve position near the target temperature.
   • Track the last stable valve position (percentage open) that maintained the target temperature.

2. Read Current Room Temperature

   • Use a temperature sensor to get the current room temperature, T_current.

3. Calculate Temperature Difference

   • Calculate the difference, ΔT = T_target - T_current.

4. Apply Hysteresis

   • If ΔT is within the hysteresis range (e.g., between -0.5°C and +0.5°C), hold the valve at the last stable position (stored from previous adjustments) rather than closing it.

5. Proportional Adjustment for Large Deviations

   • If ΔT is outside the hysteresis range:
     • Calculate the required adjustment using:
       Valve Adjustment = P * ΔT
     • Update the valve's position accordingly.
     • Track this position as the new last stable valve position if it results in the temperature stabilizing near the target.

6. Stabilize Near Target Temperature

   • Once the target temperature is reached and stabilized (within the hysteresis range):
     • Fine-tune the valve position based on the stability factor (S) to hold the temperature. This may involve keeping the valve at a slightly open position that sustains the heat without overshooting.

7. Close Valve if Temperature Exceeds Target

   • If the temperature rises above the hysteresis range (e.g., more than 0.5°C above target), fully close the valve until the temperature drops back within range.

8. Repeat

   • Continuously monitor temperature changes, adjusting based on steps 4-7.

This approach focuses on finding and maintaining a "sweet spot" valve position rather than causing frequent on/off cycles, helping to stabilize room temperature more effectively.

  @Maciej54 I have the same issue. Compared to Tp-link, I have Evohome valves in my apartment and they are super smooth. I installed Tp-link valves in my parents house, unfortunately the expected smart behaviour is not there. Tp-link basically either fully opens or fully closes the valves which is very uncomfortable since the temperature is not kept constant.

  @KariBenz thanks for backing up my observations. It's so annoying... I was hoping for an improved comfort and maybe some "side" effect in form of savings, but  now it's worse than it was with good old plain "analog" TRVs. 

I have added a summary in my first post in this thread with my issues and observations... 

I'm so angry I didn't test that in one room first.... (now I have 20+ devices from them - i'm rather happy with smart plugs etc)

I can only hope TP-Link will finally acknowledge the problem and implement solution, as this is clearly a firmware issue. 

  @Maciej54 indeed, many thanks for bringing up this issue with all the details, hopefully TP-Link will react with a firmware update. 

In my understanding smart valves should involve a learning stage that takes into account how fast a room warms up and how fast it cools down so as to control the valve accordingly. For now the algorithm behind the TP-Link valves seem to be very basic...

@KariBenz sure! If a device can adapt to its environment, that would be great, but I think even that is not absolutely necessary if there is a proper algorithm for proportional correction in place. For now I gave up in my bedroom, where these problems are most anoying (temperature variations, stone cold heaters when it's around 0 degrees outside, noise that heaters make when fully opened - at least mine do).

Here is my way to make KE100 work :

And here is the result - constant temprature throughout the night, with 0.1C acuracy, no noise (as the water flow is low and constatnt), no temp fluctuations, no cold draft from windows (as the heaters are always at least warm)

Here how flat the temperature is with a classic thermostat (that was with +2 degrees ouside, so yes, the heating was on):

Anyways.... hoping for other users to "join" this post and also someone from TP-Link....

  @Maciej54 

I fully share the extremely annoying experience. 

The valves clearly operate on a totally binary logic, open or closed with zero intermediate setting. This leads to all the aforementioned annoying effects: wildly oscillating temperature, super noisy radiators and so on. 

I am confident I could fix it if I could have access to the firmware...

I really enjoy all the "smart" features but the operation of the valves is pretty bad, and I still cannot understand why they have so many positive reviews on Amazon. 

I hope they will listen and fix this firmware...

However I have a question with your experience with the T315 sensor: in this article you already cited https://community.tp-link.com/en/smart-home/forum/topic/699248 they seem to imply that the valves are actually progressively closing in a sort of smooth operation, and they are only complaining about it missing by 0.5 degrees the set temperature (I totally would not mind a .5 degrees mismatch...). So at first I thought the "binary" operation might be due to high gradients at which the temperature sensor on the valve undergoes due to vicinity to the radiator, and that an external sensor might solve the issue. 

However you say that the external temperature sensor does not lead to any improvement. Is this the case? I would like to know about your experience before wasting more money on this...

@Matteo_T Exactly my thought - It's a shame the firmware is not on GitHub so that we could fix it, even if it voids the warranty :)

I have the following setup:

• Bathroom small - solo KE100
• Bathroom large - solo KE100
• Living Room with kitchen:: T315 + 4 x KE100
• Bedroom 1: T315 + KE100
• Bedroom 2: T315 + 2 x KE100

It's hard to see any corelation between their behaviour and using T315 external sensor or not.

Small bathroom is more keen on hiting the target temp. The large bathroom is basically always below set temprature by 0.5 - yesterday it was 0.6C below set temperature and the radiator was cold (with +2 outside). 

Bedroom with 2 valves seems to oscilate quite close to target temperature, never the less, it does "cycle" and the radiators are cold periodically (while being below set temperature). 

Regarding the partial close - it does happen, but its hard to tell when. I've noticed in the bedroom that the valve performed 1st partial close, 2nd partial close and full close. 

But that happened within just a few minutes (no more than 15 I would say). The end result was the same as if it would just close fully in one go. 

I think the amplitude of the temoperature reading depends on several factors like room size, radiatror type etc. But it fails miserably with both internal and external temp sensor - so I would say it will not fix your issues by adding external sensor. 

I would also suspect that the graph looks a bit diferent depending if its based on internal or external sensor, simply due to the location where the readout takes place. 

Closer to the radiator you will see bigger fluctuations I suppose.

I wanted to use the external sensor to get a more stable temperature, avoid any issues with TRV being behind a curtain or something else that would prevent the correct reading.

Now I have reverted back to my analog TRVs in both bedrooms and hoping for a new firmware.... 

I have two other very basic systems that work just fine with no external sensors. 

This is how close my other TRVs sit to radiators and are not affected by them (another building).

They have no issue stabilising. This one on the picture sits at 15% open. 

  @Maciej54 I agree completely with your findings, it's been driving me crazy.
I can't understand how Tapo/Tp-link can't get this right.
I feel they will just ignore this problem...