Search

The efficient solution of the AC optimal power flow (OPF) for all cases and electrical grids is not guaranteed. Heuristics, approximations and relaxations have been proposed aplenty, each with pros and cons. This work proposes to solve the AC OPF with binary decision trees (BDTs). The solution starts with the full feasible set of an AC OPF instance, and, iteratively, BDTs tighten the constraints/bounds of that set, to improve the occurring feasible set by the median of the objective function of the preceding set. The medians of the objective function of the progressively tightened feasible sets will converge to the global optimum of the AC OPF. Recent proofs for the estimate performance of top‐down BDT learning heuristics ensure convergence of the method to the global optimum, provided the feasible set is adequately sampled for the BDT training. The recursive implementation of the method and the inductive nature of BDTs may also yield dispatches at multiple optimality levels for the same AC OPF instance, to account for stochastic resources. The performance of the proposed AC OPF solution is assessed over multiple cases by NICTA and the IEEE PES task force on benchmarks for validation of emerging power system algorithms. [ABSTRACT FROM AUTHOR]

The commutation failure is the most prevalent fault in line‐commutated converter based high‐voltage direct current systems, which may result in transient overvoltage on the sending‐side system. Overvoltage level evaluation has become a crucial task for power industries to assess the tripping risk of large‐scale wind turbines and implement effective stability control measures. In this paper, a derivation of the mathematical relationship between the reactive power consumed by the rectifier and AC voltage is presented firstly, along with an analytical expression for the peak value of transient overvoltage. Secondly, decision tree (DT) model is adopted to extract the mapping relationship between transient overvoltage and massive electrical quantities of power grids. The common DT algorithm is transformed by modifying the error weight assignment, which reflects the error tolerances for different actual overvoltage regions. On this basis, an overvoltage analysis method integrating the model‐driven and data‐driven techniques is proposed, and the improved DT algorithm is applied to fast error correction, enhancing the interpretability of regression prediction results. Case studies were performed in the simulation system of Northwest China local power grid with transient overvoltage problems, and the simulation results verified the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Distributed generations (DGs) for grid‐connected applications require an accurate and reliable islanding detection methodology (IDM) for secure system operation. This study presents an IDM for grid‐connected inverter‐based DGs. The proposed method is a combination of passive and active islanding detection techniques for aggregation of their advantages and elimination/minimisation of the drawbacks. In the proposed IDM, the passive method utilises critical system attributes extracted from local voltage measurements at target DG locations as well as employs decision tree‐based classifiers for characterisation and detection of islanding events. The active method is based on Sandia frequency shift technique and is initiated only when the passive method is unable to differentiate islanding events from other system events. Thus, the power quality degradation introduced into the system by active islanding detection techniques can be minimised. Furthermore, a combination of active and passive techniques allows detection of islanding events under low power mismatch scenarios eliminating the disadvantage associated with the use of passive techniques alone. Detailed case study results demonstrate the effectiveness of the proposed method in detection of islanding events under various power mismatch scenarios, load quality factors and in the presence of single or multiple grid‐connected inverter‐based DG units. [ABSTRACT FROM AUTHOR]

Anomaly detection of steam turbines is to recognize infrequent instances within sensor data that plays a vital role in stable power supply. Machine learning models have been applied to diagnose the faults of turbine and verified useful for identifying engine problem. To detect anomalies of steam turbines with machine learning methods, here, an approach called hierarchical pre-warning strategy is proposed that combines clustering methods with classification methods. Three different clustering methods, K-means, Isolation Forest and Local Outlier Factor, are chosen to separate anomalies from normal data. Since clustering results cannot give unanimous decision, the clustering instances are labelled with three classes, real anomalies, suspected anomalies and normal data, according to their overlapping recognition. Subsequently, five classification algorithms, k-nearest neighbour, support vector machine, decision tree, random forest and gradient boosting decision tree, have been examined to train the labelled data set. The classification results illustrate that gradient boosting decision tree and random forest are much more precise to detect real anomalies of steam turbines. The real anomalies identified by clustering methods have been classi- fied into suspected anomalies by this approach that is more practicable and consistent with ground truth. [ABSTRACT FROM AUTHOR]

One of the crucial challenges of the distribution network is the unintentionally isolated section of electricity from the power network, called unintentional islanding. Unintentional islanding detection is severed when the local generation is equal to or closely matches the load requirement. In this paper, both ensemble learning and canonical methods are implemented for the islanding detection technique of synchronous machine‐based distributed generation. The ensemble learning models for this study are random forest (RF) and Ada boost, while the canonical methods are multi‐layer perceptron (MLP), decision tree (DT), and support vector machine (SVM). The training and testing parameters for this technique are the total harmonic distortion (THD) of both current and voltage signals. THD is the most important parameter of power quality monitoring under islanding scenarios. The parameter and data extraction from the test system is executed in a MATLAB/Simulink environment, whereas the training and testing of the presented techniques are implemented in Python. Performance indices such as accuracy, precision, recall, and F 1 score are used for evaluation, and both ensemble learning models and canonical models demonstrate good performance. Ada‐boost shows the highest accuracy among all the five models with original data, while RF is robust and gives the best results with noisy data (20 and 30 dB) because of its ensemble nature. [ABSTRACT FROM AUTHOR]

Islanding detection plays an important role in operation and control of distributed generations (DGs). In this study, a novel hybrid method is proposed, which combines decision tree (DT) with voltage and frequency positive feedbacks for islanding detection in DGs. In the proposed method, the feature indices of prescribed events are captured and recorded in a target location to train a DT classifier, and the classifier can then be used for categorising islanding mode or grid-connected mode. The voltage and frequency positive feedback schemes are employed to regulate the DG power outputs based on voltage and frequency deviation between DG outputs and rated values. Furthermore, the disturbances are introduced to change the feature indices significantly when DG is switched to islanding mode, which can increase the sensitivities of feature indices and improve detection accuracy. The proposed hybrid method has been fully evaluated and tested on a representative DG system with two DGs. Simulation studies demonstrate that the proposed method can detect the islanding more accurately in situations where the power is balanced between DG output and load consumption, and can also improve detection accuracy in situations where DG system operates in minimum or maximum loading states. [ABSTRACT FROM AUTHOR]

In this study, a novel intelligent relaying scheme is proposed for fault classification and faulted phase selection using magnitude of differential power (MODP). The DP is the multiplication of difference between the magnitude of the real and estimated voltage and current phasors. The MODP for each phase is used as an input data to the data mining model known as decision tree (DT). The DT is used to generate thresholds to detect and classify the faults, since determination of robust thresholds is an important challenge in protective relay engineering. The proposed method has been tested for different operating modes of single-circuit transmission line in Simulink/MATLAB software. Simulation results indicate that the proposed method is able to detect and classify the faults with acceptable accuracy, in less than half cycle. [ABSTRACT FROM AUTHOR]

Recently, power system reliability has been challenged due to the increment of electrical demand. When an outage occurs, locating the outage may take a long time because of the distribution system's radial structure and the presence of various elements. To decrease the outage detection time, this study proposes to classify the equipment‐related outage causes to diagnose the faulty equipment at the time of outage occurrence. To this end, available historical outage, load and weather data sets are integrated, and various features are defined. Then, binary classifiers are developed to classify each equipment's failures against others'. To enhance classifiers' performance, this study also proposes to use cost function and ensemble models. The results of applying proposed classifiers show the accuracy of the proposed method and improvements in outcomes. [ABSTRACT FROM AUTHOR]

The increasing penetration level of renewable energy brings new challenges to the field of islanding detection. In this study, a decision‐tree (DT)‐learning method with hardware‐in‐the‐loop (HIL) simulations is proposed to address the non‐detection zone (NDZ) issue of islanding detection in hybrid distributed generation systems including both inverter‐ and synchronous‐machine‐based distributed energy resources. This method can effectively reduce the NDZ through advanced relay training strategy and utilise the advantage of real‐time simulators on simulation efficiency. In addition, the generated DT can be programmed into a real relay and then get validated through HIL simulations. The HIL implementation adds a practical dimension to the proposed method. With the proposed method, the laboratory testing results indicate promising islanding detection performance in terms of dependability, security, reduced NDZ area and detection time. [ABSTRACT FROM AUTHOR]

Islanding detection plays an important role in operation and control of distributed generations (DGs). In this study, a novel hybrid method is proposed, which combines decision tree (DT) with voltage and frequency positive feedbacks for islanding detection in DGs. In the proposed method, the feature indices of prescribed events are captured and recorded in a target location to train a DT classifier, and the classifier can then be used for categorising islanding mode or grid‐connected mode. The voltage and frequency positive feedback schemes are employed to regulate the DG power outputs based on voltage and frequency deviation between DG outputs and rated values. Furthermore, the disturbances are introduced to change the feature indices significantly when DG is switched to islanding mode, which can increase the sensitivities of feature indices and improve detection accuracy. The proposed hybrid method has been fully evaluated and tested on a representative DG system with two DGs. Simulation studies demonstrate that the proposed method can detect the islanding more accurately in situations where the power is balanced between DG output and load consumption, and can also improve detection accuracy in situations where DG system operates in minimum or maximum loading states. [ABSTRACT FROM AUTHOR]

Power system protection and asset management present persistent technical challenges, particularly in the context of the smart grid and renewable energy sectors. This paper aims to address these challenges by providing a comprehensive assessment of machine learning applications for effective asset management in power systems. The study focuses on the increasing demand for energy production while maintaining environmental sustainability and efficiency. By harnessing the power of modern technologies such as artificial intelligence (AI), machine learning (ML), and deep learning (DL), this research explores how ML techniques can be leveraged as powerful tools for the power industry. By showcasing practical applications and success stories, this paper demonstrates the growing acceptance of machine learning as a significant technology for current and future business needs in the power sector. Additionally, the study examines the barriers and difficulties of large‐scale ML deployment in practical settings while exploring potential opportunities for these tactics. Through this overview, insights into the transformative potential of ML in shaping the future of power system asset management are provided. [ABSTRACT FROM AUTHOR]

Multienergy load forecasting (MELF) with high accuracy is crucial for the economic operation and optimal dispatch of the integrated energy system (IES). Within such systems, electrical, heat, and cold loads may exhibit complex and highly coupled relationships. The accuracy of MELF can be improved by exploiting the coupling of multienergy loads. To address this issue, the authors propose a novel framework named CAFormer (coupling auxiliary transformer) that leverages coupling auxiliary forecasting and multitask learning (MTL) to improve MELF in IES. CAFormer adopts an auxiliary forecasting coupling strategy. The aim is to construct coupling auxiliary sequences by projecting different sequences onto the same coupling space to capture the interdependencies between them. The model space structure is more complex when incorporating the coupling space. Furthermore, the authors' framework deviates from the traditional approach of treating the prediction task of multiple sequences as multiple tasks by integrating the contrastive learning task and the prediction task into a single multitask paradigm. This avoids the issue of sequence entanglement that arises when different loads directly share information in the traditional multitask prediction of multiple energy loads. MTL based on coupling space similarity helps to learn the coupling relationships between sequences while maintaining the original sequence characteristics and refining the generation of coupling auxiliary sequences. The experimental results demonstrate that the authors' proposed model can thoroughly explore the coupling relationships among energy systems and exhibit higher prediction accuracy and better prediction applicability than those of state‐of‐the‐art models. [ABSTRACT FROM AUTHOR]

The role of surge arresters is vital in power systems' protection against transient overvoltages, which can arise due to occurrences like lightning strikes or switching operations. Based on the performance of the surge arrester, assessing its condition is an essential requirement for early detection of any damage or deterioration. This article proposes innovative criteria based on the surge arresters' leakage current analysis for the classification of uniform and non‐uniform pollution levels. The leakage current signal of clean, uniform, and non‐uniform polluted surge arresters has been examined under different contamination levels, humidity, and non‐uniformity degrees. Based on the experimental results analysis, a new criterion has been proposed for assessing the surge arrester condition, which has resulted from the fifth and first‐order harmonics ratio of resistive leakage current. Analysis of the proposed criterion demonstrates that this index has a significant correlation to the pollution level. The proposed criterion ability has been established based on random forest, support vector machine, decision tree algorithm, and multi‐layer perceptron in Python language. The obtained criterion based on experimental data in different conditions has been employed for classifier training, which can be applied in the power network for better condition monitoring of surge arresters ensuring fast and accurate outcomes. [ABSTRACT FROM AUTHOR]

The increasing demand for electricity in daily life highlights the need for Smart Cities (SC) to use energy efficiently. Both technical and Non‐Technical Losses (NTL), particularly those resulting from electricity theft, present powerful obstacles; NTL alone can reach billions of dollars. Although Machine Learning (ML) based approaches for NTL detection have been embraced by numerous utilities, there is still a lack of thorough analysis of these methods. Limited research exists on NTL identification evaluation criteria and unbalanced data management in the context of SC. This research compares ML algorithms and data balancing methods to optimize electricity consumption detection. The given research applied the 15 ML techniques of Logistic regression, Bernoulli naive Bayes, Gaussian naive Bayes, K‐Nearest Neighbour, perceptron, passive‐aggressive classifier, quadratic discriminant analysis, SGD classifier, ridge classifier, linear discriminant analysis, decision tree, nearest centroid classifier, multi‐nomial naive Bayes, complement naive Bayes and dummy classifier. While SMOTE, AdaSyn, NRAS, and CCR are considered for data balancing. AUC, F1‐score, and seven relevant performance metrics were used for comparison. We have also implemented SHapely Additive exPlanations (SHAP) for feature importance and model interpretation. Results show varying classifier performance with different balancing methods, emphasizing data preprocessing's role in NTL detection for smart grid security. [ABSTRACT FROM AUTHOR]

This study proposes a methodology based on multivariate analysis and data mining techniques that credibly captures the signature of the DG‐islanding phenomenon under adverse operating conditions and network faults. This methodology produces decision trees which determine the tripping logic, protection handles and thresholds for each DG‐islanding relay within the distribution network under study. Case study results indicated that the intelligent islanding relay (IIR) produced by the proposed methodology is consistently high‐level performance in terms of dependability and security, and features reduced non‐detection zones compared with the currently used islanding devices. It is also demonstrated how to determine adaptive settings to accommodate entire ranges of system operating conditions applicable to different ranges of power mismatches (overbalance, underbalanced and near balanced) at the point of common coupling bus. Experimental validation for prove‐of‐concept of the proposed IIR using hardware‐in‐the‐loop has been conducted which was consistent with the off‐line test results. [ABSTRACT FROM AUTHOR]

Considering the high spatial complexity of signal processing methods and the low diversity between the base classifiers in the ensemble classifier, a new method with optimal multi‐resolution fast S‐transform (OMFST) with low spatial complexity and rotation forest (ROF) was proposed. Firstly, Gini importance of features is evaluated by random forest (RF), and the sequence forward search method was adopted for feature selection. Then, the intermediate matrix was constructed on the optimal feature set. The results of inverse fast Fourier transform in the main frequency points of signals based on OMFST are compressed in time domain. The intermediate matrix was used to extract features for power quality (PQ) disturbances recognition. Finally, the ROF was adopted to encourage simultaneously individual accuracy and diversity within the ensemble. The optimal ROF was applied to identify 17 kinds of PQ disturbance signals. The simulation results show that the new method can effectively compress the time–frequency matrix of the existing S‐transform (ST) method; the space complexity of the ST modular matrix is reduced significantly and has higher accuracy. Besides, the results of the experiment with real PQ data prove that the new method was effective for practical industrial applications. [ABSTRACT FROM AUTHOR]

The distributed and privacy‐preserving attributes of fine‐grained smart grid data create obstacles to data sharing. As a result, federated learning emerges as an effective strategy for collaborative training in distributed load forecasting. However, poisoning attacks can interfere with training in the federated learning aggregation process, making it challenging to ensure the accuracy and safety of the global model in load forecasting. Therefore, the authors propose a secure aggregation federated learning method based on similarity and distance (Fed‐SAD) for distributed load forecasting. The server determines approximate global model parameters based on the similarity of the model parameters of the participants and aggregates the global model parameters using a distance‐based weighting method. Fed‐SAD effectively reduces the interference of poisoning attacks by securely aggregating models in short‐term load forecasting. Experimental results show that using the Fed‐SAD results in mean absolute percentage error (MAPE) reduction for certain participants by 19% on sign flipping attacks, MAPE reduction by 15% on additive noise attacks, and MAPE reduction by 4.5% without attack, compared to the Federated Average Aggregation algorithm. Furthermore, Fed‐SAD consistently maintains robustness and has excellent attack detection accuracy. [ABSTRACT FROM AUTHOR]

This study investigates the problem of fault protection in a microgrid containing inverter‐based distributed generators (IBDGs). Owing to the low magnitude of short circuit currents generated by IBDGs, traditional protection techniques which relay on current (fuses and overcurrent relays) may fail to protect such networks. This study addresses the problem of finding suitable features derived from local electrical measurements that can be used by statistical classifiers to better discriminate fault events from normal network events. Given a series of simple electrical features, a study of feature selection and data mining techniques is conducted in the context of fault detection in isolated microgrids with IBDGs. Two statistical classifiers are compared and implemented in this framework: Naive Bayes and decision trees. The proposed approach is tested on a facility scale microgrid consisting of three IBDGs. [ABSTRACT FROM AUTHOR]

The quality of electric power has become an important issue for electric utility companies and their customers. With the extensive application of micro‐grid technologies, power quality (PQ) disturbances are more likely to affect users; thus, research on the recognition of PQ disturbances has attracted increased attention. This study presents a novel PQ disturbance‐recognition algorithm, based on time‐frequency (TF) analysis and a decision tree (DT) classifier. The proposed method requires fewer feature statistics compared to the S‐transform‐based approach for PQ disturbance identification. In this study, feature statistics extracted using TF analysis are trained by a DT classifier to perform the automatic classification of PQ disturbances. As the proposed methodology can efficiently identify PQ disturbances, the performance of the DT classifier can be ensured. In addition, the influence of noise is investigated, and 12 types of noisy disturbance, with signal‐to‐noise ratios of 30–50 dB, are considered for the classification problem. Finally, the proposed method is compared with other popular proposed disturbance‐recognition algorithms in terms of detection accuracy. The experimental results reveal that the proposed method can effectively detect and classify different PQ disturbances. [ABSTRACT FROM AUTHOR]

The social economy is growing rapidly, and the power grid load demand is increasing. To maintain the stability of the power grid, it is crucial to achieve accurate and rapid power system stability assessment. In the actual operation of the power network, data loss is an unavoidable situation. However, most of the data‐driven models currently used assume that the input data is complete, which has obvious limitations in real‐world applications. This paper suggests an IVS‐GAN model to assess power system stability using incomplete phasor measurement unit measurement data with random loss. The proposed method combines the super‐resolution perception technology based on generative adversarial network (GAN) with a time‐series signal classification model. The generator adopts a 1D U‐Net network and uses convolutional layers to complete and recover missing data. The discriminator adopts a new gated recurrent unit–attention architecture proposed here to better extract voltage temporal variation features on key buses. The result of this paper is that the stability evaluation method outperforms other algorithms in high voltage data loss rates on the New England 10‐machine 39‐bus system. [ABSTRACT FROM AUTHOR]

Voltage stability assessment based on machine learning has become an important challenge in power systems. This paper presents real‐time short‐term voltage stability (STVS) assessment based on phasor measurement unit (PMU) data and machine learning (ML). The database is created through time series of measurement data to involve system time‐temporal and dynamics. Then multiple operating states of the power system are classified through the calculation of the Lyapunov exponent and dynamic voltage index according to the database. This paper presents a weighted combination of random forest (RF) and LightGBM (LGBM) classifiers to train a time‐series database. One of the main advantages of this paper is using the gradient concept in data preprocessing, which has enhanced performance metrics and reduced the defect of data noise. Also, hyperparameter optimization is conducted to improve machine performance. Studies on the IEEE 118bus and a real local grid (RLG) demonstrate that the proposed method improves the performance metrics such as accuracy and F1‐score. Also, this approach is robust against PMU data noise and topology changes in the network. [ABSTRACT FROM AUTHOR]

Fast and accurate identification of coherent generator groups is helpful in dynamic and transient stability analysis as well as other applications such as controlled islanding. In this study, a new method is presented for predicting the generators' grouping scheme based on the data measured before and in a short time after the disturbance occurrence. To do that, a classifier model is trained using a training dataset. In the training dataset, the input is the attributes, which are obtained directly or indirectly from the data measured by phasor measurement units. On the other hand, the target in the training dataset is the generators' grouping, which in this study is calculated using a new method called subtractive technique. In subtractive technique, coherent generator groups are determined based on the generator density values. When the classifier model is built using the training dataset, it can be used for online applications. In this study, the well‐known 68‐bus, 16‐machine power system as well as the Iranian 400 and 230 kV south east regional grid are used as the test systems for investigating the efficiency of proposed coherent group prediction method. Results show that the proposed method can predict the generators' grouping scheme with high accuracy. [ABSTRACT FROM AUTHOR]

This paper presents a research work which focuses on generating synthetic data to enrich the training‐set of a deep learning (DL) based classification system to classify power system transient events using PMU frequency measurements. The synthetically improved training‐set is shown to increase the classification performance compared to the case when only the actual‐data training‐set is used. The proposed classification system helps to reveal high‐frequency transient variation information out of PMU measurements collected at a relatively much lower rate, especially when a small set of training‐data exists. Synthetic PMU frequency data is generated based on the DFT analysis statistics on the limited‐size PMU frequency data. Generation of the synthetic data is achieved by re‐synthesis of the PMU frequency data using inverse DFT, which imitates the DFT frequency and phase behaviour for each event type separately. Then the DL structure is trained to classify the power system transients using the synthetically enriched train set. The proposed method of generating synthetically supported training‐set has lower computational complexity compared to the existing methods in the literature and helps to obtain improved classification results. It can be used to increase the classification performances of power quality devices performing transient event analysis, especially for those with access to a limited set of training‐set. [ABSTRACT FROM AUTHOR]

Technical and non‐technical losses in distribution circuits result in significant economic costs to power utilities. One type of non‐technical loss is energy theft by various means including illegal tapping of feeders, bypassing the meter, and billing fraud. These losses are usually hard to detect, and can remain undetected for long periods of time. Machine learning models have been proven effective in detecting these conditions, but rely on the availability of large, good‐quality training data sets. The problem is exacerbated by the imbalanced nature of data related to these conditions—energy theft, though costly, is very rare. The available data sets generally have very few samples of theft with most of the data pertaining to normal operation. Such data sets are generally not suitable to train machine learning models. In this paper, an overview of energy theft detection techniques, the challenges with their data needs, and the limitations of current techniques to bridge such data limitations is presented. A co‐simulation framework is proposed to generate reliable training data for machine learning algorithms for theft detection. An example scenario is presented and a machine learning model is built to detect certain kinds of energy theft. [ABSTRACT FROM AUTHOR]

The classical formulation of the transmission switching problem as a mixed‐integer problem is intractable for large systems in real‐time control settings. Several heuristics have been proposed in the past to speed up the computation time, which only limits the number of switchable lines. In this paper, a real‐time switching heuristic based on neural networks that provides almost instantaneous switching actions, are presented. The findings are shown on case studies of the IEEE 118‐bus test system, and the results show that the proposed heuristic is robust to out of distribution data. Additionally, the proposed heuristic has significant computational savings while all other performance metrics like accuracy are similar to state‐of‐the‐art machine learning methods proposed for transmission switching. [ABSTRACT FROM AUTHOR]

Stochastic renewable sources of energy have been causing changes in the structure and operation of power systems. High penetration of this type of generation results in decreased inertia of a power system, increased active power fluctuations, and a higher probability of false tripping of emergency control devices. Traditional algorithms of emergency control are not adaptable and flexible enough for systems with high penetration of renewables and flexible alternating current transmission systems. Integration and development of phasor measurement units make it possible to create adaptable emergency control systems, which would require minimal pre‐defined data. The purpose of this study is to develop an adaptable algorithm for turbine fast valving control synthesis and transient stability estimation for a generator. The suggested algorithm is based on the equal area criterion in the domain synchronous generator "torque–load angle". The measurements of the generator operation under consideration are used as the input data for the steam turbine fast valving control synthesis. Thus, the algorithm becomes adaptable because no pre‐defined parameters of a power system model are required. The suggested algorithm was tested on the power system model NE39bus using Matlab/Simulink. The efficiency of the suggested algorithm is verified and demonstrated. [ABSTRACT FROM AUTHOR]

With the distribution of the large‐scale emergence of wind power connected to the power system, the stability problem of small disturbance has become increasingly prominent, especially the low‐frequency oscillation phenomenon caused by the weak or the negative damping. Aiming at the problem of probabilistic stability of small disturbance in power system connected with wind power, this paper proposes a method of probabilistic stability based on a variational Bayes and Lyapunov theory using PMU data. This method aims to minimize the computational efficiency when the number of wide‐area time series PMU data is large. Firstly, PMU data are classified by the Random Forest Gini classification index. Secondly, a variational Bayesian framework is introduced to improve the computational efficiency. And then, to obtain the probabilistic stability criterion, this framework is combined with the direct Lyapunov theory. Finally, simulation results show that the method proposed in this paper greatly improves the system probabilistic stability of small disturbance. The comparison between the traditional Monte Carlo based on the Markov chain method and the method in this paper has been achieved, which could provide the theoretical support for the study of the low‐frequency oscillation significantly. [ABSTRACT FROM AUTHOR]

Because of frequent changes in the topologies of distribution networks, the aging of lines and insufficient monitoring capacity compared to the transmission grid, the topology and line parameters are difficult to determine. Here, a hybrid data and model‐driven method is proposed for identifying the topologies and line parameters of distribution networks in the absence of voltage‐angle measurements. First, a topology identification model based on an attention mechanism and convolutional neural networks is constructed as an upper‐layer model, which requires a small number of voltage measurement snapshots of key nodes to identify the current topology. Second, a model‐driven two‐stage line‐parameter identification model is constructed as the lower‐layer model. Case studies based on the IEEE33 and PG&E69 node distribution systems are performed to validate the proposed method. The results confirm that the proposed method is effective for a stable and accurate identification of the topology and line parameters of distribution networks within a certain error range. In addition, the proposed method exhibits a reasonable generalization capability, and all models are equally applicable to both radial and ring networks. [ABSTRACT FROM AUTHOR]

Energy forecasting plays a vital role in mitigating challenges in data rich smart grid (SG) systems involving various applications such as demand‐side management, load shedding, and optimum dispatch. Managing efficient forecasting while ensuring the least possible prediction error is one of the main challenges posed in the grid today, considering the uncertainty in SG data. This paper presents a comprehensive and application‐oriented review of state‐of‐the‐art forecasting methods for SG systems along with recent developments in probabilistic deep learning (PDL). Traditional point forecasting methods including statistical, machine learning (ML), and deep learning (DL) are extensively investigated in terms of their applicability to energy forecasting. In addition, the significance of hybrid and data pre‐processing techniques to support forecasting performance is also studied. A comparative case study using the Victorian electricity consumption in Australia and American electric power (AEP) datasets is conducted to analyze the performance of deterministic and probabilistic forecasting methods. The analysis demonstrates higher efficacy of DL methods with appropriate hyper‐parameter tuning when sample sizes are larger and involve nonlinear patterns. Furthermore, PDL methods are found to achieve at least 60% lower prediction errors in comparison to other benchmark DL methods. However, the execution time increases significantly for PDL methods due to large sample space and a tradeoff between computational performance and forecasting accuracy needs to be maintained. [ABSTRACT FROM AUTHOR]

Due to the expansion of the decarbonization in modern power systems, particularly the high integration of renewable energy resources, more and more uncertainties and variabilities are brought to electricity systems. These uncertainties change the system operating conditions and face the power system with unknown issues. The secure operation of modern power systems requires developing security assessment techniques that effectively mitigate the variation of the operating conditions. This paper proposes a novel continuous learning scheme for online static security assessment (SSA) that enables the periodic renovation of the assessment model. The scheme consists of a Mondrian forest (MF)‐based training model and database generation approach conducting characteristics of operating conditions. The employed security index is a new weather‐depended performance index whose ambient temperature can affect line ratings. After the new training set is generated in the update procedure, the MF algorithm is updated using this data. Eventually, the SSA model is enriched with current circumstances in the power system. The proposed scheme is examined using real‐world data mapped to the New England 39‐bus and 118‐bus standard test systems. The effectiveness and superior performance of the proposed method are demonstrated using various criteria and measures. [ABSTRACT FROM AUTHOR]

To solve the difficulty of eigenvalue extraction of on‐load tap‐changer (OLTC) vibration signals, an eigenvalue extraction method of time‐frequency matrix based on Hilbert energy spectrum and spectral entropy is proposed in this study. By optimizing the Hilbert‐Huang transform (HHT) algorithm, the Hilbert energy spectrum of the vibration signal based on the time‐frequency distribution is obtained. The spectrum entropy of Hilbert energy spectrum was extracted to characterize the chaotic degree of vibration signal energy distribution, and avoided the problem of inaccurate feature quantity such as similarity due to the minute time difference during signal acquisition. A new time‐frequency matrix was constructed by frequency band division and singular value decomposition (SVD) was carried out. The singular value vector obtained contains the time, frequency and energy information, such as vibration frequency distribution and time‐frequency plane energy distribution, which represent the essential features of the original signal. The mechanical states of OLTC were characterized by Hilbert spectrum entropy and singular value. The features of the measured signals under the three typical faults of the driving mechanism are input into the intelligent multi‐classification support vector machine (DAG‐SVM) for pattern classification. The results show that this method can accurately identify the mechanical state of OLTC, and has practical application significance. [ABSTRACT FROM AUTHOR]

To maintain synchronous operation of interconnected systems, it is necessary to keep balance between power generation and consumption. Faults may result in large oscillations in power flows and loss‐of‐synchronism between one generator and the rest of the power system. A correlation coefficient is a useful statistic that can be used to detect voltage, frequency or angular instabilities, and to determine an asynchronous process. Hence, it is a smart technique to predict the out‐of‐step event following fault situations. In this paper, the technique based on the auto/cross‐correlation and instantaneous powers is presented to identify sudden disturbances of various electrical signals in the case of asynchronous operation. To validate the method performance, a power network with real parameters is simulated on Alternative Transient Program (ATP) platform, and the algorithm is implemented in MATLAB software. Simulation studies reveal that the proposed scheme is able to detect the out‐of‐step conditions upon which the relay issues a trip signal, yet remains inactive under normal synchronized operating conditions. Thus, the out‐of‐step is declared before the second pole slipping occurs, and an instability time can be accurately assessed. Moreover, it can develop novel quadrilateral operating characteristics to discriminate between the synchronous and asynchronous operation of the AC generator. [ABSTRACT FROM AUTHOR]

The high renewable penetration will cause the power system operation mode (PSOM) to change frequently. At present, the selection of PSOM mainly depends on the experience of relevant staff. However, selecting PSOM based on experience is difficult to comprehensively and elaborately analyze the possible PSOMs within a dispatching cycle, which inevitably invites troubles to the power system planning and dispatching. In this paper, a data‐driven analysis method is proposed, which integrates the techniques of preprocessing, dimensionality reduction, clustering, and data visualization of high‐dimensional data sets of power system operation. Simultaneously, qualitative and quantitative methods are used to analyze, evaluate the PSOM and their development laws. Finally, to verify the effectiveness of the proposed multi‐level refinement extraction method of PSOM, the normal power system operation mode (NPSOM), transition power system operation mode (TPSOM), extreme power system operation mode (EPSOM) are obtained and analyzed according to the real operation data of the urban power system. The abundant analysis results show that the proposed fine‐grained extraction method of PSOM is compelling, facilitating the power grid enterprise operation and planning. [ABSTRACT FROM AUTHOR]

Accurate detection and classification of power quality (PQ) disturbances is an essential prerequisite for PQ mitigation. To address this issue, a new PQ assessment framework based on modified empirical wavelet transform (MEWT) and light gradient boosting machine (LightGBM) is proposed here. First, the frequency estimation mechanism and band segmentation rule of empirical wavelet transform (EWT) is modified, which makes EWT suitable for analysing complex PQ signals. Second, based on the PQ disturbance analysis results of MEWT, 8 characteristic curves are defined and statistical features are extracted in both the time and frequency domain. Third, binary relevance‐based LightGBM (BR‐LightGBM) is designed for the multilabel classification of massive PQ events. Considering the impact of input features and model structure, feature selection and hyper parameter optimization are conducted for achieving better classification performance. Finally, extensive experiments based on synthetic data and two groups of measured data show the effectiveness of the proposed method on 48 types of complex PQ disturbances. Compared with other algorithms for multiple PQ signals, the proposed method is fast for computation and performs better in classification accuracy and robustness, which is a prospective alternative for the PQ monitoring system. [ABSTRACT FROM AUTHOR]

Type identification and time location of power quality disturbances (PQDs) is the key to adopting corresponding measures to suppress disturbances. More complex multiple disturbances caused by the overlapping of different micro‐grids make it a challenging task. The paper proposes a hybrid approach combing KF‐ML (Kalman filter based on maximum likelihood) with deep belief network (DBN) for dealing with PQDs. To be specific, the KF‐ML is firstly applied to reduce noise from the original distorted signal, and the innovation sequence obtained by KF‐ML can be used to locate starting–ending times of PQDs. Then, the DBN, which fuses feature extraction and classification into a single block, is capable of recognizing the type of PQDs accurately. To verify the effectiveness of the proposed method, 20 classes of PQDs with noise interference are tested, and experiment results show that the detection time of the proposed method is very close to the set time, and the absolute error of time location is less than 0.3 ms. The average classification accuracy at different noise levels reaches about 95%, and is very high even with more disturbances combined. Thus, the proposed method is immune to noise and less affected with more disturbances combined relative to other methods. [ABSTRACT FROM AUTHOR]

This paper presents a novel, sequentially executed supervised machine learning‐based electric theft detection framework using a Jaya‐optimized combined Kernel and Tree Boosting (KTBoost) classifier. It utilizes the intelligence of the XGBoost algorithm to estimate the missing values in the acquired dataset during the data pre‐processing phase. An oversampling algorithm based on the Robust‐SMOTE technique is utilized to avoid the unbalanced data class distribution issue. Afterward, with the aid of few very significant statistical, temporal, and spectral features extracted from the acquired kWh dataset, the complex underlying data patterns are comprehended to enhance the accuracy and detection rate of the classifier. For effectively classifying the consumers into "Honest" and "Fraudster," the ensemble machine learning‐based classifier KTBoost, with Jaya algorithm optimized hyperparameters, is utilized. Finally, the developed model is re‐trained using a reduced set of highly important features to minimize the computational resources without compromising the performance of the developed model. The outcome of this study reveals that the proposed theft detection method achieves the highest accuracy (93.38%), precision (95%), and recall (93.18%) among all the studied methods, thus signifying its importance in the studied area of research. [ABSTRACT FROM AUTHOR]

Sensing and measurement systems are quintessential to the safe and reliable operation of electric power grids. Their strategic placement is of ultimate importance because it is not economically viable to install measurement systems on every node and branch of a power grid, though they need to be monitored. An overwhelming number of strategies have been developed to meet oftentimes multiple conflicting objectives. The prime challenge in formulating the problem lies in developing a heuristic or an optimisation model that, though mathematically tractable and constrained in cost, leads to trustworthy technical solutions. Further, large‐scale, long‐term deployments pose additional challenges because the boundary conditions change as technologies evolve. For instance, the advent of new technologies in sensing and measurement, as well as in communications and networking, might impact the cost and performance of available solutions and shift initially set conditions. Also, the placement strategies developed for transmission grids might not be suitable for distribution grids, and vice versa, because of unique characteristics; therefore, the strategies need to be flexible, to a certain extent, because no two power grids are alike. Despite the extensive literature on the present topic, the focus of published works tends to be on a specific subject, such as the optimal placement of measurements to ensure observability in transmission grids. There is a dearth of work providing a comprehensive picture for developing optimal placement strategies. Because of the ongoing efforts on the modernisation of electric power grids, there is a need to consolidate the status quo while exposing its limitations to inform policymakers, industry stakeholders, and researchers on the research‐and‐development needs to push the boundaries for innovation. Accordingly, this paper first reviews the state‐of‐the‐art considering both transmission and distribution grids. Then, it consolidates the key factors to be considered in the problem formulation. Finally, it provides a set of perspectives on the measurement placement problem, and it concludes with future research directions. [ABSTRACT FROM AUTHOR]

This paper presents a non‐intrusive method for identifying the load state of a distribution network. The method focuses on continuously varying loads. By considering the load on‐off state switching points and the continuous features at on state, a deep convolutional method considering non‐local spatiotemporal features is proposed. The addition of an attention component to the convolutional network enhances the non‐local feature extraction capability of the convolutional network. Ultimately, the effectiveness of the method is demonstrated in an experimental setting. In addition, this paper demonstrates that the proposed method can effectively integrate switching point features as well as persistent features through neural network visualization techniques. [ABSTRACT FROM AUTHOR]

In order to extract the fault characteristics of the distribution power system automatically, a new method for fault identification based on parameter optimized variational mode decomposition (VMD) and convolutional neural network (CNN) is proposed. First, the VMD is used to decompose electric signals of the main transformer, including the current of the low‐voltage inlet line, bus voltage, and so on. The obtained intrinsic mode functions should be arranged in sequence from low to high according to the central frequency. The VMD can obtain the time‐frequency matrices. In order to decide the decomposition layers of the VMD adaptively, the partial mean of multi‐scale permutation entropy is used to optimize it. Then, the time‐frequency matrixes can be converted into pixel matrices by pseudo‐colour coding. The pixel matrices which are appropriately compressed can be used as input of CNN. The CNN can autonomously extract the eigenvectors of fault. Finally, the classification algorithms of support vector machine (SVM), naive Bayesian classifier (NBC), extreme learning machine (ELM), and random forest (RF) are used to train and test the feature vectors extracted by CNN to verify the effectiveness of the method. The experiments show that the proposed method has excellent stability, fast convergence speed, and high precision. The technique can effectively complete the fault identification of the distribution network. [ABSTRACT FROM AUTHOR]

Power outages due to cascading failures which are triggered by extreme weather pose an increasing risk to modern societies and draw attention to an emerging need for power network resilience. Machine learning (ML) is used for a real‐time selection process on preventive actions, such as topology reconfiguration and islanding, aiming to reduce the risk of cascading failures. Training data is obtained from Monte Carlo simulations of cascading failures triggered by extreme events. The trained ML‐based decision‐making process uses only predictors that are readily available prior to an extreme event, such as event location and intensity, network topology and load, and requires no further time‐consuming simulations.The proposed decision‐making process is compared to time‐consuming but ideal decision‐making and fast but trivial decision‐making. Demonstrations on the German transmission network show that the proposed ML‐based selection process efficiently prevents the uncontrolled propagation of cascading failures and performs similarly to an ideal decision‐making process whilst being computationally three orders of magnitude faster. [ABSTRACT FROM AUTHOR]

To enable reliable power delivery through transmission tie‐lines, total transfer capability (TTC) must be calculated and regulated to accommodate the transferred amount. However, the traditional optimal power flow (OPF)‐based total transfer capability calculation is computationally expensive for efficient total transfer capability control due to the inclusion of a large set of differential‐algebraic equations (DAEs) to verify transient stability constraints. In order to enable practicable total transfer capability regulation, a novel risk‐aware deep learning‐assisted paradigm is proposed here. First, a deep belief network (DBN) is employed to establish the total transfer capability predictor and surrogate the computation‐intensive differential‐algebraic equations in original optimal power flow formulas, simplifying the high‐dimensional and intractable constraints deep belief networks without loss of nonlinearity. Particularly, in order to be aware of control risk from the predictive error of the deep belief networks, prediction intervals (PIs) are produced improved by using ensemble learning and used to disclose the probability of insufficient actions, further guaranteeing the sufficient and cost‐effective control by compromising the tradeoff between cost and risk. Symbiotic organisms search (SOS) is then applied to solve the proposed risk‐aware deep belief network‐assisted total transfer capability control problem globally. The numerical studies testify that the proposed method enables economical, reliable, and full nonlinearity‐retained dynamic total transfer capability regulation control within a risk‐free surrogate‐assisted and tractable physical model‐driven hybrid framework. [ABSTRACT FROM AUTHOR]

This work presents a hybrid state estimation procedure that allows power system dynamics associated to slow and fast transient events to be accurately monitored considering a limited amount of phasor measurement units. The proposal represents an enhanced version of a previously developed static state estimation approach. This time, two main changes are introduced. First, a weighted least absolute value estimator instead the conventional weighted least square technique is used to estimate bus voltages in transient regime. This makes the estimation process more robust at fast scan rates. Second, an extended Kalman filter based dynamic state estimator is integrated. Thus, the proposed scheme is able to estimate, along with static variables, dynamic variables of all generators and motors in the power system regardless of the availability of phasor measurement unit measurements at terminal buses. This is possible thanks to a novel data‐mining based methodology for full phasor measurement unit observability restoration. Performance parameters such as accuracy, computing time and convergence properties are assessed by applying the estimation procedure to the New England benchmark system under different operating conditions. [ABSTRACT FROM AUTHOR]

This paper proposes a new islanding detection technique based on the superimposed angle of negative sequence impedance. The proposed technique is designed with the presence of different types of RES (renewable energy sources) in the system. The proposed technique discriminates the islanding events from non‐islanding events even at 0% power mismatch. It detects the islanding events within a half‐cycle even at the zero power mismatch and also restricts the undesirable tripping due to the several kinds of non‐islanding events. The technique is also able to detect islanding under network reconfiguration. For the range of power mismatches, extensive cases of islanding and non‐islanding events have been simulated. The technique has been illustrated on a 7‐bus reconfigurable microgrid test system with different types of RES in the (RTDS/RSCAD) environment. In this work, islanding has been determined considering each type of RES as targeted DG. The technique has also been demonstrated on a modified IEEE‐34 bus system. A comparison of the obtained results has also been performed. [ABSTRACT FROM AUTHOR]

This paper presents a new concise deep learning–based sequence model to detect the power quality disturbances (PQD), which only uses original signals and does not require pre‐processing and complex artificial feature extraction process. A simple gated recurrent network (SGRN) with a new recurrent cell structure is developed, which consists of only two gates: forget gate and input gate, and two weight matrices. Compared with the standard Recurrent Neural Network (RNN) model, the training process of the proposed method is more stable and the prediction accuracy is higher. In addition, this special structure retains basic non‐linearity and long‐term memory, while enabling the simple gated recurrent network model to be superior to Long Short‐Term Memory (LSTM) Network and Gated Recurrent Unit (GRU) Network in terms of the number of parameters (i.e. memory cost) and detection speed. In the light of the experimental results, the simple gated recurrent network algorithm can achieve 99.07% detection accuracy, and contains only 18,959 parameters, which indicates that our proposed method is easier to deploy in resource‐constrained internet of things (IoT) micro‐controllers. [ABSTRACT FROM AUTHOR]

Typhoons have substantial impacts on power systems and may result in major power outages for distribution network users. Developing prediction models for the number of users going through typhoon power outages is a high priority to support restoration planning. This study proposes a data‐driven model to predict the number of distribution network users that may experience power outages when a typhoon passes by. To improve the accuracy of the prediction model, twenty six explanatory variables from meteorological factors, geographical factors and power grid factors are considered. In addition, the authors compared the application effect of five different machine learning regression algorithms, including linear regression, support vector regression, classification and regression tree, gradient boosting decision tree and random forest (RF). It turns out that the RF algorithm shows the best performance. The simulation indicates that the accuracy of the optimal model error within ±30% can reach up to 86%. The proposed method can improve the prediction accuracy through continuous learning on the existing basis. The prediction results can provide efficient guidance for emergency preparedness during typhoon disaster, and can be used as a basis to notify the distribution network users who are likely to lose power. [ABSTRACT FROM AUTHOR]

In actual power system operation, there is a significant imbalance between the number of stable and unstable samples, and the misclassification costs of the two classes of samples are different. In addition to the class imbalance, there is another imbalance of regional samples in the feature space. In order to reduce the impact of class imbalance and regional imbalance on the performance of the model, a method of dual cost‐sensitivity factors for transient stability assessment is proposed. The method sets the balance factor and modulating factor in the loss function of light gradient boosting machine. The former corrects the classification bias caused by class imbalance, and the latter focuses on improving the classification accuracy of overlapping area samples. Combining the two factors, the model not only improves the accuracy of unstable samples, but also reduces the misjudgement of stable samples. In online application, a fast update scheme with memory function is proposed. In this scheme, incremental learning is used to update the model with fewer samples and less computational time, so as to achieve better evaluation performance. Case studies on three power systems demonstrate the generalisation performance of the proposed model and the effectiveness of the update scheme. [ABSTRACT FROM AUTHOR]

An affordable, stable, reliable and sustainable power supply is the need of the hour, keeping in view the journey to a smart green grid. The present study and derivation of adaptive critical angle threshold (ACAT) contributes to this effect. The ACAT serves as a key index and an indicator to grid operator in proactive mitigation of angular instability in the grid, which is seen to be one of the major initiator of blackouts worldwide. The synchrophasor technology enables accurate calculation of ACAT on real time basis by time synchronised sampling of local voltage and current phasors. The real time calculation makes the index adaptive to variation in system topologies as well as system vulnerabilities. Furthermore, the local system model formed for deriving ACAT has less computational burden while capturing the dynamics of external system, without requirement of intricate power system details. The angular separation of the station bus under supervision can be monitored with respect to the critical angle and serve as a key indicator for initiation of preventive actions during contingencies. The devised methodology is tested using standard IEEE test systems as well as a practical 400/220 kV state grid system using ETAP/MATLAB. [ABSTRACT FROM AUTHOR]

This study proposes a novel machine learning‐based methodology to estimate the air‐conditioning (AC) load from the hourly smart meter data. The commonly employed approaches for disaggregating the share of the AC load from the total consumption are either using data obtained from dedicated sensors or high‐frequency data that cannot be provided by conventional smart meters. In the present work, an alternative approach is proposed, in which a machine learning‐based pipeline is first optimised and trained using the data obtained from a set of households in a period, including both smart meter data and the AC load measurements along with corresponding weather conditions. The obtained optimal pipeline is then utilised to estimate the AC load in another set of buildings in the same period of the year, while providing it with only the smart meter data and weather conditions. As the first step of the pipeline implementation, several features are extracted from the smart meters and weather data. The most promising algorithm is then determined, the corresponding hyper‐parameters are optimised and the most influential parameters are finally determined. The proposed method leads to a lower monitoring system's cost, lower user privacy concerns and fewer data processing complexity compared to the conventional energy disaggregation approaches, while providing an acceptable accuracy. [ABSTRACT FROM AUTHOR]

The conventional first zone characteristic of distance protection is constructed mainly based on engineering experience, which limits its coverage of internal faults. The zone coverage is also adversely affected by the installation of series compensation devices with over‐voltage protection. Despite research efforts to overcome these problems, developing an algorithm for constructing an optimal first zone characteristic that encloses only all separable internal faults remains a challenging task. This study addresses this issue by presenting an optimisation‐based algorithm for constructing an adaptive optimal first zone characteristic for power lines. The algorithm takes into account different locations of the compensation devices along the line, their non‐linear protection, and the inaccuracies in measurements and considered grid model. The proposed and conventional zone construction algorithms are compared using a grid model in MATLAB Simulink. The results demonstrate that the proposed algorithm is robust to input parameters and thus able to construct an optimal zone characteristic for a specified range of grid operating points and fault parameters. Moreover, the first zone obtained by the proposed algorithm covers substantially more internal faults than the conventional zone with optimised settings. [ABSTRACT FROM AUTHOR]

For power system transient stability assessment (TSA), the consequences of misclassifications of unstable samples are more serious than those of stable samples. Therefore, the cost-sensitive method is introduced into TSA. A constant weight is set for all unstable training samples by the conventional cost-sensitive method. However, the fault severities of the training samples are different. Thus, the impacts of the samples with different fault severities on assessment rules are also different. Therefore, an improved cost-sensitive method based on the fault severity is proposed. Firstly, the fault severity of each sample is calculated by its fault clearing time. On this basis, different cost-sensitive coefficients are assigned to the samples with different fault severities. The weights of critical samples are larger than those of non-critical ones. Then, a stacked sparse auto-encoder model based on the improved cost-sensitive method is built through modifying the loss function. Finally, the proposed method is tested in an actual system. Compared with the traditional cost-sensitive method, the results show that the improved cost-sensitive method has better performance. Also, it has the advantages of the low false rate of unstable samples, high-assessment accuracy, and strong anti-noise ability. [ABSTRACT FROM AUTHOR]