Multi-source Heterogeneous Data Research Articles

A Review of Knowledge Graph-Based Reasoning Technology in the Operation of Power Systems

Knowledge graph (KG) technology is a newly emerged knowledge representation method in the field of artificial intelligence. Knowledge graphs can form logical mappings from cluttered data and establish triadic relationships between entities. Accurate derivation and reasoning of knowledge graphs play an important role in guiding power equipment operation and decision-making. Due to the complex and weak relations from multi-source heterogeneous data, the use of KGs has become popular in research to represent potential information in power knowledge reasoning. In this review, we first summarize the key technologies of knowledge graph representation and learning. Then, based on the complexity and real-time changes of power system operation and maintenance, we present multiple data processing, knowledge representation learning, and the graph construction process. In three typical power operation and fault decision application scenarios, we investigate current algorithms in power KG acquisition, representation embedding, and knowledge completion to illustrate accurate and exhaustive recommendations. Thus, using KGs to provide reference solutions and decision guidance has a significant role in improving the efficiency of power system operations. Finally, we summarize the achievements and difficulties of current research and give an outlook for future, promising roles of KG in power systems.

Main Factors on Effect of Precise Measurement and Precise Tamping Based on BP Neural Network

With the continuous development of precise measurement and precise tamping (PMPT) technology on Chinese railway conventional speed lines, the efficiency of machinery tamping operation and the quality of the track have been effectively improved. A variety of PMPT modes have been tried in the field operation, however there are some differences in the operation effect. The quality of the tamping operation is affected by multiple factors. In order to identify the key factors affecting the operation quality and to further improve the tamping operation effect, this paper establishes both the database of PMPT operation modes and the selection index system for evaluating the operation effect. Based on mega multi-source heterogeneous data and track geometry inspection data, this paper adopts the Back Propagation Neural Network (BPNN) prognosis model to quantify and sort the main factors affecting the effect of PMPT. The research results show that the initial quality of the track before tamping, whether the stabilizing operation or the tamping modes have great influence weights. It can scientifically guide the field operation to control the key factors and put forward some practical suggestions for promoting the field application of PMPT and the optimization of operation modes on the conventional speed lines.

Wind load identification of lattice towers using multi-source heterogeneous monitoring data

Wind load is one of the governing design loads for lattice tower structures. However, such information is difficult to be measured directly, and the traditional indirect identification methods highly depend on the accuracy of finite element model and are limited to fluctuating wind loads. To solve this problem, a novel wind load identification method from limited multi-source heterogeneous monitoring data is proposed. First, the complex model of a lattice tower is simplified into a reduced-order model based on the similarity principle of dynamic characteristics. And then, limited strain and acceleration data are fused to reconstruct wind-induced responses and modal parameters based on Kalman filter algorithm and mode-superposition method. Finally, the total wind loads, including both mean and fluctuating components, are obtained from identified modal parameters and modal responses. Numerical simulation of a lattice tower verified that the proposed method can reconstruct wind load and base reaction forces with high accuracy, and the anti-noise analysis shows that the proposed method has strong noise robustness. Moreover, the field test of a transmission tower during Typhoon Mangkhut was further carried out to validate the proposed method.

Ontology construction and mapping of multi-source heterogeneous data based on hybrid neural network and autoencoder

Ontology construction and mapping of multi-source heterogeneous data based on hybrid neural network and autoencoder

AutoMSNet: Multi-Source Spatio-Temporal Network via Automatic Neural Architecture Search for Traffic Flow Prediction

Recently the research of traffic flow prediction with deep learning framework has be largely developed, whereas most current methods are still faced with the following shortcomings. For spatial feature extraction, studies have shown that both local and non-local correlations exist on traffic networks. Considering the temporal dependencies, short-term impending and longer periodic components are two most critical patterns of traffic data, which further provide different information for the prediction task. Furthermore, multi-source heterogeneous external data, which naturally holds semantic gap with traffic data, also have impact on traffic flow. To solve the above problems, this paper proposes an AutoMSNet (Multi-Source Spatio-Temporal Network via Automatic neural architecture search). The AutoMSNet is composed of an encoder-decoder structure. The encoder takes neighboring data as inputs, while the decoder captures long-term periodic patterns. Thus, different functions of two temporal features are simultaneously extracted. Moreover, a neural architecture search space is designed for spatial feature extraction. Through architecture search technique, graph convolutions with different receptive fields are automatically selected and combined to form an optimal module structure. Therefore, both local and non-local spatial features can be adaptively captured. Besides, a meta learning feature fusion strategy is proposed to integrate external data, which can alleviate the semantic gap between different data sources. Extensive experiments on three real-world traffic datasets evaluate the superiority of the proposed model.

Analysis and construction of the coal and rock cutting state identification system in coal mine intelligent mining

The recognition of cutting state of coal-rock is the key technology to realize “unmanned” mining in coal face. In order to realized real-time perception and accurate judgment of coal-rock cutting state information, this paper combined the field test sampling, construction technology of complex coal seam, virtual prototype technology, bidirectional coupling technology, data processing theory, image fusion method, and deep learning theory to carry out multi domain deep fusion experimental research on multi-source heterogeneous data of coal and rock cutting state. The typical complex coal seam containing gangue, inclusion, and minor fault in Yangcun mine of Yanzhou mining area was taken as the engineering object. The high-precision three-dimensional simulation model of the complex coal seam that can update and replace particles was constructed. Based on the simulation results of Discrete Element Method-Multi Flexible Body Dynamics (DEM-MFBD), the one-dimensional original vibration acceleration signals of the key components of the shearer cutting part were determined, including spiral drum, rocker arm shell, and square head. After transforming one-dimensional original signal data into two-dimensional time–frequency images by Short-time Fourier Transform, morphological wavelet image fusion technology was used to realize the effective fusion of characteristic information of spiral drum, rocker arm shell, and square head under different working conditions. Based on the deep learning theory, the DCGAN-RFCNN (Deep Convolutional Generative Adversarial Networks-Random Forest Convolutional Neural Networks) coal and rock cutting state recognition network model was constructed. Combining convolution neural network with random forest recognition classifier, RFCNN coal and rock cutting state recognition classification model was constructed, and the recognition network model was trained to obtain the model recognition results. Through the comparative experimental analysis of the RFCNN network model with different recognition network models and different synthetic sample numbers in the recognition network, the effectiveness of the recognition network model was verified. The results show that: When synthetic samples are not included in each working condition in the RFCNN model, the average recognition rate is 90.641%. With the increase of the number of synthetic samples, the recognition rate of coal and rock cutting state increases. When the number of synthetic samples added to each working condition reaches 5000, the recognition effect is the best, and the average recognition rate reaches 98.344%, which verifies the superiority of enriching the data set by using the improved DCGAN network. Also, the RFCNN outperformed the other variants: it obtained higher recognition accuracy by 25.085, 21.925 and 19.337%, respectively, over SVW, CNN, and AlexNet. Also, the experimental platform of shearer cutting coal and rock was built, where the coal and rock cutting state recognition network was trained and tested based on the migration learning theory. Through the statistical test results, the accuracy of coal and rock cutting state recognition is 98.64%, which realizes the accurate recognition of coal and rock cutting state.

A deep learning convolutional neural network and multi-layer perceptron hybrid fusion model for predicting the mechanical properties of carbon fiber

Recently, deep learning methods have become one of the hottest topics in predicting material properties, however, one bottleneck in current research is the simultaneous analysis of heterogeneous data. In this study, a deep learning fusion model is developed for the first time to predict the material properties of carbon fiber monofilament using textual (macroscopic properties of composites and matrix) and visual (two-point statistics of microstructures) data. For this, 1200 stochastic microstructures are generated using the greedy-based generation (GBG) algorithm. Then, the statistical representations of microstructures are determined using two-point statistics and the macroscopic properties are calculated based on a micro-scale finite element (FE) simulation. Finally, the visual and textual data are fed into the convolutional neural network (CNN) and multi-layer perceptron (MLP) fusion model for predicting the mechanical properties of carbon fibers. The developed hybrid CNN-MLP fusion model achieves encouraging average testing R2 of longitudinal modulus, transverse modulus, in-plane shear modulus, major Poisson’s ratio, and out-of-plane shear modulus of carbon fibers with values of 0.991, 0.969, 0.984, 0.903, and 0.955, respectively. Thus, the proposed strategy provides a promising framework for predicting material properties via multisource heterogeneous data and is expected to accelerate the smart design and optimization of materials.

Landslide susceptibility evaluation and interpretability analysis of typical loess areas based on deep learning

Loess areas have a unique geological environment, and geological disasters occur there frequently. In this work, the typical loess area Lvliang was used as the study area. Using the historical landslide catalog, 12 influencing factors were chosen by integrating multisource heterogeneous spatiotemporal big data such as remote sensing, ground investigation, and basic geography. Based on frequency ratio (FR) and improved TabNet deep learning technology, landslide susceptibility evaluation and uncertainty analysis were performed. The results showed that the TabNet evaluation model using FR and self-supervised learning performs well and has the highest FR in extremely high-prone areas. Compared with other methods, this method has the highest scores in areas under the curve and susceptibility index distribution and the lowest uncertainty. Moreover, the SHAP method was used for interpretability analysis of the model. Therefore, this study can provide new ideas for landslide susceptibility management.

Cropland: Surplus or Deficit? From the Perspective of Meeting People’s Grain Requirement

The quantity and quality of cropland plays an important role in ensuring food security. In order to explore spatiotemporal patterns of the extent to which cropland satisfies people's grain need, we integrate multi-source heterogeneous data to investigate in which era, and in which region, the cultivated land can meet people's food demands. It turns out that in the past 30 years, with the exception of the late 1980s, the amount of cropland could satisfy people's grain needs at the nation scale. However, more than 10 provinces (municipality/autonomous region), mainly located in western China and southeast coastal areas, have been unable to meet the grain needs of local people. We projected the guarantee rate to the late 2020s. Our study concludes that the guarantee rate of cropland is estimated to be higher than 150% in China. Compared to 2019, except Beijing, Tianjin, Liaoning, Jilin, Ningxia, as well as Heilongjiang in the Sustainability scenario, and Shanghai in the Sustainability and the Equality scenarios, the guarantee rate of cultivated land will increase in every province (municipality/autonomous region) in 2030. This study has reference value for the study of China's cultivated land protection system, as well as important significance for China's sustainable development.

Deep-Learning Model for Influenza Prediction From Multisource Heterogeneous Data in a Megacity: Model Development and Evaluation.

In megacities, there is an urgent need to establish more sensitive forecasting and early warning methods for acute respiratory infectious diseases. Existing prediction and early warning models for influenza and other acute respiratory infectious diseases have limitations and therefore there is room for improvement. The aim of this study was to explore a new and better-performing deep-learning model to predict influenza trends from multisource heterogeneous data in a megacity. We collected multisource heterogeneous data from the 26th week of 2012 to the 25th week of 2019, including influenza-like illness (ILI) cases and virological surveillance, data of climate and demography, and search engines data. To avoid collinearity, we selected the best predictor according to the weight and correlation of each factor. We established a new multiattention-long short-term memory (LSTM) deep-learning model (MAL model), which was used to predict the percentage of ILI (ILI%) cases and the product of ILI% and the influenza-positive rate (ILI%×positive%), respectively. We also combined the data in different forms and added several machine-learning and deep-learning models commonly used in the past to predict influenza trends for comparison. The R2 value, explained variance scores, mean absolute error, and mean square error were used to evaluate the quality of the models. The highest correlation coefficients were found for the Baidu search data for ILI% and for air quality for ILI%×positive%. We first used the MAL model to calculate the ILI%, and then combined ILI% with climate, demographic, and Baidu data in different forms. The ILI%+climate+demography+Baidu model had the best prediction effect, with the explained variance score reaching 0.78, R2 reaching 0.76, mean absolute error of 0.08, and mean squared error of 0.01. Similarly, we used the MAL model to calculate the ILI%×positive% and combined this prediction with different data forms. The ILI%×positive%+climate+demography+Baidu model had the best prediction effect, with an explained variance score reaching 0.74, R2 reaching 0.70, mean absolute error of 0.02, and mean squared error of 0.02. Comparisons with random forest, extreme gradient boosting, LSTM, and gated current unit models showed that the MAL model had the best prediction effect. The newly established MAL model outperformed existing models. Natural factors and search engine query data were more helpful in forecasting ILI patterns in megacities. With more timely and effective prediction of influenza and other respiratory infectious diseases and the epidemic intensity, early and better preparedness can be achieved to reduce the health damage to the population.

A hierarchical blockchain-enabled security-threat assessment architecture for IoV

In Internet of Vehicles (IoV), the security-threat information of various traffic elements can be exploited by hackers to attack vehicles, resulting in accidents, privacy leakage. Consequently, it is necessary to establish security-threat assessment architectures to evaluate risks of traffic elements by managing and sharing security-threat information. Unfortunately, most assessment architectures process data in a centralized manner, causing delays in query services. To address this issue, in this paper, a Hierarchical Blockchain-enabled Security threat Assessment Architecture (HBSAA) is proposed, utilizing edge chains and global chains to share data. In addition, data virtualization technology is introduced to manage multi-source heterogeneous data, and a metadata association model based on attribute graph is designed to deal with complex data relationships. In order to provide high-speed query service, the ant colony optimization of key nodes is designed, and the HBSAA prototype is also developed and the performance is tested. Experimental results on the large-scale vulnerabilities data gathered from NVD demonstrate that the HBSAA not only shields data heterogeneity, but also reduces service response time.

Electric Vehicle Participation in Regional Grid Demand Response: Potential Analysis Model and Architecture Planning

When a large-scale random charging load is connected to the regional power grid, it can negatively affect the safe and stable operation of the power grid. Therefore, we need to study its charging load and response potential in advance so that electric vehicles can interact well with the grid after being connected to the regional grid. Firstly, after analyzing the influencing factors of regional electric vehicle ownership, an electric vehicle ownership prediction model based on the sparrow search algorithm-improved BP neural network (SSA-BPNN) is established. On this basis, an electric vehicle charging load prediction model is established based on the sparrow search algorithm-improved BP neural network and Monte Carlo algorithm (SSA-BPNN-MC). Secondly, the charging behavior of different types of electric vehicles is analyzed and modeled, and the data from a certain area are taken as an example for the prediction. Then, according to the load forecasting results, the potential of electric vehicles participating in demand response in the region in the future is deeply analyzed using the scenario analysis method. Finally, with the aim of resolving the problems of massive multi-source heterogeneous data processing and the management of electric vehicles participating in the regional power grid demand response, a basic framework of electric vehicles participating in the regional power grid demand response is developed, which provides effective support for promoting electric vehicles to participate in regional grid demand response.

A dual-attention based coupling network for diabetes classification with heterogeneous data

Diabetes Mellitus (DM) is a group of metabolic disorders characterized by hyperglycaemia in the absence of treatment. Classification of DM is essential as it corresponds to the respective diagnosis and treatment. In this paper, we propose a new coupling network with hierarchical dual-attention that utilizes heterogeneous data, including Flash Glucose Monitoring (FGM) data and biomarkers in electronic medical records. The long short-term memory-based FGM sub-network extracts the time-dependent features of dynamic FGM sequences, while the biomarkers sub-network learns the features of static biomarkers. The convolutional block attention module (CBAM) for dispersing the feature weights of the spatial and channel dimensions is implemented into the FGM sub-network to endure the variability of FGM and allows us to extract high-level discriminative features more accurately. To better adjust the importance weights of the characteristics of the two sub-networks, self-attention is introduced to integrate the characteristics of heterogeneous data. Based on the dataset provided by Peking University People’s Hospital, the proposed method is evaluated through factorial experiments of multi-source heterogeneous data, ablation studies of various attention strategies, time consumption evaluation and quantitative evaluation. The benchmark tests reveal the proposed network achieves a type 1 and 2 diabetes classification accuracy of 95.835% and the comprehensive performance metrics, including Matthews correlation coefficient, F1-score and G-mean, are 91.333%, 94.939% and 94.937% respectively. In the factorial experiments, the proposed method reaches the maximum area under the receiver operating characteristic curve of 0.9428, which indicates the effectiveness of the coupling between the nominated sub-networks. The coupling network with a dual-attention strategy performs better than the one without or only with a single-attention strategy in the ablation study as well. In addition, the model is also tested on another data set, and the accuracy of the test reaches 94.286%, reflecting that the model is robust when it is transferred to untrained diabetes data. The experimental results show that the proposed method is feasible in the classification of diabetes types. The code is available at https://github.com/bitDalei/Diabetes-Classification-with-Heterogeneous-Data.

Towards sparse matrix operations: graph database approach for power grid computation

The construction of new power systems presents higher requirements for the Power Internet of Things (PIoT) technology. The “source-grid-load-storage” architecture of a new power system requires PIoT to have a stronger multi- source heterogeneous data fusion ability. Native graph databases have great advantages in dealing with multi-source heterogeneous data, which make them suitable for an increasing number of analytical computing tasks. However, only few existing graph database products have native support for matrix operation-related interfaces or functions, resulting in low efficiency when handling matrix calculations that are commonly encountered in power grids. In this paper, the matrix computation process is expressed by a strategy called graph description, which relies on the natural connection between the matrix and structure of the graph. Based on that, we implement matrix operations on graph database, including matrix multiplication, matrix decomposition, etc. Specifically, only the nodes relevant to the computation and their neighbors are concerned in the process, which prunes the influence of zero elements in the matrix and avoids useless iterations compared to the conventional matrix computation. Based on the graph description, a series of power grid computations can be implemented on graph database, which reduces redundant data import and export operations while leveraging the parallel computing capability of graph database. It promotes the efficiency of PIoT when handling multi-source heterogeneous data. An comprehensive experimental study over two different scale power system datasets compares the proposed method with Python and MATLAB baselines. The results reveal the superior performance of our proposed method in both power flow and N-1 contingency computations.

Distributed Feature Selection Considering Data Pricing Based on Edge Computing in Electricity Spot Markets

With the rapid development of information technology, the multisource heterogeneous data containing meaningful information have been significantly generated by various edge devices in Internet of Energy, which is one of essential foundations of many knowledge discovery tasks based on edge computing. For some complicated tasks, essential features are owned by different data sellers offering data by blockchains. With limited budgets, buying features are crucial steps in knowledge discovery tasks in electricity spot markets, especially for learning-based algorithms. However, there are lack of proper data pricing mechanisms tailored to dynamic learning processes. Besides, existing methods cannot efficiently employ edge computing servers to obtain optimal policies for selecting features according to dynamic pricing with limited budgets. To overcome such drawbacks, a data pricing mechanism is proposed in this article, which consists of static and dynamic pricing parts. Based on this mechanism, given limited budgets, a feature selection (FS) algorithm considering multiple new factors is proposed, which offers near-optimal solutions for FS at different scenarios. Numeric results show the effectiveness of the proposed algorithms.

Big data classification based on improved parallel k-nearest neighbor

In response to the rapid growth of many sorts of information, highway data has continued to evolve in the direction of big data in terms of scale, type, and structure, exhibiting characteristics of multi-source heterogeneous data. The k-nearest neighbor (KNN) join has received a lot of interest in recent years due to its wide range of applications. Processing KNN joins is time-consuming and inefficient due to the quadratic structure of the join method . As the number of applications dealing with vast amounts of data develops, KNN joins get more sophisticated. The authors seek to save money on computer resources by leveraging a large number of threads and multiprocessors. Six popular datasets are used to apply the method and evaluate the sequential and parallel performance of the KNN technique. These datasets are used to compare the sequential and parallel performance of the KNN method. When compared to a matching multi-core solution, the final implementation saves computing resources. It has been optimized to utilize as little RAM as possible, allowing it to manage high-resolution photo data without sacrificing efficiency. The authors will use the technique they presented using Spark Radoop. Our performance research validates the supplied method’s efficacy and scalability.

Inferring student social link from spatiotemporal behavior data via entropy-based analyzing model

Social link is an important index to understand master students’ mental health and social ability in educational management. Extracting hidden social strength from students’ rich daily life behaviors has also become an attractive research hotspot. Devices with positioning functions record many students’ spatiotemporal behavior data, which can infer students’ social links. However, under the guidance of school regulations, students’ daily activities have a certain regularity and periodicity. Traditional methods usually compare the co-occurrence frequency of two users to infer social association but do not consider the location-intensive and time-sensitive in campus scenes. Aiming at the campus environment, a Spatiotemporal Entropy-Based Analyzing (S-EBA) model for inferring students’ social strength is proposed. The model is based on students’ multi-source heterogeneous behavioral data to calculate the frequency of co-occurrence under the influence of time intervals. Then, the three features of diversity, spatiotemporal hotspot and behavior similarity are introduced to calculate social strength. Experiments show that our method is superior to the traditional methods under many evaluating criteria. The inferred social strength is used as the weight of the edge to construct a social network further to analyze its important impact on students’ education management.

Multi-source dynamic time series data processing method based on comprehensive weighted fusion

Aiming at the fusion problem of multi-source heterogeneous dynamic data sources, based on the subjective and objective comprehensive weighting idea, an efficient multi-source dynamic data fusion algorithm is proposed. First, the outlier detection method based on linear regression model is used to preprocess the multi-source dynamic time series data. Then, determine the weight of the dynamic data. The prior weight of the data source is determined by the AHP method, cyclic scoring method, etc., and the posterior weight is determined according to the similarity between the time series data. Based on the Euclidean distance, a similarity measure between different data sources and a similarity measure between different dynamic data samples are constructed respectively. The former is used for the fusion of dynamic data standard deviation expectations, and the latter is used for the fusion of mean expectations. Finally, the comprehensive weight of each dynamic time series of each data source is obtained, and the comprehensive weighted fusion method is used to obtain the expected value and standard deviation of the dynamic response. The analysis of the calculation example shows that the proposed algorithm has higher computational efficiency than the method that treats dynamic data as several discrete static data.

Neural TV program recommendation with multi-source heterogeneous data

TV program recommendation is important for users in the face of a huge amount of information data. The existing TV program recommendation mainly relies on a collaborative filtering method to recommend through interactive data between users and programs. Although some methods utilize auxiliary information to enrich semantic features, most of them only use a single data type, which cannot capture a more diverse feature representation of the user and program. In this paper, we propose a neural TV program recommendation model with multi-source heterogeneous data, which makes full use of the multi-source heterogeneous auxiliary information. Specifically, we combine heterogeneous features derived from auxiliary information to learn a deep program representation in the program encoder module. To more accurately capture user preferences, we further utilize the personalized attention mechanism to determine the importance of different programs to the user representation based on the interaction between users and programs in the user encoder module. Extensive experiments on a real dataset of the Chinese capital show that our model can effectively improve the performance of TV program recommendations compared to the existing models.

An Adversarial Multisource Data Subdomain Adaptation Model: A Promising Tool for Fault Diagnosis of Induction Motor Under Cross-Operating Conditions

Typical domain adaptation neural network that takes multi-source heterogeneous data as input usually achieves poor diagnostic accuracy in induction motor fault diagnosis under cross-operating conditions. Aiming at this problem, the present study proposes an adversarial multi-source data subdomain adaptation (AMDSA) model. This model encapsulates three types of modules: a shared feature extractor, a label predictor and a series of domain discriminators. The joint operation of the shared feature extractor and the domain discriminators is used to perform subdomain adaptation of different types of data for obtaining domain invariant features of multi-source heterogeneous data. The label predictor is employed to fuse these domain invariant features and realize label classification. The proposed model can solve the problem of multi domain adaptation in multi-source heterogeneous data through constructing a subdomain adaptation strategy and a feature fusion strategy. The effectiveness of AMDSA is verified by a series of diagnostic experiments on faulty induction motors under cross-operating conditions. The experimental results show that the average diagnostic accuracy of all cross-operating conditions reaches 97.62%.