Display Publications by [Year] [Type] [Topic]
Advanced power generation technologies such as direct-fired fuel cell turbine hybrid systems are currently being developed in response to the rapidly increasing energy demand. However, lack of accurate system models, nonlinearities, and tight coupling between system parameters render traditional control techniques inadequate in effectively controlling these hybrid configurations. Learning based controllers trained using neuroevolution are currently being developed to address this issue. These controllers have shown great promise as they can deal with noisy sensors and actuators, and reconfigure to handle rapidly fluctuating demands. However, these systems only deal with noise and reconfigure to fluctuating demand in isolation. In real plant operation, a controller has to reliably demonstrate adaptive reconfigurability while dealing with sensor and actuator noise. In this paper, we use neuroevolution to develop neural network based controllers that demonstrate robustness and handle rapid reconfigurability concurrently. We show that our controllers are robust to noise, and concurrently exhibit dynamic reconfigurability to handle rapidly fluctuating demands.
(unavailable)
@inproceedings{tumer-khadka_powid18,
author = {S. Khadka and G. Rockefeller and D. Tucker and P. Pezzini and K. Bryden and K. Tumer},
title = {Neuroevolution of an Advanced Power Plant Controller},
booktitle = {Proceedings of the 2018 International Society of Automation Power Industry Division Symposium},
address = {Knoxville, TN},
bib2html_pubtype = {Workshop/Symposium Papers},
bib2html_rescat = {},
abstract = {Advanced power generation technologies such as direct-fired fuel cell turbine hybrid systems are currently being developed in response to the rapidly increasing energy demand. However, lack of accurate system models, nonlinearities, and tight coupling between system parameters render traditional control techniques inadequate in effectively controlling these hybrid configurations. Learning based controllers trained using neuroevolution are currently being developed to address this issue. These controllers have shown great promise as they can deal with noisy sensors and actuators, and reconfigure to handle rapidly fluctuating demands. However, these systems only deal with noise and reconfigure to fluctuating demand in isolation. In real plant operation, a controller has to reliably demonstrate adaptive reconfigurability while dealing with sensor and actuator noise. In this paper, we use neuroevolution to develop neural network based controllers that demonstrate robustness and handle rapid reconfigurability concurrently. We show that our controllers are robust to noise, and concurrently exhibit dynamic reconfigurability to handle rapidly fluctuating demands.},
month = {June},
year = {2018}
}
Generated by bib2html.pl (written by Patrick Riley ) on Wed Apr 01, 2020 17:39:43