INTERNSHIP R&D - Data Mining on Power electronics
Description
Established in 2014, SuperGrid Institute is an independent private company with over 150 employees of 21 different nationalities. The institute is a key player for energy transition in France and throughout Europe, specialising in innovative high and medium voltage direct current (HVDC and MVDC) power systems. SuperGrid Institute also develops new technologies for the integration of renewable energies into the electricity networks of the future.
SuperGrid Institute has been ranked every year since 2018, reaching the podium 3 times in the INPI prize list.
Find out more by visiting our website: www.supergrid-institute.com.
Mission
General context
Power electronics is becoming more and more important with the emergence of renewables energies, new direct current energy transport technologies and electric mobility. The power converter is at the heart of these systems.
The failure of the converter leads to a partial or total loss of functionality of the system, which leads to a loss of income for the operator and the need to organize corrective maintenance in an emergency.
The wish of operators is to move towards a greater share of predictive maintenance, that is to say to anticipate the change of components that will break before the next scheduled maintenance period.
In order to allow this, some electrotechnical systems are already "monitored" (IOT for example) and this data is periodically sent to a cloud.
Objectives / Missions
From a large amount of data from converters in commercial operation, the purpose of the internship is to determine if it is possible to implement a predictive maintenance of these converters by exploiting the data that is returned from the field to the cloud.
The goal is therefore to implement “data mining” on this data.
The following ideas should therefore be explored:
Cleaning and exploitation of the database
Search for correlation between events and failures
Search for sequences leading to breakdowns
Implementation of parameter selection method:
o Fisher method
o Entropy search
o Laplacian score
o Determination of operating modes that resemble each other, criteria of resemblance and separability
In your activity, you will be supervised by a doctor, a university professor and a teacher-researcher, all 3 specialists in this field.
Profil
Candidate profile
Technical Skills :
- Education: Master, engineering school
- Field: Data science and artificial intelligence associated with power electronics
- Software : Python
You intend to work in data science professions; you want to implement your knowledge on real data in a practical way and you want to learn and enrich yourself from the experience of the 3 experts who will supervise your work.
“Soft Skills”:
- Analytical and synthetical mind
- Desire to learn and progress
- Curiosity and creativity
- Teamwork
- Autonomy
- Problem solving
- Relationship and reporting skills
📅 6 months
📍 VILLEURBANNE (LYON)