The oil-paper insulation of high-frequency (HF) transformers frequently withstands high-frequency over-voltages at frequencies of several kHz and above. This leads to the occurrence of partial discharge (PD), which causes premature insulation failure in HF transformers. In order to investigate the effect of PD under high-frequency stresses on oil-impregnated Nomex paper, this paper analyzes the surface morphology, bond-broken types of the molecular chain, and product types formed for oil paper after PD degradation. Then, the damage mechanism of oil-immersed Nomex paper between high-frequency and AC stress is explored. The experimental results show that the branches of creepage do not exist in the oil-paper insulation during the entire discharge process under high-frequency stresses, and that their damage degree is higher than that of AC stress. This is mainly because the benzene ring of oil-impregnated Nomex paper is destroyed and opened caused by high-energy particles, the heating effect in HF discharge, and the bulk effect. These results help to improve the design theory of insulation structures and to develop PD-resistant insulation materials in HF transformers.
When the power converter connected to the high-frequency transformer breaks through the bottleneck and reaches a frequency of 100kHz or even higher, the high-frequency transformer's inter-turn insulation faces more serious high-frequency discharge and high temperature problems. In order to improve the service performance of oil-immersed high-frequency transformer insulation paper, composite K-BNNSs particles are prepared by ultrasonic stripping, heat treatment and thermomagnetic stirring. Then, K-BNNSs particles are mixed with PMIA slurry to produce composite aramid paper. And the effects of K-BNNSs particles with different contents on the thermal conductivity, dielectric properties, partial discharge properties and mechanical properties of aramid paper are explored. It can be found that when the addition of composite particles (K-BNNSs) is 10%, the comprehensive performance of composite aramid paper is the best. Compared with Nomex paper, the in-plane and through-plane thermal conductivity of composite insulating paper F-10 is increased by 668.33% and 760.66% respectively. Besides, the high-frequency breakdown voltage is increased by 48.73% and the tensile strength is increased by 2.49%. The main reason is that the composite particles form a complete thermal conductive network in the aramid paper matrix and a large number of hydrogen bonds with the matrix, which enhances the internal interface bonding force of the material and changes the charge transport mechanism.
9 December 2024 in Measurement Science and Technology
This paper proposes an innovative sensor management method utilizing deep reinforcement learning to optimize the performances of multi-target filters based on random finite sets. Within the theoretical framework of partially observable Markov decision processes, the twin delayed deep deterministic policy gradient reinforcement learning algorithm is employed to decide optimal sensor management strategies in continuous space. For each control command, predicted ideal measurement set is generated to update posterior multi-target density. The divergence between the predicted multi-target posterior density and the updated posterior multi-target density is used as a reward function to guide the decision of sensor management strategies. The simulation results demonstrate that, compared to other algorithms, the proposed method is more effective in enhancing the performance of multi-target tracking.
17 February 2026 in IEEE Transactions on Vehicular Technology
To address the time-varying direction-of-arrival (DOA) tracking problem of incoherently distributed (ID) signal sources for massive multiple-input multiple-output (MIMO) systems, which involves critical challenges of data association ambiguity and measurement uncertainty in dynamic environments, this work proposes a novel tracking algorithm for ID sources using the generalized labeled multi-Bernoulli (GLMB) filter and smoother. The algorithm first models ID sources using a uniform rectangular array (URA), and then estimates the nominal DOA, along with angular spreads, via the beamspace unitary estimating signal parameters via rotational invariance techniques (UESPRIT) algorithm. The above estimates are then used as pseudo-measurements for each time step within the GLMB filter and smoother to improve estimation accuracy. At the same time, for the uncertainty of measurement noise, Cramér-Rao bound (CRB) is used to approximate the covariance of measurement noise in the tracking process.Within the framework of the GLMB smoother, a forward filtering stage is performed to generate the association history as well as the trajectory information using the GLMB filter. Term trajectories and trajectory fragments are removed through the trajectory pruning operation. Next, the backward smoothing is applied to achieve the data association of the ID sources and the estimation of kinematic states. To verify the effectiveness of the proposed approach, Monte Carlo (MC) simulations are performed. The results confirm its superior DOA tracking capability, enhanced estimation accuracy, and high robustness.
5 May 2026 in Engineering Applications of Artificial Intelligence