Location
Hilton Waikoloa Village, Hawaii
Event Website
http://hicss.hawaii.edu/
Start Date
1-3-2018
End Date
1-6-2018
Description
In this paper, learning algorithms are leveraged to advance power system protection. Advancements in power system protection have come in different forms such as the development of new control strategies and the introduction of a new system architecture such as a microgrid. In this paper, we propose two learning schemes to make accurate predictions and optimal decisions related to power system protection and microgrid control. First, we present a neural network approach to learn a classifier that can predict stable reconnection timings for an islanded sub-network. Second, we present a learning-based control scheme for power system protection based on the policy rollout. In the proposed scheme, we incorporate online simulation using the commercial PSS/e simulator. Optimal decisions are obtained in real time to prevent cascading failures as well as maximize the load served. We validate our methods with the dynamics simulator and test cases RTS-96 and Poland.
Learning Schemes for Power System Protection
Hilton Waikoloa Village, Hawaii
In this paper, learning algorithms are leveraged to advance power system protection. Advancements in power system protection have come in different forms such as the development of new control strategies and the introduction of a new system architecture such as a microgrid. In this paper, we propose two learning schemes to make accurate predictions and optimal decisions related to power system protection and microgrid control. First, we present a neural network approach to learn a classifier that can predict stable reconnection timings for an islanded sub-network. Second, we present a learning-based control scheme for power system protection based on the policy rollout. In the proposed scheme, we incorporate online simulation using the commercial PSS/e simulator. Optimal decisions are obtained in real time to prevent cascading failures as well as maximize the load served. We validate our methods with the dynamics simulator and test cases RTS-96 and Poland.
https://aisel.aisnet.org/hicss-51/es/monitoring/6