Deep residual network based fault detection and diagnosis of photovoltaic arrays using current-voltage curves and ambient conditions

With the rapid progress in the deep learning technology, it is being used for vibration-based structural health monitoring. When the vibration is used for extracting features for system diagnosis, it is important to correlate the measured signal to the current status of the structure. The measured vibration responses show large deviation in spectral and transient characteristics for systems to be monitored. Consequently, the diagnosis using vibration requires complete understanding of the extracted features to discard the influence of surrounding environments or unnecessary variations. The deep-learning-based algorithms are expected to find increasing application in these complex problems due to their flexibility and robustness. This review provides a summary of studies applying machine learning algorithms for fault monitoring. The vibration factors were used to categorize the studies. A brief interpretation of deep neural networks is provided to guide further applications in the structural vibration analysis.

Review of vibration-based structural health monitoring using deep learning

The environmentally clean nature of solar photovoltaic (PV) technology causes PV power generation to be embraced by all countries across the globe. Consequently, installation and utilization of PV power systems have seen much growth in recent years. Although PV arrays of such systems are robust, they are not immune to faults. To guarantee reliable power supply, economic returns, and safety of both humans and equipment, highly accurate fault detection, diagnosis, and interruption devices are required. In this paper, an overview of four major PV array faults and their causes are presented. Specifically, ground fault, line-line fault, arc fault, and hot spot fault have been covered. Next, conventional and advanced fault detection and diagnosis (FDD) techniques for managing these faults are reviewed. Moreover, a single evaluation metric has been proposed and utilized to evaluate the performances of the advanced FDD techniques. Finally, based on the papers reviewed, PV array fault management future trends and possible recommendations have been outlined.

/pdf/review-and-performance-evaluation-of-photovoltaic-array-1okizxvri8.pdf

Review and Performance Evaluation of Photovoltaic Array Fault Detection and Diagnosis Techniques

This paper presents a model-based anomaly detection method for supervising the direct current (dc) side of photovotaic (PV) systems. Toward this end, a framework combining the benefits of $k$ -nearest neighbors ( $k$ NN) with univariate monitoring approaches has been proposed. Specifically, $k$ NN-based Shewhart and exponentially weighted moving average (EWMA) schemes with parametric and nonparametric thresholds have been introduced to suitably detect faults in PV systems. The choice of $k$ NN method to separate normal and abnormal features is motivated by its capacity to handle nonlinear features and do not make assumptions on the underlying data distribution. In addition, because the EWMA approach is sensitive in detecting small changes. First, a simulation model for the inspected PV array is constructed. Then, residuals generated from this model are employed as the input for $k$ NN-based schemes for anomaly detection. Parametric and nonparametric thresholds using kernel density estimation have been used to detect faults. The effectiveness of the $k$ NN-based procedures is verified using actual measurements from a 9.54-kWp grid-connected system in Algeria. Results proclaim the efficiency of the proposed strategy to supervise the dc side of PV systems.

Improved $k$ NN-Based Monitoring Schemes for Detecting Faults in PV Systems

Series arc faults can cause fire hazards in dc distribution systems. During a series arc fault, the line impedance usually increases rapidly and the arc current includes high-frequency components. To capture the arc-induced high-frequency signals, parallel capacitors are added to the circuit. The characteristics of the currents through these capacitors permit fault detection and localization. This paper determined that a Rogowski coil could be constructed with adequate bandwidth to cover the relevant frequency band of 1 kHz-1 MHz, found that the amplitude of the pulse and the difference in the integrated fast Fourier transform of the current signal are well correlated with the series fault, described experiments to demonstrate that, over the range tested, the two parameters are robust with respect to the pressure of the atmosphere in which the arc forms, the electrode material, the speed at which the electrodes separate to initiate the arc, the switch operation, the load-type and the load change, and demonstrated that within the well-known limits of traveling wave localization, such localization can be used based on the capacitor current pulse information after it has been validated by the spectrum analysis as a fault and not a spurious signal. The proposed approach was demonstrated in two low-voltage dc cable circuits.

Series Arc Fault Detection and Localization in DC Distribution System

Arc faults threaten the safe operation of photovoltaic (PV) systems. An arc fault detection and localization approach using parallel capacitors is proposed. A PV system has been analyzed and tested with five capacitors paralleled with the branches in the system. Series and parallel arc faults at nine locations have been tested in the system. When an arc occurred, current pulses were generated in the capacitors and their amplitudes and polarities were obtained through Hall current sensors. Discrete wavelet transformation was performed on the capacitor currents and the distributions of their amplitudes, frequency spectrums, and polarities are here reported. The results indicate that the distributions are unique under different fault types and locations, which could be used to detect and localize arc faults in PV systems. Moreover, the amplitudes of the capacitor currents can also help to localize a series arc fault within a PV string. Finally, the proposed approach is validated by a double-fault test.

Arc Fault Detection and Localization in Photovoltaic Systems Using Feature Distribution Maps of Parallel Capacitor Currents

The large current flowing through transformer windings during external short-circuit events causes strong vibration, and the tank emits multi-frequency sound. This paper investigates the transient acoustic signal theoretically and experimentally and proposes to monitor the winding condition with the spectrum of the acoustic signal. A two-wire model is first built as a basic unit of winding structure. The dynamic equation of the model is constructed based on the principle of least action. The coupling action between the electromagnetic force and the motion of the wires is described as the second-order term of the wire spacing in the equation. The solution shows that the spectrum of the winding vibration contains half power frequency and its multiples in some cases. If the winding is loosened and deformed, the motion space and vibration amplitude would be enlarged, and the natural frequency of winding would decrease. As a result, the vibration spectrum would contain half frequency and its odd multiples. Then, the acoustic signal of a 110-kV transformer is measured on a multiple short-circuit test platform. The acoustic time-frequency spectrums are consistent with the theoretical analysis. The half frequency ratio (HFR) of the acoustic spectrum raises along with the increase of short-circuit times. After 21 short-circuits, the HFR raises obviously, indicating serious deformation of the winding, while no obvious change is found in the short-circuit reactance (SCR). The results prove that the HFR of acoustic signal is a more sensitive indicator than SCR, which can characterize the cumulative process of short-circuit on the winding.

Multi-Frequency Acoustic Signal Under Short-Circuit Transient and Its Application on the Condition Monitoring of Transformer Winding

Investigation on vibration source and transmission characteristics in power transformers

This paper presents a detailed investigation of audible noise characterization of filter capacitors by investigating the noise radiation characteristics and analyzing the effect of voltage frequency, phase angle, and magnitude on a single capacitor unit. First, the noise radiation for capacitor surfaces was characterized theoretically using finite rectangular plate model and experimentally measured in a semianechoic room. Second, voltage frequency impact on the noise distribution was investigated through simulation. The simulation results show that noise directivity increases with increasing frequency. The excitation frequency impact on the radiation ratio and the noise level was characterized through a sweep-frequency experiment. Our investigations show that the noise frequency response below 500 Hz is quite low, while from 500 to 2000 Hz, it is significant and increases slowly. Third, the impact of the voltage phase angle on the noise level was investigated. The results show a noise-level difference of 3.9 dB for different phase angles. Finally, we present the voltage magnitude impact on the noise level. Our theoretical analysis and experimental study show a linear relationship between the sound power level and the logarithm of the voltage magnitude with a slope of 40.

Audible Noise Characteristics of Filter Capacitors Used in HVDC Converter Stations

The ultimate objective for transformer winding fault diagnosis is fault location and even severity evaluation. The issue has been studied in many works, based on frequency response analysis, using numerical indices or mathematical correlation between ladder network and winding. This paper utilizes Gauss–Newton iteration algorithm (GNIA) to diagnose winding faults for its efficient search of optimal parameters and discrimination of linearly-dependent vectors. Three key problems are handled for the application of GNIA in this paper. The first problem is the establishment of connection between ladder network components and frequency response function in GNIA. Another issue is the method to identify the faulty section to locate the defective winding position. The assessment of fault severity with GNIA data is another challenge. GNIA is applied on an air-cored model coil and a customized transformer successfully, with satisfactory diagnosis on deformation location and severity.

Yuhang Shi

Papers

Arc Fault Detection and Localization in Photovoltaic Systems Using Feature Distribution Maps of Parallel Capacitor Currents

Multi-Frequency Acoustic Signal Under Short-Circuit Transient and Its Application on the Condition Monitoring of Transformer Winding

Investigation on vibration source and transmission characteristics in power transformers

Audible Noise Characteristics of Filter Capacitors Used in HVDC Converter Stations

Application of Gauss–Newton Iteration Algorithm on Winding Radial Deformation Diagnosis