Abstract
Conventional neural networks (NNs), though efficient in rapid dc voltage calculations for medium-voltage direct current (MVDC) distribution systems with diverse converter control schemes, face accuracy challenges with untrained system parameter distributions and topology structures. This paper proposes a physics-embedded convolutional NN (PECNN) to address this issue. The PECNN is enhanced with two additional layers, including a multi-channel combination layer and a physics operation layer. The former channel combination layer strengthens the model feature extraction capability by converting all input channels constituted by initial data matrix into dc voltage, current and line conductance matrix channels. The latter physics operation layer reformulates the combined input channels by physical connections in MVDC systems. The new layers enhance the prediction accuracy and allow generalization of the model. Three MVDC distribution networks with different dc voltage levels and network layouts are used to verify the superiority of proposed PECNN compared to other NNs in different topologies.
Original language | English |
---|---|
Title of host publication | Proceedings of 2024 Ieee 10th International Power Electronics and Motion Control Conference (ipemc2024-ecce Asia) |
Publisher | IEEE |
Publication date | 2024 |
Pages | 2700-2705 |
ISBN (Electronic) | 979-8-3503-5133-0 |
DOIs | |
Publication status | Published - 2024 |
Event | 10th International Power Electronics and Motion Control Conference - Tivoli Chengdu, Chengdu, China Duration: 17 May 2024 → 20 May 2024 |
Conference
Conference | 10th International Power Electronics and Motion Control Conference |
---|---|
Location | Tivoli Chengdu |
Country/Territory | China |
City | Chengdu |
Period | 17/05/2024 → 20/05/2024 |
Keywords
- Medium-voltage dc (MVDC) system
- Dc voltage estimation
- Neural network (NN)
- Convolutional NN (CNN)