EAAI Journal 2026 Journal Article
QFTD: An efficient quantum federated learning for transformer fault diagnosis with minimal gated unit in smart grid
- Guodong Li
- Junjie Luo
- Qingle Wang
- Lin Liu
- Chunyan Wei
- Huawei Wang
- Zhichao Zhang
Quantum federated learning (QFL), as an emerging quantum algorithm, has been initially applied in fields such as healthcare and intelligent transportation. It can achieve efficient model training while protecting the privacy of power data. It is an excellent solution for addressing the challenges of data decentralization and privacy in fault diagnosis. To further enhance the generalization and efficiency of QFL, we propose an efficient quantum federated learning algorithm for power transformer fault diagnosis with the minimal gated unit (QFTD) in the smart grid, which is an initial application of QFL to power transformer fault diagnosis in the smart grid. We integrate a quantum orthogonal convolutional neural network with the classical minimal gated unit, forming a quantum minimal gated orthogonal convolutional neural network (QMOCNN) as the local model of QFTD. By adding the quantum orthogonal layer to the quantum convolutional neural network, the generalization ability and stability of the model improve. Experimental results demonstrate that QMOCNN achieves an accuracy of 99. 48% in power transformer fault diagnosis within the smart grid, exhibiting superior convergence speed and classification accuracy compared to other methods. After introducing the federated learning framework, the QFTD can also achieve an accuracy of 95. 83%. The experiment on the three types of quantum circuit noise proves that QFTD has good performance in noise resistance. Our work represents a significant exploration and advancement in applying quantum algorithms within the realm of fault diagnosis in power systems.