My setup - and the next steps

Question

question

carl-meek asked Feb 22, '23

My setup - and the next steps

Please see the attached diagram of my setup, and photographs.

I'm looking for some thoughts and suggestions on the next step, which is protecting me against grid failure. Right now - if grid fails, everything goes dead.

I do have a manual generator changeover switch which can switch my load over to the AC-OUT of the Victron, but this is mainly connected to GRID instead of VICTRON because my loads are too high (when you take into account the fact batteries need to charge from GRID/SOLAR at the same time as loads running. For example - it's quite common to see 55kW throughput just after midnight when 3x inverters are charging, and other loads such as heaters are running too.

I'm considering:

1. Replacing the changeover with an Automatic Changeover so that on grid fail it automatically flips to the Victron.

2. Installing a diesel generator, and connect that to VICTRON AC-IN-2.

3. Installing a new 3-phase armoured cable to the house to separate out ESSENTIAL Loads. Quite specifically to take the Car Chargers and Main Heat Pumps off onto another circuit. If I did this, it would mean that I could permanently run most stuff through the AC-OUT instead of being grid linked.

4. Installing a solenoid to cut off the US2000 batteries when they get down to around 11% charge, and then continue to drain from the US2000C+US5000 down to 5%.

The reason for this post - is just to start a discussion and help me make the right decision on how to achieve this. Does anyone have any thoughts?

The challenges I've battled so far:

1. Combining four stacks of Pylontech batteries that are not otherwise compatible - US5000,US2000C and US2000. All four are separately read using RS485, which is then combined together to produce a single CAN feed to the VICTRON.

2. Trying to get ESS Mode 3 working. Total fail. Regularly at least one phase would just ignore my set points. Eventually I just gave up- and decided to use MODE 2 instead, which worked perfectly. However I had to do the next step in order to achieve it...

3. Reading Eastern SDM630 meters using MODBUS and then sending this to the VICTRON by simulating a GAVAZZI meter.

4. Collecting a weather API to predict the weather and use this to choose how much to pre-charge batteries overnight.

5. Adding a nice web interface, alerts, graphing and emails.

All of the above written in Node, and running on a Raspberry PI with SIX RS485 interfaces and one CAN interface.

Diagram:

Photos:

My Raspberry PI Interface:

Victron Site:

MultiPlus Quattro Inverter Charger

diagram.png (865.5 KiB)

img-8404.jpg (441.2 KiB)

img-8405.jpg (403.0 KiB)

screenshot-2023-02-22-at-073928.png (261.8 KiB)

screenshot-2023-02-22-at-073940.png (215.3 KiB)

screenshot-2023-02-22-at-073950.png (474.8 KiB)

screenshot-2023-02-22-at-074130.png (263.3 KiB)

2 |3000

Attachments: Up to 8 attachments (including images) can be used with a maximum of 190.8 MiB each and 286.6 MiB total.

Modifications Space

Node Red Space

Answer 1 · 2023-02-22T20:07:06Z

johnone answered · Feb 22, 2023 at 08:07 PM

While some of the purists on here might frown at some of what you're doing, for me, it's great to see someone building such an ambitious system. On specifics, separating out essential (critical) loads should solve the grid fail problem - ESS Mode 2 should give you UPS functionality on critical loads.

I'm interested in your self-consumption optimising via a weather app - this doesn't seem well developed/understood in Victron world. Can you explain what system you've used, how you programmatically feed cloud data into battery SOC targets (if that's what you're doing). Thanks for sharing what you're doing.

2 |3000

Attachments: Up to 8 attachments (including images) can be used with a maximum of 190.8 MiB each and 286.6 MiB total.

Answer 2 · 2023-02-24T18:47:51Z

carl-meek answered · Feb 24, 2023 at 06:47 PM

Sorry for the slow reply.

My algorithm is going to take some time to perfect. it uses openweathermap.org API and pulls in a few factors such as cloud cover, temperature. I then come up with a guess to the % charge needed overnight, and set it automatically via MODBUS. If it's CLEAR SKIES and all good - in February it would set 40% charge overnight. If it's overcast, 100%. I'll develop the pattern monthly as we go.

2 |3000

Attachments: Up to 8 attachments (including images) can be used with a maximum of 190.8 MiB each and 286.6 MiB total.

Answer 3 · 2023-02-25T10:56:45Z

johnone answered · Feb 25, 2023 at 10:56 AM

I'm also trying to work out a usable method and have gone down the 'solar power forecasting' route (a google search returns numerous companies and orgs which sell/provide data). forecast.solar seems the most open source and others provide free data. I'm currently using solcast data. Below are forecast and actual yield data for a little 1770Wp free-standing, south facing array in the UK Midlands.

Average error over the period (Jan 14th to Feb 24th) is 9% but max is 120%. The standard deviation is 49. So far, seems like something which promises much but is hard to implement. I guess, given weather is involved, it's never going to be very accurate in the UK?

If I had unlimited time, I'd develop an app using EU PVGIS data (their 15 year or so data set) which provides a daily average yield for your array (location, size, orientation etc.). Surprised nobody has done this - it's freely available data. The data would then be the basis which you apply weather data to. The above power forecasting people say this is what they do but the forecast data they provide is often outside of the physically possible range for a given array location and orientation. Hence, a PVGIS-derived basis seems an attractive starting point.

forecasterror2023janfeb.jpg (212.0 KiB)

solaryield2023janfeb.jpg (253.9 KiB)

2 |3000

Attachments: Up to 8 attachments (including images) can be used with a maximum of 190.8 MiB each and 286.6 MiB total.

question