Open-Circuit Fault Detection and Classification of Modular Multilevel Converters in High Voltage Direct Current Systems (MMC-HVDC) with Long Short-Term Memory (LSTM) Method

MVDR-LSTM Distance Estimation Model Based on Diagonal Double Rectangular Array

Deep learning algorithms have the advantages of a powerful time series prediction ability and the real-time processing of massive samples of big data. Herein, a new roller fault distance estimation method is proposed to address the problems of the simple structure and long conveying distance of belt conveyors. In this method, a diagonal double rectangular microphone array is used as the acquisition device, minimum variance distortionless response (MVDR) and long short-term memory network (LSTM) are used as the processing models, and the roller fault distance data are classified to complete the estimation of the idler fault distance. The experimental results showed that this method could achieve high-accuracy fault distance identification in a noisy environment and had better accuracy than the conventional beamforming algorithm (CBF)-LSTM and functional beamforming algorithm (FBF)-LSTM. In addition, this method could also be applied to other industrial testing fields and has a wide range of application prospects.

Generating Low-Earth Orbit Satellite Attitude Maneuver Profiles Using Deep Neural Networks

To perform Earth observations, low-Earth orbit (LEO) satellites require attitude maneuvers, which can be classified into two types: maintenance of a target-pointing attitude and maneuvering between target-pointing attitudes. The former depends on the observation target, while the latter has nonlinear characteristics and must consider various conditions. Therefore, generating an optimal reference attitude profile is difficult. Mission performance and satellite antenna position-to-ground communication are also determined by the maneuver profile between the target-pointing attitudes. Generating a reference maneuver profile with small errors before target pointing can enhance the quality of the observation images and increase the maximum possible number of missions and accuracy of ground contact. Therefore, herein we proposed a technique for optimizing the maneuver profile between target-pointing attitudes based on data-based learning. We used a deep neural network based on bidirectional long short-term memory to model the quaternion profiles of LEO satellites. This model was used to predict the maneuvers between target-pointing attitudes. After predicting the attitude profile, it was differentiated to obtain the time and angular acceleration profiles. The optimal maneuver reference profile was obtained by Bayesian-based optimization. To verify the performance of the proposed technique, the results of maneuvers in the 2-68° range were analyzed.

A Sound and Vibration Fusion Method for Fault Diagnosis of Rolling Bearings under Speed-Varying Conditions

The fault diagnosis of rolling bearings is critical for the reliability assurance of mechanical systems. The operating speeds of the rolling bearings in industrial applications are usually time-varying, and the monitoring data available are difficult to cover all the speeds. Though deep learning techniques have been well developed, the generalization capacity under different working speeds is still challenging. In this paper, a sound and vibration fusion method, named the fusion multiscale convolutional neural network (F-MSCNN), was developed with strong adaptation performance under speed-varying conditions. The F-MSCNN works directly on raw sound and vibration signals. A fusion layer and a multiscale convolutional layer were added at the beginning of the model. With comprehensive information, such as the input, multiscale features are learned for subsequent classification. An experiment on the rolling bearing test bed was carried out, and six datasets under various working speeds were constructed. The results show that the proposed F-MSCNN can achieve high accuracy with stable performance when the speeds of the testing set are the same as or different from the training set. A comparison with other methods on the same datasets also proves the superiority of F-MSCNN in speed generalization. The diagnosis accuracy improves by sound and vibration fusion and multiscale feature learning.

Fault Detection and Diagnosis for Liquid Rocket Engines Based on Long Short-Term Memory and Generative Adversarial Networks

The development of health monitoring technology for liquid rocket engines (LREs) can effectively improve the safety and reliability of launch vehicles, which has important theoretical and engineering significance. Therefore, we propose a fault detection and diagnosis (FDD) method for a large LOX/kerosene rocket engine based on long short-term memory (LSTM) and generative adversarial networks (GANs). Specifically, we first modeled a large LOX/kerosene rocket engine using MATLAB/Simulink and simulated the engine’s normal and fault operation states involving various startup and steady-state stages utilizing fault injection. Second, we created an LSTM-GAN model trained with normal operating data using LSTM as the generator and a multilayer perceptron (MLP) as the discriminator. Third, the test data were input into the discriminator to obtain the discrimination results and realize fault detection. Finally, the test data were input into the generator to obtain the predicted samples and calculate the absolute error between the predicted and the real value of each parameter. Then the fault diagnosis index, standardized absolute error (SAE), was constructed. SAE was analyzed to realize fault diagnosis. The simulated results highlight that the proposed method effectively detects faults in the startup and steady-state processes, and diagnoses the faults in the steady-state process without missing an alarm or being affected by false alarms. Compared with the conventional redline cut-off system (RCS), adaptive threshold algorithm (ATA), and support vector machine (SVM), the fault detection process of LSTM-GAN is more concise and more timely.

Fault Diagnosis Method for MMC-HVDC Based on Bi-GRU Neural Network

The Modular Multilevel Converter-High Voltage Direct Current (MMC-HVDC) system is recognized worldwide as a highly efficient strategy for transporting renewable energy across regions. As most of the MMC-HVDC system electronics are weak against overcurrent, protections of the MMC-HVDC system are the major focus of research. Because of the insufficiencies of the conventioned fault diagnosis method of MMC-HVDC system, such as hand-designed fault thresholds and complex data pre-processing, this paper proposes a new method for fault detection and location based on Bidirectional Gated Recurrent Unit (Bi-GRU). The proposed method has obvious advantages of feature extraction on the bi-directional structure, and it simplifies the pre-processing of fault data. The simplified pre-processing avoids the loss of valid information in the data and helps to extract detailed fault characteristics, thus improving the accuracy of the method. Extensive simulation experiments show that the proposed method meets the speed requirement of MMC-HVDC protections (2 ms) and the accuracy rate reaches 99.9994%. In addition, the method is not affected by noise and has a high potential for practical applications.

Intelligent Fault Diagnosis Framework for Modular Multilevel Converters in HVDC Transmission

Open circuit failure mode in insulated-gate bipolar transistors (IGBT) is one of the most common faults in modular multilevel converters (MMCs). Several techniques for MMC fault diagnosis based on threshold parameters have been proposed, but very few studies have considered artificial intelligence (AI) techniques. Using thresholds has the difficulty of selecting suitable threshold values for different operating conditions. In addition, very little attention has been paid to the importance of developing fast and accurate techniques for the real-life application of open-circuit failures of IGBT fault diagnosis. To achieve high classification accuracy and reduced computation time, a fault diagnosis framework with a combination of the AC-side three-phase current, and the upper and lower bridges’ currents of the MMCs to automatically classify health conditions of MMCs is proposed. In this framework, the principal component analysis (PCA) is used for feature extraction. Then, two classification algorithms—multiclass support vector machine (SVM) based on error-correcting output codes (ECOC) and multinomial logistic regression (MLR)—are used for classification. The effectiveness of the proposed framework is validated by a two-terminal simulation model of the MMC-high-voltage direct current (HVDC) transmission power system using PSCAD/EMTDC software. The simulation results demonstrate that the proposed framework is highly effective in diagnosing the health conditions of MMCs compared to recently published results.

Application of a Pattern-Recognition Neural Network for Detecting Analog Electronic Circuit Faults

In this study, machine learning techniques based on the development of a pattern–recognition neural network were used for fault diagnosis in an analog electronic circuit to detect the individual hard faults (open circuits and short circuits) that may arise in a circuit. The ability to determine faults in the circuit was analyzed through the availability of a small number of measurements in the circuit, as test points are generally not accessible for verifying the behavior of all the components of an electronic circuit. It was shown that, despite the existence of a small number of measurements in the circuit that characterize the existing faults, the network based on pattern-recognition functioned adequately for the detection and classification of the hard faults. In addition, once the neural network has been trained, it can be used to analyze the behavior of the circuit versus variations in its components, with a wider range than that used to develop the neural network, in order to analyze the ability of the ANN to predict situations different from those used to train the ANN and to extract valuable information that may explain the behavior of the circuit.

Machine Learning-Based Condition Monitoring for PV Systems: State of the Art and Future Prospects

To ensure the continuity of electric power generation for photovoltaic systems, condition monitoring frameworks are subject to major enhancements. The continuous uniform delivery of electric power depends entirely on a well-designed condition maintenance program. A just-in-time task to deal with several naturally occurring faults can be correctly undertaken via the cooperation of effective detection, diagnosis, and prognostic analyses. Therefore, the present review first outlines different failure modes to which all photovoltaic systems are subjected, in addition to the essential integrated detection methods and technologies. Then, data-driven paradigms, and their contribution to solving this prediction problem, are also explored. Accordingly, this review primarily investigates the different learning architectures used (i.e., ordinary, hybrid, and ensemble) in relation to their learning frameworks (i.e., traditional and deep learning). It also discusses the extension of machine learning to knowledge-driven approaches, including generative models such as adversarial networks and transfer learning. Finally, this review provides insights into different works to highlight various operating conditions and different numbers and types of failures, and provides links to some publicly available datasets in the field. The clear organization of the abundant information on this subject may result in rigorous guidelines for the trends adopted in the future.

A Systematic Review of Key Challenges in Hybrid HVAC–HVDC Grids

The concept of hybrid high-voltage alternating current (HVAC) and high-voltage direct current (HVDC) grid systems brings a massive advantage to reduce AC line loading, increased utilization of network infrastructure, and lower operational costs. However, it comes with issues, such as integration challenges, control strategies, optimization control, and security. The combined objectives in hybrid HVAC–HVDC grids are to achieve the fast regulation of DC voltage and frequency, optimal power flow, and stable operation during normal and abnormal conditions. The rise in hybrid HVAC–HVDC grids and associated issues are reviewed in this study along with state-of-the-art literature and developments that focus on modeling robust droop control, load frequency control, and DC voltage regulation techniques. The definitions, characteristics, and classifications of key issues are introduced. The paper summaries the key insights of hybrid HVAC–HVDC grids, current developments, and future research directions and prospects, which have led to the evolution of this field. Therefore, the motivation, novelty, and the main contribution of the survey is to comprehensively analyze the integration challenges, implemented control algorithms, employed optimization algorithms, and major security challenges of hybrid HVAC–HVDC systems. Moreover, future research prospects are identified, such as security algorithms’ constraints, dynamic contingency modeling, and cost-effective and reliable operation.