217 results about "Global time" patented technology

The invention relates to the technical field of time synchronization, and discloses a hybrid time service method based on Beidou satellite navigation and a TSN, which comprises the following steps: acquiring network topology structure information and historical time service error data of each synchronization node in a target distributed system, and constructing a dynamic time delay state set for jitter identification; based on the dynamic time delay state set, a current global time reference is extracted through a time calibration algorithm, a time difference reference vector sequence is generated, and initial synchronization adjustment is carried out on the TSN main node; constructing an inter-node clock convergence function curve according to the slave node response stability, and judging whether the synchronous network is in a low jitter interval or not based on the time offset rate and the time drift trend change; in combination with the node time state after directional offset suppression, a multi-channel redundant time synchronization mechanism in a time window is applied, a weight adjustment coefficient is extracted, and an unbalanced calibration factor is introduced; and according to a regression correction result, evaluating a time service stability index in real time. The method has the advantage of improving the time service stability.

The invention belongs to the technical field of production monitoring systems, and discloses a numerical control machine tool dynamic fault diagnosis method and system. The method comprises the following steps: generating a global time reference signal through a main shaft encoder and a clock synchronization protocol; the method comprises the following steps: collecting vibration data of a main shaft bearing in each unit time, current data of an electric cabinet and process parameters, and generating a preprocessed data sequence through transmission delay compensation and multi-rate frequency raising processing; inputting the vibration data and the current data into a preset mechanical-electrical transfer function model, and calculating a time delay parameter; performing phase alignment on the preprocessed data sequence based on the time delay parameter to generate an aligned data sequence; inputting the aligned data sequence into a time sequence neural network, and outputting a fusion feature vector; calculating a cross correlation coefficient of the fusion feature vector, and generating a fault diagnosis result based on a preset cross correlation threshold value; the problem of failure of fault feature extraction caused by data asynchronization in the prior art is solved.

The invention discloses a multi-head time sequence intelligent risk control method and system based on weighted trend and fluctuation, and relates to the technical field of financial risk management. Comprising the following steps: S1, collecting multi-channel time sequence data in real time, and carrying out data preprocessing; s2, calculating the global time weight, quantifying the multi-scale fluctuation stability of the channel pair, and judging the asynchronous alignment degree of the channel pair; s3, extracting an effective frequency band interval, calculating frequency domain characteristic parameters of the signal, evaluating frequency domain energy phase characteristics of a channel signal, and quantifying fluctuation states of a channel under different scales; s4, constructing a sparse coupling relation graph, quantifying an edge weight in the sparse coupling relation graph, and obtaining a network average coupling weight; and S5, evaluating the dynamic evolution characteristics of the channel risk state, and generating risk trend prediction and control suggestions. The problem that risk control accuracy is affected due to the fact that trend and fluctuation feature extraction of multi-source heterogeneous high-noise multi-head time series data is unstable under the condition of concept drift and multi-scale coexistence is solved.

The invention discloses a command and control system resource trend prediction method based on fusion of long and short time sequence characteristics. The method comprises the following steps: acquiring a public power load or similar time sequence monitoring data set, and preprocessing the data in the data set; a deep learning network model based on a TCN-Transformer hybrid model is constructed, a TCN model and a Transformer model are adopted for parallel computing to achieve feature extraction, the TCN model extracts short-term information, the Transformer model extracts long-term features, then fusion features are obtained through a cross attention mechanism and multi-layer perceptron (MLP) weighting, and finally prediction output is generated through full connection layer mapping. Taking data in the training set as input, training the constructed TCN-Transform hybrid model, and continuously optimizing the model until convergence meets a set requirement; and performing prediction by using the trained network model. According to the method, the TCN-Transform hybrid model is constructed, so that local fine-grained features are reserved, the global time trend is effectively captured, and the accuracy of command decision making is improved.

The invention relates to the technical field of new energy power generation prediction, and discloses a t-SNE visualization and depth time sequence attention model-based wind and light power prediction method and system, and the method comprises the steps: obtaining historical power data and corresponding historical meteorological data of a wind power station and a photovoltaic station, and constructing a historical data set; inputting the preprocessed high-dimensional meteorological data into a t-SNE dimension reduction module, and mapping the high-dimensional data to a low-dimensional space through symmetric joint probability density calculation to obtain low-dimensional visual data distribution; the low-dimensional data are input into a parallel model composed of a TCN-SENet branch and a BiGRU-GlobalAttention branch, and space-time local features and global time sequence features are extracted respectively; and fusing the feature vectors output by the two branches, generating a wind-solar power prediction result through a full-connection layer, and performing evaluation. According to the method, the limitation of traditional single energy independent modeling is broken through, the adaptability of the model to a complex power generation mode of a distributed station is optimized, and the precision and generalization ability of multi-energy joint prediction are remarkably improved.

The invention provides a servo press control system and method based on a time-sensitive network, and relates to the technical field of intelligent manufacturing and industrial automation control, and the system comprises a time-sensitive network TSN which is used for providing global time synchronization and deterministic data transmission; a main controller of the multi-axis servo control unit is used for issuing a motion control instruction and a feeding and discharging control instruction through a TSN. A robot controller in the robot feeding and discharging unit is used for driving an industrial robot to execute plate taking and placing operation under the action of the feeding and discharging control instruction. The visual monitoring unit comprises an industrial camera accessed to the TSN and is used for collecting images and transmitting image data to the edge computing node through the TSN; the edge computing node is used for processing the image data to generate a compensation instruction and feeding back the compensation instruction through the TSN; and the upper monitoring system is used for subscribing real-time data published by the edge computing nodes, so that the control range of the servo pressure control system is expanded, and the convenience of data transmission and interaction is improved.

The invention discloses a self-supervised traffic flow prediction method based on multi-scale space-time-frequency fusion. The method comprises the following steps: acquiring enhanced data; performing multi-scale spatial-temporal feature coding; performing frequency domain residual filtering; generating a traffic flow prediction result; and carrying out joint target optimization. According to the invention, the multi-scale space-time frequency encoder is designed, local and global time features are captured at the same time through the mixed time encoding module in the time dimension, the multi-scale space encoding module aggregates space features under different distances in the space dimension, and the prediction precision is significantly improved. A frequency domain residual filtering module is embedded in the encoder to adaptively purify frequency domain features in an end-to-end mode, enhance key periodic features and suppress irrelevant noise, space-time features are fused through residual connection, space-time-frequency three-dimension collaborative modeling is achieved, meanwhile, frequency domain consistency loss is introduced in the optimization stage, and time-frequency three-dimension collaborative modeling is achieved. And the method is more robust when facing real traffic data containing noise.

The invention relates to the field of cloud computing and cross-border supply chain management, and discloses a cross-border supply chain real-time monitoring method in a cloud computing environment. Comprising the following steps: deploying data acquisition and access modules for supply chain nodes of different countries or regions on a cloud computing platform, and establishing a global time index based on node reporting time to realize time sequence correction of multi-source data; performing multi-modal fusion processing on the corrected data set to form a comprehensive feature vector with a unified dimension; constructing a supply chain digital twin map to associate logistics, information flow and fund flow, and realizing node state prediction and risk identification; executing cross-border data synchronization and privacy protection according to compliance requirements of each country; after the risk is identified, risk traceability and feedback adjustment are carried out through graph structure simulation, so that self-adaptive updating and periodic self-correction of system parameters are realized. According to the invention, real-time monitoring, risk prediction and compliance data synchronization of the cross-border supply chain can be realized, and the safety and stability of the system are improved.

The invention relates to the technical field of signal processing, and discloses a multichannel data acquisition system and method based on an FPGA (Field Programmable Gate Array), an acquisition module drives a global counter by using a global synchronous clock, writes acquired data into an annular buffer area with a physical address and a counter value in a linear mapping relationship, and establishes implicit time index storage. When a trigger event occurs, the system broadcasts the locked global trigger timestamp, and the acquisition module backtracks and reads historical data according to the global trigger timestamp, and packages the historical data into a sparse matrix type data packet in combination with the channel validity mask. And after receiving the data packet, the data processing module directly calculates a memory mapping address by using global time information in the data packet, and writes a data load into a corresponding position of the waveform reconstruction buffer area. According to the method, through strict binding of the physical address and the absolute time, independent time label redundancy is eliminated, high-precision synchronization of distributed multiple channels is ensured, and out-of-order automatic in-situ recombination and efficient waveform reconstruction of data of a receiving end are achieved.

The invention relates to the technical field of industrial automation control and dynamic simulation, and discloses a centralized control console and 3D model linkage virtual-real simulation system, which comprises a control signal sampling interface, a load response modulation unit, a dynamic model evolution unit and a virtual-real closed-loop feedback unit, the load response modulation unit performs energy consumption weight accumulation on all the virtual controlled objects in transient action to generate a global time scaling factor so as to correct an inertia time constant of the system in real time; the dynamic model evolution unit introduces a hard limiting boundary by using an amplitude limiting link and calculates a process variable; the virtual-real closed-loop feedback unit extracts a control deviation signal and drives a physical instrument to generate a nonlinear damping action, and by establishing a global load modulation model based on energy constraint, a discrete control system spontaneously emerges a nonlinear hysteresis characteristic conforming to physical energy conservation under the low calculation power condition without complex fluid network solution.

The invention provides a social robot detection method and system, computer equipment and a storage medium, and belongs to the technical field of social robot detection.The method comprises the steps that user multi-source data on a social platform and the social relation between users are collected; a hybrid encoder is adopted to capture local dependence and global time sequence dynamic states of behaviors, and user behavior characteristics are generated; aggregating structure attention features between the initial node features of the target account and the initial node features of the social relation account to obtain multi-modal relation aggregated structure features; introducing a multi-modal adversarial training strategy, adaptively adjusting the disturbance intensity of each modal based on gradient sensitivity, and aligning the characterization of the clean sample and the adversarial sample in combination with a content discriminator and a behavior discriminator; and outputting a target user classification result through the multi-task target function optimization model. According to the method, the problems of insufficient modal fusion and insufficient adversarial robustness of an existing method can be effectively solved, and the accuracy and stability of social robot detection in a complex and adversarial scene are improved.

The invention discloses a robot end effector trajectory planning method and device, and relates to the technical field of robots. The method comprises the steps that the first initial joint state, the pre-grabbing posture and the grabbing posture of an end effector of the robot are obtained; on the basis of a preset first constraint, the first initial joint state and the pre-grabbing pose, a feasible preparation track is obtained; on the basis of a preset first constraint, a preset second constraint, the final joint state of the feasible preparation trajectory and the grabbing pose, a feasible grabbing trajectory is obtained, and the second constraint comprises a direction constraint and a linear path constraint; and obtaining a final track based on the feasible preparation track and the feasible grabbing track. Directional-linear joint constraint of part of tracks in a grabbing task can be achieved, the efficiency bottleneck of a traditional two-step method is broken through, and the track planning efficiency is improved. The dynamics consistency of the two sections of tracks can be ensured, and the final track can be smoother. The real optimization of global time and smoothness can be realized, and the track quality is improved.

The invention provides a time synchronization method and device and a time synchronization system of an intelligent driving system, and relates to the technical field of data processing, in particular to the technical fields of automatic driving, electronic technology and time synchronization. The specific implementation scheme of the time synchronization method comprises the following steps: acquiring second-level world time and a corresponding second pulse signal; determining a second global timestamp corresponding to a predetermined edge of the pulse per second signal according to a first local timestamp of the captured pulse per second signal, a second local timestamp of the execution moment of the interrupt service function and the first global timestamp; wherein the interrupt service function is triggered by a second pulse signal; obtaining the second time of the current moment according to the global timestamp of the current moment and the second global timestamp; and obtaining the accurate time of the current moment according to the second-level world time and the within-second time. According to the invention, errors can be reduced, and high-precision global time can be provided.

The invention discloses an intelligent e-commerce data recommendation method and system based on a time sequence diagram neural network and an attention mechanism, and relates to the technical field of e-commerce data analysis, and the method comprises the steps: constructing a dynamic time sequence diagram through user behavior time sequence data, calculating the weight of an edge in the time sequence diagram through a time decay function, and calculating the weight of the edge in the time sequence diagram; meanwhile, local time sequence features among commodities are captured in combination with a gating graph neural network; the method comprises the following steps of: introducing a Star-transform network structure, wherein the structure realizes long-range information interaction of non-adjacent commodities through a relay node; meanwhile, a multi-head dynamic attention mechanism is combined, the influence of key behaviors is highlighted in the process of integrating long-range information, and global time sequence features are extracted; and designing an adaptive feature fusion mechanism, dynamically integrating local and global time sequence features to obtain fusion features, and then further fusing real-time behavior preferences of the user and commodity attribute information to generate a recommendation list meeting personalized requirements of the user. According to the method, the individuation degree, the accuracy and the user satisfaction of the e-commerce product recommendation result can be improved.

The invention discloses a space-time elastic weight consolidated edge federal continuous learning method, which is applied to a system comprising a server and a plurality of edge devices, is used for processing space-time heterogeneity time sequence data, and comprises the following steps: initializing training, broadcasting a previous time sequence global time Fisher diagonal matrix (the first time sequence is not broadcasted, the second time sequence is not broadcasted, and the third time sequence is not broadcasted) by the server; however, the global model needs to be randomly initialized and broadcasted); in the model training stage, based on local data, a global time Fisher diagonal matrix and the like, an edge device updates a local model through a loss function containing a time / space regular term, calculates a local space Fisher diagonal matrix, uploads the local space Fisher diagonal matrix, and then a server weights and aggregates the global model according to the data volume and issues the global model, and circulates until convergence; and in the global time Fisher diagonal matrix calculation stage, the equipment calculates a local time Fisher diagonal matrix based on a convergence model, and uploads and aggregates the local time Fisher diagonal matrix for the next time sequence. Historical data does not need to be stored, original data does not need to be transmitted, storage calculation / communication overhead is reduced, privacy is protected, and the model convergence speed and precision are improved.

The invention discloses a Modbus multi-axis real-time synchronous control method for a numerical control machine tool, which relates to the technical field of numerical control, and realizes multi-axis microsecond-level synchronous control by establishing innovative mechanisms such as global time reference, dynamic time slot allocation, instruction pre-caching synchronous execution, feed-forward-feedback dual-mode fault tolerance and the like. The technical limitation of a traditional Modbus protocol is broken through, the cost is greatly reduced while high performance is guaranteed, the synchronization precision reaches + / -0.8 microseconds, the roundness machining error is smaller than or equal to 9.5 micrometers, and the communication reliability reaches 99.998%. The method is particularly suitable for high-precision equipment such as a five-axis linkage numerical control machine tool and a high-speed machining center.

The embodiment of the invention provides a mobile mapping system and a time synchronization method, and is applied to the technical field of positioning and map construction. The system comprises a device to be synchronized, an industrial personal computer and a controller. The controller is used for determining a global time reference and determining a reference time point for acquiring a hardware trigger signal; the to-be-synchronized device is used for acquiring data according to the hardware trigger signal; and the industrial personal computer is used for associating the data acquisition time of the to-be-synchronized equipment with the global time reference according to the global time reference and a receiving time point for receiving the data sent by the to-be-synchronized equipment. The time service mode depends on a hardware trigger signal, and effective time service of data acquired by the multi-source sensor can still be realized in a scene with weak GNSS signals or without GNSS signals, so that the synchronization precision of data acquisition time of the multi-source sensor is improved.

The invention discloses an AI generated video detection method and system based on double-branch attention fusion. The method comprises the following steps: synchronously extracting a Patch-level feature sequence and a global-level feature sequence of a video frame by using a pre-trained visual Transform; carrying out local time sequence modeling and independent judgment on the Patch feature sequence through a local space-time artifact branch by adopting a pixel tube Transform; carrying out overall narrative modeling and independent judgment on the frame-level global feature sequence by adopting a global time sequence Transform through a global time sequence consistency branch in parallel; and finally, dynamic weighted fusion is carried out on double-branch discrimination results through an attention fusion module, and a final detection probability is output. According to the method, the AI video generated by advanced models such as Sora and Pika can be effectively identified from two complementary dimensions of local space-time details and global time sequence dynamic, and the method has strong robustness and cross-model generalization ability for interference such as video compression and fuzziness, and is suitable for security scenes such as network content auditing and media evidence obtaining.

The invention discloses a multi-experiment task production scheduling method, device and equipment and a storage medium, and the method comprises the steps: carrying out the sampling of a preset number of times for each production scheduling scheme based on the prior distribution of the duration of each experiment stage, and determining the starting time of each experiment operation, constructing a global time resource scheduling problem under the constraint conditions that the time window constraint of the experimental operation, the time period of the experimental operation belongs to the idle time period of the operation equipment and the time periods of the experimental operation distributed on the same operation equipment are not overlapped, and calculating the ratio of the infeasible times of the global time resource scheduling problem to the preset times; and if the risk value is lower than the risk threshold value, taking the production scheduling scheme as a target production scheduling scheme, and performing online scheduling of the experimental tasks according to the target production scheduling scheme, so that the method is suitable for production scheduling of high-throughput and diversified experimental tasks, and the risk of equipment resource conflict is reduced.

The application relates to the technical field of time synchronization, and discloses a mixed time service method based on Beidou satellite navigation and TSN, which comprises the following steps: acquiring network topology structure information and historical time service error data of each synchronization node in a target distributed system, and constructing a dynamic time delay state set for jitter identification; based on the dynamic time delay state set, a current global time reference is extracted through a time calibration algorithm, a time difference reference vector sequence is generated, and initial synchronization adjustment is carried out on a TSN master node; according to the response stability of a slave node, a clock convergence function curve between nodes is constructed, whether the synchronization network is in a low jitter interval is judged based on time offset rate and time drift trend change; in combination with the node time state after directional offset suppression, a multi-channel redundant time adjustment mechanism in a time window is applied, a weight adjustment coefficient is extracted, and a non-uniform calibration factor is introduced; and according to a regression correction result, a time service stability index is evaluated in real time. The application has the advantages of improving time service stability.

The invention discloses a real-time multi-modal man-machine interaction method and system based on a large model and a storage medium. Collected user interaction information is converted into interaction data in a preset format, the interaction data are input into a full-modal large model, and an output pre-generation result is obtained; secondly, semantic anchor points are marked for reply text information, absolute prediction timestamps of all the semantic anchor points are calculated, semantic time sequence windows are divided, and a global time sequence reference skeleton is constructed; combining and packaging the expression degree-of-freedom sequence and the action degree-of-freedom sequence into a unified expression frame; and locking the rigid segments to the corresponding semantic anchor point timestamps according to the anchor point tags, performing nonlinear filling on the elastic segments between the rigid segments, outputting alignment information, and finally analyzing the alignment information into control instructions to respectively drive a loudspeaker, a facial expression component and a limb motor to execute reply operation. The limitation of dependence of a single mode is broken through, the interaction stability in a complex environment is improved, and meanwhile, the real-time performance of interaction response is improved.

The invention provides a flood forecasting method and device based on an intelligent optimization neural network, and the method comprises the following steps: obtaining original hydrological time series data, and carrying out the variational mode decomposition of the original hydrological time series data based on an optimal penalty factor alpha and an optimal decomposition number K, and obtaining K intrinsic mode functions; dividing the K intrinsic mode functions into high-frequency intrinsic mode functions, intermediate-frequency intrinsic mode functions and low-frequency intrinsic mode functions based on the center frequency of each intrinsic mode function in the flood flow prediction network, and fusing the high-frequency intrinsic mode functions, the intermediate-frequency intrinsic mode functions and the low-frequency intrinsic mode functions to obtain local time sequence features; and fusing the global time sequence features and the local time sequence features to obtain global-local fusion features, and inputting the global-local fusion features into a full connection layer to obtain a predicted flood flow value. According to the scheme, the optimal penalty factor alpha and the optimal decomposition number K are globally and automatically optimized through the optimization algorithm, and low efficiency and deviation caused by manual parameter adjustment are avoided.

The invention discloses a PM2.5 complex time sequence prediction method based on a double-path fusion architecture, and belongs to the technical field of PM2.5 complex time sequence prediction methods.According to the method, a double-path fusion structure comprising a local feature extraction path and a global time sequence modeling path is constructed, and combined modeling is carried out on PM2.5 time sequence data of multiple cities, the local path captures short-term fluctuation and high-frequency disturbance characteristics by using a convolution structure, and the global path models long-term trend and multi-scale correlation by using a neural network based on an attention mechanism, so that fine prediction of a complex non-stationary sequence is realized; according to the method, a reversible normalization mechanism is introduced to dynamically adjust input distribution, self-adaptive fusion of local and global results is realized in combination with a double-prediction-head weighted fusion strategy, so that the influence of abnormal disturbance on the model is effectively inhibited, meanwhile, the overall calculation complexity is reduced through a modular design structure, and the calculation efficiency is improved. And the deployability and the real-time performance of the method in a multi-scene air quality monitoring system are enhanced.

The invention discloses an abnormal event detection method and system based on traffic data and a storage medium, and belongs to the technical field of computer network security, and the method comprises the steps: obtaining original network traffic data, and carrying out the data segmentation, and obtaining traffic sample data; performing data preprocessing on the traffic sample data to obtain processed data; respectively performing time feature extraction and spatial feature extraction on the processed data to correspondingly obtain time feature data and spatial feature data of the network traffic; and performing feature fusion based on the time feature data and the space feature data, and performing anomaly detection on the fused space-time feature data based on a space-time attention mechanism. According to the method, the spatial relationship of bytes in the data packets and the time relationship between the data packets can be captured at the same time, and the overall performance of anomaly detection is improved by utilizing the time sequence dependency between the data packets and the spatial relationship in the data packets. A space-time attention mechanism can comprehensively consider global time sequence and local space information, and the accuracy and efficiency of abnormal traffic detection are improved.

The invention discloses a VR scene control method and system, and belongs to the technical field of VR scene control, and the method specifically comprises the steps: deploying a global time controller, generating a continuously increasing global timestamp at a fixed frequency, and transmitting the global timestamp to a scene rendering, audio processing and touch driving module; the scene interaction monitoring unit continuously collects user body movement and virtual object movement state data, analyzes interaction event types and characteristic parameters and then binds the interaction event types and the characteristic parameters with a current global timestamp to form an interaction information packet with the timestamp; a multi-modal feedback instruction generation unit receives the information packet, and generates three types of instructions carrying the same global timestamp based on the interaction event type, the characteristic parameters and the bound global timestamp; the feedback instruction distribution unit forwards the three types of instructions to corresponding modules at corresponding moments according to global timestamps carried by the instructions; and after executing feedback, each module records delay and sends the delay to the global time controller, so that the actual execution moments of three types of instructions in the next round are consistent.

The technical scheme of the invention discloses a traffic flow prediction method based on a dynamic space-time diagram convolutional network. The invention provides a traffic flow prediction method fused with a dynamic space-time diagram convolutional network, which captures road network change in real time through a dynamic adjacency matrix, designs a multi-scale time attention mechanism to fuse local convolutional features and global time sequence dependence, develops an intelligent mixed precision training system to realize computing resource optimization, and improves the traffic flow prediction efficiency. And the training efficiency is improved while the prediction error is reduced. Meanwhile, traffic flow data time features and multi-section spatial features are comprehensively considered, and prediction is carried out based on a single step length and multiple step lengths. The method realizes high-precision prediction of the short-time traffic flow, and is suitable for precise flow prediction of a main line and an arterial highway of a highway network.

The application relates to the field of meteorological data processing and satellite navigation application, and provides a multi-source meteorological data fusion method and system based on a Beidou space-time reference. The method comprises the following steps: acquiring Beidou data and multi-source meteorological data to be fused, taking the coordinated universal time timestamp output by a Beidou ground-based enhancement station as a global time reference, and setting a sliding window; taking the world geodetic coordinate system corresponding to the Beidou positioning data as a space reference, resampling the multi-source meteorological data to a preset standard grid, and combining terrain elevation data to construct a unified space coordinate system; calculating the confidence degrees of fused static terrain features and dynamic change features based on Beidou satellite orbit residual and clock difference parameters, adjusting the feature weights of the fused static terrain features and dynamic change features, and generating an enhanced fusion feature matrix.

Method, apparatus and system for assigning a commit sequence number (CSN) to a WRITE transaction in a network having nodes and a global time server. The CSN is defined by a timestamp of the WRITE transaction and an error bound of the timestamp. The WRITE transaction is committed after the timestamp is issued and an amount of time equal to the error bound plus a time adjust value has passed. The time adjust value is based on round-trip times between the plurality of nodes and the global time server. The time adjust value may be the longest expected round-trip time. By waiting for an amount of time equal to the error bound plus the time adjust value, any READ transaction occurring after the WRITE transaction may receive a READ timestamp without any delay, provided the READ timestamp error bound is less than or equal to the time adjust value.

The application discloses an autonomous ship navigation induction method, which predicts the behavior of the autonomous ship and surrounding ships through a trained trajectory space-time prediction model, thereby inducing the navigation of the autonomous ship. The establishment process of the prediction model comprises the following steps: obtaining the voyage time sequence data of the ships around the autonomous ship, constructing a ship navigation situation awareness graph G θ of the autonomous ship θ , which reflects the navigation situation of the autonomous ship and surrounding ships, wherein the ship navigation posture comprises ship coordinates, a time stamp, a ground speed and a ground heading; through entity embedding of the ship navigation situation awareness graph G θ , a ship trajectory vector is obtained, and the spatial features between ships, the local time features of the ships and the global time features of the ships are captured by machine learning; the obtained features are decoded by deep learning, so as to train the trajectory space-time prediction model.

The invention discloses an electric power big data bad data detection and elimination method based on a fusion model, and the method comprises the steps: carrying out the normalization preprocessing of multi-source electric power data, so as to obtain standardized data; and carrying out dynamic denoising on the standardized data based on the K-LSTM model. And carrying out time-frequency analysis on the de-noised data based on a CT-Transform model, and extracting global and local features. And finally, rejecting bad data according to a model preset threshold. And introducing a learning feedback mechanism, and dynamically optimizing a threshold value to output safe and credible data. According to the method, the advantages that the LSTM is good at capturing a data time sequence dependency relationship, the CNN is good at extracting local time-frequency features, and the Transformer is good at capturing global time-space association are fused, so that the problem that a single LSTM is insufficient in high-frequency noise suppression capability and lacks a global view angle is effectively solved, and meanwhile, the limitation that a single CNN is insufficient in long-time dependency description and a single Transformer is insufficient in attention to local details is made up; and accurate detection and elimination of bad electric power data in a complex environment are realized.