| KEYWORD |
Non-Intrusive Load Monitoring for Energy Load Disaggregation
Thesis in external company
keywords ICT FOR ENERGY, MACHINE LEARNING
Reference persons MICHELA MEO
External reference persons Maurizio Fantino, Fondazione Links
Research Groups Telecommunication Networks Group
Description As defined by Wikipedia and many research articles, Non-Intrusive Load Monitoring (NILM) or Appliance Recognition software are algorithms that detect changes in the electrical values (power, current, voltage) going into a building to infer what appliances are used in the building as well as their individual energy consumption. The key advantage of NILM algorithms is the ability to infer and breakdown energy consumption with one meter rather than clipping current transformers to each circuit to monitor. This aims at reducing the installation complexity and cost. Nevertheless, Commercial and industrial facilities have many dynamic loads with a different mode of operations that makes almost impossible to capture signatures.
The objective of this thesis consists in the study of NILM applications for Industrial Applications starting from a deep review of the current state of the art and with the implementation of a machine learning algorithms useful for disaggregating industrial loads. The proposed algorithms will be trained using open and/or synthetic dataset. The candidate will have both the task of studying and evaluating the best algorithms to apply for the case of study.
The candidate is required to implement machine learning algorithms, with the exploratory possibility of deep learning algorithms using popular frameworks (TensorFlow, PyTorch, Keras, etc..).
Main foreseen activities
• State of the Art Analysis of NILM algorithms with particular focus on Industrial Applications
• Setting up a simple data acquisition board based on the ESP32+RaspberryPi (or similar) also capable to do some harmonic analysis, see:
o https://savjee.be/2019/07/Home-Energy-Monitor-ESP32-CT-Sensor-Emonlib/
o https://github.com/debsahu/ESP32_FFT_Audio_LEDs
o https://github.com/s-marley/ESP32_FFT_VU
• Design and Implementation of a ML-based NILM algorithm based on the SoTA analysis done at the beginning of the Thesis
• Test and Validation of the acquisition and algorithm developed
See also https://linksfoundation.com/lavora-con-noi/proposte-di-tesi/
Required skills • Knowledge of Python
• Software development skills
• Basic concepts on data science, signal processing, and machine learning, supervised and unsupervised learning
Notes Contact: send a resume with attached the list of exams to piero.macaluso@linksfoundation.com and hamidreza.mirtaheri@linksfoundation.com specifying the thesis code and title.
Deadline 25/01/2022
PROPONI LA TUA CANDIDATURA