Just try a polynomial fit
Already dit that; power_cosumption_hue_bulbs.ipynb - Colaboratory (google.com)
But the result is not really satisfactory. Looks like we’ll have to generate a few more data points after all 
Interesting link from philips - although it doesn’t compare with the measurements, if your fit is accurate - there’s a much sharper bump in the middle. With my measurements, rather than do a exponential fit, I did a polynomial fit in excel and got an R^2 of 0.99+ for all equations (although it appears I didn’t save). I don’t think it would be that complicated to do the interpolation if the equations are plugged into a template. I’ve measured for 7 values of mired, so it’s not too complex.
A crude way could be:
- If mired is between a and b, then
- calculate power for the respective a and the b mired equations with a brightness of x for both
- interpolate between the results of the two equations
- else if mired between b and c
- …
- else if mired between c and d
- …
- else if mired between n-1 and n
- …
- end if
I’d be interested in a database. I’ve done a Hue ambience and have an Hue candle RGB that I could do, but that would get pretty messy - four inputs rather than two.
No idea how to upload or share the data though
Hi
Picking this thread up again now I have time - has anyone come to a conclusion on how to do this? I appreciate the measurements for all bulbs may take time but looking at @CM000n example, do we create a sensor for every bulb
Or is this more of a script , where we can maintain a block of code that looks at all of your hue bulbs and then looks at their %brightness in order to calculate the watts used and then present the entire wattage used in the house ?
We can maintain the hue build models and their associated values by brightness level within the script ?
I just finished my code. It’s based off measurements of power by adjusting mired and brightness. I then fit a curve for each of the mired value sets, with brightness as the variable using a second order polynomial on this site Polynomial Regression Data Fit. The measured data at the bottom of this post is from a Philips Hue LTW001 bulb.
More than happy for feedback on the code.
- platform: template
sensors:
living_room_big_lamp_power:
friendly_name: Living Room Big Lamp Power
icon_template: mdi:lightning-bolt
unit_of_measurement: "W"
value_template: >
{% if is_state('light.living_room_big_lamp','on') %}
{% set mired = state_attr("light.living_room_big_lamp","color_temp") %}
{% set brightness = ((state_attr("light.living_room_big_lamp","brightness") - 3) * (100 - 1)) / (254 - 3) + 1 %}
{% set m153 = (0.000435254 * (brightness**2) + (0.002105929 * brightness) + 1.500175321) %}
{% set m203 = (0.00057964 * (brightness**2) + 0.001134825 * brightness + 1.527090668) %}
{% set m253 = (0.000785198 * (brightness**2) - 0.002785236 * brightness + 1.57216088) %}
{% set m303 = (0.000733083 * (brightness**2) - 0.002574077 * brightness + 1.573079888) %}
{% set m353 = (0.000668279 * (brightness**2) - 0.001853391 * brightness + 1.561981489) %}
{% set m403 = (0.000570268 * (brightness**2) + 0.0000671178176666897 * brightness + 1.527920206) %}
{% set m454 = (0.000487813 * (brightness**2) - 0.000161527 * brightness + 1.518884275) %}
{% if mired >= 153 and mired < 203 %}
{{ ((((mired - 153) * (m203 - m153)) / (203 - 153)) + m153)|round(2) }}
{% elif mired >= 203 and mired < 253 %}
{{ ((((mired - 203) * (m253 - m203)) / (253 - 203)) + m203)|round(2) }}
{% elif mired >= 253 and mired < 303 %}
{{ ((((mired - 253) * (m303 - m253)) / (303 - 253)) + m253)|round(2) }}
{% elif mired >= 303 and mired < 353 %}
{{ ((((mired - 303) * (m353 - m303)) / (353 - 303)) + m303)|round(2) }}
{% elif mired >= 353 and mired < 403 %}
{{ ((((mired - 353) * (m403 - m353)) / (403 - 353)) + m353)|round(2) }}
{% elif mired >= 403 and mired < 454 %}
{{ ((((mired - 403) * (m454 - 403)) / (454 - 403)) + m403)|round(2) }}
{% elif mired == 454 %}
{{ m454|round(2) }}
{% endif %}
{% else %}
{{ 0.00 }}
{% endif %}
Active Power (W)
Kelvin 6535.947712 4926.108374 3952.56917 3300.330033 2832.86119 2481.389578 2202.643172
MIRED 153 203 253 303 353 403 454
BRIGHTNESS
1 1.52 1.53 1.55 1.54 1.54 1.53 1.52
10 1.56 1.59 1.59 1.61 1.6 1.58 1.57
20 1.7 1.78 1.85 1.83 1.8 1.75 1.72
30 1.94 2.08 2.23 2.18 2.13 2.04 1.96
40 2.28 2.51 2.76 2.69 2.59 2.46 2.23
50 2.71 3.05 3.44 3.32 3.18 2.98 2.76
60 3.2 3.69 4.23 4.05 3.84 3.59 3.28
70 3.78 4.42 5.16 4.92 4.66 4.3 3.89
80 4.45 5.32 6.29 5.99 5.64 5.16 4.63
90 5.23 6.32 7.67 7.26 6.8 6.16 5.48
100 6.05 7.45 9.22 8.72 8.11 7.25 6.36
1 Like
@cloak, which device/sensor did you use to measure the power consumption of the bulb? I want to measure some other Hue lights but not sure how to measure low AC current accurately.
I used a Power-Mate Lite: https://www.systemcontrol.com.au/product-specs/11470505/PML10A
I can’t remember where i got it, or how much it cost.
If you have a smart plug with a meter in it, that may work. I have’t done a back to back comparison, so I’m not sure of the accuracy (of a Smartthings smart plug).
I’m trying to get some test rig setup using a current sensor which I can attach to a ESP microcontroller or raspberry Pi.
Want to automate setting all different brightness and mired combinations using the Hue API and than saving the current readings to a JSON file / database.
I have these sensors ACS712 and SCT-013, but I am not sure about their accuracy on low currents, need to test this soon.
Will get back to you when I get this working.
1 Like
Hi @cloak
Thanks so much for sharing that hue sensor template, although I’m not exactly sure what to do with it to work with my hue bulb?
My bulb has the entity id of light.living_room_light so i have updated your references to “ light.living_room_big_lamp” but i can’t see anything related on the UI or in the developers tools view ?
Hey @nodecentral, if youve changed only the light entity, once you reboot you should have an entity named sensor.living_room_big_lamp_power - third line of the code.
Cool approaches guys 
@cloak , did you really do all the measurement points?
With an interval of 1-100% brightness and 153-500 mired, it should “only” result in 34353 data points.
The best thing would really be if we had stored them somewhere. Then everyone can store a lookup table in their private database and query the values via an SQL sensor depending on brightness and colour temperature.
Then there would be no need for such inaccurate approximations via some formulas 
@CM000n, I would sort of like to say that I made 34353 measurements…
Mired in 50 mired(s?) steps and brighness in 10% steps, which works out to 77.
I’m all for storing these measurements, but wouldn’t know where to start.
Maybe we can agree on some sort of JSON format or CSV format and publish to a git repository.
Something like:
[
{
"model": "LTW001",
"measurements": [
{
"brightness": 1,
"mired": 153,
"watt": 1.52
},
{
"brightness": 1,
"mired": 203,
"watt": 1.53
}
]
}
]
Or maybe one json file per Hue model.
Than multiple people can contribute to having measurements for more models, and later we can build some tools to import this data into a SQL database. And maybe build a custom component or something like that.
Btw. how would we approach heaving measurements for RGBW lights? should we take measurements for all brightness, R, G, B, W combinations, which is way to much if we do all 1-255.
1 Like
pfff, only about ~1.6 million datapoints. Easy peasy 
The coolest thing would be if someone could write a Python script that iterates through all the possible combinations and writes the measurements for them away in a database.
Maybe something like this?:
# Set bulb to measure and corresponding powermeter
delay_time_secs = 30
entity_id = data.get("entity_id")
powermeter = data.get("powermeter")
for bri in range(1, 101):
for r in range(1, 256):
for g in range(1, 256):
for b in range(1, 256):
print(bri, r, g, b)
# 1.Set specified Bulb to current bri,r,g,b values
service_data = {"entity_id": entity_id, "rgb_color": [r, g, b], "brightness": bri}
hass.services.call("light", "turn_on", service_data, False)
# 2.Wait 30 Seconds to get correct power values
time.sleep(delay_time_secs)
# 3.Read power values
powerconsumption = hass.states.get(powermeter)
# 4.Write power values with bri,r,g,b values to specified database
try:
connection = psycopg2.connect(user="MYUSER",password="MYPASS",host="192.168.178.3",port="5442",database="ha_sensordata")
cursor = connection.cursor()
postgres_insert_query = """ INSERT INTO powermeasures_rgb (manufacturer, bulb, bri, r, g, b) VALUES (%s,%s,%s,%s,%s,%s)"""
record_to_insert = ('signify', 'testbulb', bri, r, g, b)
cursor.execute(postgres_insert_query, record_to_insert)
connection.commit()
count = cursor.rowcount
print(count, "Record inserted successfully into powermeasures_rgb table")
except (Exception, psycopg2.Error) as error:
print("Failed to insert record into powermeasures_rgb table", error)
finally:
# closing database connection.
if connection:
cursor.close()
connection.close()
print("PostgreSQL connection is closed")
Then we can make the measurements available, as you say, as CSV or JSON, preferably via a git repo.
Sadly Home Assistant Python Scripts doesn’t seem to support sql connections 
I might have to take a look at the Home Assistant REST API to be able to control and read entities through it. Then we could use an external Python script that allows writing to an database.
I’ve been procrastinating on doing a few measurements for my Hue candle light to see if there’s any pattern that could make additional measurements more targeted.
Later tonight I’ll have a look at the accuracy of my Smartthings Smart Plug vs the meter i originally used. I should be able to do some sort of calibration and your code could/should work…I will definitely need help making the code happen - I don’t know where to even put it!
Thanks for your input. There is a small issues with your approach.
There are just too many datapoints. It’s not 1.6 million but 1.6 billion. This will take 52 years to process and I forgot about the 30 seconds delay which would take 1577 years. Around that time we will be fully colonized on Mars
The reason I wanted to investigate taking measurements with a direct sensor on a Raspberry PI is to eliminate the need to wait for x seconds to have the Zwave/Zigbee/Wifi plug communicate the new Power measurements. I also have some Zwave smartplugs at home. Will do some testing with them as well.
To minimalize the amount of datapoints we need to measure we can do in steps of 10 as @cloak also did in his measurements. Or maybe steps of 5. This will be accurate enough imo to just pick the closest datapoint when calculating the power consumption in Home Assistant.
Also I see Hue uses the HSL color value (Hue, Saturation, Lightness). I think it would be the best to take measurements in this format.
Hue = 0 - 65535 (take in steps of 1000, which would be 65 points)
Sat = 0 - 255 (take in steps of 10 for 25 points)
Bri(L) = 0 - 255 (also in 10 steps)
For a total of 65x25x25 = 40625 measurements
Let’s say we need to wait 5 seconds between each measurement than it will take 56 hours, which will be fine.
There is no need to save this data to a relational database first. Why not directly save to a JSON file using the python file IO and json libraries? This will eliminate the need of other external dependencies.
When we have taken measurements in JSON files we can develop some python library which can load the JSON files in memory and have a function to give the power consumption for a given model and the closest HSL values. Than this library could be consumed by some custom_component or maybe in the official Hue component.
Let me know what you think.
I can setup a GIT repo in the weekend.
1 Like
I am curious if you get similar results with both meters or the results are all over the place.
You are absolutely right. Sorry for the confusion with our German number formats
Maybe I overshot the mark a little. Otherwise we’ll all be old by the time we have any consumption values.
This also makes it clear why there are so few detailed consumption figures from official sources.