Modular Memory Unit Controllers for a Hybrid Power Plant
Major Professor: Dr. Kagan Tumer
Hybrid Power systems which incorporate multiple energy sources introduce increased complexity, resulting in challenges for system control. In this project, we compare the performance of a memory based neural network architecture - Modular Memory Unit (MMU), with a simple feedforward neural network (FF). We train and test the controllers on tracking target setpoint profiles, with and without sensor noise during training and testing. Our results show that the memory based architecture - MMU is able to perform better on all tested scenarios.
Tuesday, December 10, 2019 at 10:00am
Rogers Hall, 226
2000 SW Monroe Avenue, Corvallis, OR 97331