Search

Modulation recognition plays a crucial role in the efficient management of spectrum resources. However, traditional methods have long posed challenges for researchers due to their excessive reliance on manual effort. With the advancement of deep learning, automatic modulation recognition has emerged as a promising solution. Nonetheless, most existing deep learning-based modulation recognition studies consider only single-domain feature information of the signal, such as time-domain or frequency-domain features. To further enhance recognition performance, this paper proposes a novel multi-feature input fusion network. By utilizing different representations and processing methods of the signal, the proposed approach designs distinct feature extraction networks tailored to specific processed signals, leveraging the characteristics of convolution kernels with varying sizes and receptive fields. Additionally, the attention mechanism module is improved to enhance network performance while minimizing the increase in parameter count. Experimental results on the publicly available RML2016.10a dataset demonstrate that the proposed model achieves highly efficient recognition above 2 dB, with accuracy approaching 100%. Testing with real-world signals collected under realistic channel conditions using USRP devices further confirms the model’s superior performance, achieving an effective recognition rate of 99.27% when the two USRP devices are in close proximity. These results validate the model’s high efficiency and robustness in both simulated and real-world environments.

Accurate short-term load forecasting (LF) under extreme weather is vital for the sustainable development of energy systems. This paper proposes a basic framework for future load forecasting researches of sustainable energy systems under extreme weather events and provides new direction for membrane computing model in terms of load forecasting. Inspired by nonlinear spiking mechanisms in nonlinear spiking neural P systems, the gated spiking neural P (GSNP) model is a new recurrent-like network. In this study, we develop an innovative membrane computing model, termed frequency attention temporal convolutional network-load forecasting-frequency attention gated spiking neural P (FATCN-LF-FAGSNP) model. Frequency enhanced channel attention mechanism (FECAM) is utilized to enhance the features extraction ability of temporal convolutional network (TCN) and improve prediction ability of GSNP systems. FATCN fully extracts the temporal relationship of features, the features of each channel interact with each frequency component to learn more temporal information effectively and comprehensively in frequency domain. Moreover, adding FECAM to extract features from the data fully reveals the relationship between influencing factors and the load series, which improves the quality of data features and the forecasting accuracy of the FAGSNP model. Then inspired by the interaction mechanism of impulses between biological neuronal cells, FAGSNP is able to consider the load variability and effectively predict load trends. In addition, to address load prediction challenges posed by extreme weather and promote the sustainable development of power systems, the proposed model integrates many models to solve this problem. First, optimized variational mode decomposition (VMD) is used to decompose the load series and the sub-sequences are combined with relevant features, to form the different input sequences of the prediction model. Then, FATCN-LF-FAGSNP model is developed to accurately forecast each high frequency component. Subsequently inverted Transformer model and Informer model are utilized to predict low frequency components and residual component, respectively. Finally all predicted components are reconstructed to get the final predicted results. We conducted extensive comparative experiments with ten baseline models on three real-world datasets, compared with GSNP model and TCN-GSNP model, the coefficient of determination (R2) of the FATCN-LF-FAGSNP model increases and mean absolute percentage error (MAPE), mean absolute error (MAE) and relative absolute error (RAE) reduce, the LF accuracy (measured by R2) of the proposed hybrid model gets 99.7% in seasonal LF task. In addition, the proposed hybrid model gets the best in MAPE, MAE, R2 and RAE metrics in all cases, which demonstrates the effectiveness of the proposed model in LF tasks under both extreme weather scenarios and seasonal prediction scenarios.

Skin lesion segmentation is critical in medical image processing, but the segmentation task faces numerous challenges due to the differences in size, color, shape, and texture of skin lesions between patients, as well as the blurring of the boundary between lesions and normal skin. While many models improve segmentation performance by introducing complex modules, their computational resource constraints make applications limited in a clinical setting. To cope with this problem, we propose a lightweight multi-fusion network (LMFUNet) with parameters of only 0.100M and GFLOPs of 0.106. LMFUNet uses an Efficient Multi-scale Feature Extraction block (EMFE) in deep stages, which uses grouping of features by convolution with different dilation rates to reduce model complexity and effectively capture multi-scale features. By using the Multi-level Feature Fusion module (MFF) in skip connections, different levels of information are combined step by step, realizing the first step of fusion of low-level details and high-level contextual information, which helps to accurately localize the lesion area. After MFF, we designed the Spatial-channel Fusion module (SCF). This module further optimizes the fusion of feature information output from MFF in terms of spatial and channel dimensions, significantly enhancing the recognition of lesion boundaries. We did in-depth tests on three public datasets that are typical of others: ISIC2017, ISIC2018 and PH2 datasets. LMFUNet demonstrates superior performance because it excels at segmentation and doesn’t require a lot of computing power.

Recent breakthroughs in deep learning applied to voice processing have sparked an upsurge in security and privacy concerns. This paper presents an inventive adversarial sample generation technique termed the “Ultrasonic Attack,” crafted to covertly steer downstream voice-related tasks – encompassing voice emotion classification, voice synthesis, and voice recognition processes such as biometric authentication used in sporting events. This technique is distinctive in its employment of a multi-feature fitting strategy that allows for precise targeting and alteration of key voice attributes critical for downstream tasks. Ingeniously integrating ultrasonic noise into the original vocal recordings, our method can mislead sophisticated deep learning systems while remaining undetectable to the human ear, leading to erroneous outcomes in voice-based applications. The implications of this are particularly profound in high-stakes situations like athlete identity verification or voice command integrity in sports technology systems. Rigorous experimental validations underscore the “Ultrasonic Attack” as a potent method. When juxtaposed with leading-edge adversarial sample generation techniques, our approach stands out, delivering unrivaled performance in tasks as varied as Voice Emotion Classification and Speaker Identification. Our method triumphs in creating adversarial samples that not only carry out successful attacks with enhanced efficacy but also conserve the natural features of the voice, underscoring the critical need for fortified security in speech-processing technologies.

The unified power flow controller (UPFC) based on the modular multilevel converter (MMC) requires stability and fast power-flow-regulation capabilities, whereas the linear active disturbance rejection control (LADRC) strategy precisely has the characteristics of strong disturbance rejection and fast response. Therefore, to improve the dynamic performance of UPFC when regulating the power flow, LADRC is introduced into the UPFC inner-loop controller. Meanwhile, to reduce the observed burden of the linear extended state observer (LESO) in LADRC, a disturbance observer (DOB) is introduced to compensate for the LESO’s disturbance estimation error. The frequency characteristics of the voltage inner loop in MMC-UPFC with improved LADRC are then provided, indicating that the improved DOB-based LADRC has smaller perturbation gain at low frequency, and the cut-off frequency of the transfer function is also increased. Thus, these optimizations effectively improve the dynamic and anti-interference ability of LADRC. The power flow characteristics of MMC-UPFC were also analyzed, providing theoretical guidance for UPFC to regulate the grid power flow. Finally, the UPFC power-flow control strategy and corresponding DOB-based LADRC method are verified in PLECS, showing that the improved method has the advantages of faster response speed and strong anti-interference ability under different working conditions.

System efficiency, power quality, and economic benefits have always been essential for the Unified Power Flow Controller (UPFC based on a Modular Multilevel Converter (MMC). This paper proposes a low-switching-frequency MMC-UPFC operation scheme based on the carrier phase shift pulse width modulation (CPS-PWM), which can significantly reduce switching loss and improve system efficiency. However, in medium-voltage MMC applications, the number of sub-modules in the UPFC is usually small (less than 30) exacerbating the harmonic distortions in the MMC-UPFC and reducing the switching frequency further deteriorates harmonics and increases capacitor voltage ripple. Therefore, to address the increased capacitor voltage ripple caused by low switching frequency, this paper proposes a reasonable carrier ratio (FR) design method. Analyzing and suppressing the capacitor voltage ripple at low switching frequencies can maintain capacitor voltage stability while significantly reducing the switching frequency. To address the issue of increased harmonics due to low switching frequency and few sub-modules, this paper combines the FR design method with an active LLCL filter method. An improved genetic algorithm is used to establish an objective function related to switching frequency harmonics and total inductance value, which can suppress a large number of high-order harmonics remaining from the CPS-PWM scheme, reduce the total inductance value, and compare this method with the single L filtering scheme widely used in UPFC engineering planning. Simulation on PLECS software verifies the effectiveness of the proposed method.

In traditional Wireless Sensor Network routing protocols, data collected through timed interval sensing tends to have high temporal redundancy, which leads to unnecessary energy drain. To alleviate this problem and enable sensor networks to save energy to some extent, a practical solution is to utilize prediction-based data fusion methods. To this end, this paper first proposes a Low Energy Adaptive Clustering Hierarchy-Energy-Kopt-N algorithm, an optimization algorithm explicitly designed to address the cluster-head election phase of the Low Energy Adaptive Clustering Hierarchy protocol. Then, a data collection model using data prediction techniques – the Grey Data Prediction Model is formatted. Combining these improvements, a new data fusion algorithm that relies on data prediction, Grey-Clusters-Leach (GCL), is proposed. Simulation experiments demonstrate that the network energy drain of the GCL algorithm is reduced by 18%, 35%, 21.5% and 20%, and the network operation critical period life is extended by 3%, 35%, 22%, and 5% compared to the EQDC LEACH, LEACH-E, and SEP algorithms, respectively. GCL can effectively manage the size and number of clusters and reduce the number of packet transmissions by 20%.

Exploring the complex meanings in architectural scenic images is crucial for modern intelligent systems. We propose a novel method combining multi-channel perceptual features to semantically understand scenes with intricate spatial structures. Our deep hierarchical model captures human gaze dynamics using a binary objectness measure to detect objects and details at various architectural scales. We introduce a multi-layer active learning paradigm to derive gaze shifting paths (GSPs) and calculate corresponding deep representations, tolerant to label noise via a penalty term that discards irrelevant GSP features. A manifold-regularized feature selector is used to choose high-quality GSP features, and a linear classifier distinguishes architectural scenes. Our technique outperforms traditional models, as shown in tests on three image retrieval datasets, demonstrating the discriminative capacity of our deep gaze-guided features.

In Cognitive Radio (CR) networks combined with Energy Harvesting (EH) technology, Secondary Users (SUs) are vulnerable to jamming attacks when sensing idle channels. At the same time, they may encounter numerous jamming and eavesdropping attacks during the data transmission phase. This paper examines the scenario in which SUs are susceptible to malicious attacks and energy constraints in both the sensing and transmission phases. We propose a utility function applicable to a single time slot. The blockchain uses Smart Contract (SC) technology to set rewards and punishments for users’ channel selection behavior and adjust mining difficulty. This method combines blockchain with spectrum sensing data fusion, abandons the decision-making mechanism of the traditional Cooperative Spectrum Sensing (CSS) Fusion Center (FC), and adopts a distributed structure to ensure the security and reliability of sensing data fusion. In addition, this paper uses the potential game and the Stackelberg game to study the optimal transmission channel and optimal time slot allocation strategy for SUs under malicious attacks. Considering the possible interference caused by channel switching and the greedy principle of Malicious User (MU), the proposed two-layer game method gradually optimizes the sensing detection probability and secure communication rate with time slot iteration. In order to further improve the secure communication rate, an iterative update formula for transmission power is given to make reasonable use of the remaining energy of each SU at the end of each time slot. Simulation results show that the proposed method is superior to traditional methods in both sensing performance and secure communication rate.

To promote the real-time dispatching of a power grid and balanced decision-making of power producers, accuracy and real-time forecasting are two main problems that need to be solved in ultra-short-term photovoltaic (PV) forecasting. Focusing on the problems of slow model training speed and low forecasting accuracy due to the redundancy of training data samples and insufficient long periodic capture of data in complex weather, this paper proposes a two-stage method for ultra-short-term PV power forecasting based on data-driven. In the meteorological analysis stage, the generation power samples similar to the forecast day were extracted by inputting daily meteorological features and using maximal information coefficient (MIC) weighted grey correlation degree to form the corresponding forecast data set. In the power forecasting stage, the temporal convolutional used network (TCN) extracts local features to maintain the sequence of extracted features. Then the bidirectional gating unit (BiGRU) combined with the Skip connection strategy was used to fully learn the long and the short time sequence of photovoltaic sequences, and the attention mechanism was used to pay adaptive attention to the more important historical states. This study experimented on the measured data from a photovoltaic power station in Southeastern China. The experimental results show that this method is effective in photovoltaic power short-term forecasts. In addition, compared with the latest model, this method has smaller forecasting errors and higher robustness in the time scales of 15 minutes, 30 minutes, 45 minutes, and 60 minutes. Specifically, indicator R2 increased by an average of 5.4%, while indicators RMSE and MAE decreased by 8.6% and 7.3%, respectively.

Low Earth orbit (LEO) mega-constellation satellite networks (LMCSNs) have shown great potential to extend the coverage capability of conventional terrestrial networks. The exploitation of the prevalent New Radio (NR) standard for LMCSNs can unleash the full potential toward a highly integrated satellite-terrestrial network. Despite this premise, how to systematically define, quantify, and assess the technical performance of NR enabled LMCSNs remains an open issue. In this paper, we address the challenging issue by proposing a holistic performance evaluation framework. Firstly, a key performance indicator (KPI) system including constellation-specific KPIs (i.e., $N$ -asset coverage, area traffic capacity, and service availability) and radio interface technology KPIs is devised. An efficient LEO constellation oriented performance assessment methodology is then carefully designed based on realistic beam layout and accurate interference modeling. We have carried out rigorous system-level simulations and provided extensive numerical results to assess the KPI system. It can be observed that the achieved area traffic capacity of the reference LEO constellation with 1800 satellites is less than 4 Kbps/km2. Besides, the $N$ -asset coverage and service availability range from 5.95 to 11.45, and from 24.9% to 32.7%, respectively.

The performance of cross-domain person re-identification has been greatly improved in recent years. However, there are still two problems in existing cross-domain person re-identification methods. First, most of them conduct domain adaptation on features that contain background noise. Second, they ignore the correlation between different features, including intra-domain and inter-domain. To overcome these problems, we propose a novel Graph-based Local Feature Adaptation (GLFA) framework for cross-domain person re-identification, which promotes domain adaptation by correlating intra-domain and inter-domain semantic local features with graph convolutional network. Specifically, in the feature extraction stage, we utilize a parsing model to extract semantic local features for both source and target domain samples, so as to avoid the feature shift caused by background discrepancy. In the feature adaptation stage, we apply two stacked Graph Convolutional Networks (GCN) to propagate semantic local feature information within each domain and across different domains respectively. It promotes the transfer of specific knowledge from the source domain to the target domain. Furthermore, to ensure the features from different domains updated by GCN are well aligned, a local feature distribution alignment loss is introduced on the top of GCN. The combination of GCN and alignment loss enables our framework diminish the feature shift caused by other factors effectively. Extensive experiments on Market-1501, DukeMTMC-reID and MSMT17 datasets demonstrate that our method outperforms the state-of-the-art methods evidently.

Bitcoin introduced a cryptocurrency as a form of public ledger consequently that turned into a most popular security technology, Blockchain. Its integrated mining technology lies the key security mechanism. The system allows forming a pool mining group to solve a particular job and share their revenues to their CPU usage while one of them successfully mines a block. To mine a block, a cryptographic puzzle should be solved, which requires significant compute resources that cause huge energy consumption. On the other hand, recent statistics show that low computational energy-restricted Internet of Things (IoT) devices are increasing exponentially. Although it has low energy and limited computation power, it is large in quantity when it is integrated. So we focus on a stochastic geometry theory, which resolves the issue of block mining computation via utilizing multiple mobile IoT devices, given that these IoT devices are Computation Capable Nodes (CCNs). To further normalize this issue, we propose an efficient mathematical solution that uses smart coordination of Virtual Network Functions (VNFs) for IoT devices to enable their CPU usage efficiently. At the same time, the work and credit point distribution policy is smartly handled through virtual pool mining. The proposal renders Network Function Virtualization technology to configure VNF, and Service Function Chain technology is utilized to enable the network flow of such VNFs. New algorithms are presented to solve multiple issues like node discovery, computation offloading, and work credit point distribution. Our goal is to minimize energy consumption within the given time constraint. Implementation results show that although virtual functions for block mining require extensive computations in IoT devices, dividing computation work into small fractions called tasks embedded with VNF, and offloading them to nearby CCNs, tend to minimize the cost and energy consumption of individual shared miners. The overall mining process is proved efficient and faster.

The performance of change detection between synthetic aperture radar (SAR) images mainly depends on the selection and utilization of image attributes. Nevertheless, most existing change detection approaches merely take the intensity attribute into consideration, constraining their capacities of detecting changes in complex situations. To solve this problem, this study develops an unsupervised SAR image change detection approach based on heterogeneous graph with multi-attributes and multi-relationships. First, the structural attribute, obtained by removing the complex texture, can be used to reflect the overall features of a SAR image. A heterogeneous graph is then constructed to encode the structural and intensity attributes as vertices, and two types of edges are hence connected to capture the intra-relationships and inter-relationships from these different attributes. With the support of this graph, the hyper-adjacency characteristics are proposed to quantify the multi-relationships and describe the global heterogeneous information. Constructing heterogeneous graphs respectively on bi-temporal SAR images, the changes of both intensity and structural attributes can be measured via comparing the hyper-adjacency characteristics of the bi-temporal graphs, thereby generating a difference image with good separability. Finally, the change map is obtained by cutting off the weakly related edges of heterogeneous graph on the difference image. Experiments on four real SAR datasets prove the effectiveness of the proposed graph-driven approach in improving the accuracy and robustness of change detection.

With the development of medium- and high-resolution satellites, successfully segmenting differently sized geo-objects remains a challenging issue for geographic object-based image analysis (GEOBIA). The hybrid image segmentation method is a good alternative to produce good segmentation that best matched the different sizes of geo-objects. However, the existing methods almost use segmentation parameters (SPs), such as scale, to control the sizes and shapes of segments. This will lead to two issues: (1) one single scale is impossible to segment every geo-object well due to the land cover complexity within remote-sensing imageries; (2) over- and under-segmented regions still occur in the segmentation results, whatever using any advanced segmentation methods. To solve the above problems, this paper developed a hybrid image segmentation method with local scale-guided hierarchical region merging and further over- and under-segmentation processing. First, the primitive segmentation was produced and then stratified into layers with different land covers. Then, the local scale was calculated for a more objective merging process in the separating layers. Third, the over- and under-segmentation at separating layers was recognized and re-processed for achieving a fine segmentation. To validate the proposed method, it was applied to three test images of gaofen-1 satellite with different land cover types, and ten competing methods were compared. The visual and quantitative results indicated the advantage of our method in segmenting out different sizes of geo-objects, which can effectively reduce the over- and under-segmentation error.

The policy-controlled signature (PCS) scheme uses the access policy to control signature verification permission. However public access policy that may contain private information will leak user privacy. At the same time, the expressiveness of access structures in the PCS schemes is weak. Therefore, we propose a policy-controlled signature scheme with strong expressiveness and privacy-preserving policy (PCS-PP), in which linear secret sharing schemes is to design access structure which has strong expression, the three primes composite order bilinear groups is used to hide the attribute value into the attribute name that may expose the privacy data by data distortion concept. The proposed PCS-PP scheme not only has correctness and privacy-preserving policy, but also supports fine-grained signature verification. In addition, the unforgeability is proved in the random oracle model. Compared to the related schemes, the proposed PCS-PP scheme has superiority in features, computation cost and storage.

In this paper, we proposed an integrated hybrid antenna based on spoof surface plasmon polaritons (SSPPs) which combines with omnidirectional antenna (ODA) and the frequency beam scanning antenna (FBSA). The ODA is fed by the SSPPs and the operating principle of FBSA is based on the higher-order modes of the SSPPs. The radiation mechanisms of the proposed integrated antenna are that the low-pass transmission modes of SSPPs are used to excite the ODA and the higher-order modes of SSPPs are utilized to generate the frequency beam scanning waves. That is to say, the SSPP design not only serves as the feeding structure for the ODA, but also acts as the radiating element for the FBSA. Simulated results show that the proposed integrated antenna achieves two impedance bandwidths. Within the operating bandwidth 2.32 GHz to 4.42 GHz, the proposed integrated antenna achieves omnidirectional radiation patterns, which enables the simultaneous use for wireless local area network (WLAN) and 5 G applications. Meanwhile, from 11.45 GHz to 25.5 GHz, the proposed integrated antenna acts as a wide-angle frequency beam scanning characteristic and can be used for planar integrated communication systems. Additionally, a prototype of the proposed integrated hybrid antenna is fabricated and tested. The measured results have a good agreement with the simulated ones.

The horizontal floating linear array is a new type of array used to detect deep-sea targets. Reliable target detection depends on the estimation accuracy of the array shape. So far, it has been proved that the theoretical array line of the array form is parabolic under ideal conditions. This paper constructs a segmented array estimation method based on the measurement data of the attitude depth sensor. The key of this method is to use the translation invariance of the array curve to establish and solve the underdetermined equation of the segment curve on the array state function. The array curve is segmented according to the sensor and the estimated array curve is obtained by iterative splicing. The simulation data of the horizontal floating array shows that the estimation error of this method is small, and the error optimization analysis of the parameters is carried out in this paper. In addition, this method has good universal applicability so that it can be applied to the formation estimation of other types of arrays.

The behaviors of dairy cows, such as feeding, ruminating, running, resting (standing, lying), head-shaking, drinking, and walking, can indicate their health status. In this study, a multi-sensor was used to collect data of cow’s multi-behaviors for research on behavior recognition. Firstly, a collar style data acquisition system equipped with geomagnetic and acceleration sensors to collect the behavioral data of dairy cows during their daily activities was designed. Secondly, the dairy cow behavioral recognition fusion model based on K-Nearest-Neighbors (KNN) and Random Forest (RF) models were used for behavior classification. To verify the accuracy of the fusion model, the algorithms of KNN, RF, Gradient Boosting Decision Tree (GBDT), Support Vector Machine (SVM), and Learning Vector Quantization (LVQ) were introduced for comparative recognition experiments with different algorithms. The KNN-RF fusion model had the highest average recognition accuracy of 98.51%, followed by the KNN model with an average recognition accuracy of 95.37%, and the LVQ model had the lowest average recognition accuracy of 80.81%. For the recognition and verification of each behavior, the KNN-RF fusion model had the most obvious improvement in the recognition of dairy cow feeding behavior, with a recognition accuracy of 99.34%, followed by the KNN model with a recognition accuracy of 95.07%. All six models had the lowest recognition accuracy for cow head-shaking behavior: a recognition accuracy of 89.11% with the KNN-RF model followed by the RF model with a recognition accuracy of 85.14%. The system can quickly and continuously collect cow behavior information, accurately recognize individual behaviors, and provide a scientific basis for the optimal design and efficient management of digital facilities and equipment for dairy cows.

For the uncertain linear systems described by the linear parameter varying (LPV) model, a parameter-dependent open-loop model predictive control (MPC) is proposed. The controller applies a tree trajectory to generate the vertices of uncertainty state predictions. Based on the state prediction tree, the future free control moves are parameter-dependent, whose vertices correspond to those of state predictions. The cost function penalizes the deviations of all the vertices of state/input from their steady-state target values. It is shown that the offset-free property is achieved by this method. A simulation example is given to demonstrate effectiveness of the approach.

This paper proposes a general robust model predictive control (MPC) approach for the constrained Takagi-Sugeno (T-S) fuzzy model with additive bounded disturbances. We adopt the homogeneous polynomially parameter-dependent (HPP) Lyapunov matrix with the arbitrary complexity degree and the corresponding HPP control law for the controller design. By applying the Pólya’s theorem and the extended nonquadratic boundedness property, a systematic approach to construct a set of sufficient conditions for assessing robust stability described by parameter-dependent linear matrix inequalities (LMIs) is established. The proposed approach is an improvement over the existing approaches in terms of control performance and stabilizable model range. Numerical examples are provided to show the effectiveness of the proposed robust MPC approach.

The monitoring of cutting tool wear has a great significance for the processing quality and stability. To overcome the difficulty to reflect all the wear mechanisms for the monitoring method based on finite element model or the excessive dependence on data extraction quality for the monitoring method based on sensor data, this paper proposes a model and data fusion method based on particle filter algorithm. Two different kinds of materials AISI 1045 and AISI 4340 are chosen to carry out the turning experiments. The mean absolute percentage error (MAPE) of the fusion method is 0.6%~4.1% lower and the coefficient of determination (R2) is more close to 1 compared with the finite element model method or sensor data method individually. Experimental results verify the feasibility and the superiority of the fusion method.

This paper proposes a novel dynamic latent structure with time-varying parameters for virtual sensing of industrial process with irregular missing data. The proposed latent structure is based on the linear dynamic system (LDS) model. In order to capture the time-varying process characteristics, a Karman filter based parameter updating method is developed and a virtual sensor is constructed to predict hard-to-measure quality variables. The latent variable structure of the improved model enables the virtual sensor to capture the variable cross-correlation and autocorrelation in the missing data by considering both spatial and temporal information, so that the information in the missing data can be learned from those not missing in the samples as well as the Markov process in the temporal stream. Incorporation of both spatial and temporal information renders more flexibility, resulting in a virtual sensor with higher prediction accuracy even with irregular missing data. The better performance of the proposed method is verified by two industrial applications with different ratios of irregular missing data.

Human action recognition is an important task in the fields of video content analysis and computer vision. Since the performance of most existing action recognition frameworks depends on the representation of features, many researches aim to construct more discriminative features. In this paper, we propose a manifold learning framework based on optical flow for action recognition. First, we calculate the dense optical flow field of the original video sequence, and the attention pooling layer (AP) is adopted to separate target area and background area to eliminate background interference. On this basis, motion features (MF) based on the physical characteristics of dense optical flow are developed to characterize human motion information. After that, manifold learning is introduced to calculate the motion variance features (MVF), which reflect the change rate of motion features and measure the spatial correlation between features in non-Euclidean space. Finally, fusing the MVF obtained by manifold learning and MF, feeding fusion features into two fully connected layers (FC) in series for action classification and recognition. Experiments on several classic datasets show that the proposed method achieves 0.98%, 1.86% and 0.99% performance improvement on UCF 101, HMDB51 and JHMDB.

Intra Block Copy (IBC) cannot work well for blocks with symmetry or mirror characteristic, because it is based on the translational block matching technique. In order to solve this problem, in this paper, an adequately designed IBC-Mirror mode is proposed for Versatile Video Coding (VVC), which extends our previous work. The newly proposed IBC-Mirror mode consists of two sub-modes, including IBC-Mirror Advanced Motion Vector Prediction (AMVP) mode and IBC-Mirror Merge mode. In IBC-Mirror AMVP mode, various Motion Estimation (ME) methods are applied, including previously proposed spatial search, and newly designed methods, i.e., Hash-based, Cache-based, Predictor-based and History-based IBC-Mirror search. As for IBC-Mirror Merge mode, Motion Information, composed of Block Vector (BV) and mirror flipping information, is added to the merge list to improve coding efficiency. Compared to the traditional IBC mode in VTM-10.0, the latest test model, the proposed mode can achieve 1~2% BD-rate saving. Compared to our previous work, methods in this paper can further achieve 0.5% coding gains and reduce 4% coding time. Higher coding efficiency and lower complexity show the superiority of extensions and improvements in this paper.

Cross Tensor Approximation (CTA) is a generalization of Cross/skeleton matrix and CUR Matrix Approximation (CMA) and is a suitable tool for fast low-rank tensor approximation. It facilitates interpreting the underlying data tensors and decomposing/compressing tensors so that their structures, such as nonnegativity, smoothness, or sparsity, can be potentially preserved. This paper reviews and extends state-of-the-art deterministic and randomized algorithms for CTA with intuitive graphical illustrations. We discuss several possible generalizations of the CMA to tensors, including CTAs: based on fiber selection, slice-tube selection, and lateral-horizontal slice selection. The main focus is on the CTA algorithms using Tucker and tubal SVD (t-SVD) models while we provide references to other decompositions such as Tensor Train (TT), Hierarchical Tucker (HT), and Canonical Polyadic (CP) decompositions. We evaluate the performance of the CTA algorithms by extensive computer simulations to compress color and medical images and compare their performance.

Direct decoding of speech from the brain is a faster alternative to current electroencephalography (EEG) speller-based brain-computer interfaces (BCI) in providing communication assistance to locked-in patients. Magnetoencephalography (MEG) has recently shown great potential as a non-invasive neuroimaging modality for neural speech decoding, owing in part to its spatial selectivity over other high-temporal resolution devices. Standard MEG systems have a large number of cryogenically cooled channels/sensors (200 - 300) encapsulated within a fixed liquid helium dewar, precluding their use as wearable BCI devices. Fortunately, recently developed optically pumped magnetometers (OPM) do not require cryogens, and have the potential to be wearable and movable making them more suitable for BCI applications. This design is also modular allowing for customized montages to include only the sensors necessary for a particular task. As the number of sensors bears a heavy influence on the cost, size, and weight of MEG systems, minimizing the number of sensors is critical for designing practical MEG-based BCIs in the future. In this study, we sought to identify an optimal set of MEG channels to decode imagined and spoken phrases from the MEG signals. Using a forward selection algorithm with a support vector machine classifier we found that nine optimally located MEG gradiometers provided higher decoding accuracy compared to using all channels. Additionally, the forward selection algorithm achieved similar performance to dimensionality reduction using a stacked-sparse-autoencoder. Analysis of spatial dynamics of speech decoding suggested that both left and right hemisphere sensors contribute to speech decoding. Sensors approximately located near Broca's area were found to be commonly contributing among the higher-ranked sensors across all subjects.

Finger-vein recognition technology has attracted more and more attention because of its high security and convenience. However, the finger-vein image capturing is affected by various factors, which results that some vein patterns are missed in acquired image. Matching minutiae features in such images ultimately degrades verification performance of the finger-vein recognition system. To overcome this problem, in this paper, a novel finger-vein image restoration approach is proposed to recover the missed patterns based on generative adversarial network (GAN), as the first attempt in this area. Firstly, we employ the segmentation algorithm to extract finger-vein network, which is further subject to thinning operation. Secondly, the resulting thinning image is taken as an input of a GAN model to restore the missed vein patterns. Thirdly, the minutiae points are extracted from restoration finger-vein pattern. Finally, we propose a matching approach for verification. Experimental results show that the proposed method can restore the missed vein pattern and reduce the equal error rate (EER) of the finger-vein verification system.

Due to the influence of wind speed disturbance, there are some uncertain phenomena in the parameters of the nonlinear wind turbine model with time in an actual working environment. In order to mitigate the side effects of uncertainties in speed models of wind turbines, researchers have designed a variety of controllers in recent years. However, traditional control methods require more knowledge of dynamics. Therefore, based on reinforcement learning and system state data, a robust wind turbine controller that adopts adaptive dynamic programming (ADP) is proposed. The ADP algorithm is a combination of Temporal-Difference (TD) algorithm and actor-critic structure, which can guarantee the rotor speed is stable around the rated value to indirectly adjust the wind energy utilization coefficient by changing the pitch angle in the area of high wind speed and achieve online learning in real-time. In addition, the variation of the pitch angle command of the proposed controller is relatively gradual, which can reduce the energy consumption of the variable pitch actuator, and extend the service life of the equipment. Finally, the wind speed model is simulated by combined wind speed based on Weibull distribution, the comprehensive simulation results show that the proposed controller has better control effect than some existing ones.

As the degree of nonlinearity in high-speed links becomes more and more serious, traditional decision feedback equalization (DFE) which is based on linear time-invariant assumption can no longer eliminate the inter-symbol interference effectively. In this paper, the shortcomings of traditional DFE structure for nonlinear high-speed links are first analyzed. Then, the multi-bit response (MBR) method is proposed to accurately construct the bit stream responses of the nonlinear high-speed links. Lastly, a novel DFE structure based on the MBR method is presented to improve the signal quality of the nonlinear high-speed links. The accuracy of the proposed method is verified by the simulation based on a nonlinear high-speed link model. Compared with traditional DFE, the proposed DFE structure can significantly improve the signal quality of the nonlinear high-speed links. As the degree of nonlinearity increases, the advantage of the proposed DFE becomes more prominent.

Named Entity Recognition (NER) systems have been largely advanced by deep neural networks in the recent decade. However, the state-of-the-arts on NER have been less applied to Chinese historical texts due to the lack of standard corpora in Chinese historical domains and the difficulty of accessing a quality ancient corpus. This paper addresses the respective issues and proposes an efficient automatic processing solution for tackling NER of ancient Chinese data, including the implementation of data-driven tagging and an innovative end-to-end network namely “MoGCN” (Mixture of Gated Convolutional Neural Network). A corpus consisting of three genres of Chinese historical classics is generated by our tagging approach, which is experimented for uncovering the generalization ability of proposed model. The empirical analysis demonstrates that our proposed model achieves the best results with above 1.5% F1-score improvement over other sophisticated models in this dataset, where the experimental performance shows positive dependence on the quality of corpus. Furthermore, our model can perform much better on shorter entities especially for 2-charater ones, while many long-range entities can be only identified by our model based on our auxiliary attribute analysis. This work serves as a preliminary exploitation of NER for historical data, providing unique insights and reference values for similar tasks. Future work should be focused on more exploration about NER optimization on massive Chinese traditional texts with linguistic features and learning strategies.

Cascaded boost-buck PFC (CBBPFC) converters offer a wide voltage conversion ratio and a near-unity power factor but require a large output electrolytic capacitor, leading to poor reliability and power density. In this paper, a coordinated two-stage operation and control strategy is proposed to significantly minimize the capacitor requirement without any other hardware changes. In a conventional CBBPFC converter, the boost and buck stages either operate independently nor complementally. In contrast, the proposed method operates the two stages in a concerted manner, so it is possible to use the dc-link capacitor with certain voltage fluctuation to buffer the power imbalance between the AC input and DC output. A new coordinated control strategy and a fluctuation-ratio based design consideration are developed to coordinate the operation of the two stages, further complement the system design. A 200W CBBPFC prototype based on the design concept exhibits a maximum reduction of the output capacitor by 83%, a peak efficiency of 95.8%, and a power factor of 0.99.

The asymmetric optical cryptosystems have been widely concerned due to their prominent characteristics of high security and robustness against attacks. In this paper, a resolution adaptative network is proposed to attack the asymmetric optical cryptosystems, which solves the issue of requiring a lot of experiments to change the network structure and parameters for effectively attacking to ciphertexts with different resolutions. The proposed network of fixing the network structure and parameters can be trained by inputting plaintext-ciphertext pairs with different resolutions, and then the trained network model can generate the retrieved plaintexts with corresponding resolutions. Numerical simulation and analysis show that the classical asymmetric optical cryptosystems can be attacked by the proposed network successfully. Moreover, the excellent resolution adaptability of the proposed network is verified by training and testing the images with different resolutions. Furthermore, the generalization ability and robustness of the proposed network is verified. In general, the issue of the input resolution adaptability of the cryptanalysis network is addressed firstly to our best knowledge.

Accurate and effective object detection in remote sensing images plays an extremely important role in marine transport, environmental monitoring and military operations. Due to the powerful ability of feature representation, region-based convolutional neural networks (RCNNs) have been widely used in this field, which firstly generate candidate regions through extracted feature maps and then classify and locate objects. However, most of existing methods generally use traditional backbone networks to extract feature maps with a decreased spatial resolution because of the continuous down-sampling, which will weaken the information detected from small objects. Besides, sliding windows strategy is employed in these methods to generate fixed anchors with a preset scale on feature maps, which is inappropriate for multi-scale object detection in remote sensing images. To solve the above problems, a novel and effective object detection framework named DetNet-FPN (Feature Pyramid Network) is proposed in this paper, in which a feature pyramid with strong feature representation is created by combining feature maps of different spatial resolution, at the same time, the resolution of feature maps is maintained by involving dilation convolutions. Furthermore, to match the proposed backbone, the GA (Guided Anchoring)-RPN strategy is improved for adaptive anchor generation, this strategy simultaneously predicts the locations where the center of objects are likely to exist as well as the scales and aspect ratios at different locations. Extensive experiments and comprehensive evaluations demonstrate the effectiveness of the proposed framework on DOTA and NWPU VHR-10 datasets.

In order to solve the problem when the target detection of passive bistatic radar is seriously affected by direct signal and multi-path, a clutter cancellation algorithm based on frequency domain analysis is proposed in this paper. By analyzing and processing the frequency domain correlation between the reference signal and the echo signal, the time delay and the amplitude of the direct signal and multipath in the echo signal can be estimated for clutter cancellation. This method can not only reduce the computation greatly, but also eliminate the influence of clutter echo residue on target detection. The effectiveness of the proposed algorithm is verified by simulation and real data processing.

Visual Question Answering (VQA) is a multimodal task involving Computer Vision (CV) and Natural Language Processing (NLP), the goal is to establish a high-efficiency VQA model. Learning a fine-grained and simultaneous understanding of both the visual content of images and the textual content of questions is the heart of VQA. In this paper, a novel Multimodal Encoder-Decoder Attention Networks (MEDAN) is proposed. The MEDAN consists of Multimodal Encoder-Decoder Attention (MEDA) layers cascaded in depth, and can capture rich and reasonable question features and image features by associating keywords in question with important object regions in image. Each MEDA layer contains an Encoder module modeling the self-attention of questions, as well as a Decoder module modeling the question-guided-attention and self-attention of images. Experimental evaluation results on the benchmark VQA-v2 dataset demonstrate that MEDAN achieves state-of-the-art VQA performance. With the Adam solver, our best single model delivers 71.01% overall accuracy on the test-std set, and with the AdamW solver, we achieve an overall accuracy of 70.76% on the test-dev set. Additionally, extensive ablation studies are conducted to explore the reasons for MEDAN's effectiveness.

In a cloud service supply chain composed of a cloud application developer (CAD) and two heterogeneous cloud service providers (CSPs), we study the cooperative supply decision making of the basic service unit. First, the optimization problem is modeled as a “patent labeling” game, where the labeling and supplying strategies of CAD are analyzed under different confidence indexes. Second, due to the immeasurability of the confidence index, the Shapley value contract is used to split profits at the cooperative stage, and the decision rule of CAD is presented. Finally, a cooperative subsidy mechanism is further proposed to achieve the Pareto improvement of all players when compared with the wholesale price contract. Model analysis shows that whether CAD is extremely optimistic or pessimistic in the negotiation with CSPs, it chooses to cooperate with the well-known CSP. Labeling is only selected when CAD is extremely pessimistic, and the labeling cost is low. When the confidence index fails to reach the extreme values, the supplier can make decisions referring to the hybrid contract on the basis of the Shapley value and cooperative subsidy mechanism.

In the vehicle communications, vehicle-to-vehicle interconnection has extraordinary significance for improving the quality of transportation services and ensuring effective and safe information sharing between vehicles. However in the vehicular Ad hoc networks (VANETs), messages shared between vehicles can be easily spoofed and tampered. How to ensure the authenticity, integrity and non-repudiation of the message is an important issue for VANETs. In order to solve the above problems, a privacy-preserving VANETs sharing system based on attribute-based cryptography is proposed in this paper. In the proposed scheme, the attributes of the vehicles are divided into static attributes and dynamic attributes according to the characteristics of them. And sensitive attribute values that may leak user privacy are hidden to protect user privacy. To achieve the verifiability, integrity, non-repudiation and traceability of the message, a new group signature on the ciphertext is introduced. Additionally, for the calculation limitation of the vehicle terminal, only 6 bilinear pairing operations are needed in the data recovering phase of the scheme. Finally, security analysis and performance evaluations are given to show that the proposed scheme has higher security and efficiency than those in the literature.

This paper proposes an extension of numerical P (NP) systems, called stochastic numerical P (StNP) systems. In StNP systems, a novel stochastic production function-repartition protocol is developed to evolve the variables. Compared with usual production function-repartition protocol in NP systems, there are two differences in stochastic production function-repartition protocol: (i) some stochastic/randomized factors are introduced in production function and/or repartition protocol; (ii) data updating way is relaxed and can be flexibly used. An StNP system has a nested structure and contains some variables and a group of programs. The programs consist of stochastic production function-repartition protocols and/or usual production function-repartition protocols. Therefore, StNP systems are a class of distributed parallel computing models with stochastic and dynamic characteristics. Data clustering problems are used as an application to demonstrate the availability and effectiveness of StNP systems. Based on StNP systems, a novel partition clustering algorithm is presented. Experimental results demonstrate advantage of StNP systems for data clustering problems.

Along with the rapidly increasing massive online data, recommender systems have been used as an effective approach for filtering useful information, which have been widely adopted in many web applications. In recent years, new techniques such as neural network and deep learning have demonstrated its effectiveness in studies of recommender systems. In this paper, we propose a novel approach for collaborative filtering with implicit feedback based on knowledge graph embedding. The basic ideal is to model the interactions between users and items as an interaction knowledge graph with a single relation, whose vector representation is learned through knowledge graph embedding. Base on the learned representation, the collaborative filtering problem is converted into link prediction in the interaction knowledge graph. KGECF, a neural network for knowledge graph embedding based collaborative filtering, is proposed based on this ideal and the RotatE knowledge graph embedding model. Experimental results on five datasets with different characteristics show that the KGECF model achieves the state-of-the-art (SOTA) performance on all datasets. And unlike other SOTA models that have obvious performance drops on AMusic and AToy datasets, our model's performance is very stable across all datasets.

A high efficiency and compact band-pass filter based on coplanar waveguide (CPW) is investigated in this paper, which lower and upper cut-off frequencies of the proposed filter can be tuned independently. In the proposed design, interdigital structure is used to filter the low frequency wave and a quasi-spoof surface plasmon polaritons (Q-SSPPs) structure is designed for tuning the higher cut-off frequency. The proposed Q-SSPPs structure contains only one cross-shaped element and keeps the similar properties of periodic SSPPs structure. The operating principle of the proposed design is explained by field distribution, dispersion curves, and equivalent circuits. The studies of vital parametric are carried out for better understanding the influences of the concerning parameters on the tunability. The simulated results indicate that the proposed design can obtain a wide bandwidth from 8.8 GHz to 17 GHz (about 63.6%) with high transmission efficiency (IS11I 21I > -0.2 dB). A prototype of the proposed design is fabricated and the measured results show good agreement with the simulated ones.

Long time coherent integration technique is one of the most important methods to improve the radar detection ability of a weak maneuvering target. However, the integration performance may be greatly influenced by the range migration and Doppler migration effects caused by the complex motions of maneuvering targets when long time integration is applied. Besides, the frequency hopping characteristic of the transmitted signal will also result in Doppler migration which cannot be eliminated by the conventional coherent integration method. To deal with this problem, a novel coherent integration method based on phase compensation is proposed. By using the frequency hopping pattern, the Doppler migration is eliminated by adding an extra phase in the twiddle factor of Fourier Transform in this proposed method. Since the phase error is compensated, long time coherent integration can be achieved. The effectiveness of the proposed method is verified by the simulation results and the performance analysis.

Road weather conditions are closed-related to the transportation safety and traffic capacity. With the development of road surveillance systems, weather conditions could be recognized from video. However, it is hard to be detected by machine. To address it, a deeply supervised convolution neural network (DS-CNN) is designed and trained on a self-established dataset. The traffic image dataset includes five groups labeled with “sunny”, “overcast”, “rainy”, “snowy” and “foggy”. Each group has manually labeled and selected more than 2500 images. The DS-CNN, can achieve the precision rate of 0.9681 and the recall rate of 0.9681. This practical weather detection method has been built in the surveillance system covers five freeways. The experimental result used in practice shows much worse detection results at first with disturbance of difference scenarios, worn camera, transmission failure and so on. With further improvement of DS-CNN, we found that it was much more effective than hand-crafted features in this task, and a deeper neural architecture could derive more powerful features. Moreover, the results show that dense weather information has more details in a small scale. In order to fast report the regional weather detection result, a designed visualization method is also proposed in spatiotemporal dimension to fuse with currently-used system. The high accuracy and fast detection speed with friendly visualization would lead to more precise traffic management and prompt the road weather traffic control to more intelligence levels.

The model with polytopic parametric uncertainty and bounded disturbance is controlled by the approach named dynamic OFRMPC (Output Feedback Robust Model Predictive Control). A key knob for control performance and region of attraction for this approach is the selection of Lyapunov matrix. A Lyapunov matrix, which does not have structural restriction, is proposed. In the ICCA (Iterative Cone Complementary Approach), which is invoked in optimizing the control law parameters, the starting up steps are designed as a variant CCA (Cone Complementary Approach). ICCA designs an outer loop, over CCA, for searching the minimum cost bound, while the variant CCA omits the outer loop by adding the cost bound in CCA objective function. This starting up can reduce the computational burden. The suboptimal dynamic OFRMPC (where CCA is avoided) is discussed, and a previous approach is re-formulated. A numerical example is given to show the advantages of the proposed approach.

Water distribution networks (WDNs) is crucial to ensure social operations and economic activities. However, WDNs are highly sensitive and vulnerable to disasters. The aim of this study is to mitigate the catastrophic consequences of cascading failures in WDNs. A flow-based WDN cascading failure model is built. The extended multi-objective particle swarm optimization model is developed to resist cascading failures and improve resilience. This model takes pipe diameter as the decision variable to minimize cost and maximize pressure deficit. Water balance, pressure, and standard pipe diameter are the constraints. The classical optimal scenario (COS) and the cascading failure scenario (CFS) are simulated. The model is applied to a small and medium-sized benchmarked WDN. Results show that the extended PSO can find the optimal solution on the benchmarked WDN. The Pareto fronts are obtained. Compare to the Pareto fronts between COS and CFS, the pressure deficit under CFS is significantly reduced, and the cost is reduced while the same pressure deficit increased. Different tolerance parameters are tested. The small network is not sensitive to the tolerance parameter, but the medium-sized network is sensitive. The model evaluates a variety of conflicting goals, which help designers and water managers resist cascading failures in WDNs.

Traffic congestion detection plays an important role for road management. However, it is difficult to automatically report traffic congestion when it occurs in large-scale road network. One of key challenges for rapidly and precisely identifying early congestion is huge variations in appearance caused by illumination, weather, camera settings and other traffic conditions. To address it, we proposed a traffic-oriented model to classify congestion from large dataset of ultra-low frame rate video captured from traffic surveillance system. The proposed deeply supervised traffic congestion detector has two modules: attention proposal module and deeply supervised inception network. Specifically, within the shallow layers, the binary edge/corner density features are used in attention proposal module to generate the rang of interest (ROI) mask automatically. This strategy keeps the training process focusing on the congestion features without disturbances. Following the attention proposal module, a very deep structure based on the inception network was used together to effectively extract rich and discriminative features then detect traffic congestion. The approach was tested on a self-established dataset based on empirical data, which contains images captured from 14470 surveillance cameras for monitoring 5,215 km of freeway in Shaanxi province, China. The experimental results show that the accuracy of the proposed method could reach 95.77% considering various disturbances, conditions and other limitations, which is improved than unsupervised networks.

In the grant-free massive machine-type communication (mMTC) scenario, a key challenge is the joint device activity detection and data decoding. The sporadic nature of mMTC makes compressed sensing a promising solution to the activity detection problem. However, the typical two-phase coherent transmission scheme, which divides channel training and data decoding into two separate phases, suffers performance losses, especially when only a few bits of data are transmitted by each active device. This paper follows a newly proposed non-coherent transmission scheme in which the data bits are embedded in the pilot sequences and the BS simultaneously detects active devices and decodes the embedded data bits without explicit channel estimation. To exploit statistical channel information and the specific structure of the sparsity pattern introduced by the non-coherent transmission scheme, i.e., only one row in each section can be non-zero, we propose a receiving method based on the approximate message passing (AMP) algorithm with non-separable minimum mean-squared error denoisers specifically designed for the problem. The corresponding state evolution equations, which can be used to predict the section error rate (SER) performance, is obtained and simplified under certain assumptions. We also derive closed-form expressions of the SER performance based on the state evolution results. Finally, numerical simulations are given to validate the accuracy of the performance analysis and to show the superiority of the proposed receiving method over the conventional method based on AMP with separable denoisers in the literature.

To make joints of a redundant manipulator moving automatically to a target state, a state-adjustment (SA) scheme is studied and modified in this paper. Specifically, owing to the problem of non-zero initial joint velocity in the SA scheme leading to a potential hazard to redundant manipulators, a modified state-adjustment (MSA) scheme is obtained on the basis of the SA scheme. The MSA scheme achieves the state adjustment by minimizing the differences between the joint angles and the target values. For solving the MSA scheme, a recurrent neural network (RNN) model is derived, of which the critical component is to iterate over the joint angles and joint velocities. The MSA scheme solved by the RNN model enables the redundant manipulator to adjust to the target state automatically while ensuring that the initial joint velocity is zero. Beyond that, several comparative simulations demonstrate the availability and accuracy of the MSA scheme solved by the RNN model in controlling the state adjustment of redundant manipulators.

Image texture smoothing (ITS) aims at completely removing textures while preserving as much as possible different-scale structures of an image. However, few quality assessment metrics have been formulated to objectively evaluate the ITS results, due to the lack of ground-truth texture-free images. Considering both the smoothness of textures and the preservation of structures, we make the debut of objective evaluation of ITS results, and design an intuitive Texture Smoothness and Structural Similarity Index (T3SI) based on human perception. Specifically, we first intuitively select some patches which contain relatively more texture/structure information. We then employ the Edge Preservation Index and the Structural Similarity Index Measure to evaluate the texture smoothness and the structure similarity on the selected patches, respectively. We finally formulate the novel exponential entropy function to balance the texture smoothness and the structure similarity for the quality assessment of ITS. T3SI is objective, easy to use, simple to implement, and stable. An independent User Study is performed to verify our proposed T3SI, and large experiments show that T3SI competes successfully to the state-of-the-art metrics. Our code is publicly available.

Residents clustering in different periods of load fluctuation and aggregated forecasting can increase the load prediction accuracy. But the strength of load fluctuation reflects the difference in the electricity consumption behavior of residents and affects the cluster results of residents. This paper presents a new day-ahead aggregated load-forecasting method for distribution networks based on the load fluctuation and feature importance (FI) profile clustering of residents. First, the input features are determined, the FI profile of residents is determined, and residents are clustered according to the FI profile. Then, the crow search algorithm is used to optimize the initial cluster centers for preventing the clustering results from falling into a local optimum. And the cluster verification index S_Dbw, the sum of the average scattering for the clusters and the inter-cluster density, is used to evaluate the cluster quality. The optimal clustering results of the aggregated load for different fluctuation periods are determined via statistical experiments. Finally, a random forest predictor based on ensemble learning is selected. According to the optimal clustering results in different fluctuation periods, a rolling forecasting model is constructed to realize day-ahead aggregated load forecasting in a residential distribution network.