38results about How to "Address missing data" patented technology

The invention discloses a resource management method and system for large-scale distributed deep learning, and the method and system achieve the memory resource optimization management of parameters and gradient intermediate data during the training operation of a neural network, and guarantee the reasonable configuration of distributed communication bandwidth resources. Cross-layer memory reuse is realized again, intermediate data required by iterative computation and sparse communication are migrated into a CPU main memory and then migrated back as required, and interlayer memory consumptionis reduced; on the basis of reasonable migration of CPU-GPU data, independence of intra-layer memory reuse, intra-layer calculation mining and memory access operation is achieved, and intra-layer memory consumption is reduced as much as possible. Distributed parameter communication optimization is realized while efficient utilization of memory resources is ensured. Data access in the distributedparameter updating stage is reasonably redirected, a CPU main memory serves as a mirror image access area, the data access to parameters and gradients is completed, and the problems of gradient data missing and parameter writing border crossing are solved.

The invention provides a vehicle mileage data statistical method, which comprises the steps of: receiving and analyzing vehicle data information uploaded by a vehicle-mounted terminal; acquiring a mileage record cumulative value, and replacing an original mileage record cumulative value in a first database with a vehicle mileage value when the original mileage record cumulative value is less thanthe vehicle mileage value; acquiring a mileage record cumulative value in the first database at a preset daily recording time point, and associating the mileage record cumulative values, mileage record dates and vehicle identification information; and regarding a difference value between a vehicle next-day mileage record cumulative value and a vehicle previous-day mileage record cumulative value as a mileage value run by a vehicle the day before based on daily cumulative mileage record data in a second database, and forming a piece of vehicle daily mileage statistical data. The invention further provides a vehicle mileage data statistical device. The vehicle mileage data statistical method and the vehicle mileage data statistical device can process the vehicle mileage data, and precisely count the vehicle daily mileage and monthly mileage data.

The invention discloses a power distribution network big data restoration method based on an adversarial game, and relates to the technical field of electrical engineering. Building a generator framework; building a discriminator framework; training the network; and generating missing data of the power distribution network according to the trained network. The method has the beneficial effects that a mutual confrontation game model between the generator and the discriminator formed by the deep neural network is adopted in the scheme; the generator tries to generate samples which are consistent with real data in distribution, the discriminator identifies the samples generated by the generator as much as possible, and the two networks progress together in mutual comparison. By establishing the conditional adversarial game model of gradient penalty optimization, the generation process of multi-class missing samples can be guided, and the problem of unstable training of an original adversarial game model is solved, so that the problem of data missing caused by influences of measurement faults, distributed energy and the like in the power distribution network can be solved.

The invention relates to the field of traditional Chinese medicine clinics, in particular to a cleaning method for traditional Chinese medicine clinical data. The method includes the steps that S01, electronic medical records are acquired, and according to disease names in the electronic medical records, knowledge datasets are retrieved from a knowledge base; S02, according to a knowledge code corresponding to each piece of knowledge data in the knowledge datasets, corresponding prescription information is retrieved, wherein the prescription information comprises medicine names; S03, each medicine name of each piece of knowledge data is compared with the medicine names in the electronic medical records, the knowledge data with medicine name comparison having the most times of accordance and the accordance rate exceeding a specified threshold is selected as targeted knowledge data; S04, based on the targeted knowledge data, missing data of the electronic medical records is complemented. By means of the method, the problem of data missing of traditional Chinese medicine clinical data is effectively solved; the method is widely suitable for the links of data complementation in the process of data entry and data cleaning after the data entry, and the standardization degree and data utilization capacity of the traditional Chinese medicine clinical data are effectively improved.

The invention discloses an intelligent detection management system applied to the electric power field and an intelligent detection method thereof. The method comprises the steps: scanning a first identification code through an electric power detection auxiliary terminal to obtain the instrument identification information of a test instrument, and transmitting the instrument identification information to an intelligent operation and maintenance management platform to obtain a target detection item list of the instrument; determining a target detection item of this time, obtaining device identification information of the device to be detected by scanning the second identification code, sending the device identification information to the intelligent operation and maintenance management platform to obtain target detection point information of the device for the target detection item, and prompting a tester, thereby reducing the requirements for the tester and improving the test efficiency. using the test instrument for detecting a target detection item of the equipment to be detected, sending a detection value corresponding to each detection point to the electric power detection auxiliary terminal, wherein the electric power detection auxiliary terminal associates the detection value with the corresponding detection point, associates the detection value with identity informationof a tester this time and then uploads the detection value to the intelligent operation and maintenance management platform. The accuracy in the detection data collection process is improved.

The invention relates to a real-time GNSS data processing method and system. The method comprises the steps that a task management scheduling server obtains GNSS data streams from observation base stations in real time, and the task management scheduling server determines a resolving server for processing the GNSS data streams according to the load state of the resolving server; the resolving server obtains the GNSS data stream and the last processing result of the observation base station to which the GNSS data stream belongs; the resolving server carries out task processing according to the GNSS data stream and the last processing result; the resolving server monitors task processing; when the monitoring thread receives a feedback task processing completion message, the tracking information cached in the buffer area is emptied; and when the monitoring thread receives a feedback task processing failure message, processing is carried out again until a task processing completion message is obtained, and a GNSS data processing result is output. The task processing is monitored through the monitoring thread, so that the reliability of the task processing is ensured, and the problem of data missing is solved.

The present invention provides a power line communication signal reconstruction method. With the power line communication signal reconstruction method adopted, a power line communication signal sequence with no missing data can be generated by means of reconstruction. The method includes the following steps that: a power line communication signal sequence is collected and is converted into a two-dimensional signal; the two-dimensional signal is decomposed, so that a power line communication signal decomposition coefficient is obtained; the actual edge probability distribution of each parameterin the decomposition coefficient as well as the mean and mean square error of the parameters are calculated; and the power line communication signal decomposition coefficient is adjusted according tothe actual edge probability distribution of each parameter in the decomposition coefficient as well as the mean and mean square error of the parameters; and a power line communication signal sequencewith no missing data can be generated according to the adjusted power line communication signal decomposition coefficient. The method of the invention belongs to the electric power field.

The invention discloses a method for batch mounting of cross-platform electronic files. The method comprises the following steps: S1, data description; S2, digital acquisition; S3, dual-computer proofreading; S4, data entry; S5, format conversion; S6, data checking; S7, directory association, wherein S4 is specifically divided into three steps: scattered data input, archived data import and OCR recognition of scanning text. The import process is simulated and the data is checked after the import by adding the steps of digital collection and data inspection on a conventional mounting method, the problems of data import errors and data missing due to the improper mounting mode of the existing electronic archives are solved, and the integrity and standardization of the imported data are guaranteed.

The invention discloses a dynamic time delay adjusting device and method for unginned cotton foreign fiber sorting system. The device comprises a first linear array CCD camera, a second linear array CCD camera, a cotton falling cavity, a gas spray valve, a spray valve driver and an embedded type FPGA image processor; the first linear array CCD camera and the second linear array CCD camera are positioned at two sides of the cotton falling cavity with vertical height difference h existing between the two cameras, images of unginned cotton lumps falling in the cotton falling cavity are collected, the first linear array CCD camera and the second linear array CCD camera are connected with the embedded type FPGA image processor via CameraLink wire cables, the spray valve driver is connected with the embedded type FPGA image processor via a difference signal wire cable, and the gas spray valve is connected with the spray valve driver. According to the dynamic adjusting device and method, the embedded type FPGA image processor is used for image processing operation and feature matching operation, unginned cotton flow falling speed can be accurately measured and calculated, foreign fiber striking time delay of the unginned cotton foreign fiber sorting system can be dynamically adjusted, and accuracy of the unginned cotton foreign fiber sorting system can be improved.

The invention provides a power line communication signal adaptive reconstruction method which can reconstruct and generate a power line communication signal sequence without lost data. The method comprises the steps of: acquiring a power line communication signal sequence and converting the power line communication signal sequence into a signal matrix; according to the signal matrix obtained by conversion, constructing a transformation operator matrix; constructing a measurement matrix; according to the obtained transformation operator matrix and the constructed measurement matrix, iterativelyupdating the signal matrix, until current iteration times are equal to a length of the power line communication signal sequence; and carrying out conversion on the currently obtained signal matrix togenerate the power line communication signal sequence without lost data. The invention relates to the field of communication.

The invention discloses a data breakpoint resuming method for a new energy centralized control system, which comprises the following steps that: when communication is interrupted, a power station caches real-time data, copies and stores a real-time library of a centralized control master station as a real-time library copy at the moment, and records a corresponding communication interruption eventin an inversion control library resumeDB; after the communication is recovered, the compressed file is uploaded to a centralized control main station; the centralized control master station fuses thecompressed file and the real-time library copy to form an inversion packet and complements the communication interruption event; the centralized control main station consumes data in the inversion packet in a periodic micro-batch processing mode according to the completed communication interruption event record and inversion packet, and performs time sequence and time section calculation on the basis of the data to obtain cooked data; and the cached real-time data and the cooked data are stored into a historical library. According to the method, seamless fusion of continuous transmission dataand real-time monitoring historical data is realized, the problem of data loss in a new energy centralized control system is solved, and high requirements of operation and maintenance and analysis services on historical data integrity are met.

The invention discloses a dam safety monitoring data completion method based on space-time multi-view fusion, which comprises the following steps: abstracting view models on a global space view, a global time view, a local space view and a local time view respectively according to the characteristics of dam safety monitoring data; fusing the four models by using lasso regression to generate a space-time multi-view fusion model; and generating complemented data by using the space-time multi-view fusion model. Under the condition of strong correlation of spatial-temporal characteristics, the method can well solve the problems of block missing, local missing and the like in dam safety monitoring data, and through verification on real dam safety monitoring data, the method has smaller errors and better completion effects compared with a conventional classical algorithm and a conventional spatial-temporal model.

The embodiment of the invention discloses a geofence data extraction method, a device and equipment. The method comprises the steps of acquiring a target geofence image, wherein the target geofence image comprises reference points with known geographic coordinates; preprocessing the target geo-fence image to obtain a binarized image with a closed contour; converting the binarized image into an ordered point sequence set; and determining the geographic coordinates of each pixel point of the point sequence set according to the geographic coordinates of the reference points.

The invention discloses a method and a device for processing simulation base period data. The method comprises the following steps: acquiring national economy statistical data; dividing the national economy statistical data into a plurality of sub-data items according to national economy industry classification; according to a preset data processing principle, processing the national economy statistical data to obtain simulation base period data, and inputting the simulation base period data into an economy simulation system for simulation, wherein the step of processing the national economy statistical data according to the preset processing principle comprises the following steps: converting the national economy industry classification into simulation system industry classification according to an industry conversion principle, further classifying the national economy statistical data again, and classifying the national economy statistical data into corresponding categories of the economic simulation system industry classification; and supplementing missing data items of enterprises in each industry in the economic simulation system industry classification according to a total quantity decomposition principle. By implementing the embodiment of the invention, the problems of poor systematicness and data missing of the existing simulation base period data can be solved.

The invention discloses a wind profile radar signal processing method. The method comprises the following steps: S1, collecting original power spectrum data of a wind profile radar; s2, performing basic processing on the original power spectrum data acquired in the step S1; s3, the power spectrum data subjected to basic processing in the step S2 are subjected to wave crest position identification and weather significance judgment, and the wave crest position and intensity of a signal spectrum with weather significance are recorded; and S4, turbulence signal identification is carried out on the signal spectrum with meteorological significance in the step S3. The wind profile radar signal processing method provided by the invention not only solves the problem of data missing in the rainfall period in the existing algorithm, but also verifies the reliability of the method for inverting the horizontal wind field.

The embodiment of the invention provides a flow data processing method and device based on RPA and AI, equipment and a medium. The method comprises the following steps: S1, reading a data table corresponding to a current process in a server; and S2, according to a preset data writing rule, the operation data in the current process execution process is written into the data table, and the preset data writing rule corresponds to the data structure of the data table. By adopting the technical scheme, the problem of data missing in the execution process of the RPA robot is solved, and a basis is provided for review of subsequent operation records.

The invention discloses a ship track monitoring system based on a block chain. The ship track monitoring system comprises a main communication node A0 and n replica communication nodes, the main communication node A0 and the n replica communication nodes form a block chain; the block chain stores AIS information; the replica communication node obtains AIS information from the block chain, and track backtracking query is completed; the invention relates to a ship trajectory monitoring system based on a block chain technology, which mainly solves the problem of accurate backtracking of a channel supervision mechanism on the whole process of a ship moving trajectory, and effectively prevents data missing caused by ship trajectory data tampering and node abnormity. The strong-collisiveness cryptographic algorithm used in the invention ensures the security and confidentiality of the data, the load of all nodes is reduced by using a node marginalization calculation summarization mode, the resource waste is reduced, and hacker attacks can be prevented.

The invention discloses a backup system data synchronization method based on a power distribution main station, which specifically comprises the following steps: S1, detecting backup data when main data is damaged or lost, and checking whether the integrity and consistency of the backup data can be achieved or not. The method relates to the technical field of power distribution main station systemsynchronization. According to the standby system data synchronization method based on a power distribution main station, when main data is damaged or lost, backup data is detected firstly, the replaced system data can be immediately detected after the system data is replaced; in this way, the same problem can be prevented from occurring subsequently. The standby system data synchronization methodbased on the power distribution main station greatly facilitates workers, enables standby system data to be not replaced frequently, reduces use cost, and solves the problem that when a problem occurs in an existing standby system data synchronization method based on the power distribution main station, a problem source cannot be found after the standby system data is replaced when a main systemcannot be used.

The invention provides a power line communication signal reconstruction method, which can reconstruct and generate a power line communication signal sequence without missing data. The method includes: collecting a power line communication signal sequence and converting it into a two-dimensional signal; decomposing the two-dimensional signal to obtain the decomposition coefficient of the power line communication signal; calculating the actual marginal probability distribution of each parameter in the decomposition coefficient and the distribution of each parameter Mean value and mean square error; according to the actual marginal probability distribution of each parameter in the calculated decomposition coefficient and the mean value and mean square error of each parameter, the power line communication signal decomposition coefficient is adjusted; according to the adjusted power line communication signal decomposition coefficient, no Power line communication signal sequence with missing data. The present invention relates to the field of electric power.

The invention discloses a method and a system for synchronizing historical data of a database, electronic equipment and a storage medium, and the system comprises a measuring point data self-checking module, a synchronization task information generation module, a database interaction module and a data synchronization execution module. According to the method, the starting time and the ending time of data missing are obtained, the synchronization task information is generated, the database is synchronized, the purpose of automatically synchronizing the missing data after the data are lost is achieved, and a large amount of disk space does not need to be occupied. According to the system, the problem of data missing in the data synchronization process can be solved, and high manpower resource and financial resource cost can be saved for the whole project.

The invention discloses a method for predicting mortality in an ICU based on deep learning. The present invention predicts the death in the ICU by using deep learning, not only using the patient's dynamic sign change data, but also adding important demographic characteristics such as age and disease type that affect the mortality rate; Based on the long-short-term memory model) network, three time gates are added to control the time increment change to solve the common irregular sampling and data missing problems of clinical data; the attention mechanism is introduced to fuse the hidden state at each moment , the model converges faster and is less likely to lose the information contained in the initial stage.

The invention relates to an Internet of Things edge gateway data caching method and device. The method comprises the steps that an edge gateway judges whether the communication state with an Internet of Things platform is normal or not; if yes, the edge gateway obtains sensing layer data and sends the sensing layer data to the Internet of Things platform; if not, the edge gateway obtains sensing layer data and a cache triggering rule of the sensing layer data; and the edge gateway caches the sensing layer data according to the cache triggering rule. According to the invention, data uploading or caching can be carried out based on the communication state with the Internet of Things platform, the problem of platform data missing caused by communication faults of the edge gateway and the Internet of Things platform is effectively solved, the data integrity is ensured, and the problem that subsequent control strategies are affected or the prediction precision is poor due to data missing is avoided.

The present invention provides a power line communication signal adaptive reconstruction method which can reconstruct and generate a power line communication signal sequence without missing data. Themethod comprises the steps of: collecting the power line communication signal sequence, and converting the power line communication signal sequence to a signal matrix; according to the signal matrix obtained through conversion, constructing a conversion operator matrix; constructing a measurement matrix; determining a reconstruction weight, iterating and updating the signal matrix according to theobtained conversion operator matrix, the reconstruction weight and the constructed measurement matrix until the number of the current iteration times is equal to the length of the power line communication signal sequence; and converting the current obtained signal matrix to generate a power line communication signal sequence without missing data. The power line communication signal adaptive reconstruction method relates to the field of communication.

The invention discloses a method and system for solving the problem of network element concurrent access limitation and a computer readable storage medium, and the method comprises the steps: 1, traversing network elements one by one, querying all tunnels taking the network elements as source nodes, and adding the queried tunnels to a to-be-detected tunnel queue; 2, according to the residual concurrent number leftNum of the current equipment, removing leftNum tunnels from a to-be-detected tunnel queue, issuing detection of the tunnels to the equipment, and adding the tunnels into the tunnel queue which does not obtain a detection result; and 3, removing the tunnel with the obtained result from the tunnel queue without the obtained detection result, releasing the occupied leftNum, and storing the detection result on a controller; repeating the above steps, thereby completing the collection of the tunnel performance information on the controller, and completing the control of the concurrent access network element of the controller. Concurrent access of sudden increase of the controller to the network elements can be effectively solved, and healthy and stable operation of the controller is guaranteed.

The invention discloses a data-driven power system attacked measurement data two-layer restoration method, and relates to the field of power system operation safety maintenance, and the method comprises the steps: building a measurement regression model under each topological structure according to the network parameter information of a regional power grid, and simulating an attacked measurement vector under an FDIA working condition; separating attacked data on the upper layer by using a multi-dimensional scaling method for dimension reduction in combination with a federal forest algorithm; inputting the separated normal data into a generator G of the improved CBEGAN at the lower layer to generate data G (z); the G (z) and a historical measurement regression model are injected into a discriminator D, and authenticity is distinguished by calculating a Hellinger distance between reconstruction error distributions of a real sample and a generated sample; convergence is judged through the target M, and finally data restoration is achieved. According to the method, the measurement data can be effectively restored on the basis of the residual normal data, the self-healing capability of the power grid after being attacked is realized, and the method has relatively good measurement convergence, control distribution diversity and practicability.

The invention discloses a robot life cycle management method, a device and a storage medium. The method comprises the steps: enabling robot equipment to be in butt joint with an enterprise managementsystem, and distributing a corresponding robot equipment certificate to the robot equipment; performing data interaction with a cloud server based on the distributed robot equipment certificate, monitoring the operation state of the robot corresponding to the robot equipment certificate in real time, obtaining all operation data corresponding to the operation state, and uploading all the obtainedoperation data to the cloud server; based on the robot equipment certificate, performing multi-dimensional data analysis and calculation on all data associated with the robot equipment certificate, and managing the life cycle of the robot corresponding to the robot equipment certificate; the problems of insufficient informatization level and data loss of the existing robot in each life cycle linkare solved, a set of management mode for the full life cycle of the robot is formed, the scheduling of the robot is optimized, and the life cycle of the robot is favorably prolonged.

The invention discloses a quantum chemistry method capable of calculating gas-phase hydroxyl radical reaction rate constants of various perfluoroalkyl / polyfluoroalkyl compounds. The gas phase kOH value of the PFASs can be directly obtained by utilizing the most stable configuration of the known PFASs and applying the constructed quantum chemical calculation method. The construction process of the quantum chemistry method comprises the following steps: collecting 15 PFASs stable configurations of known kOH, searching a reaction path of PFASs and. OH by using a density functional theory, carrying out subsequent structure optimization, calculating single-point energy by using different DFT methods, and calculating thermodynamic parameters and kOH. And performing linear fitting on results of different calculation methods and experimental values, and screening an optimal method. The quantum chemical calculation method developed by the invention can accurately and efficiently calculate the kOH of the unknown PFASs, provides a necessary tool for evaluating the atmospheric durability of the PFASs, and has very important significance.

The invention provides a missing feature re-representation method and system based on a graph convolutional network. The method comprises the steps: extracting the features of a training sample with a marked category, taking the features of the training sample as nodes, obtaining the similarity relation between the nodes through distance measurement, constructing a connection edge between the nodes according to the similarity relation, and carrying out the re-representation of the missing features of the training sample. Obtaining a graph network of the training sample; Training a graph convolutional network according to the feature information of the adjacent samples in the graph network and the annotation categories corresponding to the nodes in the graph to obtain a feature re-representation model, inputting the features of the to-be-classified samples into the feature re-representation model, reconstructing the features of the to-be-classified samples, and classifying the reconstructed features to obtain a classification result. And obtaining a classification result of the to-be-classified sample.

The invention discloses a data breakpoint resume transmission method of a new energy centralized control system. When communication is interrupted, the power station caches real-time data, copies and saves the real-time database of the centralized control master station at that time as a real-time database copy, and resumeDB in the inversion control database Record the corresponding communication interruption events in the system; after the communication is restored, upload the compressed file to the central control master station; the centralized control master station fuses the compressed file and the real-time library copy to form an inversion package, and completes the communication interruption event; According to the completed communication interruption event records and inversion packets, the master station consumes the data in the inversion packets in the form of periodic micro-batch processing, and performs timing and time-section calculations on the basis of the data to obtain cooked data; the cached real-time Data and cooked data are stored in the historical database together. The invention realizes the seamless integration of continuous transmission data and real-time monitoring historical data, solves the problem of missing data in the new energy centralized control system, and satisfies the high requirements of operation and maintenance and analysis services on the integrity of historical data.

The invention provides a total radiation data correction method based on photovoltaic power generation historical information. The method comprises the following steps: dividing photovoltaic power station historical data into to-be-corrected data and non-missing historical data; substituting power data in the to-be-corrected data into a pre-trained optimization model to obtain total radiation data corresponding to the to-be-corrected power data; fitting the total radiation data in the non-missing historical data and the total radiation data corresponding to the to-be-corrected power data to obtain total radiation data; wherein the historical data of the photovoltaic power station comprises power data and total radiation data, and the optimization model is obtained by training a sample constructed by non-missing historical data. Complete total radiation sequence data is provided for a photovoltaic power station, and the problem of data missing in total radiation monitoring is solved.