Against the backdrop of renewable energy integration and the large-scale integration of power electronic devices into power systems, the coupling of power quality issues within the system has become increasingly complex, leading to the formation of complex power quality disturbances (PQDs) characterized by the combined occurrence of multiple disturbances. Reliable and effective identification of disturbance types forms the foundation for the classified mitigation strategies of PQDs. This article proposes a data processing method, the variable-length distribution transfer field (VLF), which processes variable-length PQDs signal inputs, overcoming the limitations of fixed-length signal processing. VLF can process variable-length PQDs signals in a single pass, enhancing the algorithm’s tolerance to fluctuating computation times and enabling effective utilization of signals of arbitrary lengths. It thereby resolves the discontinuity issues caused by signal overwriting or loss due to fixed-length constraints. Concurrently, targeting the characteristics of PQDs, this article designs an edge-enhanced encoding method capable of achieving enhanced boundary discrimination and noise-robust encoding for disturbances. Furthermore, this article proposes a high-generalization self-balanced multitask classification architecture that enables loss balancing among different tasks within a multilabel and multitask classification framework. Through the above optimization, simulation results under 30 dB to 50 dB noise environments and twentyfold downsampling intensity demonstrate a classification accuracy of 99.57%. Meanwhile, experimental simulations combined with real-world grid validations still achieve a high accuracy of 99.43%. The classification algorithm exhibits strong robustness and generalization capability under highly downsampling and noisy conditions.

Developing robust Alzheimer's Disease (AD) classification models necessitates extensive training data, but aggregating multi-center medical data poses privacy risks. Although Federated Learning (FL) and Swarm Learning (SL) allow training generic models without data sharing, their performance is limited by variations in AD pathology features and sample class imbalances across centers. To address this issue, we propose a novel Harmonized Swarm Learning framework with Guided Optimization (HSGO) to enhance multi-center collaboration while preserving data privacy. Our framework employs a class-balanced loss function to train a robust generic model and guides the optimization of personalized models towards the generic model, eliminating extra AD pathology feature extraction steps. Furthermore, we design a dynamic feature similarity storage mechanism to facilitate personalized training. Experiments performed under two different multi-center data partitioning scenarios demonstrate that HSGO achieves competitive performance when compared with five baseline methods. Additionally, Layer-wise Relevance Propagation (LRP) analysis indicates that HSGO may help identify potential key brain regions in AD by integrating local and global features compared to traditional SL.

Complex power quality disturbances (PQDs) are caused by the combination of multiple single disturbances. There exist complex mutual exclusion and combination relationships among these single disturbances, so-called mutually-exclusive labels mean that the disturbances corresponding to these labels cannot occur simultaneously (such as disturbance C2:Sag and C3:Swell). However, in the multilabel classification neural network, the fully connected layer and activation function corresponding to each label are identical, and there is no design to accommodate the mutually-exclusive and combination relationships, which allows the multilabel classification output mutually-exclusive disturbances simultaneously, which could mistakenly activate two incompatible governance devices at the same time. This is one of the reasons why this method cannot be applied in the power system, which requires an ultra-safe identification approach. A multilabel and multitask classification method with robust generalization ability and universal applicability is proposed to realize the mutually-exclusive output of PQDs mutually-exclusive labels. In principle, this method cannot output mutually-exclusive labels at the same time, to achieve an ultra-safe output. Meanwhile, this article proposes a faster and more accurate calculation method for PQDs envelope by using fast plug Hilbert transform, and uses Landau level transition field proposed in this article to achieve further feature extraction of the envelope and convert it into a two-dimensional (2-D) image, and then uses the proposed shallow double-path neural network DPN-17 to achieve classification and identification of PQDs. This methodology has strong universality and can be extensively utilized in various 1-D signal identification and feature extraction fields.

Complex power quality disturbances (PQDs) is the concomitant occurrence of multiple single disturbances, which is a common disturbance in modern power system. In this article, a new plug Hilbert transform method is proposed, and the rotation vector modulus corresponding to the complex PQDs is obtained by using the back projection. Then, the characteristic signal of the complex PQD is obtained by tiling the modulus according to the time passing direction. Based on the Markov transfer chain and fuzzy logic, a new dimension conversion scheme, fuzzy transfer field (FTF), is proposed, and the disturbances characteristic signals are converted into two-dimensional (2-D) images by using this FTF, and the 2-D images covering complex PQDs information are classified and recognized by depth residual network. The accuracy and practical effect of this method is far superior to similar research. At the same time, the abrupt change and edge problem of the Hilbert transform is solved in principle, and the problem of ignoring the dependence of time sequence signal amplitude in the existing image conversion methods is pointed out and solved. These methods have strong universal applicability, not only for the recognition of PQDs signal, but also for other industrial signal and medical signal processing recognition.

Along with the widespread use of face recognition systems, their vulnerability has become highlighted. While existing face anti-spoofing methods can be generalized between attack types, generic solutions are still challenging due to the diversity of spoof characteristics. Recently, the spoof trace disentanglement framework has shown great potential for coping with both seen and unseen spoof scenarios, but the performance is largely restricted by the single-modal input. This paper focuses on this issue and presents a multi-modal disentanglement model which targetedly learns polysemantic spoof traces for more accurate and robust generic attack detection. In particular, based on the adversarial learning mechanism, a two-stream disentangling network is designed to estimate spoof patterns from the RGB and depth inputs, respectively. In this case, it captures complementary spoofing clues inhering in different attacks. Furthermore, a fusion module is exploited, which recalibrates both representations at multiple stages to promote the disentanglement in each individual modality. It then performs cross-modality aggregation to deliver a more comprehensive spoof trace representation for prediction. Extensive evaluations are conducted on multiple benchmarks, demonstrating that learning polysemantic spoof traces favorably contributes to anti-spoofing with more perceptible and interpretable results.

This paper investigates the robust distributed cruise control problem of multiple high-speed trains under external disturbances. First, by modeling each train as a cascade of point masses connected by spring-like couplers, the longitudinal interaction between adjacent cars are represented by the connected topological graph. Then, under the framework of the communication-based train control technology, the interaction of desirable speed information among trains and the wayside control center is described by the directed topological graph. Next, a distributed cruise controller is designed by taking advantages of the graphic theory such that the multiple trains track different target speeds, and both the distance of neighboring cars and the headway of successive trains are kept in appropriate ranges. Finally, to eliminate the influence of external disturbances, we adopt the disturbance observer to approximate the perturbations, and present a sufficient condition for the existence of the distributed control strategy and the observer gain parameter in form of the linear matrix inequality (LMI). Numerical experiments illustrate that the composite control law is effective in inhibiting the external disturbances, and guaranteeing the safety, efficiency and comfort of high-speed trains' movement.

Recently, researchers set an ambitious goal of conducting speaker recognition in unconstrained conditions where the variations on ambient, channel and emotion could be arbitrary. However, most publicly available datasets are collected under constrained environments, i.e., with little noise and limited channel variation. These datasets tend to deliver over-optimistic performance and do not meet the request of research on speaker recognition in unconstrained conditions.In this paper, we present CN-Celeb, a large-scale speaker recognition dataset collected ‘in the wild’. This dataset contains more than 130,000 utterances from 1,000 Chinese celebrities, and covers 11 different genres in real world. Experiments conducted with two state-of-the-art speaker recognition approaches (i-vector and x-vector) show that the performance on CN-Celeb is far inferior to the one obtained on Vox-Celeb, a widely used speaker recognition dataset. This result demonstrates that in real-life conditions, the performance of existing techniques might be much worse than it was thought. Our database is free for researchers and can be downloaded from http://project.cslt.org.

This paper proposes a highly efficient approach for power disturbance waveform (PDW) compression in the view of Heisenberg uncertainty. The key idea is to represent each signal component of PDW using as few nonzero coefficients as possible by the uncertainty principle restriction. PDWs are projected in a union of bases (UB), and each signal component of the PDWs can be represented very sparsely. The UB decomposition is solved by orthogonal matching pursuit. The features and cross correlation of subbases of the UB guarantee the PDW compression with high a compression ratio and recovered accuracy. With various simulated and field PDWs tests, the compressed data size of the new method is proven with good characteristics such as low sensitivity to sampling frequency increment and types of signal components contained in PDWs. Moreover, it is found that the new method and methods that employ wavelet techniques share the similar effect of noise for PDW compression. The proposed method is also applied at a 220-kV power substation for field PDW compression from fault recorders. The comparisons, analyses, and experiments indicate that the proposed method has a high PDW compression efficiency for future power grid monitoring.

Power quality events are usually associated with more than one disturbance and their recognition is typically based on multilabel learning. In this study, we propose a new method for recognizing multiple power quality disturbances (MPQDs) based on variational mode decomposition (VMD) and a random discriminative projection extreme learning machine for multilabel learning (RDPEML). First, VMD is employed to decompose the MPQDs into several intrinsic mode functions and the standard energy differences of each mode are extracted as features that form the input vectors of the classifier. Second, a novel multilabel classifier called RDPEML is constructed by combining a random discriminative projection multiclass extreme learning machine (ELM) and a thresholding learning method-based kernel ELM. In order to obtain better classification performance, a tenfold cross-validation embedded particle swarm optimization approach is utilized to search for the optimal values of the structural parameters. Finally, a test study was conducted using MATLAB synthetic signals and real signals sampled from a three-phase standard source under different noise conditions. Compared with the several recent state-of-the-art multilabel learning algorithms, RDPEML achieved better classification performance with superior computational speed.

In this paper, a new classification approach for detection and classification of complex power quality disturbances (PQDs) using a three-level multiply connected Bayesian network is proposed. First, the model consisting of features evidence layer, disturbances state layer, and circumstance evidence layer is established, which represent the features extracted from the sample signal, the state of each single label of PQDs and the circumstance factors that may affect the PQDs, respectively. Second, the parameters of three-level multiply connected Bayesian network (TLBN) are studied from statistical data and Monte Carlo simulations. Finally, the classification is determined by computing the posterior marginal probabilities of each event given observed evidences. The new method not only utilizes the existing features extracting methods, but also takes the historical data, and other surrounding factors into account. Simulation results and real-life PQ signal tests show that the performances of TLBN classification of complex disturbances are better than the other approaches in existing literatures.

In this paper, with recent advances in cooperative and cognitive wireless networks, we propose a CBTC system to enable train-train direct communications. In addition, the proposed system is optimized with the cognitive control method. Unlike the exiting works on cooperative wireless networks, in this paper, train control performance in CBTC systems is explicitly used as the performance measure in the design. Reinforcement learning is applied to obtain the optimal handover decision and adaption policy of communication parameters. Simulation result shows that the performance of train control can be improved significantly in our proposed CBTC system.

The safe operation of Chinese Train Control System Level 3 (CTCS-3) is of great significance, particularly with respect to the increasing operational speed and expanding railway networks of the Chinese high-speed railway system, which has drawn deep concern from both customers and strategic makers. A hazard and operability (HAZOP) study has been recognized as an effective systematic examination to identify any potential problems existing in various industrial processes or operations. This paper presents a process of applying a HAZOP study to identify the hazards of a CTCS-3 onboard system for the first time, which is composed of two major parts: system models and hazard identification on the examination session. To better reflect the structure and functions of a CTCS-3 onboard system, the following models are developed, i.e., a reference model, a function hierarchical model, a state diagram, and a sequence diagram. To demonstrate the effectiveness of the HAZOP study, hazard identification on the basis of the functions of the CTCS-3 onboard system and a scenario of temporary speed restriction is considered and employed. The results indicate that existing hazards can be dug out at express speed, which allows the relevant actions to be proposed and implemented to prevent the hazards from spreading to a wide range in the whole CTCS-3 onboard system.

The Onboard system is a typical safety-critical system, in which any fault can lead to huge human injury or wealth losing. Function testing method which is mainly focus on the conformance relation between the specification and the SUT has been widely used in testing the Onboard system in the past few years. However, most of the test cases are manually generated which can't be reused and leads to repeat works when the specification is changed. To improve the testing efficiency and quality, Model-based testing method is introduced. We use a tool chain to generate test case automatically based on Timed Automata theory and apply in function testing of the Onboard system. EBD-TR timed automata network model is established using tool Uppaal. And based on the EBD-TR model, two kinds of coverage criteria (all-location coverage, and all-edge coverage) are used in tool of CoVer to generate test case automatically. Different test suits of the Onboard system are acquired and a complete model transition function test suit is derived which is proven very useful for testing the Onboard system.

The importance of the specification of train control system is increasingly recognized and gained more attention in signalling field in China as the specification is the basis to ensure that the signalling system supplied by manufacturer meet the requirements of the railway administration, for example, the requirement for the interoperability of the system. The specifications which are described in natural language are probably deficient and it is inadequate that the specifications are checked only based on the experience of experts. In this paper, a modelling method was applied on the description and the verification of the System Requirement Specification, SRS, of train control system. The Specification and Description Language, SDL, was used to describe the functional behavior of the onboard equipment which is defined in the SRS. First, we defined the principles of modelling which are summarized for the purpose of the interoperability validation. Then, we applied a top-down hierarchical approach to model the functional behavior. The model started from the system level to describe the interface and refined in the block level to show the interaction of system scenarios and detailed the state transition and the working processes of the system in the process level. Debugged the SDL model, we validated the model in Telelogic Tau tool to find the problems of the SRS. The results showed that the ambiguous terms and the incompatible descriptions of the SRS can be found. It can be helpful for the modification of the SRS and the quality of train control system further.