The proposed work accomplishes detection and classification of simple and complex power quality disturbances occurring in the realm of wind-grid integrated system using fast time-time transform and small residual-extreme learning machine. Taking the advantage of time-time transform, this work has further tailored time-time transform by accommodating the dyadic scaling to make it a computationally less complex and faster technique. Here, threephase power quality signals are first segmented into single phases to be analyzed with fast time-time transform for getting the event information in timetime domain. The event information is further passed through two-stage feature extraction process to bring out twenty two characteristic curves; each of them is further used to compute five statistical features to produce 107 features after ignoring three redundant features. Thus, a high-dimensional feature set of size 321 is obtained corresponding to three phases of voltage signal. Because, high-dimensional feature set demands a robust classifier that can accurately classify the disturbances even in the presence of noise, small residual-extreme learning machine has been trained to classify the power quality signal into 12 classes (symmetrical sag, asymmetrical sag, swell, unbalance, harmonics, notch, momentary interruption, normal voltage signal, sag with harmonics, swell with harmonics, sag and notch with harmonics, and swell and notch with harmonics). This simulation-based study has been proved effective in dealing with noisy power quality signals also with reduced computational complexity and higher classification rate. KEYWORDS fast time-time transform, power quality, small residual-extreme learning machine classifier, wind-grid integrated system List of symbols and abbreviations: h(t), a time signal; w(t), Gaussian window; σ, standard deviation; f, frequency; β, output weight of SR-ELM network; G, output of hidden layer of SR-ELM network; X, input to SR-ELM; L, number of neurons in hidden layer of SR-ELM; M, number of neurons in output layer of SR-ELM; e, error; C. B.

Recently, power quality (PQ) issues have drawn considerable attention of the researchers due to the increasing awareness of the customers towards power quality. The PQ issues maintain its preeminence because of the significant growth encountered in the smart grid technology, distributed generation, usage of sensitive and power electronic equipments with the integration of renewable energy resources. The IoT and 5G networks technologies have a number of advantages like smart sensor interfacing, remote sensing and monitoring, data transmission at high speed. Due to this, applications of these two are highly adopted in smart grid. The prime focus of the paper is to present an exhaustive survey of detection and classification of power quality disturbances by discussing signal processing techniques and artificial intelligence tools with their respective pros and cons. Further, critical analysis of automatic recognition techniques for the concerned field is posited with the viewpoint of the types of power input signal (synthetic/real/noisy), pre-processing tools, feature selection methods, artificial intelligence techniques and modes of operation (online/offline) as per the reported articles. The present work also elaborates the future scope of the said field for the reader. This paper provides valuable guidelines to the researchers those having interest in the field of PQ analysis and exploring the better methodologies for further improvement. Comprehensive comparisons have been presented with the help of tabular presentations. Although this critical survey cannot be collectively exhaustive, still this survey comprises the most significant works in the concerned paradigm by examining more than 300 research publications.

Summary This work introduces an optimized wavelet designed by fractionally delaying the coefficients of the unit delay filter. The wavelets produced by optimized fractionally delayed filter coefficients are named as fraclets. Fraclet has been utilized in the real‐time power quality events' detection and classification which has been hitherto addressed with the help of unit delay filters. Normal signal, voltage sag, swell, harmonics, sag with harmonics, swell with harmonics, sag and swell with harmonics, and sag and swell with interruption are considered in this work to validate the performance of proposed algorithm based on fraclet. Along with real‐time generation of various power quality events with TMS320C6748 DSP board, different events have also been simulated by using parametric equations. The upper hand of fraclets over wavelets has been highlighted by maximally flat frequency response of fraclets. Moreover, fraclets facilitate better multiresolution analysis by providing low energy compaction ratio (ECR). Further, 11 characteristic features have also been extracted from each decomposition level of PQ signal up to fifth level and fed to the modified probabilistic neural network (MPNN) for validating the proposed power quality event detection and classification algorithm. MPNN has outperformed the support vector machine (SVM), both with fraclet and wavelet. The proposed algorithm relying on fraclets has shown better results as compared with wavelet.

In this work, a new fractionally delayed biorthogonal wavelet is designed for recognition of voltage sag causes. This work addresses the classification of voltage sag causes into three categories, i.e., fault events (namely line-to-ground fault, line-to-line fault, double-line-to-ground fault and symmetrical fault), induction motor starting and transformer energization. Fractionally delayed biorthogonal Coiflet wavelet of order 2 (named as Coiflet fraclet) is designed using Lagrange interpolation and employed for the multiresolution analysis of voltage sag signals. To achieve higher classification accuracy, event information is derived from both time-frequency domain and frequency domain. Frequency information is obtained by applying multiple signal classification (MUSIC) algorithm on voltage sag signals. Both the time-frequency domain and frequency domain features are concatenated to implement the feature level fusion for strengthening the feature set. The feature set obtained after feature level fusion is used for training the recursive reduced kernel based extreme learning machine which addresses the problem of highly complex structure due to huge dataset of large dimensions. The classifier has shown robustness to the presence of noise also and classified the voltage sag causes with high accuracy.

No abstract

To combat cyber threats in the smart grid, an intrusion detection system can be integrated into the advanced metering infrastructure. Anomaly-based intrusion detection can detect even the tiniest changes in the parameter under investigation, whereas signature-based intrusion detection only recognises known attacks. The growing usage of smart grids necessitates the classification, identification, and implementation of countermeasures to threats. At the absolute least, smart grids must be protected against cyberattacks; thus, the highest level of information security must be offered. As a result of digitisation and the usage of more smart applications, the research looked at a variety of attack types, smart grid assaults, and major cyber threats on the voltage regulation. Machine learning techniques that analyse data in real time and formulate patterns to recognise an attack and scan through huge data for anomalies can be implemented into the advanced metering infrastructure (AMI) for intrusion detection for anomaly-based intrusion detection. The comparative test study done for the research found that the proposed method, median absolute deviation for anomaly identification in smart metering datasets, produced the most accurate and precise differentiations with the highest accuracy and precision. The median absolute deviation (MAD) algorithm model is trained using test data, and raw predictions are made, before true data are used to derive final test result parameters, precision, recall, and F1 scores. The methodology of the entire study is discussed in this paper, as well as how the MAD algorithm is best suited for anomaly-based intrusion detection, as well as result comparisons of other machine learning algorithms.

No abstract

No abstract