Due to the particularity of their process, petrochemical enterprises have high requirements for the reliability of power supply. If a large-scale blackout occurs due to a grounding fault, it will pose a huge threat to safe production. When the resonant grounding system of petrochemical enterprises faults, due to the complex fault process and weak fault signal, it is difficult to accurately detect the faulty feeder by traditional methods. This paper presents a new method of grey correlation degree based on adaptive frequency band. Firstly, the transient zero-sequence current of each feeder is decomposed by coif5 wavelet, and the low frequency band a5 (power frequency component) and high frequency band d1, d2 (noise signal) are removed. By stacking all of the remaining frequency band signals to construct the wavelet area matrix, the faulty feeder detection characteristic scale and the first faulty feeder detection result are obtained. Secondly, based on the faulty feeder detection characteristic scale, the second faulty feeder detection result is obtained by the average grey correlation degree matrix, which detects the faulty feeder according to the waveform correlation degree. Finally, the final faulty feeder detection result is obtained by equal weight voting. In MATLAB/Simulink, the 10 kV resonant grounding system of petrochemical enterprises is modeled. A large number of simulation results show that the faulty feeder detection method is not affected by the initial phase angle (0°, 45° and 90°), transition resistance (10 Ω, 100 Ω and 1000 Ω), fault distance (1 km, 8 km and 15 km) and overcompensation degree (5%, 8% and 10%), and has good sensitivity.
The dielectric loss angle can better reflect the overall insulation level of mining cables, so it is necessary to implement reliable and effective online monitoring of the dielectric loss angle of mining cables. In order to improve the monitoring accuracy of the dielectric loss angle tan δ of mining cables, a low-frequency dielectric loss angle online monitoring method combining signal injection method and double-end synchronous measurement method is proposed in this paper. Firstly, the superiority of the low-frequency signal in improving the detection accuracy of dielectric loss angle is explained, and the feasibility of the low-frequency signal injection method is analyzed. Secondly, the cable leakage is calculated using the double-terminal synchronous measurement method to measure the core current at the first and last ends of the cable, and the phase sum of the voltage at the first and last ends is selected as the reference phase quantity to realize the effective calculation of the dielectric loss angle tan δ of the cable. Then, the simulation model for online monitoring of dielectric loss angle of mining cable is built, and the feasibility of the online monitoring method proposed in this paper is verified by combining the simulation results. Finally, the theoretical and simulation analysis of the monitoring error of dielectric loss angle of mining cable is carried out.
Facial Action Unit (AU) detection is a crucial task for emotion analysis from facial movements. The apparent differences of different subjects sometimes mislead changes brought by AUs, resulting in inaccurate results. However, most of the existing AU detection methods based on deep learning didn't consider the identity information of different subjects. The paper proposes a meta-learning-based cross-subject AU detection model to eliminate the identity-caused differences. Besides, a transformerbased relation learning module is introduced to learn the latent relations of multiple AUs. To be specific, our proposed work is composed of two sub-tasks. The first sub-task is meta-learningbased AU local region representation learning, called MARL, which learns discriminative representation of local AU regions that incorporates the shared information of multiple subjects and eliminates identity-caused differences. The second sub-task uses the local region representation of AU of the first sub-task as input, then adds relationship learning based on the transformer encoder architecture to capture AU relationships. The entire training process is cascaded. Ablation study and visualization show that our MARL can eliminate identity-caused differences, thus obtaining a robust and generalized AU discriminative embedding representation. Our results prove that on the two public datasets BP4D and DISFA, our method is superior to the state-of-the-art technology, and the F1 score is improved by 1.3% and 1.4%, respectively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.