Low‐frequency noise, generated inherently by the number or mobility fluctuation of carriers, is a crucial concern for the design of analog and digital circuits. Unified modelling based on experimental validation of near‐DC noise in amplifiers is a long‐standing open problem. This article develops a model for low‐frequency noise by deriving new bounds for carrier capturing and releasing. According to the proposed model, a measurement system is suggested that operates in a wide frequency range and even at very low frequencies. The system is noise‐tolerant, since the amplifier is selected based on acceptable noise levels. Among the advantages are the independence from specialized structural noise models for each component and the low cost of the measurement system. The evaluation results show that the proposed method leads to a promising improvement in the low‐frequency noise measuring and is superior to conventional models in the normalized root mean square error indicator. Findings reveal that the proposed measurement method can estimate the flicker noise around the DC frequency, and the proposed model agrees reasonably with the proposed measurement circuit.

A customizable measurement system for a wide frequency ranges from 0.1 Hz. A roadmap to select an appropriate Op‐amp (low noise amplifier) is based on device under test output resistance. The low‐frequency noise (LFN) modelling is developed by deriving new bounds for carrier capturing and releasing. The LFN model agrees reasonably with the proposed measurement circuit.image

This paper proposes a feature extraction method combining adaptive variational mode decomposition (AVMD) and singular value decomposition (SVD) for electric shock fault‐type identification. The AVMD algorithm is utilized to adaptively decompose the electric shock signal into intrinsic mode components, each containing distinct frequency information. Subsequently, the correlation coefficient is employed to extract the intrinsic mode component with amplitudes greater than or equal to 0.1 ( ≥ 0.1). Feature extraction is then performed using SVD on the ≥ 0.1 intrinsic mode component, based on its maximum singular value and singular entropy. This approach effectively overcomes the limitation of the traditional VMD that necessitates manual K value setting. Moreover, it achieves dimensionality reduction and feature extraction of the intrinsic mode components through SVD, resulting in enhanced computational efficiency and fault identification accuracy. Extensive simulations demonstrate the remarkable recognition rates of electric shock fault types in animals and plants using the proposed AVMD‐SVD method, achieving a recognition rate as high as 99.25%. Comparative performance analysis further verifies the superiority of AVMD‐SVD over similar empirical mode decomposition‐SVD feature extraction techniques.

This paper proposes a feature extraction method combining adaptive variational mode decomposition (AVMD) and singular value decomposition for electric shock fault‐type identification. The number of mode components (K) in VMD is adaptively determined through singular entropy relative increment. The optimal AVMD modal components are selected through the correlation coefficient, constructing the Hankel matrix and extracting the maximum singular value and singular entropy from the Hankel matrix as characteristic phasors for animal and plant electric shock.image

In order to monitor the state of bushing online, an intelligent monitoring system for transformer bushing was developed. A four‐in‐one sensor integrating hydrogen sensing technology using palladium nickel alloy, pressure sensing technology, wide range temperature sensing, and micro water measurement technology was developed. A three‐in‐one integrated sensor based on micro current detection technology was developed to realize online monitoring of bushing dielectric loss, capacitance, and partial discharge. The test results show the hydrogen measurement range of sensor is 0 to 10,000 μL/L, and the measurement uncertainty is lower than 10% or 10 μL/L. The pressure measurement range is 0 to 1.0 MPa, and the uncertainty is lower than 0.3%. The temperature measurement range is −40°C to 85°C, and the uncertainty is lower than ± 1°C. The micro water measurement range is 0 to 1000 μL/L, and the measurement uncertainty is lower than ± 5% or 10 μL/L. The dielectric loss and capacitance error increased by one order of magnitude compared to current standards. The resolution of partial discharge is 5 pC. The performance of the device fully satisfies the requirements for online monitoring of transformer bushing. It has been installed in dozens of 330 and 750 kV substations, providing a reliable guarantee for safe operation of transformer bushing.

A four‐in‐one sensor integrating hydrogen sensing technology using palladium nickel alloy, pressure sensing technology, wide range temperature sensing, and micro water measurement technology was developed. A three‐in‐one integrated sensor based on micro current detection technology was developed to realize online monitoring of bushing physical, chemical, and electrical parameters.image

A large number of monitoring sensors are introduced in the power grid. However, the traditional trust models commonly used for edge‐side security management are weak in detecting large‐scale malicious interactions and collusion attacks. For that, a lightweight and anti‐collusion trust model combined with nodes’ dynamic relevance for the power Internet of Things (IoT) is proposed. Firstly, a global trust management system is constructed according to the working mechanism of sensors in the power grid. After that, trust feedback and contact frequency of the devices are combined to build an adaptive dynamic weight vector based on relevance volatility. Fluctuations in trust values are reduced and the trust difference between normal and malicious nodes is widened. An anti‐collusion algorithm based on contact set awareness is also designed to effectively detect collusion attacks. The checksum local broadcast is established in the trust model to counteract the risk of intelligent terminal failure. The results show that the trust model achieves 100% accuracy of node discrimination when the maximum proportion of malicious nodes is 20% in a 50‐node network scale. In addition, the calculation time of the overall model is 211 ms and the memory consumption is 161 kb, which is suitable for power IoT sensor networks.

Currently, a large number of monitoring sensors are introduced into the power grid. However, the traditional trust models commonly used for edge‐side security management are weak in detecting large‐scale malicious interactions and collusion attacks. For that, a lightweight and anti‐collusion trust model combined with nodes’ dynamic relevance for the power IoT is proposed.image