[ECA1] Improving learning methods for hidden markov models for the recognition of generic consumption patterns | BEAMS

[ECA1] Improving learning methods for hidden markov models for the recognition of generic consumption patterns

Project information
Project type: 
Master thesis
Academic year: 
2017-2018
Status: 
Proposal
Research unit: 
Electrical Energy
BEAMS supervisors
Academic promoter
Supervisor
Promoter: Prof. Jean-Claude Maun
Supervisor: Jacobs Gilles (contact: gjacobs [at] ulb [dot] ac [dot] be)
 
 
1 Context
People want to know more about their energy consumption and especially their electricity consumption.
One solution to achieve this goal is to instrument totally a house. This solution costs a lot and is very intrusive for people living in the house.
The ideal solution is to put a unique device at the power meter location and recover detailed information about electrical consumption. This is the NIALM concept (Non-Intrusive Appliance Load Monitoring). Of course this approach requires more computational complexity than the intrusive solution with sensors.
The ECA project (Energy Consumption Advisor) initiated by the BEAMS service attempts to implement the NIALM concept. Hence, the idea consists in measuring the voltage and currents at the global connection point and, by disaggregation algorithms, retrieve relevant information contained in the global signals.
The system is implemented by You Know Watt, a ULB spin-off. It is currently evolving with new algorithms, hardware implementation, data,… This master thesis will contribute to the development of this spin-off.
 
2 Objectives
In that general context, targets of this master thesis are:
  • To understand the learning methods linked to the hidden markov model approach
  • To implement learning algorithms with prior generic data
  • To test and quantify the results given by the implemented learning algorithm

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer