HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Distribution Network Reconfiguration Including OLTC Limits and Variable Shunt Reactors
No full textInternational audienceNetworks often suffer from overvoltage problems. The On-Load Tap Changers (OLTCs) may not always be able to regulate the voltage. Distribution network reconfiguration and variable shunt reactors can help mitigate these issues. In this paper, OLTCs are modeled and incorporated into the optimization. The results show an improvement in the accuracy
Volume reduction of magnetic components in DC/DC converters for fuel cell vehicles
No full textThis paper was presented at the EPE 2025 conference and was subsequently selected for publication in the special issue of Elsevier's PEDC journal. Therefore, only the abstract appears on the conference website.International audienceIn power electronic converter systems, the magnetic components are often the bulkiest and heaviest components. This is particularly disadvantageous in automotive applications, where volume, weight and costs are particularly important. Customized core geometries are a promising option to significantly reduce the volume of magnetic components compared to the use of standard core geometries. In this paper, customized magnetic components for an eightfold interleaved boost converter for fuel cell vehicles is presented. The customized core geometries are compared to an equivalent stacked standard core design in terms of enveloping volume and total losses. In addition, further possibilities for improvement in the form of customized coupled inductors that replace the discrete components are shown and discussed
PWM-Less Utility Frequency Output Capacitive Wireless Power Transfer System
No full textInternational audienceThis paper presents a capacitive wireless power transfer (CPT) system designed to provide utility frequency output, specifically targeting onboard chargers (OBCs) for electric vehicles (EVs) in automated parking systems. The system functions by directly injecting high frequency into a fully rectified sinusoidal waveform, thus removing the need for DC link capacitors. On the secondary side, the waveform is converted back into a fullwave rectified sinusoid using a diode rectifier and filter. The unfolder on the secondary side provides the utility frequency output without pulse width modulation (PWM). This topology reduces switching losses and supports system miniaturization. Experimental results from a prototype show AC voltage output at a utility frequency of 50 Hz. The analysis reveals a total harmonic distortion (THD) of 4.10% for the input current and 5.43% for the output current
Bidirectional Single-Input Multiple-Output Power Converters: Model and Performance Analysis
No full textInternational audienceThis paper introduces a small-signal transformation model based on the cantilever approximation for transformers commonly used in LLC converters. The proposed transformation is evaluated towards extending resonant converters in multiport applications. The transformation is compared in a traditional LLC multi-active bridge against a cascade multilevel inverter configuration on the secondary side. The objective is to analyze the differences and similarities in their implementation and assess their impact on system efficiency based on topology, with a focus on comprehensively describing losses in resonant multioutput converters. Two low-power prototypes were developed to validate the proposed model and approximation. The case study examines the efficiency and cross-regulation trade-offs between the configurations. Simulation and experimental results verify the accuracy of the analytical approach, offering valuable insights for designing and implementing high-frequency transformers in multiport applications
Grid Impedance Estimation with Large SCR Disturbances based on Grid-Forming Converter
No full textInternational audienceGrid-Forming (GFM) converters face significant synchronization challenges when suddenly exposed to ultra-weak grid conditions, primarily due to the lack of prior knowledge about grid impedance. This paper proposes a passive method for estimating grid impedance during large disturbances caused by a substantial reduction in the grid short-circuit ratio (SCR). The proposed method relies solely on local measurements, specifically the point of common coupling (PCC) voltage and output power. By dynamically adjusting the active power reference based on these estimates, GFM converters can maintain synchronization even under extremely low SCR conditions, well below 1. This method has been validated through experimental results
Robust Classification in Bayesian Neural Networks
No full textInternational audienceBayesian neural networks (BNNs) are competitive multiclass classifiers with reasonably sound probabilistic characteristics. In this paper, we consider BNNs as classifier ensembles obtained through sampling, and define the inference problem in BNNs as 1. computing a median probability distribution from the ensemble of distributions, under a given distance, and 2. finding a Bayes-optimal prediction (BOP) under the evaluation metric given the median. With this (re)formulation, all the results related to computing medians of sets of probability distributions can be leveraged to strengthen the predictive performance of BNNs. We shall recall a generic formulation of the problem of computing medians and provide empirical evidence to illustrate the potential impact of the choice of distance regarding accuracy metrics and calibration errors.</div
Joint Group Profiling and Recommendation via Deep Neural Network-based Multi-Task Learning
No full textInternational audienceGroup recommender systems aim to generate recommendations that align with the collective preferences of a group, introducing challenges that differ significantly from those in individual recommendation scenarios. This paper presents Joint Group Profiling and Recommendation via Deep Neural Network-based Multi-Task Learning, a framework that unifies group profiling and recommendation tasks within a single model. By jointly learning these tasks, the model develops a deeper understanding of group dynamics, leading to improved recommendation accuracy. The shared representations between the two tasks facilitate the discovery of latent features essential to both, resulting in richer and more informative group embeddings. To further enhance performance, an attention mechanism is integrated to dynamically evaluate the relevance of different group features and item attributes, ensuring the model prioritizes the most impactful information. Experiments and evaluations on real-world datasets demonstrate that our multi-task learning approach consistently outperforms baseline models in terms of accuracy, validating its effectiveness and robustness
Influence of vent distribution on the violence of a gas explosion
No full textInternational audienceThe development of new energies has led to their implementation in ISO maritime containers, raising the risk of flammable gas accumulation and explosion. Vent panels are commonly used to release excess gas produced by combustion and limit explosion overpressure. However, explosion discharge orifices are generally concentrated in one area. Little research has been done on the impact of vent distribution across the enclosure's surface. This article presents the results of an experimental study in which 1.2 m2 of vent area was distributed over the surface of a 37 m3 blast chamber. Four vent surface distribution configurations are studied. Two flammable mixtures, 15.5% and 17.4% hydrogen-air, respectively, were used, with two ignition source locations (backwall, central). An experimental study found that vent distribution reduces internal overpressure in the case of backwall ignition but has little influence when the ignition source is central. However, vent distribution plays a significant role in reducing external pressure effects
Deep evidential fusion with uncertainty quantification and reliability learning for multimodal medical image segmentation
No full textInternational audienceSingle-modality medical images generally do not contain enough information to reach an accurate and reliable diagnosis. For this reason, physicians commonly rely on multimodal medical images for comprehensive diagnostic assessments. This study introduces a deep evidential fusion framework designed for segmenting multimodal medical images, leveraging the Dempster-Shafer theory of evidence in conjunction with deep neural networks. In this framework, features are first extracted from each imaging modality using a deep neural network, and features are mapped to Dempster-Shafer mass functions that describe the evidence of each modality at each voxel. The mass functions are then corrected by the contextual discounting operation, using learned coefficients quantifying the reliability of each source of information relative to each class. The discounted evidence from each modality is then combined using Dempster's rule of combination. Experiments were carried out on a PET-CT dataset for lymphoma segmentation and a multi-MRI dataset for brain tumor segmentation. The results demonstrate the ability of the proposed fusion scheme to quantify segmentation uncertainty and improve segmentation accuracy. Moreover, the learned reliability coefficients provide some insight into the contribution of each modality to the segmentation process
Live Monitoring of the Blocking Behaviour of an IGBT Module in the HV-H 3 TRB Test
No full textInternational audienceOne of the main electrical characteristics of IGBT modules influenced by humidity-related degradation is the blocking behaviour. This paper presents an approach for online monitoring of the blocking behaviour during a accelerated ageing test based on the high voltage, high humidity, high temperature reverse bias (HV-H 3 TRB) test by injecting a constant current into the device under test