Big Data and Security

In view of the continuous improvement of the current energy consumption data of many types of enterprises, the effective monitoring of enterprise energy consumption and the early warning of different reward energy use will be the top priority. This paper proposes a data-driven energy efficiency evaluation and energy anomaly detection method for multi-type enterprises based on energy consumption big data mining. This method uses K-means Clustering algorithm identifies the energy consumption patterns of different enterprises, which is convenient for the evaluation of enterprise energy efficiency. Then, on the basis of pattern division, the outlier detection of enterprise energy consumption data is completed by CEEMDAN-LOF algorithm, and the abnormal energy consumption detection and research of enterprises are realized. The example uses the real energy consumption data of power grid enterprises, and the simulation results show the effectiveness of the proposed method.

With the high penetration of random energy such as wind power and photovoltaic in the power grid, the influence of the accuracy of regulation of traditional thermal power units on the operation of the power grid is gradually increasing. Aiming at the problem of the deviation between the actual output of thermal power units and the AGC command of the power grid, this paper proposes a data-driven AGC command execution effect identification method. Firstly, based on Kernel Principal Component Analysis (KPCA), a data preprocessing method is proposed, which maps feature datasets into low-dimensional vectors to achieve dimensionality reduction. Secondly, the Independent Recurrent Neural Network (IndRNN) is used to process and predict the dimensionality reduction data, so as to realize the accurate perception of the adjustment effect of the unit execution command. Finally, the real power grid data is used to simulate and verify the proposed method. The results show that the model can effectively reduce the deviation of instruction execution.

Traditional power system scheduling optimization methods cannot fully deal with the massive data brought by the increase of new energy penetration. Aiming at the above problems, this paper proposes a day-ahead scheduling optimization method based on big data clustering and filling for power system includes wind power plants and photovoltaic power stations. Firstly, according to the collected historical data of power grid, the K-means clustering method of big data is used to generate representative load, wind power and solar illumination scene sets. Secondly, a missing value filling method based on historical data assisted scene analysis was proposed. Finally, the day-ahead scheduling model of power system with wind power plants and photovoltaic power stations is established and solved by improved particle swarm optimization algorithm. The simulation results show that this method can improve the filling accuracy of the missing value of the day-ahead dispatching historical data of the power system, and meet the development demand of the power system with new energy.

With the massive application of new energy, the contradiction of power grid regulation has become increasingly prominent. How to effectively predict the range of the power grid is a huge challenge faced by the day-ahead dispatch of power system. Aiming at this problem, this paper proposes a method for generating day-ahead scenarios for renewable energy based on historical data analysis. First, the deep embedding clustering (DEC) algorithm is used to analyze historical data, and periods with similar characteristics are divided into one group. Then the conditional deep convolutions generative adversarial network (C-DCGAN) model generates a day-ahead scenario set for renewable energy. At last, the Belgian Elia renewable energy data is used for simulation analysis, and the results show that the proposed method can accurately describe the uncertainty of renewable energy.

The modeling and prediction of stock prices is the core work in securities investment, and it is of enormous significance to reducing decision-making risks and improving investment returns. Existing research mainly focuses on mid or low-frequency stock price prediction, which is challenging to apply to intraday high-frequency trading scenarios. Meanwhile, the model accuracy face limitation due to the neglect of the influence of random noise and the refinement of the price sequence law. This paper proposes a high-frequency stock price prediction method based on mode decomposition and deep learning to improve intraday stock price prediction accuracy. Firstly, this method stabilizes the stock price series through empirical mode decomposition to tackle the issue of random noise interference. Then the convolutional neural network is introduced to extract the high-dimensional data features hidden in the stock price series by using multiple convolution kernels. Furthermore, the gated recurrent unit is used to process time-sequential data and to predict the stock prices at the minute level. The experimental result indicates that the proposed high-frequency stock price prediction method can achieve a significant forecasting effect, and its accuracy outperforms the existing methods.

To solve the multi-platform user association problem of complex trajectory matching process and high time cost in cross-platform association positioning of instant messaging users, and at the same time make full use of the information in user trajectories, this paper proposes a supervised learning-based cross-platform instant messaging user association positioning method. The algorithm firstly places probes in the area where the target may appear to obtain user information; then gets user trajectories through the obtained user distance information and time information; selects user features through the classification algorithm of supervised learning, and designs a cross-platform instant messaging user association localization method based on supervised learning, so as to increase the association efficiency and accuracy of cross-platform instant messaging user association. The method conducts specific experiments for the most commonly used instant messaging tools in China, WeChat and Stranger users, and the results show that the method can achieve efficient and reliable association for these two types of instant messaging users.

The application of big data requires effective tracing of data transmission and flow processes, so as to achieve effective determination of data authenticity and security. If network data is not effectively supervised, unexpected events such as data loss, leakage or tampering will occur, resulting in network data threats and risks that cannot be traced and responded to. The traditional data tracing methods are found difficult to meet the processing needs of massive data. Hence, data tracing in Hadoop environment is considered to better deal with the risk of data loss, tampering and leakage in the process of multiple data distribution. In this paper, the Hadoop environment system and its application in the field of data watermark tracing is explored. By analyzing the data tracing model and its implementation, a data watermark tracing system in Hadoop environment is established and the tracing process is examined. The experiments are designed and the efficiency of the proposed system is validated.

At present, cable labels have problems that labels are easy to knock and fall off during the storage and allocation of equipment. The binding falls off affected by the external environment and time, which cannot effectively support the unified coding and full life management of the equipment. A cable fault positioning algorithm based on big data is designed, which combines the positioning algorithm of uhf UHF RFID electronic label with the positioning algorithm of active label to realize the work safety monitoring management of inspection personnel and the automatic positioning reporting management of circuit barrier problems. The read-write conflict and interference problem in the passive UHF RFID electronic tags and the active UHF electronic tags are introduced in the original algorithm. The simulation of the algorithm is analyzed, and the simulation results prove that the present algorithm can solve the problem better than the original one. The algorithm can effectively locate the cable fault, and has certain engineering and theoretical value.

[Objective] Due to the 14th Five-Year Plan and the “30·60” double carbon targets advocates energy reform, the number of articles on virtual power plant (Abbreviation is VPP) has surged, the annual number of articles published has maintained a positive growth trend. In the context of big data, the research track of VPP is sorted out by literature measurement method and visual knowledge graph. It will provide a reference direction for the following scholars’ research and accelerate the combination of emerging technologies and VPP. [Methods] Through the data collection of CNKI related literature on VPP in China in the past 20 years, Using CiteSpace to process and analyze big data, rapidly form a visual map to reveal the development and evolution trajectory of VPPs, and make it easier to track research hotspots and core frontier technologies. [Result/Conclusion] The knowledge map generated by big data reveals the research hotspots of VPP and the evolution trajectory of research hotspots, and the conclusion obtained is more reasonable and realistic significance. Through the map, the main technological research directions of VPP were controllable load, distributed energy resource, communication and energy storage. It aims to achieve the stability of massive distributed energy resource grid-connection, fast and safe communication, expansion of energy storage capacity and overall economic benefits, etc. Combined with the current research on VPP in China, the research on VPP in the future will closely follow the trend of national policies, and the emergence of new energy will gradually disperse and diversify the research on VPP.

Transverse cracks dominated by reflective cracks are one of the most common diseases of expressways in our country. In order to improve the structural integrity and driving comfort, this study relied on big data stored in the PMS of Jiangsu Province to analyze the development law and influencing factors of transverse cracks in semi-rigid base asphalt pavement. Two index types, node index and development index, were proposed, and the evaluation results and significant influencing factors of these two evaluations indexes were analyzed respectively. The fitting model function in JMP software was used to statistically analyze the traffic and structural influence factors of the transverse crack in the total mileage of 854 km of 291 original road sections. Then the SCB test was carried out to obtain the fracture energy of each sublayer of the sections. At last, a TCS prediction model was constructed by the significant influencing factors and the composite fracture energy representing the overall crack resistance level of the asphalt layer. The results of this study show that 85% of the road sections cracked for the first time within 3–9 years of opening to traffic. The first 5 years after opening to traffic was the slow development stage of transverse cracks, and the 5–10 years was the stage of rapid development of transverse cracks. 10–15 years later, the development speed of transverse cracks tended to slow down. The main factors affecting the generation and development of transverse cracks are traffic volume, the gradation type of each layer, the thickness of modified asphalt layer and the type of base material.

As a disruptive technological change in IT industry, big data has changed the traditional information asymmetry and physical area barrier, which has brought profound changes to the strategic management, organizational structure, business decision-making and operating model of modern enterprises. Data and information have become important emerging factors of production, changing the management and financing methods of new energy industry. At the same time, ecological theory is extended to the field of new energy industry financing, and it becomes a new trend to study the new energy industry financing ecosystem. Based on the in-depth analysis of big data and new energy industry financing ecosystem, we combine the life cycle theory and construct logistic model to analyze the evolution process of new energy industry financing ecosystem in the context of big data, and find that China's new energy industry financing ecosystem has entered the growth stage, but it has not yet achieved efficient collaborative evolutionary development. By reshaping the new energy industry financing ecosystem with big data as the core resource, constructing a big channel of data exchange for the new energy industry financing ecosystem in all aspects, innovating the key business and processes in the system with big data as the core, and obtaining the competitive advantage of the new energy industry financing ecosystem, building a big data information sharing platform to promote open collaborative innovation in the new energy industry financing ecosystem. We can provide strong support for the continuous optimization and collaborative evolution of the new energy industry financing ecosystem in the context of big data.

Recommender systems (RSs) based on big data have been shown to be very powerful tools for solving the information overload to assist the choice-making when dealing with the massive amount information in the age of big data and artificial intelligence. This paper presents an overview of the state-of-art RS that can be classified into four categories: content-based algorithms (CR), collaborative filtering-based algorithms (CF), and knowledge-based algorithms (KR), as well as hybrid recommendation-based algorithms (HR). The popular CF-based recommender algorithms are especially focused by classifying them into the memory-based algorithms, and model-based algorithms as they show the advantages of great rating prediction without contextual features compared to the rest of RS approaches. By reviewing the current RS and understanding their limitations, the emerging solutions or possible extensions that would improve recommendation capabilities involving deep learning, knowledge graphs, and parallel computing techniques are significantly discussed for future RS research direction. At the same time, by identifying current problems, some possible solutions will be shown in the last part.

DNNs (Deep Neural Networks) have been proved to be a successful technique in many areas. Understanding the reasons behind DNN is, however, quite important in assessing trust, which is fundamental if one plans to take action based on leaning results, or when choosing whether to deploy a new model. Lack of insights into the model can be an obstacle that hinders the development and application of the DNN. AdaBoost is a well-known boosting learning technique with better interpretability, combining many relatively weak and inaccurate rules to increase the performance. And as it modified sample weights in the training process, it shows excellent adaptability even in complex cases as it can alleviate overfitting. In the paper, we propose a method training the nets by the AdaBoost based on the soft label, which we call flexibly soft-labeling AdaBoost (FSL AdaBoost). Our soft labels are made in a novel and sequential way, adding further interpretability and adaptability to the learning process. Experimental results on several well-known datasets have validated the effectiveness and novelty of FSL AdaBoost.

Aiming at the problem of single characteristic quantity and low accuracy of traditional on-load tap changers, this paper presents a fault diagnosis algorithm of on-load tap changers based on IPSO-LSSVM. Firstly, the time-frequency signal extracted from the vibration signal and the improved scale permutation entropy form a multi-feature fusion matrix; Then, considering the redundancy and high computational complexity of the multi-dimensional feature matrix, the principal component analysis is used to screen the feature subset and obtain the sensitive feature subset; Finally, IPSO-LSSVM is used to select and classify the initial feature subset. Experimental results show that the classification accuracy of IPSO-LSSVM is improved by at least 5% compared with other methods.

This paper identifies the microstructural characteristics of the research teams of academic institutions using graphlet-based measures. The results provide references for the evaluation and development of research teams. Scientific collaboration networks in the Top 20 institutions were evaluated using papers published in the past 6 years on computer image recognition in the field of artificial intelligence. The structural features of 3–5 node graphlets were extracted and analyzed. Significant graphlet structures were distinguished, and graphlet-based measures were used to determine the similarities and differences in the scientific collaboration networks. It was found that the graphlet structures contained significant information, and the graphlet correlation measures could be used to distinguish the similarities and differences of scientific research teams. The data can be used to investigate collaboration efficiency and develop and expand research teams.

In recent years, massive multi-source heterogeneous South China Sea data have been widely used in the construction of South China Sea digital resources, such as the South China Sea Sovereign Evidence Chain Project. Due to the data sparsity, a large number of isolated data are generated, which seriously affects the analysis effect of the South China Sea Big Data. In this paper, we proposed a novel data association method. We collected data from the South China Sea Library Digital Resources as South China Sea evidence data, which is a sentence or paragraph containing time, place, people, institutions and events can prove the sovereignty of the South China Sea. According to the definition of the evidence weight by the International Court of Justice, the logical relationship of South China Sea evidence data was constructed. Firstly, we randomly selected 3068 data from 21174 evidence data to label the logical relationship. Secondly, we used the BERT pre-training model to extract the logical relationship of evidence data. Finally, the Knowledge Graph technology is used to retrieve and visualize the logical relationship of evidence data. In this paper, we applied the BERT to extract the logical relationships of evidence data with an accuracy of 0.78, which indicates that the model has some feasibility. This paper could help to improve the correlation of the South China Sea Big Data and to enhance the ability of data processing.

War archive is a quintessential big data issue about national history and military data security in urgent need of exploitation. Knowledge graph is one of the core technologies of knowledge engineering in the era of big data. With the ability of deep knowledge reasoning and progressively expanding cognition, knowledge graph has become a key technology for the construction and application in the field of military big data. Most of the existing knowledge graph is general knowledge graph for general fields, but there is no mature method of knowledge graph construction and application for the military archival big data. Taking the archival data of the War to Resist U.S. Aggression and Aid Korea as an example, this paper, based on the special needs for military archive fields, explores the construction path of knowledge graph from the aspects of knowledge modeling, knowledge extraction, knowledge fusion and knowledge management. At the same time, the application of knowledge retrieval, archive resource linking, knowledge Q & A, knowledge recommendation and other scenarios are explored.

By upgrading the existing distribution network fault statistical analysis system of Shanghai Institute of Electrical Science and Technology, comprehensive research is needed to investigate the hardware configuration and system functions of the system, optimize the system structure, meet the needs of the company’s distribution network business department, and provide strong support for the daily operation and management of the distribution network. In this paper, we present a semi-supervised learning enabled fault analysis method for power distribution network based on LSTM autoencoder and attention mechanism. We make the LSTM autoencoder’s loss function more robust so that it may be affected by both labeled and unlabeled input. Next, by minimizing the loss function, we can learn how the distribution of both types of data is distributed. We also added an attention mechanism to make the model performance more stable as the weight of the marked data changes. We apply the improved experience of LSTM autoencoder to the LSTM prediction model and realize the LSTM prediction model under semi-supervised learning. The proposed algorithm can effectively solve the problems of strong dependence of time series data and high cost of marking, so as to obtain better fault detection results of power distribution network.

The State Grid Shanghai Institute of Electrical Science and Technology fault statistics system requires implementation of data analysis and index control capability improvement. This requires the purchase of corresponding hardware and ongoing algorithm and rule optimization in accordance with the application’s actual needs. By enhancing the loss function of the unsupervised LSTM prediction model based on the enhanced experience of the semi-supervised time series anomaly detection algorithm based on LSTM autoencoder and attention mechanism, a semi-supervised LSTM prediction model is proposed in this paper to address the issue that the prediction model is affected by the abnormal data in the training set. This paper further proposes a semi-supervised anomaly detection algorithm based on ensemble learning after verifying the enhanced semi-supervised LSTM prediction model. The algorithm consists of two parts: an anomaly detection model based on LSTM autoencoder and attention mechanism, and a semi-supervised LSTM prediction model. In order to create a semi-supervised model with superior performance, we combined the semi-supervised LSTM prediction model with the semi-supervised LSTM autoencoder model after applying the improved experience of the LSTM autoencoder to the LSTM prediction model. Both algorithms can successfully address the issues of high marking costs and significant time series data reliance in order to produce improved time series anomaly detection results.

The China’s knowledge payment industry has entered the “pan-knowledge payment” period, and the overall situation of industry is gratifying. O2O paid knowledge platforms, which are not under the public media spotlight, are also developing slowly. The non-standardized knowledge services provided by this type of platform have a unique charm that standardized online knowledge payment products do not have because they can be personalized and socially interacted on site, and their business value is worthy of attention. The research issue of this paper is what factors influencing users’ willingness to pay can be identified and verified using statistical models from the user behavior and content information available on a typical China’s O2O paid knowledge platform. From the perspective of Information Adoption theory, we construct a model of the factors influencing the willingness to pay for O2O paid knowledge products, which includes two aspects: information quality and source credibility. We collected a total of 9,907 actual behavioral data and topic data from users of the China’s O2O paid knowledge platform, “Zaih.com”, and used log-linear regression model to verify the hypotheses. The results show that: in terms of information quality, content integrity, information relevance, information applicability and interactive empathy have significant positive effects on users’ willingness to pay; in terms of credibility of information sources, the expertise and attractiveness of knowledge providers indicated by integrity of their profile and their high level residence have significant positive effects on users’ willingness to pay. The above findings have important guiding implications for the three parties of O2O knowledge payment, especially for the service improvement and personal development of knowledge providers of this type of platform.

With the widespread promotion of smart grid, the power time series data collected by smart meters also increases rapidly. How to collect these data safely and effectively, analyze and utilize them, and provide better power supply service has become a hot topic of current research. The federated learning technology has attracted much attention from researchers in recent years and various federated learning-based applications have been utilized due to its characteristics of distributed, security, encryption, and reliability. In the development of smart grids, federated learning has been applied for data analytics, privacy preserving, energy management, and so on. This paper is aimed at exploring the feasibility of applying the federated learning framework to the area of smart grids. We conclude the analysis of power time series data, discussing the tribulations and solutions in the process of privacy preserving in the smart grid, and highlighting different challenges of federated learning with the smart grid. We present a summarization among federated learning-based methods with the smart grid for a variety of purposes, with the aim to draw a comparison among federated learning-based methods in the smart grid from different aspects.

The distribution network storage is sensitive to the channel reliability and delay, especially with the requirements of the new power system construction, the real-time control requirements of the secondary system of the distribution network are further improved. Due to the diversity of technology applications, distribution network communication naturally has multiple heterogeneous networks. The noncorrelation of medium voltage power line carrier and 5G is used. Using the multi-frequency adaptive medium ballast carrier and its fusion access technology with 5G can effectively avoid the impact of the single network channel, such as the wireless random channel index degradation on the service carrying performance. The overall service delay performance, reliability and anti-attack capability of the network are improved through the multi-channel self-operation and dynamic switching mode, so as to meet the new service communication needs of the distribution network in the construction of the new power system.

The application of medical big data and artificial intelligence algorithms are majorly popular in biomedical field. In this paper, BP neural network optimized by genetic algorithm was used to study the classification of muscle fatigue. Although BP neural network has a strong nonlinear mapping ability by using the gradient descent search method, it is easy to fall into the local minimum during the search process because of the randomness of the initial weights and thresholds generated, which would affect the training rate and the accuracy of muscle fatigue classification. the genetic algorithm was used to complete the configuration of the initial population parameters and the design of fitness function, and the optimal weights and thresholds that met the conditions were output to BP neural network. Finally, the classification results of muscle fatigue were output. The experimental results showed that the GA-BP neural network had a stronger ability to jump out of the local optimization compared with the classification effect of BP neural network. The maximum recognition rate of fatigue state reached 90.4%, and the model running time was 17.1 s, which was relatively reduced by 4.5 s.

As the construction of the new type of power system, the high proportion of the distributed power grid, the electric car charging load is increased, make the distribution network scenarios cannot adapt to the current operation mode, in view of the current distribution network scenarios cannot adjust to the problems in operation mode, put forward an improved strategy based on multiple load evaluation model and the distribution network of clustering method to generate scenario. First, building the appraisal model of multiple load distribution network, through the model for regional assessment and scoring load, will score results using K-means clustering initial scenario, secondly, as the guidance, can use different way to design of initial scenario, can get different use type set in the middle of the scene, finally by improving strategy clustering method to analyze the middle scene set, The typical scenario set of distribution network is obtained. Simulation results show that the proposed method can effectively provide support for regional distribution network planning of new power systems with multi-type energy participation.

In the future construction of new power system, in order to provide support and guidance for distribution grid planning, it is necessary to understand the economy of distribution grid. Therefore, this paper uses the multi-level fuzzy comprehensive evaluation method base on indicators to evaluate the technical economy of distribution grid. Then, a deep learning model is built, trained by a large amount of evaluation data, makes the network get the evaluation thought of each expert. Using this method can reduce the subjectivity in the process of evaluation, and can increase the rate of fault tolerance, the example analysis shows that using the evaluation results of this method is feasible and has practical significance.

With big data growing rapidly in importance over the past few years, Virtual Reality (VR) display technology has gained widespread attention. Especially, virtual avatars with unique identification are becoming increasingly important. Traditional generation methods are complicated, time consuming and low image quality. To improve the quality and rendering speed of generated virtual avatars, we present a set of optimized parameters of GAN based on deep learning. The experimental results showed that the best image quality is achieved in the case of ADAM (Optimization Function), BCE (Loss Function), l-r = 0.00075 (Learning Rate), the epoch = 200. The questionnaire survey showed that the recognition accuracy could be up to 84.29%. The conclusion of the questionnaire survey is consistent with our experimental results.

This paper considers the ordering, pricing, and coordination for a closed-loop supply chain (CLSC) made up of a risk-neutral manufacturer, a risk-neutral third-party collector, and a risk-preference retailer. We introduce the M-CVaR criterion depending on two parameters to describe the retailer's risk-preference behavior. The optimal order quantity, wholesale price, and acquisition price of decentralized and centralized decision conditions are found using the backward induction approach and the influences of the risk-averse and pessimistic coefficients on the optimal solutions are investigated using numerical simulations. The findings demonstrate that the optimal order quantity goes up with the risk-averse coefficient and falls down with the pessimistic coefficient in both centralized and decentralized conditions. The more risk-averse the retailer, the higher the manufacturer's wholesale price. The retailer's increased risk-taking and risk aversion tendencies could lead to the market exit. The manufacturer prefers the retailer who is more risk-taking. In the case of risk-taking, the CLSC's expected profit under the centralized condition may be smaller than under the decentralized condition, but in the case of risk-neutrality, the expected profit under the centralized condition is always bigger than that under the decentralized condition. The CLSC can achieve the expected profit under centralized conditions with risk neutrality through revenue-sharing and cost-sharing (RS&CS) contracts.

Vibration out-of-tolerance is a common condition of unqualified engine inspection, and the cause of vibration out-of-tolerance is closely related to the quality of the engine assembly. When the engine is found to vibrate out of tolerance, it needs to be reassembled. This process is very time-consuming and labor-intensive. In order to improve the assembly quality and reduce the vibration value during the engine assembly process. An explainable optimization algorithm is proposed, which combines LightGBM, SHAP and PSO. First, a vibration value prediction model is trained using the LightGBM algorithm. Second, SHAP is used to explain the vibration value prediction model, and obtain the importance order of each assembly process parameter, which is used as the subsequent optimization order. Finally, according to the optimization order, PSO algorithm is used to iteratively optimize the assembly process parameters and it uses the vibration prediction model as the fitness function. It has been verified by experiments that the optimized assembly process parameters can effectively reduce the vibration value of the engine, which has positive guiding significance for the assembly of the engine.

Stock price prediction generates interesting outputs for investors. In recent years, stock technical indicators (STI) have played an important role in stock price prediction. However, the current performance on high-dimensional data remains problematic due to the error rate in the correlational analysis. In this paper, we proposed a new correlational strategy to tackle such problems with STI and deep neural networks. We designed this strategy based on the Pearson correlation coefficient with a close index. we took eight companies’s stock data for our analysis. The experimental results demonstrate that BiLSTM’s performance (0.892% of $$R^2$$ R 2 ) outperformed GRU and LSTM in various factors for stock price prediction.

Privacy-preserving computing is an important topic in epidemic prevention, credit report and education resource sharing. Data analysis is the basis for realizing the overall speed increase of returning to work and production. This paper comprehensively summarizes the current academic research on privacy-preserving computing. In this work, we systematically summarize the four aspects of the related schemes on the attribute-based encryption schemes: keyword search, symmetric search, PSI, homomorphic encryption. We classify the schemes and compare the same-type schemes accordingly. Finally, we present the directions and challenges for our future work. This paper is intended to provide convenience for the subsequent research on privacy-preserving computing.

With the development of power digitization, BIM based modeling has been widely used in the process of power construction, and the challenges of privacy and big data security have also become an increasing trend. According to the requirements of digital management of power safety production, the big data security technology in the process of digital management modeling of power safety production is studied based on BIM Technology. Firstly, the adaptability of BIM in power safety production management and the basic model of big data safety analysis are introduced, including BIM data acquisition and visualization methods, as well as evaluation methods. Among them, the accuracy and effectiveness of big data security are further improved through the combination of high-precision positioning data based on satellite positioning and BIM data. Secondly, the modeling method and examples based on BIM big data security technology are analyzed from four aspects: risk identification, safety plan, safety inspection and safety training. Finally, through a group of practical applications, the verification results show that the proposed scheme is feasible.

The Internet of Things brings security issues as well as high connectivity. This paper proposes a security risk decision based on the bounded rationality of users, aiming at the security problems of power Internet of things. First of all, a sparse node cognitive network is constructed for each user. Based on this simplified cognitive network, each user establishes his own security decision by minimizing his own security cost in the real world. These two stages constitute a game-to-game framework. Then the concept of a structured Nash equilibrium (GNE) solution is proposed to solve the game decisions of users in security management under this bounded rationality. At the same time, an iterative algorithm based on the nearest point is designed to calculate GNE. Finally, we analyze the case of intelligent power station in the Internet of Things, and the results show that this algorithm can successfully identify key users. Other users need to consider the decisions of these key users in the security decision-making process, and their own security decisions also reduce each other’s security management costs.

5G has become an important driving force for the digital transformation of the power industry. Especially today, with the increasing security risk of big data, the safe and available power 5G business plays an important role in improving power business innovation and user experience. This paper analyzes the demand level differences of different businesses for big data security in the application scenario of power 5G business, and establishes the SLA classification and classification model. On this basis, the power 5G business objective is divided into three processes: business perception, business execution and business SLA evaluation, which are assigned to the corresponding slices respectively. A power 5G slice perception collaborative optimization model is proposed and solved iteratively by multi-objective particle swarm optimization algorithm. Simulation experiments show that compared with the traditional 5G application mode, the proposed scheme can search and optimize the network resource allocation through the cooperation between slices, effectively schedule the slice resources and improve the operation efficiency and performance of power 5G service.

As the cloud network collaborative operation platform becomes the start and end point of network traffic and the number of service types carried by the data center increases, traffic interaction between data centers (east-west traffic) increases gradually. In this paper, a virtualized network functions placement scheme is proposed to meet the demand of large bandwidth, low delay and differentiated load for the increasing east-west traffic, which mainly carries east-west traffic between data centers in different geographical locations. The proposed scheme provides fast, automatic, diversified and differentiated carrying service for the above traffic. The proposed scheme mainly includes three parts: 1) L3 virtual private network services, namely the cloud network backbone for customers in different network room or L3 virtual private network services cloud pool of resources, ensure data security; 2) differentiation bandwidth service, i.e. the cloud network backbone network according to the different quality of service requirements of different customers and money demand reserved link bandwidth resources; 3) self-service automation, choreographer and controller as the cloud network backbone to achieve business automation opened and deployment, administrators through management portal implementation business rapid deployment and adjustment, the customer can realize self-help resources through services portal application and business plan. Evaluation results have validated the effectiveness of the proposed method and show its advantages over counterparts.

Grounded theory can help scholars solve complex management problems and build a link between practice and theory and has received much attention in recent years. This paper takes CNKI as primary data source, constructs professional retrieval strategy, realizes initial collection of academic papers related to grounded theory, and then combines content analysis methods to manually annotate 2358 papers from titles, keywords, abstracts, and full texts. To realize whether one paper uses grounded theory approach to conduct research. Using the scientific knowledge graph drawing tool CiteSpace, combined with information visualization analysis and social network analysis methods, to identify characteristics of knowledge production driven by grounded theory as a knowledge carrier. The research indicates: In terms of research focus, related research focuses on the multidisciplinary application of grounded theoretical analysis methods and their integration with qualitative research, mixed research, case study and other methods, and focuses on revealing influencing factors of various research issues; in terms of time, grounded theory became the central node of complex networks in 2004, and began to focus on the integration and application of various research tools in 2006, with apparent characteristics of temporal changes; in terms of the features of scientific research collaboration, the grounded theory-driven scientific research collaboration process has formed multiple research sub-networks oriented to the fields of management and education. In the institutional cooperation network, it is found that the large-scale institutional cooperation network driven by the grounded theoretical analysis method has not yet formed and the formed ones were driven by grounded theory can be well used in key links such as innovation management, management consulting decision making, etc.

This paper faces the typical application scenarios of power grid equipment operation and maintenance, maintenance management, emergency disposal, etc., and establishes the array camera spatial deployment strategy of under the constraint condition of the number of acquisition terminals. The automatic fitting and three-dimensional correlation of real scene images are proposed to meet the requirements of single data association and real scene visualization service configuration capability of the power grid equipment model. The selected digital twin technology must conform to the demand characteristics of economic, compatibility, efficient, safe and sustainable development of the current power grid equipment management. The research results of this paper promote the development of DT technology and the application of data science in engineering.

Compression algorithm can drastically reduce the volume of spatiotemporal big data. However, lossy compression techniques are hardly suitable due to its inherently random nature. They often impose unpredictable damage to scientific data, making them unsuitable for data analysis and visualization that require certain precision. In this paper, we propose a tree-based indexing method using Hilbert curve. The key idea of this method is that it divides the space into minimum bounding rectangles according to the similarity of the data. Our algorithm is able to select appropriate minimum bounding rectangles according to the given maximum acceptable error and use the average value contained in each selected MBR to replace the original data to achieve data compression. We propose the corresponding tree construction algorithm and range query processing algorithm for the indexing structure mentioned above. Experimental results emphasize the superiority of our method over traditional quadrant-based minimum bounding rectangle tree.

The serious consequences of the leakage of personal privacy information spread in social networks have reached the point where we have to pay attention. Conventional research on social network privacy measurement usually measures the risk of user privacy leakage from different perspectives, such as personal data, privacy settings, information dissemination, etc. The factors that affect privacy risks considered in the research content are not comprehensive enough. This paper constructs an effective privacy measurement method based on the specific situation of measuring the risk of user information authorization behavior in the social network environment. To take full advantage of many factors to more accurately measure user privacy risk, We propose a new global privacy metric scheme to measure user privacy risk. The personal privacy risk is measured by the proposed Attribute Frequency Inverse User Frequency (AFIUF) algorithm, and then the importance of users in the social network is estimated according to the social network user importance algorithm. Finally, considering users’ common privacy leakage, the global privacy score is calculated by combining the similarity between users. We apply our metric scheme to a large dataset of real Facebook social networks. Experimental results demonstrate the effectiveness and efficiency of our solution.

The number and types of network access devices and functions in the cloud computing and big data environments are continuously increasing, so the ability to actively expand network services is needed to ensure that the security and privacy protection needs of network practices are met on a continuous basis. The traditional physical isolation network security protection architecture is facing challenges in new technology scenarios such as big data and mobile Internet, and urgently needs to be changed. Based on this, this paper proposes a 5G-based zero-trust network security platform to better adapt and serve the dynamic network application environment. The security architecture and model are firstly studied to provide security principles of the proposed platform. The implementation scheme of proposed platform is then further analyzed where security algorithms are suggested. The results show that the proposed platform demonstrates better performance in identity, program, configuration and behavior detection. Therefor, it is expected to better deal with the network risks under the new technology scenarios.

Mobile Internet is an emerging field that is developing rapidly. However, the frequent data leakage events caused by the security problems of mobile Internet have brought great impact to enterprises and individuals. Cryptography is promising way for data security and is widely used in the industries such as power systems. How to effectively use data encryption algorithms in the field of mobile Internet to meet the security requirements of data sharing is a problem to be solved. Therefore, this paper proposes a mobile data leakage prevention system that adopts the encryption algorithms SM3 and SM4 that are approved by the State Cryptography Administration of China. The cipher-text retrieval model is proposed for secure data sharing between data owner and data user via cloud servers. The system model that is built based on the studied encryption algorithms and models is describe. Additional functions are also designed for mobile data leakage prevention such as data security management and security data analysis. The proposed system is expected protect the data transmission, storage and sharing between data owner, data user and cloud server, providing an architecture to build a security mechanism to lower the risks from data theft, eavesdropping and tampering.

Cloud computing is an Internet-based virtualization application model that has been in development for these years. Cloud computing technology provides users with a large number of virtualized computing and storage resources. As a continuously developing emerging technology, some theoretical and technical systems have been formed in countries all over the world. However, the data security in the virtualization environment of cloud service providers has attracted extensive attentions of users over the years. In the cloud computing environment, data access control mainly includes two aspects: one is the safe transmission of required information resources in the network, and the other is the reasonable control of user permissions. In this paper, we analyze the cloud computing platform based on OpenStack and the related data security access control model is proposed. The data access control process is studied where trust of the cloud user is evaluated. An access control system based on the studied model is then designed whose architecture and its components are described. The proposed model and system is expected to provide a new technical method for data security protection so as to lower the risks of data leakage.

Data security storage technology is an important part of information management. The security of data storage is the key to user information security, system operation efficiency and economic benefits. The implementation of data security storage technology plays an important role in improving the efficiency of data management and reducing the operating costs of enterprises. Due to the data leakage caused by intentional or unintentional wrong operation, which infringes the user’s personal privacy, the traditional data storage methods face greater risks in the operation process. These disadvantages affect the use effect of the system, and even lead to system paralysis and data loss in serious cases. Distributed database has the characteristics of high degree of structure, good openness and small storage space requirements, and plays an important role in data management. In this paper, a data security storage model is studied combining the symmetric encryption algorithm, asymmetric encryption algorithm and the hash algorithm. A data security storage system is then proposed based on distributed storage while its functional composition, data transmission framework and database design are described. The proposed system in this paper is expected improve the data storage security so as to reduce the risks of information leakage.

Nowadays, with the rapid development of network technology and the rapid increase of information, the importance of data is becoming more and more obvious, and the requirements of data security are also improving. Using the detection and tracking technology based on data watermarking can quickly and effectively obtain the information of data watermarking, and trace the source of possible attacks, so as to improve the security of data. The traditional passive data watermarking detection and tracking technology has some limitations in the real-time tracking of possible attacks. Therefore, this paper proposed an active data watermark detection system. The related watermark embedding and detection technologies are firstly studied. The system architecture is proposed where primary modules and submodules are described. The proposed system is designed to be integrated into the network nodes and servers in order to embed and detect the watermark information in real time. In addition, the watermarking parameters can be adjusted according to the sensed network environment. The proposed system can be implemented in the industrial departments that own a large amount of sensitive data so that the data transmission can be monitored and traced.

With the applications of big data and cloud computing technologies in industries, data mining technologies have been developing rapidly in these years. However, privacy issues have been attracting attentions for users and researchers since the laws and regulations of protecting personal information are issued. How to appropriately apply data mining technologies while meeting the privacy protection requirements become an important problem to address. In this paper, the privacy preserving data mining technologies are studied including K-means, Support Vector Machine, decision tree and association rule mining. In addition to their principles, the corresponding privacy protection methods for them are discussed. Furthermore, the commonly used privacy protection methods are studied including restricted release, searchable symmetric encryption, homomorphic encryption and digital envelope. Finally, the suggestions are given that the data processing algorithms need to be improved to obtain the better balance between data mining efficiency and privacy protection, and the system could be designed to provide privacy protection measures to meet personalized demands. The studies in this paper are expected to provide technical ideas to various service providers such as personal recommendation to implement privacy protection strategies.

Existing methods for face de-identification often cause inevitable damage to the utility of facial information. The anonymized facial images can hardly be applied in practical applications. In this work, we propose a cluster-based generative model that conceals the identity of images while preserving the utility of facial images. We extract facial features in the first stage, then classify images into several clusters. Four naive protection methods, blindfold, mosaic, cartoon and mosaic, are adopted to form facial image inputs without private information. Along with the four de-identified images, a random facial image in the same cluster is also chosen as another input for better preservation of facial features. We train a novel model with multiple inputs, called Multi-stage Utility Maintenance-Variational AutoEncoder (MsUM-VAE), generating a facial image using the mentioned multi-inputs. The output of the model retains a large portion of the facial characteristics, but cannot be distinguished from the original image dataset, avoiding the disclosure of privacy. We perform numerous evaluations on the CelebA dataset to showcase the effectiveness of our model, and the findings indicate that the model surpasses conventional techniques for obscuring identity and maintaining the utility of images.

Cables in 10 kV distribution networks play a key function in ensuring the secure operation of distribution network systems. The traditional methods of diagnosing partial discharges in 10 kV cables need to be improved at the levels of cost, efficiency, diagnostic accuracy, and anti-interference capability. Therefore, this paper studies and designs a security diagnosis platform that can effectively extract, identify and diagnose the signal characteristics of cable discharge models. The architecture and diagnostic process of the proposed system are described in details. Through the security diagnostic verification of the partial discharge capability of 10 kV cables in distribution networks, the effectiveness and efficiency of the proposed platform is validated. The application results show its capability in quickly locating and eliminating hidden cable insulation problems in distribution networks. The proposed platform, through the integration of different modules, is expected to significantly improve the security and reliability of power system.

The generation of partial discharges in distribution networks may have a significant negative impact on the secure and stable operation of the power grid. Therefore, it is of great practical value to design a security system that can monitor the partial discharge status of the distribution network in real time, remotely and online. The security monitoring system is expected to effectively diagnose the health status of the power grid system and predict the development of its status. Although the traditional distribution network security monitoring system is capable of monitoring partial discharge to an extent, there is still much room for improvement in terms of signal immunity, data transmission efficiency and intelligent identification of partial discharge patterns. Hence, this paper designs an electromagnetic sensor-based partial discharge security monitoring system to achieve rapid identification, filtering, transmission and processing of local discharge signals. The effect of the proposed security system is verified by applying it to distribution lines, the results show that the security monitoring system based on electromagnetic sensors proposed can effectively meet the actual monitoring requirements of the power grid.

The scale of the power grid and the degree of informationization and intelligence of the power equipment are increasing, and the end devices and users are becoming more diversified and complex, which makes higher requirements on the stability and security of the electrical system. On the other hand, the facilities of the power grid system are widely distributed, and the relevant equipment is deployed in a more dispersed manner, which puts forward higher requirements on the real-time and remote nature of the security monitoring means of the distribution network. In this paper, through the analysis of the components of ultrasonic-based online security monitoring system, especially based on the mechanism of partial discharge and the mechanism of action between partial discharge and ultrasonic waves, the online security monitoring device of partial discharge is designed. The system contains operation modules including acquisition, amplification, filtering and processing of partial discharge signals. The simulation results show that the proposed device may improve the identification efficiency and the anti-interference capability.

Millions of intelligent terminals are connected in the IoT (Internet of Things) networks providing services for industrial applications around the world. The distributed deployment of terminals increases the exposure to cyber attackers. The detection of intelligent IoT terminals is the prerequisite for effective terminal security protections. Traditional detection methods have shortcomings such as low accuracy and difficult to detect heterogeneous terminals. Hence, a detection system for intelligent IoT terminals from the aspect of information security is proposed in this paper. Critical methods in the proposed system including comprehensive detection model and IoT protocol analysis method are studied. The system architecture including functional modules is introduced. The proposed system may operate in terminal mode and access point mode and work flows of both modes are described. Finally, the IoT terminal security is discussed from the levels of system security hardware, terminal file system and IoT network security. The proposed system is expected to detect terminals including hidden terminals in the networks of targeted area so as to effectively monitor and protect the terminals in target area.

Intelligent IoT (Internet of Things) terminals are widely deployed in IoT systems and complete tasks such as data collection, data processing and analysis. The intelligent services provided by IoT systems are implemented based on various application software installed in terminal side and server side. Hence, the behaviors of intelligent IoT terminals depend largely on the interactive behaviors of application software. It is necessary to monitor and identify IoT terminal behaviors so as to achieve security state awareness and abnormal detection. The identification system in the environment of limited hardware configuration should not affect the performance of the target system. Therefore, non-invasive instruction and data tracing method is studied in this paper where the supported tools are described. Based on the studied method, an intelligent IoT terminal software identification system is proposed and the functional modules are designed. The proposed system can operate in offline training mode and online identification mode. The processes of both modes are described in details. Finally, the terminal identification technology based on multi-source information is discussed. The proposed system is expect to identify the software and even the terminal in the IoT system so that the security mechanism can be strengthened.

The continuous access to various intelligent terminals in the power system has greatly improved the functionality and intelligence of the system, while also affecting the security of the whole system to a large extent. Through designing the identification model of power terminals from the aspects of security, convenience, applicability and efficiency, and verifying the identity of the terminals connected to the power system, it is an important premise and guarantee to effectively prevent malicious terminal access from causing damage to the whole system. In this paper, an identity identification model of power terminals is proposed and the discrete feature coding is optimized. The implementation of the proposed model in the power information network is then suggested. Experiments including training data sets, verification data sets and interaction environment are established. The results show the advantages of the proposed model in the aspects of feature selection, abnormal identity identification and identity classification of terminals.

Power IoT (Internet of Things) has been developing for a few years where various types of terminals are deployed. Since the power IoT devices need to be connected to the public network, the security situation is more severe, and it is imperative to develop an efficient and reliable vulnerability mining model for the device firmware in the power IoT field. Based on this, this paper analyzes the common mining means of power IoT device firmware vulnerabilities including static and dynamic analysis methods. By comparing the characteristics of different mining techniques and their applicability, an IoT device firmware vulnerability mining model applicable to the power system environment is proposed and its process and associated methods are designed. Finally, a test system is established to verify the effectiveness of the proposed model compared to the common static and dynamic analysis tools. The test results show that the proposed model demonstrates better performance in terms of execution time and code coverage efficiency.

The existing research on ethical issues in big data governance mainly focuses on organizational analysis, stakeholder analysis, industrial supply chain, and other fragmented management analysis, and pays less attention to the participation of non-profit organizations as third parties, which makes it difficult to solve the integration and systematic social problems of big data governance. The research takes the third-party social work service as the analytical framework. First, the ethical problems of big data governance are attributed to the decentralized production and intensive use of data and the separation of producers and users. Then it demonstrates and analyzes the multi-dimensional isomorphism of social work as a third party and big data governance in regards to work objects, ethical connotation, and technical requirements. Then, according to the professional advantages of social work, it is proposed that social workers should assume the roles of educators, coordinators, organizers, and advocates, and promote the data-rich represented by enterprises, hospitals, and governments to actively protect the rights and interests of the data-poor, to achieve a win-win situation for individuals, groups, and society in big data governance.

Distributed new energy consumption scenarios, such as photovoltaic, energy storage, charging stake, etc., are facing the needs of processing massive real-time data, large-scale distributed new energy and access to diverse loads. Based on the business characteristics such as business peak-valley dynamic change, network connection and time-delay differential demand of different business in energy and power business, a reasonable and effective integrated resource scheduling model of computing resources suitable for distributed new energy consumption scenario is studied to support power planning and dynamic dispatch application. In this article, we propose an arithmetic planning strategy based on federal learning. Specifically, we first introduce a computing priority network scheduling framework in edge cloud computing environments. Secondly, we process the absorption data of corresponding energy nodes by random forest algorithm, adjust the connection relationship between a large number of internal nodes, control and dispatch the nodes, and then conduct integrated training through federal learning to dispatch the computing power of the overall network, so as to achieve fast and accurate algorithm dispatch. Then, under the same environment conditions, the simulation experiments of deep neural network and random forest algorithm are compared. A large number of simulation results show that the system can effectively assist the smart grid in reasonable algorithm dispatch and improve the resource utilization efficiency.