Magnetic components are essential parts in many power electronic applications. Their characteristics deeply impact the performances of the applications. This article proposed a 2-D calculation method for frequency-dependent winding losses and leakage inductance of magnetic components of round conductors. The method does not have any limitations on the winding arrangement and considers the impact of magnetic cores and air gaps. The method is compared with several analytical methods and the 2-D finite-element method (FEM). Measurements and 3-D FEM are also used to validate the method. The results show that the proposed method generally has more than ten times shorter computational time than 2-D FEM and comparable accuracy, which can speed up the magnetic component design.@en

Winding loss calculation is essential for inductor and transformer design. In this article, a revised 1-D Ferreira's formula is proposed, which considers the interaction between conductors. Then, a 2-D loss calculation approach is proposed based on the analytical solution of round conductors under a uniform external field. An equivalent external magnetic field is calculated to estimate the winding losses, considering the impact of eddy current. The proposed approach is compared with the 2-D FEM with three types of windings and shows good accuracy with less than 10% error. 3-D FEMs and samples are built based on two simulated windings to validate the loss calculation.@en

In this article, an approach combining semi-empirical equations and the method of images is proposed for round conductor layer windings with un-gapped core. The new equation for proximity effect can convert the constant field strength from the magneto-motive force (MMF) across the core window into a frequency-dependent uniform background magnetic field strength, which can take partly the interaction between conductors into account. Geometric factors are introduced by fitting the finite element method (FEM) results to improve the accuracy. The method of images is used to calculate the field strength in order to counteract the impact of the 2-D edge effect. The new method is compared with the 2-D FEM, analytical methods, and is also validated by measurements with EE core transformers. The proposed method shows good accuracy (< 10% error) compared with 2-D FEM for both high and low porosity factor windings. Therefore, it can handle more winding configurations than other 1-D analytical methods.@en

The correct identification of partial discharges (PDs) is instrumental for the maintenance plan in gas-insulated systems (GIS). However, onsite PD measurements are difficult, especially in HVDC systems, where partial discharges can be confused with interference. This paper proposes a method to discern PDs from interferences based on the GIS characteristic impedance. The characteristic impedance is measured using very-high frequency electric and magnetic sensors, and it is calculated using four approaches based on the PD charge magnitude, peak value, peak-to-peak value, and frequency spectrum. The method is first tested with a PD calibrator in a matched and open-circuited GIS testbench. Then, the identification of PDs and interference is tested in a full-scale GIS, where the measurements are subjected to pulse overlapping and noise. Five types of interferences and PDs are injected into the GIS in two positions and measured in multiple mounting holes. The results show that all four approaches can precisely calculate the characteristic impedance in a matched testbench. In the full-scale GIS, these approaches show more deviation, with the peak approach being the most accurate. A practical application of the method is demonstrated using a calibrator in the full-scale GIS. The proposed method contributes to a more reliable PD monitoring system for HVDC/AC GIS and allows better maintenance planning, reducing unnecessary costs, notably for offshore substations.@en

Traditional methods such as Steinmetz's equation (SE) and its improved variant (iGSE) have demonstrated limited precision in estimating power loss for magnetic materials. The introduction of Neural Network technology for assessing magnetic component power loss has significantly enhanced accuracy. Yet, an efficient method to incorporate detailed flux density information—which critically impacts accuracy—remains elusive. Our study introduces an innovative approach that merges Fast Fourier Transform (FFT) with a Feedforward Neural Network (FNN), aiming to overcome this challenge. To optimize the model further and strike a refined balance between complexity and accuracy, Multi-Objective Optimization (MOO) is employed to identify the ideal combination of hyperparameters, such as layer count, neuron number, activation functions, optimizers, and batch size. This optimized Neural Network outperforms traditionally intuitive models in both accuracy and size. Leveraging the optimized base model for known materials, transfer learning is applied to new materials with limited data, effectively addressing data scarcity. The proposed approach substantially enhances model training efficiency, achieves remarkable accuracy, and sets an example for Artificial Intelligence applications in loss and electrical characteristic predictions with challenges of model size, accuracy goals, and limited data.@en

Traditional methods such as Steinmetz's equation (SE) and its improved variant (iGSE) have demonstrated limited precision in estimating power loss for magnetic materials. The introduction of Neural Network technology for assessing magnetic component power loss has significantly enhanced accuracy. Yet, an efficient method to incorporate detailed flux density information—which critically impacts accuracy—remains elusive. Our study introduces an innovative approach that merges Fast Fourier Transform (FFT) with a Feedforward Neural Network (FNN), aiming to overcome this challenge. To optimize the model further and strike a refined balance between complexity and accuracy, Multi-Objective Optimization (MOO) is employed to identify the ideal combination of hyperparameters, such as layer count, neuron number, activation functions, optimizers, and batch size. This optimized Neural Network outperforms traditionally intuitive models in both accuracy and size. Leveraging the optimized base model for known materials, transfer learning is applied to new materials with limited data, effectively addressing data scarcity. The proposed approach substantially enhances model training efficiency, achieves remarkable accuracy, and sets an example for Artificial Intelligence applications in loss and electrical characteristic predictions with challenges of model size, accuracy goals, and limited data.@en

Traditional methods such as Steinmetz's equation (SE) and its improved variant (iGSE) have demonstrated limited precision in estimating power loss for magnetic materials. The introduction of Neural Network technology for assessing magnetic component power loss has significantly enhanced accuracy. Yet, an efficient method to incorporate detailed flux density information—which critically impacts accuracy—remains elusive. Our study introduces an innovative approach that merges Fast Fourier Transform (FFT) with a Feedforward Neural Network (FNN), aiming to overcome this challenge. To optimize the model further and strike a refined balance between complexity and accuracy, Multi-Objective Optimization (MOO) is employed to identify the ideal combination of hyperparameters, such as layer count, neuron number, activation functions, optimizers, and batch size. This optimized Neural Network outperforms traditionally intuitive models in both accuracy and size. Leveraging the optimized base model for known materials, transfer learning is applied to new materials with limited data, effectively addressing data scarcity. The proposed approach substantially enhances model training efficiency, achieves remarkable accuracy, and sets an example for Artificial Intelligence applications in loss and electrical characteristic predictions with challenges of model size, accuracy goals, and limited data.@en

In this paper, air discharges for different frequencies are tested in a point-plane geometry with different air gaps. The stress frequencies vary between 50 Hz to 5 kHz, and air gaps vary from 2.5mm to 10mm. It is observed that the partial discharge inception voltage of the positive corona increases with the increase of frequency, while the inception voltage of the negative corona stays nearly the same. The breakdown voltage increases with the increase of frequency in our test set, and breakdowns happen during negative half-cycle, especially at 5kHz, which is supposed to be caused by the space charges.@en

With widely use of power electronics in medium voltage level, there is increasing concern about behaviour of insulation material under arbitrary voltage stress. Solid-gas insulation is a general choice in medium voltage level. Therefore, it’s unavoidable to encounter surface discharge problem. In this paper, the surface discharge on epoxy-air interface is investigated, which is a commonly used insulation material. Point to plate electrode was used to test the surface flashover voltage and partial discharge (PD) behaviour at 50Hz, 500Hz, 1kHz, 2kHz and 5kHz. As the result, with increasing frequency, PDIV keeps roughly constant and flashover voltage first increases and then decrease, and the frequency obtaining highest flashover voltage is related to gap length.@en

In this study, the immediate time to breakdown for pure epoxy resin is measured by applying a ramp sinusoidal voltage signal at certain frequency (50Hz, 500Hz and 5000Hz) until breakdown occurs. The HV testing procedures, epoxy resin sample and experimental set-up preparation and Weibull analysis of obtained results are elaborated. The experiments were carried out on epoxy samples with the thickness between 0.1mm and 0.2mm and the thickness was divided into three groups for more accurate statistical analysis.@en

Litz wire, which is used to suppress eddy current, always have complex structure. Solving its 3-D finite element model (FEM) requires high computational resources. This article presents a 2.5-D loss calculation method for round Litz wires, which do not need mesh. One pitch of Litz wire is set as an object. The exact structure is constructed by a recursive method and then is sliced into several sections per pitch. Each section is represented by a cross section area. Two-dimensional problems are solved based on an analytical method, which is based on magnetic vector potentials in quasi magneto-statics situation. One pitch of the Litz wire is approximately represented by the summation of 2-D problems. The proposed method is compared with 3-D FEM results, which shows the proposed method has good accuracy and fast computational speed.@en