Search

Arc fault detection devices are mandatory worldwide for mitigating DC series arc faults in photovoltaic systems. However, they are prone to nuisance tripping. Artificial intelligence-based approaches can be a solution, but they are “black boxes” and challenging to modify. This paper proposes an explainability and attention-based method to investigate the intensive details of such an algorithm. The contributions of an arc feature to the proposed model can be visualized by the proposed interpretable methodology so that insensitive arc features can be removed to reduce the quantity of input data. Additionally, the structure of the proposed model can be optimized by cutting the redundant layers. Thus, an accuracy of 99.63% is achieved with only 48.48% of the parameters compared to the original model. Finally, the optimized model is implemented by a Cortex M7-based microprocessor with a runtime of only 7.8 ms, making it ready for industrial application.

Automatic fault detection is currently combined with deep networks owing to the possibility of dispensing signal processing, which significantly accelerates reactions to faults. Changing the type of defect or object forces repetition of the network learning process and its implementation. The design of universal systems for detecting different faults can be developed using transfer learning techniques. This paper presents the application of the transfer learning of a convolutional neural network to the fault diagnosis of permanent magnet synchronous motors. The crucial point of this research was to develop accurate diagnostic applications based on the data obtained from the motor field model and to use their functionality for a real object. This study compares the accuracy of diagnostic systems using three currently known techniques: neural network-based, instance-based, and mapping-based transfer learning. The experimental verification of the systems was carried out on an experimental bench with a 2.5 kW motor.

The reliability of the model significantly affects early detection and accurate classification of electrical faults. In this study, a Long Short Term Memory based fault classification model was developed for the Power System Machine Learning benchmark dataset, focusing on improving reliability by increasing interpretability. First, novel metrics are introduced to measure model interpretability. These interpretability metrics are uniquely defined based on the disentanglement of the fault classification factors. Subsequently, hyperparameter optimization was performed using multi-objective Bayesian Optimization to determine the optimal model architecture. The objective of optimization is to maximize interpretability and classification accuracy. The Pareto-optimal solution presents different model architectures with varying accuracy and interpretability trade-offs. Finally, the manifestation of interpretability in terms of subsequences is studied using Shapley Additive Explanations. The impact of class representation and architectural parameters on interpretability was also analyzed. Furthermore, the most accurate model in the Pareto front achieved highly competitive accuracy for the benchmark data.

The traditional manual periodic screening method of grading electrode sediments is prone to cause the equipment failure of high voltage direct current converter valves. Therefore, we propose to use ultrasonic time-domain reflection method to detect the sediments. However, the ultrasonic echo signals are characterized by nonlinearity and nonsmoothness, which makes it very difficult to extract effective features for sediment detection. To address this issue, we propose an intelligent detection method based on multiscale hybrid entropy characteristics in the time-frequency domain. First, a multiscale decomposition of the signal is performed. Second, the weighted form factor index is proposed to select the noise modes. Moreover, we propose to calculate the hybrid entropy in the time-frequency domain of each mode as the characteristic input bidirectional long and short-term memory network model, and verified that feature enhancement can be achieved by noise modes noise reduction. Finally, the experimental validation shows that the proposed method can achieve nondestructive testing and intelligent identification of graded electrode sediment with a correct identification rate of 94.25%.

To further improve the detection ability of residual current in low-voltage distribution networks, an adaptive residual current detection method based on variational mode decomposition (VMD) and dynamic fuzzy neural network (DFNN) is proposed. First, using the general $K$ -value selection method of VMD proposed in this study, the residual current signal is decomposed into $K$ intrinsic mode functions (IMFs). By introducing the cross-correlation coefficient $R$ and the time-domain energy entropy ratio $E$ as two classification indexes, IMFs are divided into three categories: effective IMFs, noise IMFs and aliasing IMFs. Then, the aliasing IMFs are denoised by recursive least squares (RLS), and the denoised IMFs are superimposed with the effective IMFs to obtain the reconstructed signal. Finally, the dynamic fuzzy neural network (DFNN) is adjusted by the minimum output method to achieve the detection of the reconstructed residual current signal, and the network is used to predict the residual current according to the detection results. The detection results of the simulation and measured data show that the proposed algorithm has high detection accuracy and is superior to the wavelet neural network, empirical mode decomposition-thresholding, and wavelet entropy-auto encoder-back propagation neural network methods in terms of mean square error, goodness of fit and running time. This method provides a reference for further research on new adaptive residual current protection devices.

This paper proposes a non-contact method for fault location in transmission lines, which is based on magnetic fields produced by current signals measured using magnetoresistive sensors installed only at transmission line terminals, under the phase conductors of the first transmission tower at both terminals or at the substations portals. The proposed method uses the Extended Kalman filter to process these measurements and is based on a travelling wave approach in order to perform the fault localization. This paper also describes the implementation and testing of the method, firstly introducing its overview, followed by an analysis of the magnetic fields produced by the current signals, as well as considerations on their measurement; secondly, detailing the Extended Kalman filter and the travelling wave approach; and, finally, presenting the results of the method with regards to simulations built using EMTP/ATP to evaluate its robustness under different conditions varying the fault resistance, fault inception angle, phases involved and fault location. The results indicate that the proposed method is robust and accurate.

In recent years, DC fault arc detection has been an electrical engineering research hotspot. At present, most proposed detection methods do not analyze the effects of fault arc electrical characteristics on both line current and supply voltage. Therefore, this study extensively analyzes variations of the line current and supply voltage because of DC arc faults based on the volt-ampere characteristics of DC arc faults. Then, a DC series arc fault detection method is proposed that comprehensively uses information on line current and supply voltage. An experimental platform for DC fault arc generation and detection was established using a DC-DC converter and a photovoltaic power supply as DC power supplies, and the proposed method was confirmed by experiments using this platform. Experimental results demonstrate that the proposed method can effectively distinguish arc faults and has the characteristics of clear physical meaning while maintaining a low amount of calculation.

The present paper proposes a genuine, simple to implement, robust and reliable method for current sensors fault detection and compensation used for 3-phase inverters. Its functionality is based on an algorithm programed into a Field Programmable Gate Array (FPGA) as general controller. Usually, 3-phase inverters are controlled using field oriented control (FOC) which is generally triggered to sample the measured currents on the established switching frequency. As the latter is much smaller than the base clock of the FPGA, this allows hundreds of free time samples to perform other calculations in-between two FOC samplings. In this paper, a method that uses these free and unused time samples is presented. This method performs calculations for fault detection and compensation ensuring that at each new FOC sampling, this will receive the correct current data to reach continuous operation despite faults. The interleaving principle of the FOC with the fault detection method, as it will be proven in the paper, ensures high reliability of the complete controller diminishing the possibility of undesired or faulted readings to disturb the FOC's calculations. The fault detection philosophy is based on continuous comparison of the instantaneous measured currents (sensor's response) against reference values. The experimental results presented in the paper prove reliable operational performances of the method in both steady state and transient conditions. The added value of the paper consists in its genuine approach to handle the fault detection and compensation in-between two PWM ticks, ensuring that no faulted measurements can reach the control unit. This added value is based on functionality divided on several functions triggered by an internal generated clock synchronizing and handshaking their operations towards one goal: fault detection and compensation.