Faric Kadri's suggestions are good. At third year level, you need to show you understand an advanced topic. It need not be completely new (leave that for the PhD students), however, a little bit of newness will earn you some credit.
In power electronics this can be topics such as: Space Vector Control of an Induction Machine, DC machine control with H-bridge drive and Unipolar Switching, Bidirectional Buck-boost converter etc.
In order to determine where one stands with respect to newness, a god measure of recent publications on the respective topics needs to be read. Reading makes up a large part of research, in general.
Adding a specific suggestion to what Mr. Kadri and Mr. Johnson stated, work out something related to "Wide Area Monitoring Systems based on Synchrophasor Measurement"
hey now if you are interested in power system cum energy then you shift in renewable energy with power system such as solar power generation ,wind power generation because this is future of electrical power system so I suggested you that you are in 3rd year it is a nice time to switch in electrical cum renewable energy ....
Power electronics is one of the leading research fields in power engineering. If you can combine this with renewable energy it's a winner in my opinion. But still, what is even more important is what YOU like. If you don't like a subject it won't matter if it's the best in the world, you will not be good at it. So, you have to decide for yourself in the end...
Much of the effort from the answers above involves embedded control of the equipment. Embedded controllers by their very nature present a different form of instability that is based on unexamined libraries to construct the controllers. There is the deliberate attempt to suborn the system and the accidental or incidental occurrence. How about using one of these ideas that have been given to you but then begin a fault tree analysis of the possible of incursions into the reliability of the controllers? This isn't really my field, but you might find that the increased dependence on Industrial Control Systems (ICS) presents a substantial reliability issue. To use the controllers requiring some network connectivity to provide nominally improved load sharing amongst disparate resources could provide a reliability trap. Not really my field, but that's the immediate answer of an outsider.