[Energy Solutions] Insulation Intelligence in the AI era

Blogs

June 9, 2026

Tl;DR

The kind of simulation that people dont ask before commiting 300k+

Energy Solutions↗ building-to-blender↗

Intro

How confy is going to be a flat that just exists over a conceptual drawing?

Coming from this post with a solar experiment

git clone https://github.com/JAlcocerT/RPi

The Sun is interesting

With all that push of energy to us via radiation

What about blender?

I made some simulations recently on how sun rays are hitting our buildings:

sudo snap install blender --classic --channel=5.1/stable
#cd ./poc/building-to-blender
#make help

Then, this happened: https://github.com/JAlcocerT/poc/tree/main/building-to-blender

IoT, Walls, Sun and Heat Transfer

Previoulsy I thought that Linux was tricky compared to W11 to push code towards microcontrollers.

It seems that we dont even need arduino GUI, and a proper data cable + arduino CLI does the job:

cd ./poc/iot-rpi-dht
arduino-cli board list #make serial-list

Fix serial access for this session:

sudo setfacl -m u:$USER:rw /dev/ttyACM0

Edit WiFi in: iot-rpi-dht/scripts-microcontrollers/firmware-esp32/esp32-dht11-mqtt-emqx-deepsleep.cpp

const char* WIFI_SSID     = "your-wifi";
const char* WIFI_PASSWORD = "your-password";
const char* MQTT_BROKER   = "192.168.1.14";

Upload:

cd /home/jalcocert/Desktop/poc/iot-rpi-dht/scripts-arduino-setup
./upload-deepsleep.sh /dev/ttyACM0

EMQX is already running on this laptop at 192.168.1.14:1883.

EMQX Docker Config 🐋 ↗

So you just need the storage layer:

#make sqlite-count
# cd ./poc/iot-rpi-dht/scripts-microcontrollers/backend-node-sqlite
sqlite3 data/readings.sqlite "select * from readings order by received_ms desc limit 10;"
2|esp32/humidity/dht11|humidity|45.0|2026-06-04T14:42:47.351Z|1780584167351
1|esp32/temperature/dht11|temperature|25.3|2026-06-04T14:42:47.251Z|1780584167251

I made it better: because hardcoding wifi/pwd is not nice

#make serial-sketch
make serial-fix PORT=/dev/ttyACM0
make flash

See that the ESP32 is having proper logs while still connected and that emqx is getting the messages:

make serial-log PORT=/dev/ttyACM0 #see whats flowing
#http://192.168.1.2:18083/#/
make emqx-remote #admin/public.18083 #this one can work not only with the local emqx as the previous ones
make mqtt-listen MQTT_HOST=192.168.1.2 MQTT_TOPIC='esp32/#

Thensimilarly for the picow DHT22:

make serial-fix PORT=/dev/ttyACM0
make flash-picow

make mqtt-listen MQTT_HOST=192.168.1.2 MQTT_TOPIC='pico/#'

So with both pushing data:

#git clone /poc
cd ./poc/iot-rpi-dht/

We can have this view:

alt text

And improve it with these:

alt text

make mqtt-listen MQTT_HOST=192.168.1.2 MQTT_TOPIC='pico/#'
#  make arduino-compile
# make serial-fix PORT=/dev/ttyACM0
# make flash # make flash PORT=/dev/ttyACM0
make mqtt-listen MQTT_HOST=192.168.1.2 MQTT_TOPIC='esp32/#'

alt text

Running this at the homelab for a few days: yea, there is a cool UI

cd ./poc/iot-rpi-dht-insulation

alt text

⚠️

The ESP32 deepsleep mode and pushing every 60s instead of 5s made the same battery to last instead of 12h/24h

We need the MLX

The DHT’s are great, but they only measure ambient temperature.

We need the MLX90614 to measure the surface temperature of the wall, because that is what will determine the heat transfer from the wall to the inside.

Of course there is a lot of work on this topic, as you may guess:

https://github.com/JAlcocerT/poc/tree/main/iot-rpi-mlx

cd ./poc/iot-rpi-mlx/

im levereging the public scripts for this sensor here

And the final dashboard

https://github.com/JAlcocerT/poc/tree/main/iot-rpi-mlxdht-insulation

The one that combines all the goodies of previous steps, with the UI/X ive polished:

cd ./poc/iot-rpi-dht-insulation
#docker compose up -d --build dashboard
MQTT_HOST=192.168.1.2 docker compose up -d --build dashboard #so that emqx at homelab is reachable to sqlite and webapp

Run this any time to check: sqlite3 ingester/data/readings.sqlite
“SELECT device, metric, COUNT(*), ROUND(AVG(value),1) FROM readings GROUP BY device, metric;” Or see the latest readings: sqlite3 ingester/data/readings.sqlite
“SELECT device, metric, value, received_at FROM readings ORDER BY received_ms DESC LIMIT 10;”

alt text

I also tried it from the homelab: MQTT_HOST=192.168.1.2 docker compose up -d --build

Conclusions

Measure, dont model:

FAQ

Solar vs Wind

The cost of a small wind turbine for a home ranges from $3,000 to $8,000 per kilowatt installed.

In contrast, an equivalent 1 kW solar panel system would cost about $2,770 on average in the U.S. (SRT 00:07:46 - 00:08:04).

Source: https://www.youtube.com/watch?v=nPvTH7Siclg Channel: Undecided with Matt Ferrell

Compliant mechanisms and MultiBodySystemsDynamics dot com