Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

35results about How to "Good clustering effect" patented technology

Power system load data identification and recovery method

The invention discloses a power system load data identification and recovery method. Firstly, according to user historical load data, the number of clusters and initial cluster centers of sample data are determined on the basis of the hill climbing method; secondly, the final cluster center and the characteristic curve of the historical load data are obtained on the basis of the fuzzy C-means clustering algorithm; thirdly, each kind of load characteristic curve is processed, and the feasible region interval where normal data of the load curve is located is obtained; fourthly, according to correlation coefficients with the load characteristic curves, the category to which a to-be-tested load curve belongs is determined; finally, on the basis of the feasible region interval and the to-be-tested load curve whose category is judged, bad data of to-be-tested load data is identified and corrected. According to the method, the fuzzy C-means algorithm serves as the basis, the hill climbing function method is used, the number of clusters and the initial cluster centers are determined at the same time to improve clustering efficiency, and the initial cluster center determination problem and identification effect judgment randomness problem of bad data are solved.
Owner:TIANJIN UNIV

Driving behavior identification method based on intelligent mobile terminal

The invention discloses a drive behavior identification method based on an intelligent mobile terminal. The method includes the steps of S1, using the intelligent mobile terminal to collect and screenvehicle original state data which comprises the acceleration and angular speed information of three axes; S2, preprocessing vehicle original motion state data; S3, using a main component analysis method to acquire driving behavior comprehensive feature vectors from the preprocessed data; S4, using a k-means clustering algorithm to perform cluster partition on the driving behavior comprehensive feature vectors to obtain optimal clustering number; S5, using an FCM algorithm to cluster the driving behavior comprehensive feature vectors according to the optimal clustering number to obtain the deblurred final clustering result; S6, collecting real-time vehicle state data, and identifying the vehicle driving behaviors according to the final clustering result. By the method, refined driving behavior data clustering is achieved, and vehicle driving behavior features are effectively clustered into turning, speed changing and lane changing.
Owner:WUHAN UNIV OF TECH

Method for clustering network-based short texts

The present invention discloses a method for clustering network-based short texts. The specific implementation process comprises: firstly acquiring a network-based comment; pre-processing the acquired network-based comment, wherein the pre-processing comprises performing word segmentation on the network-based comment, then removing the word that is not used, segmenting a keyword, and performing weighted calculation on the keyword; and clustering the pre-processed texts. The method for clustering network-based short texts, as compared with the prior art, implements collection and analysis of massive data over the network, such that a user conveniently searches for valued information. With this method, the precision in clustering the network-based short texts is high, thereby accommodating practical needs of the user. Therefore, the method according to the present invention has great practicability and can be simply promoted.
Owner:QILU UNIV OF TECH

Data clustering method and system, and data processing equipment

The invention is applicable to the field of data processing, provides a data clustering method, a data clustering system and data processing equipment. The method comprises the following steps: inputting a data set consisting of n objects with a block data feature required to be clustered and an expected class number k; selecting k block data objects from the data set to serve as an initial class center; calculating the distance from each object to the initial class center; distributing each block data object to the center closest to the block data object according to the calculated distance to form k disjointed classes; calculating the center of each class to serve as a new class center; repeatedly executing the step of distributing each block data object to the center closest to the block data object according to the calculated distance to form the k disjointed classes and the step of calculating the center of each class to serve as the new class center until the algorithm is converged; obtaining the division result of the data set. By the data clustering method, the data clustering system, and the data processing equipment, the data with the block feature can be processed directly without compressing the block data, so that the loss of information is avoided, and the obtained clustering result is better than the clustering effect obtained after the block data is compressed.
Owner:SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI

Group concept-based improved Fast-Newman clustering method applied to complex network

InactiveCN102571431AAccurate portrayalClustering accuracy avoids the problem of not being the highest when reaching the global maximum avoidanceData switching networksModularityData mining
The invention discloses a group concept-based improved Fast-Newman clustering method applied to a complex network. According to the invention, the group concept is introduced; the adjacent cluster concept is confined according to the characteristics of complex network cluster structure; a modularity evaluation function proposed by Newman is improved, the maximal modularity evaluation functional value is saved, and the problem that the clustering precision is not highest at global maximum is solved, so the clustering result can more accurately reflect the real network cluster structure. Compared with a conventional FN clustering method, according to the method provided by the invention, the precision of the cluster analysis for the large scale complex network is greatly improved; and especially for the familiar complex network with large size, sparse connection and uneven relation, the clustering effect is more remarkable.
Owner:河南众诚信息科技股份有限公司

Invasion detection method and device

The invention is applicable to the field of an information safety, and provides an invasion detection method and an invasion detection device. The method comprises the steps of preprocessing primary data sets; carrying out distance measurement on the preprocessed data; obtaining the number of clusters based on a preset algorithm and the distance measurement; calculating the density indexes of preprocessed data points based on the distance measurement; calculating the distance indexes of the data points based on the distance measurement and density indexes; calculating the product of the density indexes and the distance indexes of the data points and sorting; selecting the former k data points as the center points of all clusters; distributing the rest of data points to the clusters which are closest to the data points and have the density indexes higher than the center points; sorting the clusters distributed according to the number of the data points, and judging the cluster with most data points in the cluster to be a normal cluster, and judging the rest of clusters to be abnormal clusters. According to the invasion detection method provided by the invention, the problems that the operation cost is high and the clustering result is affected by the setting of an initial value in the prior art can be solved effectively.
Owner:南方电网互联网服务有限公司

Gene classification method and device

The invention relates to a gene classification method and device. An LLE algorithm and an AP clustering algorithm are combined, and a proposed mixed kernel function is utilized to improve a similaritymeasurement function. According to the method, first, the LLE algorithm is adopted to map an original high-dimensional gene expression dataset to a low-dimensional space to achieve the purpose of dimension reduction; second, a new global kernel function is proposed as an F-type function, the F-type function and a Gaussian kernel function are linearly combined into a new mixed kernel function, theproposed mixed kernel function is utilized to calculate similarity measurement, and a new similarity matrix S is constructed; third, data is clustered through the AP clustering algorithm and the similarity matrix, and a final clustering result is obtained through iteration; and finally the effectiveness and accuracy of the method are verified through comparison with other clustering methods.
Owner:HENAN NORMAL UNIV

Doppler frequency-based 77G vehicle-mounted radar data quick clustering method

The invention discloses a Doppler frequency-based 77G vehicle-mounted radar data quick clustering method, and belongs to the field of radar data processing. The invention provides a method for performing three-dimensional clustering by using Doppler frequency information and reducing the operation time consumption by using a threshold-based optimal clustering number determination method, so that the clustering accuracy is improved and the operation time is reduced. Clustering process, clustering is carried out by using the target spatial position and Doppler frequency three-dimensional information; a relatively close position can be effectively distinguished; different targets with different Doppler frequencies exist; the optimal clustering number is searched; clustering thresholds according to settings, when the clustering number is traversed to calculate the clustering result and the evaluation function, corresponding calculation is stopped if the number of points in a certain classin the clustering result is smaller than a threshold value, the optimal clustering number and the optimal clustering result are obtained according to the maximum value of the evaluation function, increase of calculation time caused by increase of operation dimensions is reduced, and clustering accuracy and operation efficiency are improved.
Owner:UNIV OF ELECTRONICS SCI & TECH OF CHINA

Spectral clustering region division-based power system stability judgment method and system

The invention relates to a spectral clustering region division-based power system stability judgment method and system. The method comprises the steps of performing region division on the whole powersystem according to a spectral clustering algorithm; acquiring total capacity of a power generator in each region, a current frequency value of a node, a rated frequency value, a total mechanical power of the power generator and a total electromagnetic power of the power generator, and calculating a regional equivalent inertia time constant; correcting the regional equivalent inertia time constantof each region, and acquiring a region correction equivalent inertia time constant of each region, and calculating a system correction equivalent inertia time constant according to region correctionequivalent inertia time constants of all regions; and judging the power system stability according to the system correction equivalent inertia time constant. Compared with the prior, the method has the advantages that the influence of node frequency inconsistency during a disturbance period on inertia estimation of the power system is reduced, so that more accurate equivalent inertia time constantis obtained and is used for judging the stability of the power system.
Owner:SHANGHAI UNIVERSITY OF ELECTRIC POWER

Adaptive coloring method, system, storage medium, terminal for infrared image

The invention discloses an adaptive dyeing method for an infrared image, a system, a storage medium and a terminal. The method comprises the following steps: gray histogram statistics: calculating theproportion of pixel points of each gray value of the infrared image to the total pixel points; condition Judgment and Selection: judging whether the number of pixel points distributed on the number of gray scale above X% accounts for less than Y% of the total pixel points of the whole infrared picture, if so, entering K-Means clustering; Pseudocolor Discrete Transform of Rainbow Codes: the numberof clusters completed by K-Means clustering or Mean-Shift clustering is used as discrete color segments to divide the rainbow color from red to blue into several segments, and the center color of each segment is used for pseudo-color mapping. The invention adopts different clustering methods for images of different qualities and adopts interval adaptive rainbow code pseudocolor transformation, which has good effect on infrared pictures with low thermal contrast which need to obviously distinguish layers.
Owner:UNIV OF ELECTRONIC SCI & TECH OF CHINA

Method and system for selecting number of face clustering samples

The invention relates to the technical field of face recognition, and discloses a method for selecting the number of face clustering samples, and the method comprises the steps: constructing a face test set, and constructing a plurality of face training sets, wherein the number of face images of each face training set is different; clustering the plurality of face training sets to obtain a plurality of corresponding clustering centers; calculating a cosine distance between each clustering center and the feature vector of each face image in the face test set, and obtaining a mean value and a root mean square value of the cosine distance corresponding to each clustering center; and according to the cosine distance mean value and the root-mean-square value corresponding to each clustering center, obtaining the number range of the face clustering samples. Correspondingly, the invention further discloses a system for selecting the number of the human face clustering samples. The invention provides a method for selecting the number of face clustering samples, and a good clustering effect can be ensured.
Owner:上海铼锶信息技术有限公司

Ultra-short-term prediction method for power load of industrial park factory

An ultra-short-term prediction method for an industrial park factory electrical load comprises the steps of obtaining D-day historical electrical data of the industrial park factory electrical load, performing clustering analysis on the historical electrical data to obtain NCh1-class load electrical data, and then based on existing electrical data of now time periods before a to-be-predicted day, calculating the total correlation degree of the obtained typical power consumption curve of each type of load in the same time period; and then determining the power consumption load prediction basic values of the next npre time periods of the day to be predicted according to the calculation result of the total correlation degree; finally, predicting an error expected value of the same type of daily electricity consumption data by using historical daily electricity consumption data of various loads, and correcting the obtained electricity consumption load prediction basic value to obtain an electricity consumption load prediction result of the next npre time period of the to-be-predicted day. According to the design, the ultra-short-term prediction precision of the power load of the factory is effectively improved.
Owner:STATE GRID CORP OF CHINA +1

Abnormal signal semi-supervised classification method and system, and data processing terminal

The invention relates to the technical field of deep learning and wireless communication spectrum signals, and discloses an abnormal signal semi-supervised classification method and system, and a data processing terminal. The method comprises the following steps: establishing a deep clustering model, taking abnormal signal data as input of a CNN model, and then extracting compression features of input data as input of a K-means clustering algorithm for clustering; meanwhile, inputting the features extracted by the CNN into a classification layer of the CNN for classification; and finally, calculating the loss between the output of the K-means and the output of the CNN, and updating the parameters of the CNN until the iteration process is converged, so as to achieve the purpose of using the clustering result to assist in training the classifier. In order to enable the model to have better performance on a data set, optimization methods of a pre-training model, determining an initial centroid of clustering, constructing a category mean value Memory, replacing a pseudo tag and the like are introduced; in addition, the adopted semi-supervised learning method can enable spectrum management personnel to classify abnormal signals under the condition of small user interaction.
Owner:XIDIAN UNIV

Cable layout method for multiple substations and multiple types of fans, and computer storage medium

The invention provides a cable layout method for multiple substations and multiple types of fans, and a computer storage medium. The method comprises the steps that S1, acquiring initial parameters toform a data set; S2, minimizing the total square distance of the data set for all data points in the data set to obtain a cluster set; S3, according to the distance between the center of each clusterin the cluster set and the transformer substation, enabling the transformer substation to be in one-to-one correspondence with each cluster; S4, obtaining the total power generation capacity of the fans contained in each cluster, and obtaining the relation between the total power generation capacity and the maximum power generation capacity of the transformer substation corresponding to the cluster; S5, redistributing the data of the clusters according to the relationship, if the total capacity of the fans of all the clusters is smaller than or equal to the maximum power generation bearing capacity of the corresponding transformer substations, obtaining a clustering result, and otherwise, redistributing the data of the clusters; S6, calculating the distance between the fan of each clusterand the transformer substation corresponding to the cluster where the fan is located, and connecting the nearest fan point with the transformer substation; and S7, carrying out independent cable connection layout on the fans in each cluster.
Owner:GUANGDONG ANHANCE ELECTRIC POWER TECH CO LTD

Behavior identification method based on AP cluster bag of words modeling

The invention discloses a behavior identification method based on AP cluster bag of words modeling. The method comprises the following steps: detecting time-space interest points of videos; obtaining combined feature vectors by describing all the detected time-space interest point by use of a 3D HOG and 3D HOF descriptors; generating a visual dictionary by performing AP clustering on all the feature vectors, and re-describing the feature vectors by use of the visual dictionary; describing feature vectors of test videos by use of the visual dictionary; and obtaining behavior types of the test videos by learning and classifying features obtained through the precious two steps by use of a support vector machine. According to the invention, a proper visual dictionary capacity can be obtained at a time, multiple tests carried out for a conventional bag of words model are unnecessary, the clustering time can be greatly reduced, the method has a better clustering effect for multiple local features of combined description, and the behavior recognition rate is improved.
Owner:ZHEJIANG UNIV OF TECH

Underground logistics network node grading and site selection system and method

The invention discloses an underground logistics system node grading and site selection system and method. Underground logistics nodes are classified into first-level nodes and second-level nodes; twofactors of node distance and freight volume are considered; the Euclidean distance weighted quadratic sum of a traditional fuzzy C clustering method is corrected; a genetic algorithm and a simulatedannealing algorithm are combined, a good clustering effect is obtained, and the grading of the nodes and the difference of grading attributes are considered at the same time, so that the network transportation state of the underground logistics system can be reflected more truly, and a good data mining model can be provided for subsequent research of the underground logistics network.
Owner:ARMY ENG UNIV OF PLA

Adaptive spectral clustering method of extracting network node community attribute

An adaptive spectral clustering method of extracting a network node community attribute is disclosed. The method comprises the following steps of 1, setting the number of nodes in a mobile group intelligence perception network to be M and defining modularity of one community formed by the M nodes to be Qmax, wherein an initialized community number N equals to 1, a community attribute of a marked node v is Cv and the Qmax equals to 0; 2, acquiring a similarity matrix through an intimacy vector of the node v relative to the whole M nodes; 3, arranging characteristic values of the similarity matrix from large to small, clustering characteristic vector spaces constructed by the first N characteristic values and marking a community attribute of each node; 4, using the community attributes of all the clustered nodes to calculate the modularity Q, if the Q is greater than or equal to Qmax, making the Qmax equal to the Q and making an optimum community classification number Nop equal to the N, otherwise, directing entering into step 5; 5, making N equal to N+1; 6, repeating step3 to step5 till that the N equals to the M, the Nop value is the optimum community classification number and the nodes in the community possess the optimum community attribute. By using the method, accuracy of the network node community classification can be increased.
Owner:保定笙墨信息科技有限公司

Dynamic streaming data clustering method

The invention discloses a dynamic streaming data clustering method, which comprises the steps of converting structured data into time field streaming data, sorting the time field streaming data according to time fields so as to acquire time slices, and solving a union set; building a training model, and building HMM prediction for the missing data; checking the data validity, and adding time slices for repeated data points; eliminating abnormal data, checking whether data with abnormal fluctuations exists or not according to all of the time slices; and performing mass center data clustering. According to the invention, special optimization is performed in allusion to characteristics of the data, an HMM is adopted to perform prediction in allusion to the missing data, and processing is performed in allusion to repeated data with the same identification in the same time slice, so that time-varying characteristics of the data can be reflected more accurately, abnormal data can be distinguished, the number of clustering categories is optimized automatically, and a high-quality clustering result is acquired.
Owner:CHENGDU SEFON SOFTWARE CO LTD

Integrated unsupervised student behavior clustering method

The invention provides an integrated unsupervised student behavior clustering method aiming at the limitation of a questionnaire method in the aspect of data collection and the serious dependence of astatistical method, a supervised learning method and a semi-supervised learning method on student tags. The method comprises the following steps: firstly, extracting characteristics of student behavior data, dividing the characteristics into three parts, describing a centralized trend of the data by utilizing a mode, an average value and a range, expressing a discrete situation of the data by utilizing a minimum value, a first quantile, a median, a third quantile and a maximum value, and measuring a law degree of behavior occurrence time and a law degree of a behavior place by utilizing Shannon entropy; then, selecting the optimal behavior characteristics through variance and correlation analysis; and finally, carrying out initial clustering on the behavior characteristics of the studentsby utilizing DBSCAN, and further subdividing the super-large cluster by adopting K-means to obtain a final clustering result. The method does not depend on student tags, clustering is completed onlyby analyzing behavior data, and a foundation is laid for refined service and management of students.
Owner:BEIJING UNIV OF TECH

Keyword vectorization method based on topic semantic information and application thereof

PendingCN114298020AReduce vector dimensionAddressing missing semanticsSemantic analysisCharacter and pattern recognitionText categorizationSemantic feature
The invention discloses a keyword vectorization method based on topic semantic information and an application thereof. The keyword vectorization method specifically comprises the following steps: firstly, generating a vector with document semantic information for each document by utilizing a Sension-BERT model; dimension reduction is carried out on the generated document vector through a UMAP dimension reduction algorithm, and local semantic features are highlighted; then, HDBSCAN topic clustering is carried out on the document vectors after dimension reduction, and each document is classified into one or more topics; and finally, calculating a subject term frequency-inverse subject frequency (TTF-ITF) score of each keyword in the subject by using a relationship between the document and the subject, and merging the keyword and the subject term frequency-inverse subject frequency (TTF-ITF) score of each subject to generate a final keyword vector. According to the method, high-precision keyword vectorization of topic semantic information is realized, and the method can be applied to topic word extraction, text classification and document retrieval.
Owner:NANJING UNIV OF POSTS & TELECOMM

Resident load ultra-short-term prediction method

A resident load ultra-short-term prediction method comprises the following steps: S1, acquiring daily historical electricity consumption data of resident electricity consumption loads, performing clustering analysis based on daily electricity consumption data of a day-ahead time period to obtain class load electricity consumption data, S2, selecting load power consumption data of the same category as the day to be measured, and performing wavelet decomposition to obtain three components; and S3, training the sum component by using LSTM to respectively obtain prediction results of the three components of the day to be measured, and superposing the prediction results of the sum component and the three components to obtain a prediction result of the day to be measured. According to the design, the precision of ultra-short-term prediction of the power consumption load of a single household is effectively improved.
Owner:ECONOMIC & TECH RES INST OF HUBEI ELECTRIC POWER COMPANY SGCC

Gesture detection method and system based on space-time sequence diagram

The invention provides a gesture detection method based on a space-time sequence diagram, which belongs to the technical field of computer vision, and comprises the following steps: constructing the space-time sequence diagram of hand articulation points, and extracting a feature relationship between each articulation point and an adjacent articulation point; performing position coding operation on the feature relationship to obtain a position coding vector; coding by combining the position coding vector and the feature relationship to obtain an action vector; performing time sequence coding on the action vectors to obtain time-space relation vectors of the articulation points and other articulation points; and carrying out clustering analysis on the action vector and the space-time relation vector to realize classified recognition of gestures. Hand key points are extracted from input video frames according to unnecessary parameters such as viewpoints and illumination, and abnormal behaviors in gesture actions are effectively detected; and the model is trained in a semi-supervised mode, behaviors are clustered and judged in a clustering mode, a good clustering effect is obtained, and recognition and anomaly detection of gesture actions in different environments are achieved.
Owner:SHANDONG NORMAL UNIV

Power distribution network single-phase earth fault line selection method based on feature fusion and clustering

The invention discloses a power distribution network single-phase earth fault line selection method based on feature fusion and clustering, and the method comprises the steps: 1, extracting a zero-sequence current waveform of each line in a power distribution network based on a set single-phase earth fault condition, and determining a fault feature matrix of each line according to the zero-sequence current waveform; step 2, reconstructing the fault feature matrix of each line, and calculating a correlation coefficient matrix of the reconstructed fault feature matrix; step 3, according to the correlation coefficient matrix and a preset characteristic equation, calculating a characteristic database matrix formed by characteristic values of the correlation coefficient matrix; and 4, performing clustering operation on the principal component feature vectors in the feature database matrix based on a K-means clustering algorithm to obtain a fault clustering center model so as to determine the single-phase earth fault line of the to-be-detected line. Through the technical scheme in the invention, the problems of low accuracy and large characteristic calculation amount in the existing single-phase earth fault line selection process are solved.
Owner:DATONG POWER SUPPLY COMPANY OF STATE GRID SHANXI ELECTRIC POWER

Keras model-based spectrum identification and classification method

The invention belongs to the technical field of information classification methods, and particularly relates to a Keras model-based spectrum identification and classification method. The method comprises the following steps: 1, establishing a standard sample library for a sample set; 2, randomly dividing the standard sample library into a training sample set and a test sample set; 3, constructing a model; 4, compiling the model; 5, training the model; and 6, outputting a classification result. According to the method, the spectral text file is converted into the two-dimensional matrix, high-precision spectral recognition and classification are carried out by building the classification model, and the model structure can be continuously optimized to improve the classification effect; and the spectrum is classified and recognized through the algorithm, manual intervention is reduced, and the classification efficiency is improved.
Owner:BEIJING RES INST OF URANIUM GEOLOGY

Risk model construction method and device, risk detection method and device and computer equipment

The invention relates to a risk model construction method and device, a risk detection method and device and computer equipment. The risk model construction method comprises the steps that behavior log data and risk log data are obtained, the behavior log data comprise log data generated by behaviors using a system, and the risk log data comprise log data recorded when the system generates risks; integrating and screening the behavior log data and the risk log data to obtain risk behavior data; determining risk feature dimensions according to the triggering conditions of the risks; determining a behavior deviation degree according to the behavior log data and a predetermined behavior clustering model; and constructing a risk detection model according to the risk feature dimension, the behavior deviation degree and the risk behavior data. By adopting the method, the risk confirmation difficulty and time cost can be reduced.
Owner:深圳竹云科技股份有限公司

Unit grouping method and device based on time sequence clustering analysis of wavelet transform

The invention provides a coherent unit grouping method based on time sequence clustering analysis of wavelet transform, which comprises the following steps: acquiring a power angle swing curve delta deltai(t) of each generator through data acquisition after each generator breaks down and after the fault is removed, and performing Haar wavelet transform on each swing curve for multiple times, obtaining characteristic statistics of the time sequence, and combining the low-frequency trend signals of the time sequence with the statistics to form a characteristic matrix; and normalizing the formed characteristic matrix, and clustering by using a fuzzy c-means clustering method based on particle swarm optimization to realize clustering of coherent units. According to the method, the trend information and the statistical characteristic quantity of the time sequence are utilized in clustering, the effectiveness of coherence clustering is guaranteed, and the method has a good clustering effect and very high calculation efficiency and is suitable for coherence clustering of a large-scale system.
Owner:STATE GRID TIANJIN ELECTRIC POWER +1

Intelligent networked vehicle lane changing method combining emergency degree and game theory

The invention discloses an intelligent networked vehicle lane changing method combining the emergency degree and the game theory. Basedon a machine learning algorithm principle, clustering analysis is adopted to carry out emergency degree definition on a vehicle lane changing data set. The method innovatively predicts the emergency degree of the intelligent networked vehicle at the lane changing moment through an RBF radial basis function neural network, obtains an emergency degree factor, introduces the factor into a game matrix, calculates a profit value, and finally obtains a lane changing decision result. Compared with an existing intelligent networked vehicle lane changing method only depending on the game theory, the method is high in real-time decision-making performance and convenient to achieve through a computer, the emergency degree of the vehicle lane changing moment can be quantified, the execution efficiency is higher and the decision-making accuracy is better during lane changing decision making, the defect that a traditional game theory lane changing model is not high in adaptability is overcome, theoretical and technical support can be provided for an intelligent networked vehicle lane change collision early warning system and an autonomous lane change decision system.
Owner:SHIJIAZHUANG TIEDAO UNIV +1

Power grid temporary splitting method and system for blocking multi-direct-current commutation failure linkage

The invention discloses a power grid temporary splitting method and system for blocking multi-direct-current commutation failure linkage. The method comprises the steps of designing an index which canevaluate the mutual influence of commutation failures among a plurality of direct currents, and carrying out the quantification of the mutual relation among the plurality of direct currents; mappingthe plurality of direct currents into points in the graph, describing a mutual relation between the direct currents by connecting lines between the points and weights of the connecting lines, and constructing an incidence relation graph containing the plurality of direct currents; based on the association relationship graph, establishing a graph cutting mathematical model; and solving the graph cut mathematical model by adopting a clustering algorithm, and obtaining a power grid temporary splitting strategy on the basis of comprehensively evaluating the effects of various clustering algorithms. According to the invention, starting from weakening the mutual influence of a multi-direct-current system through an alternating-current power grid, the risk of a novel cascading failure caused by commutation failure of multiple direct currents is reduced by means of the operation means of temporary disconnection of the power grid, and the problem of formulating a temporary disconnection strategy, namely the problem of answering disconnection among the direct currents, is mainly solved.
Owner:SHANDONG UNIV

Cancer subtype identification method and system based on self-attention deep learning

The invention provides a cancer subtype identification method and system based on self-attention deep learning, and the method comprises the following steps: firstly, carrying out the preprocessing of multi-omics data of a cancer, then employing a deep learning Dense network to learn the low-dimensional features of each omics, and carrying out the preliminary integration of different omics features through a splicing mode; and then a similarity matrix between samples is constructed by using self-attention, and feature fusion is carried out according to matrix weight and splicing features to obtain final integrated feature representation. And using a decoder to minimize an error between the fusion feature and the original omics feature, and performing adversarial learning of integrated feature distribution through a discriminator. And finally, clustering the learned integrated feature distribution through a Gaussian mixture model to identify cancer subtypes. According to the method, multiple omics data can be effectively integrated, meanwhile, the relation between samples is modeled in a self-adaptive mode, better feature representation is learned, a better clustering result is obtained, and accurate recognition of cancer subtypes is achieved.
Owner:XUZHOU MEDICAL UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products