Using AI for prepping

[GeraldTheBonzai]

Thought I'd share something I'm working on. I've been looking to see how I can use AI to help me prep. Disclosure : my last job was developing LLMs and the data systems they use, so I know how flaky LLMs can be when used incorrectly but also how to mitigate the issues.

An LLM like ChatGPT is limited by the data it has been trained on, and the data it can access. This can causes it to hallucinate - it extrapolates to far when trying to answer a question and basically starts to make things up. There are technical reasons for this that i'm not going to go into.

What you want to be able to do is control the data that the LLM uses, so that you know that this data is "good" or at leat provide it with meaningful context for the questions you ask it. There is a way to do this: Retrieval Augmented Generation (RAG) - this is a way to use curated documents to steer the LLM. It's tricky to do with ChatGPT but much easier with the Google tools, Gemini and NotebookLM and that's what I use.

Right, into the weeds:
I loaded all of the user manuals for all of my kit into NotebookLM. Inverter specs, panels, my Aferiy and Ecoflow , breakers, cables - everything.
I loaded into NotebookLM a circuit diagram of how all my kit is wired up, both for my main solar rig and my smaller solar rig.
So, that's all my source reference material.

I then linked Gemini to my Notebooklm. Gemini now "knows" about my kit, my support equipment like my generator etc. I then asked Gemini to generate a master architecture specification, based on my kit, which gets loaded into NotebookLM. Gemini now "understands" my systems. I can ask it questions such as how long will my battery last if I am pulling 200 W through my inverter, and it uses the actual spec to calculate it.

So now things get interesting. In the event of a power outage, my plan was to run my fridge and freezer off of my inverter. I asked Gemini what to do. Its suggestion was to run the appliances off of my ecoflow. It also said that i don't have to power them all the time. Instead, run the appliances for 1 hour at 21:00. Then disconnect. The thermal mass would keep everyone cold until morning. At 8:00 run the appliances for an hour. Then recharge the Ecoflow from the main solar rig. But don't use the inverter - instead use the XT input connected to the bus bars. The voltage will be within range for the ecoflow and it will recharge either from the battery or panels. Brilliant.

Another suggestion. I was going to use the generator to recharge my battery, then power my base load via the inverter. Gemini spotted that my generator had two AC outputs, so it recommended charging the battery from one AC output but powering the base load from the other AC output, negating the need to use the inverter.

At the moment t I am focusing on energy, but I am going to expand out into water management and then food. Next step is to get it to generate a ConOps - give it some scenarios and work out how to manage them. I'm also looking to create a user manual, so that SHMBO knows how to power everything up/down etc.

[Yorkshire Andy]

Just to throw a spanner on the works does a fridge freezer work more efficiently if constantly powered ??

It's ok for it to graft an hour solidly to regain the thermostat set temperature but is a big long run to "get it back " after x hours of no power better than the constant supply and shorter runs to maintain the temperature Vs inverter losses running constantly especially at night when its usually cooler so less loss / smaller gradient of temperature change

Know mid summer my 1600 delta max will run form just shy of 48 hrs Samsung fridge freezer and the old swan freezer

[GeraldTheBonzai]

Just to throw a spanner on the works does a fridge freezer work more efficiently if constantly powered ??

It's ok for it to graft an hour solidly to regain the thermostat set temperature but is a big long run to "get it back " after x hours of no power better than the constant supply and shorter runs to maintain the temperature Vs inverter losses running constantly especially at night when its usually cooler so less loss / smaller gradient of temperature change

Know mid summer my 1600 delta max will run form just shy of 48 hrs Samsung fridge freezer and the old swan freezer

Thats exactly what I thought until I did the sums. The freezer has a substantial thermal mass. As such, leaving it unpowered overnight, it will start to warm up, the rate being dependant on the ambient temperature. But in a 9 hour period its not going to be a lot. On the other hand, leaving it powered, even when it is just ticking over, the inverter is essentially a parasitic drain on the battery. Doing some calcs, it is more energy efficient to leave it unpowered over night and then "re-freeze" it in the morning. The net Wh drain is significantly less.

Point is, Gemini worked this out, not me.

[Frnc]

That's like the question should I leave my central heating on for the next few hours while I'm out, or have it go off and then have to use loads of gas in the first half hour. Never an easy one to answer, with or without AI. The problem with AI is it's likely to give you one answer one time, and the other answer another time, and both times will insist it is certain.

Perhaps the answer is in the fridge itself. How ofted does it run a compressor and other things? You might find out by observing an IHD showing current electricity use.

[GillyBee]

It is all down to the physics

Solar energy is stored into the battery. We lose some on the way in but dont generally worry about that.
On the way out, some energy goes to the fridge for cooling and some is wasted by the invertor.
The fridge slowly warms because of door opening and imperfect insulation. The fridge cooling system cuts in intermittently to top it back up, drawing energy from the battery. Or is topped up manually at 8am.

The amount of heat the fridge loses is fastest at the beginning and slowest at the end. (Like a cooling cup of coffee in reverse)
e.g 10 units in hour 1 down to 1 unit in hour 10.

So leaving the fridge as long as is safe will use less energy in total than topping it up more frequently. (10 lots of 10 units = 100 units or 10+9+8+7+6+5+4+3+2+1 = 50 units)

NB The actual figures in this are not real I made them up to illustrate the point. And once the power is off overnight it would be worth wrapping the fridge in a duvet to slow the losses even further.

[Frnc]

It is all down to the physics

Solar energy is stored into the battery. We lose some on the way in but dont generally worry about that.
On the way out, some energy goes to the fridge for cooling and some is wasted by the invertor.
The fridge slowly warms because of door opening and imperfect insulation. The fridge cooling system cuts in intermittently to top it back up, drawing energy from the battery. Or is topped up manually at 8am.

The amount of heat the fridge loses is fastest at the beginning and slowest at the end. (Like a cooling cup of coffee in reverse)
e.g 10 units in hour 1 down to 1 unit in hour 10.

So leaving the fridge as long as is safe will use less energy in total than topping it up more frequently. (10 lots of 10 units = 100 units or 10+9+8+7+6+5+4+3+2+1 = 50 units)

NB The actual figures in this are not real I made them up to illustrate the point. And once the power is off overnight it would be worth wrapping the fridge in a duvet to slow the losses even further.

That's a point. What I said tells you normal fridge activity, not what you can get away with. Heat loss as you say is proportional to the difference between what you are interested in, eg the temperature of the food, and the surround, ie the room temperature.

Obviously by the same logic, room temp is a factor.

[GeraldTheBonzai]

The model I used assumed an ambient temperature of 20C. Most freezers run at around -18C and ideally you dont want them to rise above -5C. Doing the math, the estimated heat transfer is around 0.7C/hour. so after 9 ish hours, the internal temperature would be around -11C, well below the limit of -5C