CN111626324A - Seabed observation network data heterogeneous analysis integration method based on edge calculation - Google Patents
Seabed observation network data heterogeneous analysis integration method based on edge calculation Download PDFInfo
- Publication number
- CN111626324A CN111626324A CN202010284486.1A CN202010284486A CN111626324A CN 111626324 A CN111626324 A CN 111626324A CN 202010284486 A CN202010284486 A CN 202010284486A CN 111626324 A CN111626324 A CN 111626324A
- Authority
- CN
- China
- Prior art keywords
- data
- observation
- edge
- matching matrix
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to the technical field of data processing, and relates to a heterogeneous analysis integration method for data of a submarine observation network. The method comprises the following steps: the method comprises the steps that a local client acquires a direct data source of a submarine observation network and sends the direct data source to an edge server; the edge server carries out data preprocessing, calculates the characteristic value of each type of data and transmits the obtained characteristic value to the cloud management server; the cloud management server inputs the received characteristic values into a pre-trained data classification model for classification, and a matching matrix is generated; the cloud management server transmits the generated matching matrix to an edge server; and the edge server continuously transmits the matching matrix to the local, performs conversion of the matching matrix, and realizes analysis and integration of data at the local client. Based on the edge computing technology, when massive, multi-source, multi-type and complex storage structure direct data sources obtained by a submarine observation network are integrated into the existing ocean observation integrated system, the network data transmission quantity can be reduced, and the transmission efficiency and accuracy are improved.
Description
Technical Field
The invention belongs to the technical field of data processing, and relates to a heterogeneous analysis integration method for data of a submarine observation network.
Background
The submarine observation network consists of submarine observation nodes and shore base stations on land, and the submarine photoelectric composite cables connect the observation nodes to form the submarine observation network, so that the limitation of energy and storage of the traditional marine observation sensor is broken through, and long-term, real-time and continuous observation data acquisition is realized. The seabed observation network is a huge observation system, and can be connected into thousands of scientific observation instruments through armored seabed photoelectric composite cables to obtain various ocean hydrological observation elements. With the improvement of the current multi-source heterogeneous ocean information access and sharing requirements and the convenience of realizing the comprehensive and unified research on the ocean floor network data, the direct data source acquired by the ocean floor observation network needs to be integrated into the existing mature observation integrated system to realize the interconnection, intercommunication and interoperation of the ocean data.
An observation integration system commonly used by a submarine observation network is a Neptune (Neptune) submarine observation network observation integration system (hereinafter referred to as a NEPTune integration system), the NEPTune integration system has the characteristics of multiple observation element types and complex database table structures and naming methods, data in a direct data source acquired by the submarine network also has a series of characteristics of high quantity, multiple sources, multiple types, complex storage structures and the like, and when the direct data source acquired by the submarine network is analyzed and integrated into the NEPTune integration system by heterogeneous data, the data conversion calculation amount is large, so that the problems of long time consumption, high error rate and the like of operations such as analysis and matching are caused. Therefore, how to design an efficient heterogeneous analysis integration method for mass data is a key problem to be solved urgently in the construction of the current submarine observation network.
At present, methods for realizing heterogeneous analysis and integration of a direct data source of a submarine network include manual mapping integration and a machine learning analysis and integration method based on a cloud computing platform. These two approaches have the following disadvantages: (1) the manual matching process is complex, the workload is large, and the error rate is high; (2) when the analysis integration algorithm is completely handed over to the cloud computing platform, the data transmission distance is long, the network congestion and the delay time are long, and the real-time performance of data analysis integration is poor.
Disclosure of Invention
In order to solve the problems of large operation amount, high error rate and the like of the existing ocean observation network data heterogeneous integration analysis, the invention provides a seabed observation network heterogeneous integration analysis method based on edge calculation. In addition, a grouping weighting KNN classification algorithm improved aiming at the characteristics of large ocean data volume and unbalanced samples is provided, known kinds of data sample grouping and characteristic value variances are used as weight improvement algorithms, the classification analysis efficiency and accuracy are improved, the problem of real-time concurrent heterogeneous analysis integration of mass data is solved, and interconnection, intercommunication and interoperation of submarine network data are achieved.
The technical scheme adopted by the invention for solving the technical problems is as follows: the heterogeneous analysis integration method of the submarine observation network data based on edge calculation comprises the following steps:
the method comprises the steps that a local client acquires a direct data source of a submarine observation network and sends the direct data source to an edge server;
the edge server carries out data preprocessing, calculates the characteristic value of each type of data and transmits the obtained characteristic value to the cloud management server;
the cloud management server inputs the received characteristic values into a pre-trained data classification model for classification, and a matching matrix is generated;
the cloud management server transmits the generated matching matrix to an edge server;
and the edge server continuously transmits the matching matrix to the local, performs conversion of the matching matrix, and realizes analysis and integration of data at the local client.
As a preferable mode of the invention, the data classification model is constructed by adopting a classification algorithm, and historical observation data in a NEPTUNE integrated system is taken as a training set to carry out model training.
Further preferably, the classification algorithm is a group weighted KNN algorithm improved for marine data characteristics.
Further preferably, the improved grouping weighted KNN algorithm is to decompose the training set into groups according to the categories, calculate the position of the center of each group, represent the group of characteristic information by the center point, calculate the distance from the center of each group by the test set, select several groups with the closest distance, and then apply the KNN algorithm to the selected groups.
Further preferably, the historical observation data is preprocessed, and a characteristic value is extracted to be used as a training input of the classification model.
The invention has the beneficial effects that: based on an edge computing technology and a grouping weighting KNN classification algorithm improved aiming at characteristics of ocean data, when a seabed observation network obtains massive, multi-source, multi-type and complex storage structure direct data sources to be integrated into an existing ocean observation integrated system, (1) a local-edge-cloud framework is adopted, and when an edge server interacts with a local part and a cloud end, the edge server is closer to the data sources, so that data preprocessing is facilitated, only characteristic values obtained in a specified time period need to be transmitted to the cloud end, data transmission quantity is reduced, and transmission delay is reduced. (2) The improved grouping weighting KNN classification algorithm groups data samples, increases the variance of the characteristic values as weight values, and improves the classification analysis efficiency and accuracy.
Drawings
FIG. 1 is a schematic diagram of a data source integration process;
FIG. 2 is a flow chart of subsea network data source heterogeneous analytics integration;
FIG. 3 is a schematic diagram of classification model training;
FIG. 4 is a block diagram of a local-edge-cloud architecture data heterogeneous parsing function;
fig. 5 is a schematic flow chart of the subsea network data heterogeneous analysis integration process using the local-edge-cloud architecture according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described relatively clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment provides a method based on edge calculation to establish a matching model of a direct data source acquired by a submarine network and a NEPTUNE integrated system. As shown in fig. 1, the matching model is actually a mapping relationship between an observation element in a direct data source and an observation element in the NEPTUNE integrated system, and after the mapping relationship is established, when a data source is newly accessed, the heterogeneous data source can be quickly integrated and accessed into the NEPTUNE integrated system through the matching model. The specific implementation inside the matching model comprises two modules of element matching and element conversion, the element matching module needs to classify the accessed direct data source based on a classification model obtained by historical data training to form a matching matrix, and the element conversion module analyzes the generated matching matrix, so that the conversion integration of the observation elements of two different systems is realized.
Based on the above principle, the flow of the heterogeneous analysis integration method for seafloor observation network data based on edge calculation in this embodiment is shown in fig. 2, and the method includes the following steps:
1. training data classification model
The classification model is a core part in the element matching module, the classification model is obtained by training according to historical data, historical observation data in the NEPTUNE integrated system is used as a training set, the historical observation data is subjected to data preprocessing and feature acquisition, and then feature values are extracted.
The sensors in the submarine observation network may be affected by factors such as surrounding environment and nonlinearity of the sensors themselves, and the like, so that data loss or data distortion occurs, some data may be invalid abnormal values caused by random interference, and also may be valid reference values caused by continuous change of marine parameters, therefore, data preprocessing is to adopt some algorithms to control data quality and eliminate abnormal values. For example, the improved 53H algorithm may be used for data preprocessing, and the specific steps are as follows:
(1) let x (i) be the measured online data sequence. To construct a new sequence x from x (i)1(i) The method is to take the intermediate value of x (1), x (2) … and x (5) as x1(3) Then, x (1) is truncated, x (6) is added, and the intermediate value is taken to obtain x1(4) (ii) a And so on until the last data is added. Obviously, x1(i) Is 4 less than x (i).
(2) In a similar way at x1(i) Form a sequence x by selecting intermediate values from three adjacent numbers2(i)。
(3) Finally by the sequence x2(i) X is constructed as follows3(i)：
x3(i)＝0.25x2(i-1)+0.5x2(i)+0.25x2(i+1)
(4) If the following formula holds, use x3(i) In place of x (i),
|x(i)-x3(i)|≥k
wherein k is a predetermined value.
(5) And (3) reversely arranging the 8 points at the beginning and the 8 points at the end of the x (i) sequence to generate a sequence x' (i), namely:
x(8)，x(7)，x(6)，x(5)，x(4)，x(3)，x(2)，x(1)，x(9)，…，x(n-8)，x(n)，x(n-1)，x(n-2)，x(n-3)，x(n-4)，x(n-5)，x(n-6)，x(n-7)。
(6) repeating the previous 4 steps for the x' (i) sequence to form a new x3' (i) sequence, using x in the new sequence3’(5)、x3’(6)、x3’(7)、x3’(8)、x3’(n-7)、x3’(n-6)、x3’(n-5)、x3' (n-4) substitutes for x (4), x (3), x (2), x (1), x (n-1), x (n-2) and x (n-3), respectively.
The classification algorithm is the core of the classification model, which determines the accuracy of the whole integration process, the training phase of the whole model can be regarded as a process of learning the mapping function, the mapping function is finally available after data training, namely, the classification rule is an important parameter of the classification model, as shown in fig. 3, the classification algorithm provided by the embodiment improves the classification efficiency and accuracy by aiming at the problems of huge ocean data volume and unbalance of sample data caused by the sampling time frequency of the sensor. The overall description of the algorithm is: the basic idea of the KNN algorithm is to calculate the distance between a test set and each sample characteristic value in a training set, find k samples with the closest distance, and the class of the test set depends on the large classes in the k samples. The calculation method needs to calculate the distance between the test set and each sample in the training set, and the calculation amount is huge. The improvement method is that the training set is decomposed into groups according to the categories, the position of the center of each group is calculated, the center point represents the characteristic information of the group, the distance between the test set and the center of each group is calculated, a plurality of groups with the nearest distance are selected, and the KNN algorithm is applied to the selected groups. In addition, the distance between the neighbors is calculated by determining all feature values of the sample according to the same metric, which may cause that the distance between the neighbors may be dominated by a large number of irrelevant features, so that it is also necessary to give a smaller weight to the features with strong relevance and small dispersion degree in the classification, and a higher weight to the features with weak relevance and high dispersion degree in the classification. The specific algorithm is described as follows:
(1) historical data is obtained to obtain a training set X, and a certain feature vector in the training set is represented as
The training set X has m samples, m is the total number of samples (j is more than or equal to 1 and less than or equal to m),
representative sample xjN feature values in total;
(2) solving the center point coordinate C of each type of feature value, wherein the center point coordinate of a certain type is expressed as
Assuming that the number of the types of the observation elements is p, the central point coordinate set C has p (j is more than or equal to 1 and less than or equal to p) in total;
(3) obtaining a test set T obtained after data preprocessing and characteristic value extraction of a direct data source, wherein any characteristic vector in the test set is represented as
Computing a test set tjEuclidean distance l from each class of training set observation element central point coordinatec(tj,ck) Wherein:
(4) to lcSorting from small to large, and selecting the top q types (q)<p) observation elements, generating a new training set N, and expressing any one feature vector in N as
In the training set N, there is s(s)<J is more than or equal to m and 1 and less than or equal to s) samples;
(5) calculating the mean value mu of each characteristic value in the new training set NjVariance DjAnd carrying out normalization processing on the square difference:
(6) will Dj' weight ω as each featurejAnd calculating the Euclidean distance between the test set T and the training set N by adopting the following weighted KNN distance calculation formula:
(7) and sorting the Euclidean distances after the weight is introduced, classifying the test set T according to the selected main categories of the first k neighbors, and generating a matching matrix.
2. When a data source directly acquired by a submarine network is accessed, the accessed data source is stored and backed up, then the data is preprocessed (the processing method is the preprocessing of the historical observation data) to obtain the characteristic value of each type of data, then the characteristic value of the newly accessed data is classified by using the constructed classification model parameters to generate an element matching matrix, and finally the element matching matrix is analyzed to finish the integration of observation elements in the data source directly acquired by the submarine network into a NEPTUNE integrated system.
In the above whole process, when the classification algorithm is used for model training, the process needs to perform a large amount of repeated calculation, and the requirement on the performance of the server is also high, and in addition, the direct data source acquired by the submarine network also has the characteristics of large observed data amount and many observed elements, so the process of storing the partial data and extracting the characteristics also has a high performance requirement on the performance of the server.
In order to relieve the data operation processing pressure of each step, the present embodiment adopts a local-edge-cloud architecture to perform data heterogeneous analysis, and the architecture uses the characteristic that an edge node is close to a data source, uses the edge node to perform data preprocessing, reduces the data transmission amount, improves the transmission efficiency, divides the whole task of integration analysis, and is responsible for completing different integration processes according to the advantages and characteristics of each part, and the division is as shown in fig. 4.
In this embodiment, a flow of performing data heterogeneous analysis integration based on a local-edge-cloud architecture is shown in fig. 5, and the specific steps are as follows:
(1) the cloud management server reads historical observation data of the NEPTUNE integrated system, and training of a classification model is carried out based on the historical observation data;
(2) storing the classified training model parameters obtained after training into a database of a cloud management server;
(3) the method comprises the steps that a local client acquires a direct data source of a submarine observation network and sends the direct data source to an edge server;
(4) the edge server stores the received data source in a database;
(5) the edge server carries out data preprocessing and calculates the characteristic value of each type of data;
(6) the edge server transmits the calculated characteristic value to a cloud management server;
(7) the cloud management server classifies the received characteristic values by using the classification model parameters stored in the database in the step (2) to generate a matching matrix;
(8) the cloud management server transmits the generated matching matrix to an edge server;
(9) and the edge server continuously transmits the matching matrix to the local, performs conversion of the matching matrix, and realizes analysis and integration of data at the local client.
By adopting the heterogeneous integrated analysis method for the submarine observation network data based on the edge calculation, the technical scheme of the invention is further explained by integrating the dissolved oxygen and chlorophyll data analysis directly acquired by the submarine observation network into NEPTUNE.
(1) And the cloud management server reads the historical observation data of the NEPTUNE integrated system, and based on the historical observation data, an improved grouping weighting KNN classification algorithm is adopted for training a classification model.
A plurality of kinds of clear historical data are stored in the NEPTUNE integrated observation system, and if four kinds of data including dissolved oxygen, chlorophyll, turbidity and CDOM are provided, part of the historical data is selected as a data set to carry out algorithm training. For example: 156 times of observation data are selected, wherein in 156 times of observation lasting for 24 hours each time, 126 times are selected as a training set, 30 times are selected as a testing set, and dissolved oxygen, chlorophyll, turbidity and CDOM are the factors of classified observation.
The specific training steps are as follows: the data from 126 observations were subjected to a 53H data preprocessing algorithm, which, after completion of the preprocessing, selecting maximum value, minimum value, mean value, variance, zero crossing point, maximum positive slope and maximum negative slope (or selecting other characteristic values by combining with actual curve characteristics) as data characteristic values of each observation for 126 times of observation, obtaining the characteristic values, the 126 observations are processed as a training set of the improved grouping weighted KNN algorithm (the 126 observations have definite characteristic values and definite types), then 30 observations are processed as a test set to carry out 53H algorithm data preprocessing and extraction of maximum values, minimum values, mean values, variances, zero crossing points, maximum political rates and maximum negative slope characteristic values, after the extraction of the characteristic values is completed, and (3) carrying out verification classification on the 30-time test set by using an improved grouping weighting KNN algorithm, and training to obtain a k value with the highest classification accuracy.
(2) And storing the k-value parameter with the highest classification accuracy rate obtained by training into a cloud database.
(3) The local client acquires the actual data of dissolved oxygen and chlorophyll of the submarine observation network and sends the actual data to the edge server.
(4) And the edge server stores the received actual data of the dissolved oxygen and the chlorophyll into a database.
(5) The edge server carries out data preprocessing on the actual data of the dissolved oxygen and the chlorophyll and calculates the characteristic value of each type of data;
the edge server performs data preprocessing of 53H algorithm on each type of received data, and performs characteristic value extraction according to a certain time interval.
(6) The edge server transmits the calculated characteristic value to a cloud management server;
the edge server is used as a link of intermediate processing, original data with large data volume does not need to be directly transmitted to the cloud for processing, and only a characteristic value containing data type characteristics needs to be transmitted to the cloud server, so that network transmission flow is reduced, and transmission efficiency is improved.
(7) The cloud management server classifies the received characteristic values by using the classification model parameters stored in the database in the step (2) to generate a matching matrix
And (3) the cloud management server classifies the received feature value data of unknown classification by using an improved grouping weighting KNN algorithm and the k value and the training set stored in the step (2) to generate a matching matrix. The matching array distance is the data type of the actual observation data of the seafloor observation network in the NEPTUNE integrated observation system.
(8) The cloud management server transmits the generated matching matrix to an edge server;
(9) and the edge server continuously transmits the matching matrix to the local, performs conversion of the matching matrix, and realizes analysis and integration of data at the local client.
The edge server is still only a link of intermediate transmission, the edge server is sent to the local client, the local client analyzes according to the information of the matching matrix to obtain the type information of the original observation data, and then the type information is matched and integrated with the data in the NEPTUNE.
Claims (5)
1. The heterogeneous analysis integration method of the submarine observation network data based on edge calculation is characterized by comprising the following steps:
the method comprises the steps that a local client acquires a direct data source of a submarine observation network and sends the direct data source to an edge server;
the edge server carries out data preprocessing, calculates the characteristic value of each type of data and transmits the obtained characteristic value to the cloud management server;
the cloud management server inputs the received characteristic values into a pre-trained data classification model for classification, and a matching matrix is generated;
the cloud management server transmits the generated matching matrix to an edge server;
and the edge server continuously transmits the matching matrix to the local client, performs conversion of the matching matrix, and realizes analysis and integration of data at the local client.
2. The seafloor observation network data heterogeneous analysis integration method based on edge calculation as claimed in claim 1, wherein the data classification model is constructed by a classification algorithm, and model training is performed by taking historical observation data in a NEPTUNE integration system as a training set.
3. The method for integrating data heterogeneous analysis of the seafloor observatory network based on edge calculation as claimed in claim 2, wherein the classification algorithm uses a group weighted KNN algorithm improved for ocean data characteristics.
4. The method as claimed in claim 3, wherein the improved grouping-weighted KNN algorithm is to decompose the training set into groups according to the categories, calculate the position of the center of each group, represent the group of characteristic information by the center point, calculate the distance between the test set and the center of each group, select the groups with the closest distance, and apply the KNN algorithm to the selected groups.
5. The method for integrating data isomerism analysis of seafloor observatory networks based on edge calculation as claimed in claim 2, wherein the historical observation data is preprocessed to extract characteristic values as training inputs of the classification model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010284486.1A CN111626324A (en) | 2020-04-13 | 2020-04-13 | Seabed observation network data heterogeneous analysis integration method based on edge calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010284486.1A CN111626324A (en) | 2020-04-13 | 2020-04-13 | Seabed observation network data heterogeneous analysis integration method based on edge calculation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111626324A true CN111626324A (en) | 2020-09-04 |
Family
ID=72259679
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010284486.1A Pending CN111626324A (en) | 2020-04-13 | 2020-04-13 | Seabed observation network data heterogeneous analysis integration method based on edge calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111626324A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114050978A (en) * | 2021-11-11 | 2022-02-15 | 安徽电信规划设计有限责任公司 | Communication load big data adjusting method and system |
| CN114385601A (en) * | 2022-03-24 | 2022-04-22 | 山东省计算中心（国家超级计算济南中心） | Cloud-edge collaborative high-throughput ocean data intelligent processing method and system based on super computation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770263A (en) * | 2017-10-16 | 2018-03-06 | 电子科技大学 | A kind of internet-of-things terminal safety access method and system based on edge calculations |
| CN109240821A (en) * | 2018-07-20 | 2019-01-18 | 北京航空航天大学 | A kind of cross-domain cooperated computing of distribution and service system and method based on edge calculations |
| US20200007409A1 (en) * | 2018-06-29 | 2020-01-02 | Electronics And Telecommunications Research Institute | INTELLIGENT INTERNET OF EVERYTHING (IoE) EDGE COMPUTIING SYSTEM FOR HIGH RELIABLE INTERNET OF THINGS (IoT) SERVICE |
-
2020
- 2020-04-13 CN CN202010284486.1A patent/CN111626324A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770263A (en) * | 2017-10-16 | 2018-03-06 | 电子科技大学 | A kind of internet-of-things terminal safety access method and system based on edge calculations |
| US20200007409A1 (en) * | 2018-06-29 | 2020-01-02 | Electronics And Telecommunications Research Institute | INTELLIGENT INTERNET OF EVERYTHING (IoE) EDGE COMPUTIING SYSTEM FOR HIGH RELIABLE INTERNET OF THINGS (IoT) SERVICE |
| CN109240821A (en) * | 2018-07-20 | 2019-01-18 | 北京航空航天大学 | A kind of cross-domain cooperated computing of distribution and service system and method based on edge calculations |
Non-Patent Citations (2)
| Title |
|---|
| 胡桐 等: "《基于边缘计算的海洋观测要素自匹配方法》", 《海洋技术学报》 * |
| 陈杰 等: ""海底观测网络数据管理系统设计"", 《山东科学》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114050978A (en) * | 2021-11-11 | 2022-02-15 | 安徽电信规划设计有限责任公司 | Communication load big data adjusting method and system |
| CN114050978B (en) * | 2021-11-11 | 2023-08-29 | 安徽电信规划设计有限责任公司 | Communication load big data adjusting method and system |
| CN114385601A (en) * | 2022-03-24 | 2022-04-22 | 山东省计算中心（国家超级计算济南中心） | Cloud-edge collaborative high-throughput ocean data intelligent processing method and system based on super computation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
| CN109886464B (en) | Low-information-loss short-term wind speed prediction method based on optimized singular value decomposition generated feature set | |
| CN110232434A (en) | A kind of neural network framework appraisal procedure based on attributed graph optimization | |
| CN112185104B (en) | Traffic big data restoration method based on countermeasure autoencoder | |
| CN112231305A (en) | Digital power grid system and method based on digital twinning | |
| CN111626324A (en) | Seabed observation network data heterogeneous analysis integration method based on edge calculation | |
| CN111507504A (en) | Adaboost integrated learning power grid fault diagnosis system and method based on data resampling | |
| CN111861023A (en) | Statistical-based hybrid wind power prediction method and device | |
| CN116362325A (en) | Electric power image recognition model lightweight application method based on model compression | |
| CN112784920A (en) | Cloud-side-end-coordinated dual-anti-domain self-adaptive fault diagnosis method for rotating part | |
| CN116090757A (en) | Method for evaluating capability demand satisfaction of information guarantee system | |
| CN110751201B (en) | SAR equipment task failure cause reasoning method based on textural feature transformation | |
| CN113449920A (en) | Wind power prediction method, system and computer readable medium | |
| CN109657907B (en) | Quality control method and device for geographical national condition monitoring data and terminal equipment | |
| CN111027771A (en) | Scenic spot passenger flow volume estimation method, system and device and storable medium | |
| CN115809747A (en) | Pyramid cause-and-effect network-based coupling information flow long-term prediction method | |
| CN115389955A (en) | Differential pressure prediction method and device for total pressure test inside and outside battery | |
| CN115631771A (en) | Sound event detection and positioning method based on combined convolutional neural network | |
| CN115598459A (en) | Power failure prediction method for 10kV feeder line fault of power distribution network | |
| CN110739030A (en) | Soft measurement method for small sample in ethylene production process | |
| CN116359762B (en) | Battery state of charge estimation method based on deep learning and network compression | |
| CN116564423B (en) | Method and system for constructing microbial metagenome database | |
| CN117118689B (en) | Unbalanced data Internet of things intrusion detection method and device | |
| CN112650770B (en) | MySQL parameter recommendation method based on query work load analysis | |
| CN117196996B (en) | Interface-free interaction management method and system for data resources |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200904 |
|
| RJ01 | Rejection of invention patent application after publication |