Partial discharge (PD) detection is critical for diagnosing and maintaining insulation systems of high-voltage equipment. In this article, a novel broadband three-directional (3-D) omnidirectional magnetic induction-based sensor module is proposed for noncontacting PD detection of high-voltage equipment in the industrial Internet of Things (IIoT) application. The equivalent circuit models of a bandpass filter and planar spiral coil are comparatively investigated. Both broadband and filtering performance are realized by combining multiple spiral coils associated with the lumped capacitors. Moreover, five coils are wrapped around the dielectric cube to construct a 3-D omnidirectional sensor. Benefiting from the subcoil interactions, the magnetic field vector is redistributed to produce a transverse component that can detect the fields parallel to the coil’s plane, thus contributing to omnidirectional sensing performance. To verify the proposed measurement technique, an IIoT-based real-time monitoring system prototype is manufactured, and the designed sensor is fabricated by using 3-D printing technology. The experiments show that the designed sensor module can achieve an ultrawide impedance bandwidth ranging from 4 to 70 MHz (|$S_{11} {\unicode {0x007C}}$¡ −10 dB) while the out-of-band interference can be suppressed. Besides, the proposed system has been demonstrated for 3-D omnidirectional real-time monitoring and intelligent analysis for PD events, showcasing its significant potential in IIOT applications.

This paper presents a hardware-software co-designed system for real-time online monitoring of Partial Discharge (PD) in the Power Grid without human interaction. PD is a critical indicator of insulation degradation, which can lead to premature failure of components and disrupt reliable electric supply. The proposed system utilizes an SoC-based Edge Computing Unit for long-term monitoring activities. It incorporates a wavelet denoising module, an auto management program, and an AI-based detection model to enable automatic PD alarm generation. The system is tested and iterated in a controlled lab environment, capturing standard PD signals. The AI classification model is trained using manually labeled PRPD pattern datasets. Subsequently, the system is deployed in a real power grid for long-term, real-time PD monitoring. The effectiveness of the proposed system is demonstrated through experimental setups and result analysis. The SoC-based real-time online PD monitoring system offers a proactive approach to safeguarding power equipment by detecting and addressing insulation degradation in the Power Grid.

Partial discharge (PD) is an essential indication of insulation degradation in high-voltage power equipment like gas-insulated switchgears (GIS). However, in certain applications that are exposed to intense external noise, traditional PD detection methods often encounter numerous challenges due to their vulnerability to noise and interference. This paper proposes a low-noise ultra-high frequency (UHF) sensing system for PD detection and classification. In analog front end, the noise performance is optimized using a broadband noise-shaping network (BNSN) integrated with a wideband printed monopole antenna (PMA). In digital back end, the combination of noise cancellation, wavelet time scattering (WTS) based features extraction and a support vector machine (SVM), yields 95% correct classification. Simulation and Comparative experimental results validate superior noise performance, effectiveness and accuracy of this UHF sensing system.

State estimation with sensors is essential for mobile robots. Due to different performance of sensors in different environments, how to fuse measurements of various sensors is a problem. In this paper, we propose a tightly coupled multi-sensor fusion framework, Lvio-Fusion, which fuses stereo camera, Lidar, IMU, and GPS based on the graph optimization. Especially for urban traffic scenes, we introduce a segmented global pose graph optimization with GPS and loop-closure, which can eliminate accumulated drifts. Additionally, we creatively use a actor-critic method in reinforcement learning to adaptively adjust sensors’ weight. After training, actor-critic agent can provide the system better and dynamic sensors’ weight. We evaluate the performance of our system on public datasets and compare it with other state-of-the-art methods, which shows that the proposed method achieves high estimation accuracy and robustness to various environments. And our implementations are open source and highly scalable.