Search

DC series arc faults are known as a great threat to photovoltaic (PV) systems. Due to component aging and the high level of dc voltage, the occurrence of dc series arc fault is a serious concern in PV generation units. This article deals with driving effective dc series arc models from the original Nottingham arc model using the recorded practical data in a PV system. The original Nottingham arc model has three constant parameters and suffers significant inaccuracy for modeling the series arc faults. To avoid the inaccuracy in the modeling, at first, a Nottingham model is developed that is relevant to voltage and current of the fault with three parameters including one constant model order for exponents of the current and two time-series coefficients. Also, to further improvement in the model's accuracy, a two-fold Nottingham model is proposed that is relevant to voltage and current of the fault with five parameters including two constant model orders for exponents of the current and three time-series coefficients. Finally, a full-time-variant Nottingham model is proposed that mathematically contains three time-series parameters including one time-variant model order of the current exponent. In the full-time-variant Nottingham model, the optimal parameters corresponding to each sliding window of data are utilized. In all the proposed models, the arc coefficients are estimated using a new formulation using the least-squares error technique. Verifying the effectiveness of the proposed models through actual data, the accuracy of the proposed models is compared with the original Nottingham model. [ABSTRACT FROM AUTHOR]

One of the major challenges in grid-connected wind farms is the flicker emission caused by the reactive power extremely short time variations. A general used solution is installing a static volt-ampere reactive (VAR) compensator (SVC) as the most cost-saving device, which still holds the risk of incomplete reactive power compensation subject to the operating delay time. To this end, an accurate extreme short-term reactive power forecasting structure is essential. Although there are many studies related to the forecasting of wind power in long, medium, and short terms, however, the extremely short-term reactive power prediction of the wind farms is only addressed in few studies. The main aim in this article is to develop a method based on the convolutional neural network (CNN) to directly learn nonstationary and complex features from the raw wind farm reactive power time series and to contribute a predictive controller to mitigate the voltage flicker through an SVC connected to the wind farm. However, high sensitivity of typical mean-squared-based loss functions might lead to considerable errors in the forecasting. To resolve this issue, a modified loss function is proposed to enhance the performance and, consequently, an improved CNN is presented. The evaluation of the proposed method is based on the actual recorded data of a wind farm in Manjil, Iran. The data are directly used in the modeling of the wind farm. A current source, which is based on the actual records its amplitude and phase angle change every 0.01 s, is utilized to model the wind farm. The numerical results in terms of flicker sensation and short-term flicker perceptibility (Pst) measurement are used to verify the performance of the proposed method through comparison with the wind farm performance without SVC and SVC with a common control system. [ABSTRACT FROM AUTHOR]

Differential protection, as the key protection element in the power transformers, has always been threatened with sending false trips subjected to external transient disturbances. As a result, differential protection needs an additional block to distinguish between internal faults and external transient disturbances. The protection system should, first, be able to perform based on raw data, second, be able to learn fully temporal features and sudden changes in the transient signals, and, third, impose no assumption on noise. To address these challenges, a fast RNN, namely fast gated recurrent neural network (FGRNN). By removing reset gate in the gated recurrent unit (GRU), the proposed network is capable of learning abrupt changes in addition to significantly reducing the computational time. Furthermore, a loss function based on an information theory concept is formulated in this article to enhance the learning ability as well as robustness against non-Gaussian/Gaussian noises. A generalized form of mutual information is also adopted to form a noise model-free loss function, then incorporated with the designed deep network. Simulated and experimental examinations engaging various external factors, in addition to comparison between the proposed FGRNN, GRU, and seven firmly established methods indicates the faster and more reliable performance of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Series arc faults (SAFs) are the most common faults in photovoltaic systems. If a SAF is not detected in time, it may lead to some catastrophic damages to the system. This article presents a new method for detecting SAFs in photovoltaic systems based on determinism in the voltage signal. The determinism of the signal is derived using the recurrence plots method. The recurrence plots provide a two-dimensional plot of the signal, from which the degree of determinism in the signal is derived as a criterion for detection of SAFs. The performance of the method is evaluated using plenty of experiments carried out in different SAF and normal conditions. The results confirm the high accuracy, high speed, and low computational burden of the proposed method. [ABSTRACT FROM AUTHOR]

This study deals with a new approach to tackle the challenges encountered in the filters-based reactive power calculation techniques utilized in the control system of static VAr compensators. In this approach, step and time-varying responses are improved by a combination of differential and filter-based methods. Then, optimum values of the main parameters of the low-pass and band-stop filters are determined considering three criteria: accuracy, speed, and capability for following the time-varying loads. To assess the performance of the proposed method, a new objective function is introduced, calculated for two sets of signals, captured from a real field in Mobarakeh Steel Company in Isfahan/Iran. Then, the proposed and the other reactive power calculation methods are assessed in practice and simulation. Also, the algorithm is implemented using a TMDSCNCD28335 as the processor to assess the real-time feasibility of the method. To do so, the reactive power variation of an electrical arc furnace is emulated in a prototype test bench and the algorithm is run in real-time by the processor during the system operation. [ABSTRACT FROM AUTHOR]

This article puts forward a new algorithm based on Kullback–Leibler divergence (KLD) for discriminating inrush and internal fault currents in power transformers. Specifically, the main idea of this algorithm is to utilize the current signal discrepancy of the distribution with ideal fault current waveforms. To such an aim, the proposed index reproduces the differential current signal. Utilizing fast modified least squares technique, the differential current signal is generated. Applying the reconstructed differential current and ideal sinusoidal waveforms to the KLD indicator, the distribution discrepancy of the current signal from ideal sinusoidal waveform discriminates inrush and internal fault currents. Also, the internal/external identification index is introduced that utilizes the phase content of the signals from current transformers (CTs) on both sides to distinguish internal and external faults, especially the faults accompanied with CT saturation. As a result, the proposed method can distinguish inrush currents, internal and external fault currents with/without CT saturation, and internal fault currents during transformer energization. The performance of the proposed index is evaluated using practically recorded data and is further compared with the state of the art. [ABSTRACT FROM AUTHOR]

Stochastic nature of some input variables dictates the requisite of probabilistic analysis in power systems operation and planning. Wind generation is considered as a main source of intermittency in power systems due to the uncertain nature of wind speed. The proposed probabilistic optimal power-flow (POPF) method investigates spatial correlation among sources to attain more practical output distributions. The method established reduced-discrete point estimate method (RDPEM) along with the Latin hypercube sampling (LHS) in order to attain the stochastic characteristic of optimization’s outputs. Despite needing less computational effort, highly accurate results can be obtained, while there is no prerequisite for probability distribution of the input random variables. In order to more validate the efficiency of the proposed method, the Gram–Charlier (GC) expansion is used to compare the outputs’ cumulative distribution functions (CDFs) that are obtained from Monte Carlo (MC) with RDPEM methods. The performance and precision of the proposed solution are ascertained by comparison with those of Monte Carlo with discrete LHS (MCDLHS) in a hybrid IEEE 14-bus test system. [ABSTRACT FROM AUTHOR]

This study tries to outline a comprehensive approach established on the kinetic energy of the distributed generations (DGs) for islanding mode detection in micro-grids. More specifically, the proposed index calculates the rate of change of kinetic energy over reactive power (ROCOKORP) for each DG at the point of common coupling (PCC). Despite the islanding detection being the most challenging issue during complete match between the load and generation of DGs, the proposed method can swiftly deal with such an issue, therefore confirming its superiority. Other challenging conditions where islanding detection can either be difficult or confused with other events such as loads inside the non-detection zone (NDZ), various fault types, and loads at various power factors have also been considered to investigate the effectivity of the proposed approach. The proposed method’s performance has been verified and evaluated in comparison to the state-of-the-art via a simulation testbed. [ABSTRACT FROM AUTHOR]

Power delivery interruption and catastrophic failures are some of the potential consequences of undetected faults in photovoltaic (PV) arrays. A voltage protection scheme with the minimum number of sensors which is capable of detecting, classifying, and locating line-to-line (intrastring and cross string), line-to-ground, and open-circuit faults in a utility-scale PV array is proposed in this article. The proposed protection scheme consists of three steps. In the first stage, by a simple and efficient algorithm occurrence of any disturbance in the system is detected using some voltage sensors. At this stage, faulty strings in the PV array are also identified. In the second step, using a new least squares-based method, fault and shading conditions are discriminated from each other. Finally, in the third step, the faults inside the faulty string are located using a new method based on multiclass artificial neural networks. Fault location methods that have been developed so far require a large number of sensors or additional equipment for the exact location of faulty modules inside the fault string, but in the proposed method, only one sensor is used per string for fault localization. The proposed scheme for detection and location of faults is evaluated by simulating a typical system in different conditions. [ABSTRACT FROM AUTHOR]

So far, the overall operation time of relays has been used as the objective function in all directional overcurrent relays (DOCRs) optimal coordination problems. However, it cannot guarantee that in an optimal coordination of DOCRs at presence of distributed generations (DGs), the reliability indices will reach the minimum value. The reason is the existence of DGs voltage protection at their connection point to the network. This protection plays a key role in determining participation of DGs in improving network reliability during fault conditions. However, in the previous studies in which the overall operation time of relays is considered as the objective function, the voltage protection of DGs is not taken into account. In this article, coordination between DOCRs and their interactions with the DGs voltage protection is considered in calculation of the energy not supplied (ENS). Also, instead of the overall operation time of the DOCRs, the ENS value is considered as the optimization objective function. A probabilistic hourly seasonal model has been used to model the renewable DGs and the network loads. Various scenarios have been simulated to demonstrate the efficiency of the proposed method over a standard 33-bus radial distribution network. [ABSTRACT FROM AUTHOR]

DC series arc faults are known as a great threat in photovoltaic (PV) systems that may lead to fire and electric shock hazards. This article concentrates on driving effective models based on Cassie and Mayr equations for dc series arc using actual recorded data. Despite Cassie and Mayr equations that use invariant parameters during arc modeling, the proposed models employ the combinations of both Cassie and Mayr equations in which the time-variant nature of arc’s coefficients is taken into account. While the original Cassie and Mayr equations have two influential parameters for arc modeling, the first model presents a developed Cassie–Mayr equation with three parameters. Also, to further improvement in the model’s accuracy, the second model presents a developed Cassie–Mayr equation with five parameters. While the first and second models provide more prevision comparing with the original Mayr and Cassie–Mayr modeling, a third model is presented that treats with parameters of the original model as time-series parameters. To obtain the arc’s coefficients, a new formulation for coefficient estimation based on Levenberg–Marquardt (LM) algorithm is presented. Evaluating the proposed models with the actual recorded data, the results show the proposed models have more accuracy in comparison with the original Mayr and Cassie–Mayr. [ABSTRACT FROM AUTHOR]

Fast reactive power variation in wind farms creates flicker. Static VAR compensators (SVCs) are the most commonly used devices for flicker mitigation. The reactive power calculation unit is the main component in the control system of SVCs. This article deals with two new approaches for reactive power calculation. In both methods, the main parameters of a filter-based technique are optimized using a novel optimization approach. The first method has a non monotonic step response which is converted to a desirable monotonic response by the second method. This results in a better response to a time-varying reactive power. The problem formulation and definition of a new objective function are presented. A data pool is gathered from Manjil wind farm, Iran, for different conditions. This data pool is used in the objective function to derive the optimal setting. The performance of the proposed methods is assessed using this dataset and is compared with other techniques. Also, real-time implementation of the algorithm is evaluated using TMDSCNCD28335 as the processor. To do so, the reactive power variation in a wind generation unit is emulated in a prototype test bench and the algorithm is run in real-time by the processor during the system operation. [ABSTRACT FROM AUTHOR]

Tools such as short-term load forecast (STLF) play an ever-important role in the operation and planning of power systems. Improving STLF accuracy could reduce heavy penalties and missed financial opportunities. In this article, a new hybrid forecast method is proposed to improve STLF accuracy. The proposed hybrid STLF method is based on artificial neural network (ANN) and artificial bee colony (ABC) algorithms. The ABC algorithm is used to optimize the learning procedure of ANN. New load modeling is presented based on historical and weather data. Bad data elimination and calendar effects are considered in the STLF procedure. The proposed hybrid STLF method is verified by forecasting the Bushehr province demand. The results proved that the proposed method significantly improved STLF accuracy. Résumé—Les outils tels que les prévisions de charge à court terme (STLF) jouent un rôle de plus en plus important dans l’exploitation et la planification des réseaux électriques. Améliorer la précision des STLF pourrait réduire les lourdes pénalités et les opportunités financières manquées. Dans cet article, une nouvelle méthode de prévision hybride est proposée pour améliorer la précision des STLF. La méthode STLF hybride proposée est basée sur les algorithmes de réseau neuronal artificiel (ANN) et de colonie d’abeilles artificielle (ABC). L’algorithme ABC est utilisé pour optimiser la procédure d’apprentissage des ANN. Une nouvelle modélisation de la charge est présentée sur la base de données historiques et météorologiques. L’élimination des mauvaises données et les effets de calendrier sont pris en compte dans la procédure STLF. La méthode STLF hybride proposée est vérifiée en prévoyant la demande de la province de Bushehr. Les résultats ont prouvé que la méthode proposée améliore considérablement la précision de la méthode STLF. [ABSTRACT FROM AUTHOR]

A novel scheme is suggested for change detection and fault classification in ac microgrids (μGs) with different operating modes. The proposed method utilizes a Teager–Kaiser energy operator based approach for analyzing the current-based signal to detect changes in the μG. The current-based signal is a summation of squared three-phase currents (SSC), measured at one end of a line in the μG. Since μGs have different operating modes, a proper signal should be defined to have some valuable signatures of the changes in different conditions. The SSC signal is constant during normal conditions of the grid-connected and islanded μG, whereas in any possible change condition experiences considerable variation. After the detection, the faulty phase is determined using three similar indices, derived from the squared currents. These indices are the ratios of squared currents for each phase with half-cycle window length, before and after the change detection. The proposed protection scheme is able to handle both operation modes of the μG. The proposed technique is evaluated in a MATLAB/Simulink simulated network and laboratory small-scaled test bench. The results confirm high accuracy and quickness of the proposed approach. Furthermore, the performance of the proposed methodology is compared with some other similar methods. [ABSTRACT FROM AUTHOR]

Photovoltaic networks are outdoor infrastructures, faced to different harsh conditions, which may experience various failures such as parallel arc fault (PAF) and series arc fault (SAF). Although PAF is more severe than SAF, the detection of SAF is more problematic, and its hazards including fire and the risk of staff electrocution are more serious. This paper proposes a new method for timely and reliable detection of SAF in photovoltaic systems. In this method, one of the blind-source separation algorithms called the principal component analysis (PCA) is employed. This method separates the nondependent components of some measurable quantities such as voltage and current using eigenvectors of their covariance matrix. In a normal condition, these signals include the dc component, switching components, and the network disturbances. When a SAF occurs, some new components by the electrical arc also add to the system. Using PCA, a suitable index is derived to discriminate signatures of SAF from the other components. Performance of the method is evaluated using plenty of experiments in different conditions. [ABSTRACT FROM AUTHOR]

Series arc fault (SAF) is one of the most important causes for failure of photovoltaic (PV) systems. Conventional protections could not detect SAF at initial moments. In this paper, a new method is presented for detection of SAF. The method is based on extracting SAF signatures from high frequency contents of the normalized dc terminal voltage. Inverter switching signatures in the signal have a periodic nature, which are rejected by subtracting the resized data-windows based on the lag relevant to maximum cross-correlation value. The utilized criterion is defined as ratio of power of the low frequency components to power of the arc signal (or signal-to-noise ratio). A set of possible scenarios, whether in practice or simulation, is considered to evaluate the performance of the method. They include different arc lengths, different fault locations in the PV system, various partial shadings, different switching frequencies for the inverter, and different types of environmental noise. [ABSTRACT FROM AUTHOR]

System reliability enhancement of distributed generations (DGs) without aid of fast switches (FSs) is not possible. FSs are responsible for disconnecting unfaulted islands from faulty network to reduce energy not supplied (ENS). Directional overcurrent relays (DOCRs) are employed to send trip signals to FSs. In most of the DGs optimal allocation studies, the switches are assumed to operate correctly. This assumption might not be correct due to the delayed operation of the DOCRs. Few studies offer restricting minimum allowable size of DGs or allowable location of DOCRs to guarantee correct operation of DOCRs. These additional constraints limit the search space which causes nonoptimal answers. Here, the impact of actual operation of DOCRs is considered directly in calculation of ENS and hence in the final objective function. Also, a modified formulation for ENS in order to include cases where DGs capacity is not sufficient for supplying island loads is developed. Simultaneous allocation of DGs, FSs, and DOCRs to minimize cost of energy losses, ENS, and environmental costs are proposed. A multistate model that combines all possible operating conditions of DGs and load levels is utilized. Different cases are simulated in 33-bus, 69-bus, and 118-bus systems to evaluate the proposed method. [ABSTRACT FROM AUTHOR]

This paper presents a novel real-time algorithm to find unknown maximum power of photovoltaic (PV) systems. The maximum power of a typical PV system is continuously changing with variations in solar irradiance and PV module temperature. There is no precise mathematical equation to show relation between PV output power and environmental variables. Since power of a PV cell/module is unknown nonlinear and time-varying function of irradiance and temperature, real-time approaches are required to find maximum power values through time. Based on strong concavity feature of PV output power, one can consider this measurable signal as an unknown objective function that needs to be maximized. Hence, an estimation-based extremum-seeking control algorithm is proposed to solve this nonlinear and uncertain optimization problem. First, gradient of the objective function with respect to the PV voltage is treated as an unknown time-varying parameter. An adaptive estimation technique is utilized to identify this parameter. Then, a direct gradient-based adaptive controller is designed to find optimum voltage value and the corresponding maximum power point at uniform and uncertain weather condition in real time. Unlike most model-free approaches, the proposed algorithm has proof of convergence with precise steady-state and fast-transient performance. [ABSTRACT FROM AUTHOR]

Because of considerable importance of induction motors in different industries, fault detection of these motors has gained importance in the literature. One of the attractive approaches in this regard is motor current signature analysis. Thanks to simplicity and its lower cost, this method has been preferred compared to the other methods. In this paper, first, a model in which stator turn fault was taken into account is presented. Then, by stator current analysis using empirical mode decomposition (EMD), intrinsic-mode functions of the signals are derived. By selection of the calculated IMF in the first row of the EMD, its instantaneous frequency (IF) is estimated using adaptive IF estimation. Variation of the estimated IF is utilized for detection of different stator turn faults. The method is evaluated using plenty of simulations and some experiments. [ABSTRACT FROM AUTHOR]

Directional relays have a key role for protection of transmission and distribution networks. They are used as supervisory element for distance and overcurrent protections. Therefore, accuracy and performance of these protections are dependent on the performance of directional relays. In this paper, a high-speed current-based directional algorithm is presented. For a specified interval, the fault current samples are multiplied by corresponding prefault current samples and summation of the results is calculated as a criterion. Sign of the summation and direction of power flow at normal condition determine fault direction. In spite of most high-speed techniques, just accurate current transformers are needed to obtain the criterion and determining normal power flow can be relegated to simple voltage transformers at steady-state condition. The method is evaluated using plenty of simulations and experiments. Also, performance of the proposed algorithm is compared with some similar approaches. The results confirm reasonable accuracy, speed, and simplicity of the proposed method. [ABSTRACT FROM AUTHOR]

When a fault occurs in transmission lines, first, it is detected, then classified, and finally, located by a distance protection. Having a fast fault detection approach helps saving the required time for the whole protection procedure. In this paper, a novel method for fault detection of transmission lines based on the summation of squared three-phase currents (SSC) and a moving average technique is proposed. The SSC is constant during a normal condition of the network, while it has considerable variation under a fault condition. Using a moving average concept, a criterion is defined, by which fault occurrence can be detected. The proposed fault detection method is evaluated in four different systems: a typical 5-bus system, a typical double-circuit transmission line, WSCC 9-bus system, and a laboratory small-scaled system. The results confirm a high degree of accuracy and speed of the proposed approach. Moreover, the performance of the method is compared with some other similar methods from different aspects. [ABSTRACT FROM AUTHOR]

This paper proposes an efficient strategy for control of multiple grid connected microgrids (MGs) through converters, to improve network management during normal, islanding, and fault conditions. In normal condition, the converters control the bidirectional power flow between the MGs (downstream) and the main network (upstream). When the upstream is disconnected from downstream, if one of the MGs cannot handle its local loads, the deficit of power is provided by the other MGs. Transferred powers of the MGs are controlled regarding power quality (PQ) preserving criteria of the network. If PQ in a point of common coupling of an MG degrades, its converter goes into semi-isolated mode. Furthermore, in case of fault occurrence in the upstream, the converters limit fault current contribution of the MGs, in order to preserve the MGs PQ. However, in case of fault occurrence in the MG, the converters are bypassed by the relevant static switches. Therefore, short circuit current contribution of the upstream will be maximized to improve the MGs PQ. This strategy can also overcome the problems related to the system short circuit capacity increment in case of a newly added distributed generation unit, such as loss of the over current relays coordination. [ABSTRACT FROM PUBLISHER]

In electrical arc furnaces (EAFs), an unstable arc leads to unfavorable operation of the EAF, which can be rectified by using series reactors. Installation of the series inductor affects some electrical parameters of the EAF, such as arc stability, power factor, arc length, and the EAF efficiency. Furthermore, it reduces the effective transferred power to the EAF. To achieve desired transferred power after installation of the series reactors, the secondary voltage of the EAF transformer should be increased. However, after any tap changing of the EAF transformer, arc stability and other relevant parameters are affected. So the impact on the EAF electrical magnitudes caused by changes in the series reactor and in the EAF transformer tap should be simultaneously studied. This research deals with a new method for optimal setting of the series reactor and EAF transformer tap to achieve the EAF best performance. Using a nonlinear model for the EAF, its electrical characteristics, at nominal transformer current and for all of the possible transformer and reactor taps, are calculated and the results are tabulated. Using these tables, the optimal setpoint for transformer and series reactor taps is determined in different conditions. Finally, the method is investigated from an economic point of view. [ABSTRACT FROM AUTHOR]

The ability of static VAr compensators (SVC) to reduce the flicker caused by electric arc furnaces and other variable VAr loads depends on their speed, which is limited by delays in reactive power measurement and thyristor ignition. In this paper, a predictive method is proposed to compensate the delay time and, hence, to improve the SVC performance. Previous samples of the load reactive power are used to predict its future values. Predicted VAr is then applied as a reference for reactive power compensation. [ABSTRACT FROM AUTHOR]