Skip to content

Latest commit

 

History

History
101 lines (78 loc) · 4.44 KB

piNILM.md

File metadata and controls

101 lines (78 loc) · 4.44 KB
Raw

The piNILM Project

Copyright (c) 2016 by Stephen Makonin

Last modified: May 3, 2016.

This project allows SparseNILM to run on a Raspberry Pi 2B+ and 3 and disaggregate your smart meter reading using a Rainforest EMU2 in real-time.

RPi Setup & Config

I recommend flashing the SD card with Raspbian JESSIE LITE. Here are the steps to get SparseNILM running on your RPi:

  1. Once you boot the RPi for the first time run sudo raspi-config and reboot.
  2. Setup wifi using sudo wpa_cli and use the following commands:
    • list_networks
    • add_network
    • set_network 0 ssid "<replace with the ssid>"
    • set_network 0 psk "<replace with the psk>"
    • enable_network 0
    • reconnect
    • list_networks
    • status
    • save_config
    • quit
  3. Apply general OS updates:
    • sudo apt update
    • sudo apt upgrade
    • sudo apt install vim (optional)
  4. SparseNILM uses Python 3.x, so remove Python 2.7 and install 3.x and needed modules/libraries:
    • sudo apt remove python2.7
    • sudo apt-get autoremove
    • sudo apt install python3 python3-rpi.gpio python3-serial python3-pip python3-pandas
  5. Finally, we need to setup git to download SparseNILM from GitHub:
    • sudo apt install git
    • git config --global user.name "<insert your name here>"
    • git config --global user.email [email protected]
    • git clone https://github.com/smakonin/SparseNILM.git
  6. Optionally if you have a GitHub account run:
    • ssh-keygen -t rsa -b 4096 -C "[email protected]”
    • THEN copy and paste the key into GitHub account settings

Building a Model for SparseNILM

First you must build a model using the train_SSHMM.py. Create a dataset file in `~/SparseNILM/datasets’ with the prefix “BCH”; (e.g., BCH_datafile1.csv). The dataset file should be comma seperated value or CSV with he first line having header info. For example:

TimeStamp,Mains,Load1,Load2,…
1333263600,60.7,0,0.7,…
1333263660,60.7,0,0.9,…
1333263720,60.6,0,0.9,…
1333263780,44.7,0,0.9,0.1,39.1,0,0,0,0,0.4,1.1,1.3,0.5,0.2,0,0.2,0,0.5,0.1,0.1

The columns named Load1 and Load2 can be names of your choosing — to identify/name each load. Do not use spaces in the names.

For creating a dataset, it is recommented that you create the CSV file on your local computer and then secure copy that file to the RPi. For exmaple, scp BCH_datafile1.csv [email protected]:./SparseNILM/datasets/

Here is an example of how to create a model:

cd ~/SparseNILM
./train_SSHMM.py bch1 BCH_datafile1 1 24000 denoised 8 1 l1+l2,l3

In the above example we create a model named bch1 from the dataset called BCH_datafile1 with a precision of 1, max observed value of 24000W, only using the loads we specify, each loads having a max of 8 states with only 1 testing fold. We are building a model that disaggregates 2 appliances. We combine the l1 and l2 columns in the dataset because it is the same appliance that run on both L1 and L2. Appliance l3 is only on one leg. We use a single fold because we are not testing against another dataset. We are building a model to disaggregate using live data from the smart meter via an EMU2.

Here is a description of command-line parameters for train_SSHMM.py:

USAGE: ./train_SSHMM.py [modeldb] [dataset] [precision] [max obs] [denoised] [max states] [folds] [ids]

       [modeldb]    - file name of model (omit file ext).
       [dataset]    - file name of dataset to use (omit file ext).
       [precision]  - number; e.g. 10 would convert A to dA.
       [max obs]    - The max observed value, e.g. 200.1 A.
       [denoised]   - denoised aggregate reads, else noisy.
       [max states] - max number of states a each load can have.
       [folds]      - number usually set to 10, 1 means data same for train/test.
       [ids]        - e.g. CDE,FGE(,...) -- case sensitive!

Running SparseNILM Live

Once a model is build we run SparseNILM live. For example:

cd ~/SparseNILM
./disagg_EMU2.py bch1 1 W SparseViterbi /dev/ttyACM0

Here is a description of command-line parameters for disagg_EMU2.py:

USAGE: ./disagg_EMU2.py [modeldb] [precision] [measure] [algo name] [device]

       [modeldb]       - file name of model (omit file ext).
       [precision]     - number; e.g. 10 would convert A to dA.
       [measure]       - the measurement, e.g. A for current
       [algo name]     - specifiy the disaggregation algorithm to use.
       [device]        - usually /dev/ttyACM0 in a RPi.