how do i ... send a delta load value using the API?

[ragg987]

Hi, I'm trying to send a load increment to predbat using automations, but I think I'm not fully understanding the options correctly. Any guidance would be appreciated...

Scenario:
I calculate the energy required to heat my house based on weather etc. I then use Octopus Target timeframes to find the optimum slot(s) to turn on my heat pump based on rates and external temperature. This all works fine. What I'd like to do is to tell predbat of this load through an automation.

What I've tried:
I've tried using the select to perform this, it works, however it requires me to concatenate together ALL the 30-minute slots required for the heating. Simple (manual) example:

action: select.select_option
data:
  option: >-
    17:00:00,17:30:00,18:00:00
target:
  entity_id: select.predbat_manual_load_adjust

I then looked at the predbat API example and it seems more suitable, in that I can feed a date, start and end times. However I don't understand how to send it a load increment as a fixed value - e.g. 0.4kWh. Simple example that works, but using load_scaling:

action: select.select_option
data:
      option: >-
        rates_export_override?date=2025-11-11&start=18:00:00&end=19:00:00&load_scaling=2.0
target:
    entity_id: select.predbat_manual_api

Have I missed something here on the API method? If a discrete load increment load is not supported how could I use the select method to create a concatenated list between start and end times?

[gcoan]

I don't think you have missed anything, but what you are asking for doesn't exist as an option in predbat at the moment

the predbat manual API using rates_export_override allows you to set the same parameters for the rate period as can be in apps.yaml https://springfall2008.github.io/batpred/energy-rates/#manually-over-riding-energy-rates - which is a load scaling factor not a specific new load amount

the manual load adjust selector requires you to select all 30 minute time slots that you want the heating on for. Its possible to do this in template code, google for similar examples of iterating between two time periods with 30 minute increments, or give chat GPT a go to create the code for you

As an aside, general advice is to run your heat pump on a low and slow temperature all the time, that's by far the most efficient way of heating your home. And since you have batteries controlled by predbat, let predbat charge your batteries on the cheapest periods that your heat pump can then use when it needs to

[ragg987]

Thank for the response, ChatGPT came good and I've managed to cobble up a sensor that generates the string for all the hours in question. I'll test it with the load adjust selector and share on this posting in case it is of use to anyone else in future.

As an aside, general advice is to run your heat pump on a low and slow temperature all the time, that's by far the most efficient way of heating your home. And since you have batteries controlled by predbat, let predbat charge your batteries on the cheapest periods that your heat pump can then use when it needs to

I've used my heat pump's weather compensation and set an appropriate compensation curve for 9 winters, 24hrs per day. It seems far from optimal. My observations:

• in milder weather the compensated target flow temperature was low (in the mid-20s) and the heat pump was constantly cycling on and off at the lowest scroll setting (~650W input) as there was too much output energy to maintain the low temperature. Each time the pump cycled on the COP would take a hit for the initial 20-30 minutes.
• in very cold weather the compensation curve would ask for ~35C flow which my heat pump could not sustain. This led to excessive defrost cycles as the HP went to maximum input power (~2.5 to 3kW), plus the COP took a massive hit.
• The HP compensation controls are reactive - they cannot anticipate that tomorrow is very sunny and solar gains will be substantial, hence reduce heating tonight else the house will overheat
• My batteries are not sized to sustain the HP through cold or gloomy days - this was a deliberate choice I made in sizing my batteries, I calculated the returns vs investment and for me the sweet-spot was 10kWh storage with 6.4kWp of solar.
• "Low and slow" does not factor in electricity pricing, so while heating efficiency will be good bills will not be lowest - I'm happy to trade away some efficiency for lower bills
• As context mine is a very low energy self-build house, built to passiv standards so it does not cool down rapidly once I turn heating off. The UFH and screed also serves as a large heat buffer (equivalent thermal capacity to about 3,000l of water), which helps a lot.

So more recently I managed to connect my HP to home assistant and can now control the various parameters through automation. I am currently experimenting with the control to improve on the built-in weather compensation. So far it is working well.

I've discovered that a flow temperature of 30C lets the HP run continuously and with a good COP of 4.5 to 6 when external temperatures are 6C or above. So I now run the HP for x hours per day based on the calculated demand using temperature and wind forecast, solar gain based on PV forecast, set and current temperature as inputs.

Once I've calculated x, I use target timeframes and split it into a maximum of 2 slots:

• Time-slot 1 to coincide with IOG off-peak
• Time-slot 2 to coincide with highest daytime forecast temperature, factoring in any additional IOG dispatches and free slots. My logic here is that HP efficiency will be at its best during the warmest part of the day.

Time-slot 2 is only required if x exceeds 6 hours - in these shoulder months it is seldom needed. On very cold days 1 and 2 could well merge into a single time-slot of 24hrs.

I've yet to experience very low external temperatures with this setup, so testing remains incomplete; however it looks very promising so far. Here's an example of the coldest night so far @ just below 6C outside. COP is hovering around 5. I had 3 defrost cycles.

[gcoan]

thanks for all your thoughts and explanation of the heat pump logic, and once you get your template automation working, do share it, and I can add it to the documentation.

I too have a heat pump (in fact 2 9kW heat pumps), but generally only run one heat pump except for the really worst weather. Running one heat pump means that its demand will usually fit into my battery output (2x GivEnergy Gen 1 hybrid inverters, so 4.8kW charge/discharge).

I follow your logic about heat loss of a passiv house, and with UFH that has a high thermal mass you can run into overheating the house. Not a problem I have!

I've had my heat pumps for 3 years and progressively worked on the best way to operate them and reduce electricity consumption. I don't have any HA integration for my heat pump, it's on the list, but not straight forward to do. So I have no COP figures to work from, it's been very much trial and error of setting the temperature curve and adjusting it bit by bit when too cold or too hot. Interesting your experience with the heat pump at the upper and lower temperature extremes, I don't have the same with mine; here it is today's power and temperature chart, its sitting quite happily at 20-25 degrees

I was on Agile from October 2023 through to January 2025, and Predbat served me well, filling the battery on the cheapest periods. I only moved off to Cosy after more than a month of high Agile import costs; Cosy is much easier to handle with predictable cheap periods. My battery storage is only 13.5kWh which isn't enough to run the house and heat pump for a whole day either, would love to have more but economically it's not worth it.

There's a good Heat Geek video on the cost of running your heat pump with different energy tariffs and operating strategies. Worth a look

Cheers

[ragg987]

Here is the first-cut automation. It relies on a target timeframes entity for the start and end times of heating. I am using a static 0.4kWh load increase for now - I will evolve this into a dynamic value (helper) once I understand my HP behaviour in colder weather. The automation does not take into consideration the 18hr forward looking constraint on setting a load over-ride - predbat seems to throw a wobbly if more than 18hrs worth of hours are set by this method.

My preferred approach would be to create a stand-alone helper to generate the long string for the delta load and consume that from the automation, however I discovered that HA has a limitation of 255 characters for a sensor value.

alias: predbat_calculate_energy_for_ashp_heating
description: "automation takes forecast heating run time and converts it into an energy load for predbat manual load forecast"
variables:
  times: >
    {% set t_list =
    state_attr('binary_sensor.target_timeframes_octopus_energy_import_myclimate_intelligent_heating','target_times')[:]
    | map(attribute='start')  | list %} {% set t_count = t_list | count %}

    {% set t_start = '00:00'%} {% set t_end = '00:00'%} {% set times = 'none' %}

    {% if t_count > 0 %}
      {% set t_start = t_list[0] %}
      {% set t_end = t_list[t_count-1] %}
      {% set t_start = as_timestamp(t_start, '%H:%M') | int %}
      {% set t_end = as_timestamp(t_end, '%H:%M') | int %}
      {% set step = ( 30 * 60 ) %}  {# 30 min interval in seconds #}
      {% set times = range(t_start, t_end + 1, step)
       | map('timestamp_custom', '%H:%M:%S')
       | list %}

      {% set times = times | join(',') %}
    {% endif %}

    {{ times }}
triggers:
  - trigger: state
    entity_id:
      - >-
        binary_sensor.target_timeframes_octopus_energy_import_myclimate_intelligent_heating
    attribute: target_times
conditions: []
actions:
  - choose:
      - conditions:
          - condition: template
            value_template: "{{ times != 'none' }}"
        sequence:
          - action: input_number.set_value
            target:
              entity_id: input_number.predbat_manual_load_value
            data:
              value: 0.4
          - action: select.select_option
            data:
              option: "{{times}}"
            target:
              entity_id: select.predbat_manual_load_adjust
mode: single

Thanks for sharing your experience @gcoan. I had a look at the Heat Geek video and they have some good background info. I'm trying to take my optimisation a step further than low and slow by including tariff incentives and my approach would be more suitable for a house with a large thermal store and low energy loss.

I might have to wind some of this back if comfort suffers - e.g. over- or under-heating. At present I'm getting complaints from the OH that it is too warm first thing in the morning, even though the room thermostat shows it is at a comfortable temperature. I suspect this is part of the human psyche I had not allowed for - we tend to feel warmer first thing in the morning (having just popped out from under a warm duvet!) and cooler late in the day.

BTW I came across this when searching for the youtube videos - have you seen it? https://www.youtube.com/watch?v=XdgIoLXrvPU

Some of the comments by the guy in the black shirt resonated with me as inputs to where I am trying to get to - though he does keep pulling out the car analogies which are a bit naff. It tries to go below the "low and slow" headline message and pulls out a few nuances.

Just a comment on connection heatpump to HA: mine is a Hitachi Yutaki unit and I needed to create some custom modbus entries as well as install an ethernet to modbus converter. If you have a Daikin then some chap is building and selling an ESP32-based converter and also supplies the yaml etc to get it to work - nice guy, I think based in France. Fully local control no reliance on cloud. https://github.com/Arnold-n/P1P2MQTT