This article is concerned with developing a privacy-preserving distributed-learning-based fault diagnosis approach for power transformers. Due to the constraints of data privacy, it is not possible to have enough labeled samples for training. Recently, the emergence of federated learning (FL) has provided a secure and distributed learning framework. However, the unbalanced data from multiple power stations may reduce the overall performance of FL while an untrusted central server can threaten the data privacy and security of clients. To address such challenges, a privacy-preserving FL scheme is developed for transformer fault diagnosis, where a multistep data-sharing strategy and an adaptive differential privacy technology are proposed. Specifically, amounts of shared data and noise perturbation will be designed according to the quantity of local data by the central server. The experimental results on the dataset generated according to IEC publication 60599 show that the proposed method has high diagnostic accuracy across various categories of transformer faults and even on training datasets with extremely unbalanced data quantity where the average accuracy is as high as 95.28%.

Low-cost inertial measurement units (IMUs) suffer from low sensitivity and high random walk noise, which makes it challenging to use them directly for dead reckoning. Regular model-based inertial navigation methods require accurate modeling of IMU noise to get better results, while learning-based methods need plentiful datasets. In this article, a novel low-cost IMU dead-reckoning algorithm for wheeled mobile robot is introduced by integrating model-based and learning-based approaches, which inherits the merits of both methods. It achieves the dead reckoning by using invariant extended Kalman filter (InEKF) and IMU error model, and computes the noise parameters of the model with the aid of a deep-learning-based method. Our deep-learning-based strategy is designed to obtain noise-reduced inertial information of robots from low-cost IMU data such that the InEKF can converge. The experimental results show that the proposed method can accurately estimate the attitude, velocity, and position of the wheeled mobile robot, and can compete with vision algorithms. In addition, our proposed method consumes few computational resources to satisfy the needs of low-cost applications.

This article is concerned with modeling of a class of permanent-magnet-synchronous-motor-based drive systems with dead-zone input nonlinearity. The experimental testing method based on recursive least squares (RLS) is combined with the mechanism analysis method based on physical laws. With the structure of the analytical model first determined by the mechanism analysis method, an online RLS-based data-driven method is then exploited to obtain the unknown parameters of the analytical model. Since the unknown parameters are decoupled in the proposed method, all parameter estimates are available. An auxiliary model and a variable forgetting factor are constructed to ensure the high accuracy and the fast convergence of the proposed method, respectively. Besides, the asymptotic convergence property of the proposed method is analyzed, and the upper bound of the parameter estimation error is also presented. The effectiveness of the proposed method is verified by simulations and experiments on a self-designed computer numerical control engraving system.

Global Navigation Satellite System Real-Time Kinematic (GNSS-RTK) serves as a vital tool for providing absolute positioning for autonomous systems. However, its performance suffers considerable degradation in urban canyon environments due to the well-known challenges caused by multipath effects and Non-Line-of-Sight (NLOS). Light Detection and Ranging (LiDAR)/Inertial Odometry (LIO) offers high-precision local pose estimation in structured urban settings, but it tends to accumulate drift over time. Recognizing their complementary strengths, this paper proposes an adaptive integration of GNSS-RTK with LIO to achieve continuous and precise global positioning for autonomous systems in urban environments. The raw data is modeled and optimized within the framework of a factor graph. At the same time, double difference carrier phase and ambiguity are added to the estimated states. Finally, RTK-LIO is evaluated on public datasets. It greatly exceeds the benchmarks (LIGO, GLIO and RTK) in both accuracy and smoothness. To benefit the community, the implementation is open-sourced at https://gitee.com/bryantaoli/rtk-lio.

Recently, data-driven methods have demonstrated excellent performance in diagnosing faults in power transformers. However, concerns about data privacy and computational resources arise due to the requirement of collecting and training large numbers of labeled samples in a centralized manner. Although centralized federated learning (CFL) provides optimal solutions, the dependence on central servers engenders issues associated with reliability and increased communication expenditures. Meanwhile, existing decentralized federated learning (DFL) frameworks are not specifically designed for industrial applications with unbalanced data and high communication costs. To tackle these issues, this paper proposes a novel parallel sequential DFL (PS-DFL) framework that exchanges information directly between client pairs, thus overcoming the limitation caused by the central server in CFL. In particular, by combining parallel training with sequential communication techniques, the impact of unbalanced data on model performance is mitigated, as well as communication costs are reduced. The experiment conducted on IEC TC 10 and other public datasets demonstrate that, despite handling tasks involving unbalanced data, the diagnostic accuracy is maintained at 97.09%. Furthermore, the communication overhead is significantly lower than that of existing methods, thereby proving its feasibility for industrial deployment within the power grid.