Non-Intrusive Load Monitoring for Home Assistant

What you need

• A working Home Assistant OS or Home Assistant Container installation.
• A mains power sensor already available in Home Assistant.
• Recorder history for the time range you want to use for training.
• At least 4 GB of RAM for Home Assistant and the NILM apps.

What to expect

Direct metering and NILM

Home Assistant OS quick start

1. Open Home Assistant.
2. Go to Settings, then Apps, then Install App.
3. Add this repository: https://github.com/lgarciamarrero92/ha-nilm
4. Install the NILM app.
5. Choose a training server option: install the NILM Training Server app, or run the ha-nilm-trainer Docker container on a reachable Docker host.
6. Start the training server first.
7. Start NILM.
8. Open the NILM interface.
9. Select and save your mains power sensor.

Training server selection

1. Open the Training interface.
2. Use the autodetected internal training server if you installed the Home Assistant OS Training Server app.
3. Choose Custom External Server if the trainer runs outside Home Assistant.
4. For an external trainer, save the URL in this form: http://<host-or-ip>:<port>/train
5. Confirm the training server is ready before preparing a job.

Home Assistant Container quick start

1. Create a long-lived access token from your Home Assistant user profile under Security.
2. Create a folder for the NILM Docker Compose stack.
3. Add the .env file and docker-compose.yml shown below.
4. Start the stack with docker compose up -d.
5. Open http://<docker-host>:8099/
6. Select and save your mains power sensor in NILM Dashboard.
7. Open the Training interface, choose Custom External Server, and save http://nilm-trainer:8080/train

.env for Home Assistant Container

HA_TOKEN=replace_with_your_home_assistant_long_lived_access_token
HA_REST_API_URL=http://YOUR_HOME_ASSISTANT_HOST_IP:8123/api
HA_WS_URL=ws://YOUR_HOME_ASSISTANT_HOST_IP:8123/api/websocket

docker-compose.yml

services:
  nilm-trainer:
    image: ghcr.io/lgarciamarrero92/ha-nilm-trainer:latest
    container_name: nilm-trainer
    restart: unless-stopped
    ports:
      - "8024:8080"
    networks:
      - nilm

  nilm-edge:
    image: ghcr.io/lgarciamarrero92/ha-nilm-edge:latest
    container_name: nilm-edge
    restart: unless-stopped
    ports:
      - "8099:8099"
    volumes:
      - nilm_data:/data
    environment:
      SUPERVISOR_TOKEN: "${HA_TOKEN}"
      HA_REST_API_URL: "${HA_REST_API_URL}"
      HA_WS_URL: "${HA_WS_URL}"
    networks:
      - nilm

volumes:
  nilm_data:

networks:
  nilm:
    driver: bridge

Service roles

Confirm the training server

1. Use the internal Home Assistant OS training server if it was autodetected.
2. Use Custom External Server for Docker-based trainers and save the trainer URL.
3. Confirm the training server is ready before selecting data or sending a job.

Choose an annotation method and the appliance name

Choose the mains interval

1. Load the mains history range that will be used for training.
2. Select a range you can annotate completely, not partially.
3. The training range is limited to the previous 7 days to avoid very large training datasets.

Define labels

1. Manual interval annotation: draw ON intervals for the appliance in the selected mains range.
2. Sensor annotation: select the appliance power sensor used as ground truth.

Prepare and send

1. First review the Training Overview section and confirm appliance name, supervision mode, labels, and mains range.
2. Click the button: Prepare and Send to Training Server.
3. NILM validates the configuration, prepares the package, creates the local job, and sends it to NILM Training Server.
4. Follow progress in the Training Jobs table until the model is done.

Main dashboard elements

What the dashboard is good for

• Saving the mains power sensor used by NILM.
• Inspecting the historical mains signal.
• Running offline disaggregation for one appliance or for all available models.
• Debugging predicted power, ON/OFF state.

Appliance share diagram

• NILM also plots an appliance share diagram to help you understand how energy is distributed across the currently plotted predictions.
• The diagram uses only the predictions that are currently visible on the chart.
• A dedicated Base Load share is also shown; it represents the estimated background portion of mains consumption.
• Other is the remaining mains portion not explained by the visible appliance predictions plus Base Load.
• The diagram updates automatically whenever predictions are added or removed.

Live entities

• sensor.nilm_<appliance>_power
• sensor.nilm_<appliance>_energy_consumed
• binary_sensor.nilm_<appliance>_on

What they mean

• The power sensor is the estimated appliance power at runtime.
• The energy consumed sensor is the cumulative estimated appliance energy for dashboards and energy workflows.
• The binary sensor is the estimated appliance ON/OFF state at runtime.

Appliance types that are usually harder

• Very low-power appliances such as lighting, phone chargers, and small standby loads.
• Appliances that stay ON for long periods with flat consumption, such as some TVs or computers.
• Very short-cycle appliances or events that are brief and easy to hide inside aggregate mains noise.
• Multiple similar appliances running at the same time, which can produce overlapping signatures.

Where this method is usually stronger

• The current method was trained with representative, but not exhaustive, appliance types such as fridge, dishwasher, washing machine, kettle, and microwave.
• Because of that training exposure, performance is often better for appliance signatures similar to those categories.