Electrical network transient classification


When sudden events occur in electricity networks, this causes small signals to echo throughout the network for a few milliseconds. These signals can be caused by faults, such as when two powerlines touch, or harmless events such as reconfiguring the network to spread power more evenly. Being able to identify faults quickly reduces the cost associated with blackouts, and also the risk of fires. It also paves the way for new protection techniques that will be required when the level of rooftop solar increases.


This project will apply machine learning to a database of thousands of these signals to classify them as either “fault” or “non-fault”. To add to the challenge, some signals are caused by “incipient faults”, which are network problems not severe enough to be considered faults, but are warnings of likely faults in the future.


This will expose students to a range of supervised and unsupervised machine learning techniques, as well as the smart grid.


Expected background Familiarity with machine learning


Preferred background: Good mathematical skills and a knowledge of basic statistics would be an advantage.


Supervisors:Lachlan Andrew, with Reza Razaghi (Monash electrical engineering)