Automatic Power Quality Events Classifier based on hybrid CNN–LSTM network and multisine fitting algorithm

This paper presents a novel approach for automatic power quality (PQ) event detection and classification based on Stockwell transform (S-transform) and wild goat optimization (WGO)-tuned extreme learning machine (ELM). The distinctive features associated with PQ event signals have been extracted by S-transform to obtain the feature vectors characterizing the signal nature. Considering these feature vectors as input, a classifier based on ELM optimally tuned with modified WGO technique is proposed. The WGO technique originated from the social hierarchy and strategic planning to reach at peak by the wild goats in nature is adapted to formulate an effective ELM model by parameter tuning for better classification. To justify the enhanced performance of the proposed approach, it is tested on a wide range of extracted synthetic PQ event data by MATLAB simulation. To ensure the real-time implementation, the PQ event data with the addition of 20, 30, and 50 dB to the synthetic signals are considered. The analysis of results presented reveals a very high performance for both PQ event recognition and classification, ensuring the efficiency of the proposed approach.

Widespread use of renewable energy resources and electronic equipment has aggravated the power quality issues. Delivery of the undistorted sinusoidal rated voltage and current at rated frequency are the main expectations from electrical power system. Thus, continuous detection and localization of power quality disturbances has become a major concern for transmission and distribution networks to identify sources of disturbance and to circumvent equipment damage. However, automatic Power Quality (PQ) event localization is a challenging concern for both utility and industries. A novel technique based on Discrete Wavelet Transform (DWT) has been proposed for online PQ event detection and offline near-by area localization in distribution networks. Line current signal features based on DWT are extracted and the proposed scheme is evaluated by MATLAB simulation for a typical 415 V benchmark Low Voltage AC (LVAC) microgrid. The algorithm performance like steady-state, sensitivity and selectivity is validated in presence and absence of distributed energy resources (DERs) for various PQ events. Thus, manual effort involved in PQ disturbance location and source rectification could be minimized once a defined near-by area in the entire distribution network is identified.

In this paper, Hilbert Huang transform (HHT) and weighted bidirectional extreme learning machine (WBELM) are integrated to detect and classify power quality events (PQEs) in real time. Empirical mode decomposition is used to decompose the nonstationary PQEs into the monocomponent mode of oscillation, known as intrinsic mode functions (IMFs). The efficacious features are extracted by applying the Hilbert transform (HT) on the IMFs. An efficient WBELM computational intelligence technique is proposed to recognize the single, as well as multiple PQEs and its performances are compared with the recently developed classifiers such as support vector machine, least-square support vector machine, extreme learning machine, and bidirectional extreme learning machine. The recognition architecture of HHT integrated with WBELM (HHT-WBELM) method is tested and compared with the empirical wavelet transform associated with HT and WBELM method, and tunable-Q wavelet transform along with HT and WBELM method. The faster learning speed, lesser computational complexity, superior classification accuracy, and short event detection time prove that the proposed HHT-WBELM method can be implemented in the online power quality monitoring system. Finally, a hardware prototype is developed based on the digital signal processor to verify the cogency of the proposed method in real time. The feasibility of the proposed method is tested and validated by both the simulation and laboratory experiments.

The new power generation systems, the increasing number of equipment connected to the power grid, and the introduction of technologies such as the smart grid, underline the importance and complexity of the Power Quality (PQ) evaluation. In this scenario, an Automatic PQ Events Classifier (APQEC) that detects, segments, and classifies the anomaly in the power signal is needed for the timely intervention and maintenance of the grid. Due to the extension and complexity of the network, the number of points to be monitored is large, making the cost of the infrastructure unreasonable. To reduce the cost, a new architecture for an APQEC is proposed. This architecture is composed of several Locally Distributed Nodes (LDNs) and a Central Classification Unit (CCU). The LDNs are in charge of the acquisition, the detection of PQ events, and the segmentation of the power signal. Instead, the CCU receives the information from the nodes to classify the PQ events. A low-computational capability characterizes low-cost LDNs. For this reason, a suitable PQ event detection and segmentation method with low resource requirements is proposed. It is based on the use of a sliding observation window that establishes a reasonable time interval, which is also useful for signal classification and the multi-sine fitting algorithm to decompose the input signal in harmonic components. These components can be compared with established threshold values to detect if a PQ event occurs. Only in this case, the signal is sent to the CCU for the classification; otherwise, it is discarded. Numerical tests are performed to set the sliding window size and observe the behavior of the proposed method with the main PQ events presented in the literature, even when the SNR varies. Experimental results confirm the effectiveness of the proposal, highlighting the correspondence with numerical results and the reduced execution time when compared to FFT-based methods.

In recent years Power Quality issues vital performance measurement in smart grids like as voltage disturbance occurring in switching events or faults on power system. This paper proposes an automatic and Real time Power Quality disturbance classification voltage leakage, Transients, Harmonics and Interruptions based on statistical moments in the time domain and support vector machines (SVM) classifier. An online algorithm is utilized to update the statistical moments in real time. The proposed method has the advantage of giving faster response time and lesser memory requirements when compared to the wavelet-based methods. A standard model comprising of seven classes of Power Quality disturbances is used. The proposed method provides better results. An analysis is presented for various simulation measures demonstrate with superiority of the proposed approach. The use of SVM simplifies the power quality classification task and is implemented by MATLAB/Simulink.

The paper presents a comparative study to investigate the optimum feature selection using three signal processing techniques for automatic clustering of power quality events. The techniques include the wavelet transform, the S transform, and the newly introduced forward Clarke transform. The last method has the advantage for monitoring all three phases of a three-phase signal simultaneously. The paper provides unique features for each transformation, and then offers a comparative study that is based on the abilities of selected pairs of features to distinguish power quality events. In the paper, the performance of each signal processing technique is studied and an optimum combination of the most useful features is identified.

Automatic Power Quality Events Classifier Based on Hybrid Cnn-Lstm Network and Multisine Fitting Algorithm

This paper implements a deep learning‐based modulation pattern recognition algorithm for communication signals using a convolutional neural network architecture as a modulation recognizer. In this paper, a multiple‐parallel complex convolutional neural network architecture is proposed to meet the demand of complex baseband processing of all‐digital communication signals. The architecture learns the structured features of the real and imaginary parts of the baseband signal through parallel branches and fuses them at the output according to certain rules to obtain the final output, which realizes the fitting process to the complex numerical mapping. By comparing and analyzing several commonly used time‐frequency analysis methods, a time‐frequency analysis method that can well highlight the differences between different signal modulation patterns is selected to convert the time‐frequency map into a digital image that can be processed by a deep network. In order to fully extract the spatial and temporal characteristics of the signal, the CLP algorithm of the CNN network and LSTM network in parallel is proposed. The CNN network and LSTM network are used to extract the spatial features and temporal features of the signal, respectively, and the fusion of the two features as well as the classification is performed. Finally, the optimal model and parameters are obtained through the design of the modulation recognizer based on the convolutional neural network and the performance analysis of the convolutional neural network model. The simulation experimental results show that the improved convolutional neural network can produce certain performance gains in radio signal modulation style recognition. This promotes the application of machine learning algorithms in the field of radio signal modulation pattern recognition.

With an increasing usage of sensitive electronic equipment power quality has become a major concern now. One critical aspect of power quality studies is the ability to perform automatic power quality data analysis and categorization. The objective of this paper is to present a technique based on fuzzy logic to categorize power quality events and to simulate the impact of power quality on sensitive equipment. Inherent features are extracted from recorded waveforms using Fourier and wavelet analyses and fed into a fuzzy expert system. The categorization technique has been implemented using the Fourier, Wavelet and fuzzy logic toolboxes in MATLAB and tested with real power quality measured data. The impact of power quality on the operation of sensitive equipment has been illustrated through simulations in MATLAB SIMULINK. Such study is essential to predict the performance of modern loads and also to be able to explain why a specific load fails during a power quality event. The findings are reported in detail in this paper.

Analyzing power quality data manually requires a lot of time and special expertise. In this paper, a new approach is shown for automatic identification power quality disturbance on mobile phone by combining the S-transform, a support vector machine (SVM), wireless markup language (WML) and the Matlab web server (MWS). The S-transform is used to generate a set of features for the SVM to classify various power quality disturbances; while the WML and MWS are used to integrate the graphical and computational process through the mobile phone. Results obtained proved the robustness and effectiveness of the proposed method.

The automatic detection and classification of power quality disturbances has become a significant issue in modern power industry, because of electric load sensitive to power transient signal. This paper presents a novel approach for detection and location of power quality disturbances based on wavelet transform and artificial neural network. The wavelet transform is the projection of a discrete signal into two spaces: the approximation space and a series of detail spaces. The implementation of the projection operation is done by discrete-time subband decomposition of input signals using filtering followed by downsampling. The wavelet transform is utilized to produce representative feature vectors that can accurately capture the characteristics of power quality disturbance, exploring feature extraction of disturbance signal to obtain dynamic parameters. The feature vector obtained from wavelet decomposition coefficients are utilized as input variables of neural network for pattern classification of power quality disturbances. The training algorithm shows great potential for automatic power quality monitoring technique with on-line detection and classification capabilities. The combination performance of wavelet transform with neural network is evaluated by simulation results, approving that the proposed method is effective for analysis of power quality signal.

Microgrids operating in islanded mode are more prone to fundamental frequency variations and power quality distortions. To mitigate power quality problems, it is essential to first identify the type of distortion using a proper classifier. There are many classifiers in the literature. However, they are tested assuming that there is no fundamental frequency variation. In this paper, we in the effect of fundamental frequency variations on classification accuracy. For that purpose, a well-known classifier is tested with data sets of different fundamental frequencies. Then, accuracies are compared using statistical tests. To the best of our knowledge, this is the first work adopting this approach. The results of the comparison show that changes in fundamental frequency greatly affect classification accuracy. A large decrease occurs even with moderate frequency deviations. Therefore, future studies should consider this effect, since non-adapted classifiers may perform poorly in weak microgrids operating in islanded mode.

The automatic sound event classification (SEC) has attracted a growing attention in recent years. Feature extraction is a critical factor in SEC system, and the deep neural network (DNN) algorithms have achieved the state-of-the-art performance for SEC. The extreme learning machine-based auto-encoder (ELM-AE) is a new deep learning algorithm, which has both an excellent representation performance and very fast training procedure. However, ELM-AE suffers from the problem of unstability. In this work, a bilinear multi-column ELM-AE (B-MC-ELM-AE) algorithm is proposed to improve the robustness, stability, and feature representation of the original ELM-AE, which is then applied to learn feature representation of sound signals. Moreover, a B-MC-ELM-AE and two-stage ensemble learning (TsEL)-based feature learning and classification framework is then developed to perform the robust and effective SEC. The experimental results on the Real World Computing Partnership Sound Scene Database show that the proposed SEC framework outperforms the state-of-the-art DNN algorithm.

Breast cancer is a deadly disease having high mortality rate from several years. It is second and fourth leading disease in the world and India respectively as per the WHO. The conventional techniques are unsupervised to classify breast cancer that involves erroneous, laborious and demanding inevitable presence of clinician. It is also experimented on small dataset and the accuracy of the previous classifier methods was unsatisfactory. To overcome these problems, we have experimented on large dataset and extracted several features such as area, convex area, bounding box, eccentricity, orientation, solidity, and perimeter, contour based fractal dimension etc. These feature set describes the size and geometrical shape of the tumor. The increase in feature set leads to increase in the accuracy of the classification. The automatic classification is based on multilayer back propagation artificial neural networks (ANN) algorithm. The breast cancer tumors have an important clue in its boundary, hence analysis of that plays a vital role for better identification of disease. The dataset is split into training and testing data on an around 1700 samples using 80-20 rule with different neural network architectures. Hence the accuracy of 98.11% has been achieved in the classification rate. The successful classification depends on the quality of the enhanced mammograms, localization of tumor and accurate segmentation. The image samples from MIAS, DDSM and local hospitals had been involved in the experiment.

Breast cancer is a deadly disease having high mortality rate from several years. It is second and fourth leading disease in the world and India respectively as per the WHO. The conventional techniques are unsupervised to classify breast cancer that involves erroneous, laborious and demanding inevitable presence of clinician. It is also experimented on small dataset and the accuracy of the previous classifier methods was unsatisfactory. To overcome these problems, we have experimented on large dataset and extracted several features such as area, convex area, bounding box, eccentricity, orientation, solidity, and perimeter, contour based fractal dimension etc. These feature set describes the size and geometrical shape of the tumor. The increase in feature set leads to increase in the accuracy of the classification. The automatic classification is based on multilayer back propagation artificial neural networks (ANN) algorithm. The breast cancer tumors have an important clue in its boundary, hence analysis of that plays a vital role for better identification of disease. The dataset is split into training and testing data on an around 1700 samples using 80-20 rule with different neural network architectures. Hence the accuracy of 98.11% has been achieved in the classification rate. The successful classification depends on the quality of the enhanced mammograms, localization of tumor and accurate segmentation. The image samples from MIAS, DDSM and local hospitals had been involved in the experiment.