48 results about "Time dynamics" patented technology

The invention discloses a continuous time dynamics prediction method, system, terminal and medium fusing a diffusion model and a graph frequent differential equation, and relates to the technical field of dynamics prediction.The method comprises the steps that a multi-node time sequence is obtained, network structure inference is conducted on the multi-node time sequence through the diffusion model, and a potential graph structure between nodes is obtained; carrying out continuous time dynamic modeling by adopting a Scheng ordinary differential equation, and predicting a node state at any time point; diffusion reconstruction loss, dynamic prediction errors and structure sparsity constraints are constructed, total loss is established, network structure inference based on a diffusion model and dynamic modeling based on a Shenzheng differential equation are coupled based on the total loss, and collaborative training optimization is achieved. According to the method, the potential graph structure of the system can be stably recovered in a complex noise environment, high-precision and continuous prediction can be carried out on dynamic evolution of the potential graph structure, and the limitation that structure inference and continuous time modeling cannot be considered in the prior art is overcome.

The invention discloses a time enhanced knowledge tracking method based on dual-channel deentanglement, and belongs to the technical field of education data mining and cognitive modeling. According to the technical scheme, the method comprises the steps that time dynamic features and behavior reaction features in a learning interaction sequence are extracted and coded through a time domain encoder and a behavior domain encoder respectively; separating long-term trends and short-term fluctuations in the input features by using a multi-scale decoupling layer based on causal convolution; a time perception dual-channel attention module is adopted to independently decouple time and behavior characteristics after decoupling, and a nonlinear attenuation item based on a real interval is introduced to simulate memory forgetting; and finally, integrating dual-channel information through a gating fusion mechanism and predicting future answering performance of the learner. According to the method, optimization conflicts are effectively relieved, the robustness to a complex learning mode is enhanced, and knowledge state modeling which better accords with a cognitive law is realized.

A technique is disclosed to use time series phasor data to perform real-time dynamic line rating of electric power transmission lines. A variety of techniques are used to generate well-poised solutions to the determination of transmission line parameters from phasor data. Line health information can also be determined from changes to transmission line parameters, such as galloping, icing, vegetation encroachment, imperfect splicing, and conductor corrosion.

The invention provides a power grid state prediction method based on a graph neural network and a time convolutional network, and the method comprises the steps: firstly carrying out the modeling of a power grid into a graph structure (nodes are power equipment, and edges are physical connection), and constructing a space-time input tensor of continuous time window operation data; generating a dynamic adjacency matrix to capture topological dynamic changes; node space topology dependence is learned through a multi-layer GCN, and space features are output; extracting a time evolution trend through multi-layer TCN (expansion convolution), and outputting time features; multi-scale convolution and an attention mechanism are adopted to fuse spatio-temporal features, key nodes and time steps are weighted, and finally future node-level states, regional-level loads or system-level abnormity early warning is predicted. According to the method, power grid space topology and time dynamic change are fused, and prediction precision, stability and real-time performance can be improved.

The invention discloses an icing thickness long-time sequence prediction method, system and device and a medium, and the method comprises the following steps: obtaining an icing state data set, carrying out the preprocessing of the icing state data set, and obtaining a standardized data set; constructing a time-aware neural network model according to the standardized data set; based on the time-varying data and the fixed data in the standardized data set, constructing a multi-channel fusion network by using a time-aware neural network model to obtain a fusion prediction model; training the fusion prediction model through a training strategy to obtain a trained prediction model; and predicting the icing thickness by using the trained prediction model. According to the method, the time-sensing LSTM network is adopted to process the problem of time discontinuity of SAR inversion data, and the time dynamic characteristics of icing thickness data can be accurately modeled by introducing a time interval weight mechanism.

The invention relates to the technical field of flux modeling, in particular to an intertidal zone greenhouse gas flux modeling method fusing spatial variation and time series, which comprises the following steps: S1, acquiring spatial variation data of an intertidal zone target area, and outputting a spatial variation feature set; s2, acquiring time sequence data of an intertidal zone target area, and outputting a time sequence feature set; s3, calculating a spatial weight factor based on the spatial variation feature set, and outputting spatial weight distribution; s4, calculating a time dynamic factor based on the time sequence feature set, and outputting a time dynamic sequence; s5, constructing a greenhouse gas flux model, and outputting a flux estimation value; and S6, outputting a greenhouse gas flux distribution diagram of the target area of the intertidal zone according to the greenhouse gas flux model. According to the method, a coupling mechanism of a spatial variation factor weight field and a time dynamic factor sequence is introduced, the flux estimation model of space-time joint regulation is constructed, and the estimation precision of the model in different regions and different meteorological conditions is improved.

The invention discloses an attitude estimation method based on graph convolution and double-branch fusion, and the method comprises the steps: employing a double-branch structure, a Mama branch and a Transform branch to work in parallel, enabling the Mama branch to process long-distance time dependence information through employing an efficient state space model, capturing a long-time-span correlation mode, and enabling the Transform branch to carry out the parallel processing of the long-distance time dependence information; the Transform branch strengthens modeling of local and global attention through a self-attention mechanism and pays attention to interaction relations of different time steps, and after the two branches are output and fused, dynamic changes of human body joint points on a time sequence can be more accurately expressed, a complex time sequence mode can be flexibly and effectively processed, different human body actions can be better adapted, and posture estimation accuracy is improved. The GCN is used for extracting the spatial topological features of the human skeleton, generating the preliminary feature representation, subsequently performing spatial structure optimization on the fused time features by using the GCN, and generating the structured associated features, so that the mode of combining the spatial information and the time sequence information can more comprehensively understand the posture of the human body and consider the dynamic change of the time and the relative position and the connection relationship of the space.

The invention discloses a landslide early warning method, device and equipment and a medium, and the method comprises the steps: obtaining historical landslide three-dimensional displacement time sequence data of a historical landslide event; constructing a landslide state evolution model based on a Gaussian mixture model, a hidden Markov model and historical landslide three-dimensional displacement time series data; determining a current displacement state of the target landslide body based on the real-time three-dimensional displacement time sequence data of the target landslide body and a landslide state evolution model; obtaining the remaining time of the target landslide body from the current displacement state to the landslide occurrence state; and carrying out landslide early warning on the target landslide mass based on the current displacement state and the remaining time. By analyzing the displacement time sequence data of the historical landslide event and optimizing the prediction performance of the HMM model by using the GMM model, a scientific and reliable geological disaster monitoring and early warning system is constructed, real-time dynamic monitoring and accurate early warning of the landslide geological disaster are realized, and the timeliness and accuracy of early warning of the landslide geological disaster are remarkably improved.

The invention discloses a high-order multi-agent system control method with unknown input time lag and quantization, which comprises the following steps: S1) under the condition of unknown time-varying time lag and input quantization, carrying out system description on an uncertain high-order multi-agent system and introducing a lag uniform quantizer; s2) converting an uncertain high-order multi-agent system tracking control problem with unknown time-varying time-lag and input quantization into a bounded problem of a differential equation solution with time-lag; s3) constructing a dominant expression of the system and feasibility conditions for realizing bounded tracking according to the bounded problem of the time delay differential equation; and S4) performing control design according to a dominant expression of the system and bounded tracking feasibility conditions. The invention designs an adaptive finite time dynamic surface control scheme for a high-order multi-agent system with unknown input delay. Meanwhile, the finite time stability theorem is utilized to prove that all signals in the closed-loop system are semi-globally practical and stable in finite time.

The application discloses a kind of equipment dynamic health degree fusion evaluation methods based on multi-source time series data, comprising: collecting and aligning working condition monitoring data and environmental monitoring data in equipment running cycle, and constructing multi-source time series training sample set with health label and environmental label;Introduce the causal representation model including environment encoding subnetwork and degradation encoding subnetwork, use the time window as contrast sample with the same health label but different environmental label, obtain the degradation representation with lower sensitivity to environmental change;Then input the degradation representation sequence into health degree regression network to generate dynamic health degree sequence, and set uniform health degree threshold on multiple devices and multiple working condition distribution, to give stable degradation degree evaluation results from real-time dynamic health degree sequence.

The invention discloses an available time pushing method and device and a storage medium, and the method comprises the steps: firstly receiving request information used for requesting to push target available total time, and then obtaining the predicted available time of a target object for carrying out a target action at each candidate place in each unit time period according to the request information, obtaining reachable information between every two candidate places; according to all the predicted available time and all the reachable information, the total available time of the target is obtained through calculation, dynamic time planning is carried out by comprehensively considering the space-time relation between time and places, and the sum of the available time of the target object for carrying out target actions in multiple unit time periods can be calculated more accurately and comprehensively. According to the method, the processing efficiency of time planning and the reasonability of time planning are effectively improved, so that a more accurate time planning scheme with a better optimization effect can be pushed to the target object, and the satisfaction degree of the target object for time planning is improved.

The invention belongs to the technical field of large freight vehicle weighing, and discloses a large freight vehicle-mounted weighing method and system based on a time sequence large model, and the method comprises the steps: 1, fusing multi-source data collected by a plurality of sensors through employing a timestamp alignment algorithm, and generating and storing time sequence coded data with spatial-temporal characteristics; 2, preprocessing the original time sequence data, and constructing a multi-modal training data set; 3, performing pre-training and fine tuning by using the data set to construct a time sequence large model based on a Transform architecture; 4, inputting newly collected load data with time sequence characteristics into the time sequence large model to carry out real-time dynamic weighing calculation; 5, constructing an error correction hybrid system to accurately correct the weighing result; and 6, outputting an accurate weighing result with a time sequence characteristic in real time, and carrying out reliability prediction on a load change trend.

The invention discloses a dynamic feature space construction method for time dimension coupling multi-modal data, which comprises the following steps: time synchronization modeling is realized through time dynamic correlation calculation and time step relation weight generation; time granularity alignment is realized through dynamic time warping optimization and time scale unification; and dynamic feature space construction is realized through global time feature extraction and local-global feature fusion. The invention discloses a dynamic feature space construction method for time dimension coupling multi-modal data, and the method comprises the steps: capturing a time synchronization relation between modals through dynamic correlation calculation and time step weight generation; unifying a multi-modal time resolution and a time step relationship through time granularity alignment; and finally, through fusion of local and global time features, a dynamic shared feature space is generated, and the multi-modal time dynamic state is comprehensively expressed.

The embodiment of the invention discloses a method for calculating a visible window of a satellite. The method comprises the following steps of: establishing a time dynamic stepping function about a distance and stepping time between a simulated earth station and the satellite; judging whether the current distance is greater than the farthest distance; if not, controlling the satellite to move according to the preset stepping time when the satellite is in the visible range at the current moment T0 until the distance between the earth station and the satellite at the first moment T1 is greater than the farthest distance, and subtracting the current moment T0 from the first moment T1 to obtain the time range and duration of the visible window; if yes, the current distance is substituted into the time dynamic stepping function to obtain the corresponding stepping time when the satellite is not in the visible range at the current moment T0, and the satellite is controlled to move according to the corresponding stepping time until the distance between the earth station and the satellite at the second moment T2 is smaller than or equal to the farthest distance.

A system for sensing a state of a device is provided. The system includes an autoencoder including an encoder, a potential sub-network, and an extension decoder. An encoder encodes each input data point of input data from an input state space into a potential space to produce potential data points, and propagates the potential data points using a Sheng differential equation (ODE) to estimate a potential dynamic initial point of the device in the potential space. The potential sub-network propagates the initial point up to the temporal index of interest using the neural ODE to produce a potential dynamic state of the device at the temporal index of interest. The extension decoder decodes the potentially dynamic state of the device into an output state space different from the input state space to produce output data including the state of the device at the time index of interest.

The invention discloses a time sequence prediction method based on multi-dimensional feature fusion and a computer program product. The prediction method is constructed based on a lightweight architecture, and complex module stacking is avoided. The method comprises the following steps: firstly, designing a multi-dimensional convolution extraction module, respectively capturing short-term fluctuation, long-term trend and timestamp association by utilizing three-dimensional convolution operations, and realizing comprehensive coverage of local details, global trend and time dynamic; secondly, a mask mechanism is designed, and masks are used on the dimension based on periodic modeling so as to improve the prediction effect and the model stability. And finally, the features of the three dimensions are sent to a prediction module and weighted fusion is carried out to obtain a final result. Therefore, the features of different time dimensions can be effectively captured, the prediction efficiency is improved, and the computing power demand is reduced. In the face of different service data in the time sequence field, the model can actively capture the core law and potential association of different time sequence data without complex adjustment, so that accurate prediction is realized.

The invention discloses a short video real-time dynamic recommendation method and system based on deep learning, and relates to the technical field of Internet, through overall execution of the steps, a user behavior set Beh can be divided and modeled under different time period sets Tsg, a time period feature set Tfe and a training set Trn are generated, and the time period feature set Tfe and the training set Trn are subjected to real-time dynamic recommendation. And finally obtaining an interest model set Mod and a time period weight set Twg. When a user request arrives, the corresponding interest model set Mod can be called immediately according to the time period set Tsg, and the recommendation result Rlt is dynamically generated in combination with the time period weight set Twg, so that the recommendation process not only avoids the problems of fuzzy interest description and static stiffness of the recommendation result caused by modeling of a single time dimension in a traditional method, but also improves the user experience. And the recommendation result Rlt can be adaptively updated along with the switching of the time period set Tsg, so that the timeliness, the matching degree and the user experience of the real-time recommendation of the short video are remarkably improved.

The invention discloses an airport scene event autonomous prediction method based on hierarchical time sequence characteristics and a dynamic map, and belongs to the technical field of airport scene intelligent operation control and video processing. The method comprises the following steps: collecting image data containing an airport scene event, and carrying out manual labeling and preprocessing; extracting short-time dynamic features, medium-time interaction features and long-time evolution features based on a multi-granularity sliding window strategy to form a multi-level feature sequence; constructing a dynamic graph structure fusing spatial topology and business linkage; performing high-order feature aggregation on the nodes by using a graph attention network, and introducing a difficult sample weighting mechanism to improve the discrimination capability of high-risk events; constructing a historical memory library, and dynamically enhancing extreme scene discrimination; according to the method, under the condition of extremely low memory consumption and computing power consumption, the problems of high dynamics and diversity of event prediction are solved through hierarchical time sequence characteristic decomposition and dynamic graph construction, the airport scene autonomous prediction is realized, and the intelligent operation and control energy of the airport scene is effectively improved.

The invention discloses a user potential intention modeling method based on continuous time dynamics, which comprises the following steps of: firstly, performing time sequence perception embedding on an original behavior sequence of a user; then, the embedded sequence is processed through a sequence encoder, and a state vector representing the initial potential intention of the user is obtained through variational inference; thirdly, performing dynamic evolution on the initial potential intention state in continuous time by using a neural network parameterized Sheng differential equation so as to solve the state at any time point in the future; and finally, mapping the evolved future potential intention state to a project space through a decoder module, and generating a recommendation result. According to the method, the discrete behavior sequence is mapped to the continuous potential intention evolution trajectory, and the problem that a traditional discrete time model is difficult to process irregular interaction intervals and simulate smooth evolution of user interests is solved.

The invention discloses a bus bunching prediction method and system based on a time-aware network. The method comprises the following steps: preprocessing bus arrival data, fusing multi-source spatial-temporal characteristics, introducing time period codes, and processing abnormal and missing values; targeted denoising is performed on data with different characteristics based on Fourier analysis, so that the data quality is improved; a time perception neural network is established, local features and long-term dependence are captured through a CNN-LSTM architecture, a time decay factor mechanism is introduced, dynamic modeling of a time sequence information importance decay process is realized, and a long-term rule of data and short-term disturbance after time calibration are effectively integrated; and a dual attention mechanism of time dimension and feature dimension is fused, key features are extracted and integrated, and multi-step prediction of bus bunching is realized. According to the method, the non-linear relation and the space-time dynamic rule of the bus data can be fully captured, and the problem of unequal intervals in a bunching time sequence is effectively solved.

The invention provides a social situation factor modeling and intelligent fusion method and system based on time dynamics to solve the problems that in existing network information demand prediction, social situation factors lack time dynamics modeling, news recognition precision is low, a fusion mechanism is static, and multi-source factors do not have a unified framework. According to the method, three types of situation factors are collected in a sampling period of one hour, and classification completion is carried out according to an Ingwersen framework; traversing [24, 24] hours through the CCF to determine the optimal lag time, and extracting the time dynamic characteristics in combination with Granger causality test (plt; using a BiGRU + CRF classifier to identify sensitive news and quantify emotion intensity; and according to factor types, a differential attenuation function is matched to establish a dynamic weight mechanism, and feature vectors are fused and output. The method can be used as an independent module for a time sequence prediction model; in the BAI data set emergency scene, the MSE is reduced by 47.8%, the news identification accuracy is 91.2%, the community scene prediction precision is improved by 32.1%, the operation and maintenance response time is shortened by 57%, and the prediction precision and robustness are effectively improved.

The embodiment of the specification provides a kind of nonlinear dynamic reservoir network timing prediction system and implementation method, wherein the system includes: microcontroller, built-in pulse width modulation PWM module, for receiving input signal, the input signal is preprocessed, obtain the input signal after preprocessing, the input signal after processing is encoded by the pulse width modulation PWM module, obtain PWM encoding signal;Multiple linear resistance-capacitance R-C network, wherein each R-C network has different resistance value and capacitance value;For realizing reservoir layer with distributed nonlinear characteristics and diversified time dynamic characteristics, each R-C network in the storage pool is connected in parallel, the PWM encoding signal is refitted, and the final output signal is obtained;Linear regression network is used to obtain the final output signal, and the classification and prediction task of time series data is executed.

The invention discloses a power system space-time hybrid mode uncertainty modeling method and system based on aggregation-decomposition, and aims to solve the problem of mode uncertainty caused by multiple factors such as new energy output and load change. The method comprises the following steps: firstly, collecting time sequence operation data of a power system, and performing stationarity test on the time sequence operation data to obtain a stationary time sequence; then, disassembling a time dynamic dependency and space correlation dependency relationship, and extracting multi-dimensional spatio-temporal data characteristics to obtain a data basis suitable for spatio-temporal modeling; then, on the basis of an aggregation-decomposition framework, performing aggregation processing on the original data, performing modeling on time sequence correlation characteristics of the aggregated data, and calculating spatial correlation of the objects by taking the time sequence data of the objects as characteristics of the objects; and finally, establishing a space-time hybrid mode uncertainty modeling evaluation index, and evaluating a space-time hybrid modeling effect. According to the method, the uncertainty of the space-time hybrid mode can be efficiently represented, and a reliable basis is provided for stability analysis, risk assessment and decision making of a power system.

The disclosure relates to a method (600) of providing a representation of temporal dynamics of a first system (200) comprising sensors (212) by utilizing a middleware system (300), the middleware system (300) comprising two or more network nodes (355), wherein a first set of the two or more network nodes (355) are connectable to the sensors (212), the method comprising: receiving (620) activity information from the sensors (212) indicative of the temporal dynamics of the first system (200), wherein the activity information evolves over time; applying (630) a set of unsupervised learning rules to each of the one or more network nodes (355); learning (640) a representation of the temporal dynamics of the first system (200) by organizing (645) the middleware system (300) in accordance with the received activity information and in accordance with the applied sets of unsupervised learning rules; and providing (650) the representation of the temporal dynamics of the first system (200).The disclosure further relates to a middleware system, a controller system, computer program products and non-transitory computer-readable storage media.

The invention discloses an industrial process remaining time prediction method and device and a storage medium, and belongs to the field of data processing. The method comprises the following steps: acquiring a track prefix sequence, inputting the track prefix sequence into an activity completion model, and outputting a complete track prefix sequence after activity completion; the semantic feature vector and the time interval feature vector of each activity in the complete track prefix sequence are extracted and spliced, and after fusion feature vectors are generated, the fusion feature vectors are arranged according to the activity sequence to form a feature sequence; inputting the feature sequence into a remaining time prediction model, and outputting predicted remaining time of the industrial process instance; wherein the activity completion model is obtained on the basis of a BERT architecture in combination with comparative learning training, and the remaining time prediction model is based on a Transform architecture and introduces a time perception attention mechanism. Effective repair of activity missing of the industrial process, deep fusion of multi-dimensional features and accurate modeling of time dynamic characteristics are realized, and the accuracy of predicting the remaining time is improved.

The invention discloses an adaptive continuous learning time-varying distribution parameter system space-time modeling method, device, equipment and medium, an initial space-time prediction model is constructed through off-line data, after entering an online stage, the space-time non-stability degree of a system is quantified in real time, and the time-varying distribution parameter system space-time modeling method is realized. A space-time forgetting factor is adaptively generated by using a multi-criterion inference mechanism, so that the updating rhythm of the model is dynamically matched with the variable-scale time-varying rhythm of system dynamics, and the problem that a fixed learning rate is difficult to adapt to multi-scale time-varying characteristics is solved; through a space-time collaborative replay learning mechanism, a disastrous forgetting phenomenon in space-time continuous learning is overcome by recall consolidation of a historical core space-time dynamic mode, and the steps of online distributed parameter system data acquisition and model dynamic updating are repeated until an online stage is ended. Accurate tracking and long-term knowledge maintenance of time-varying space-time dynamics are realized, and prediction precision and robustness of a distributed parameter system in a complex non-stationary environment are remarkably improved.

The invention provides a Bi-GRU and ARIMAX-based hybrid prediction method for solving the problems that in network information demand prediction, a single model is difficult to give consideration to both a linear period and nonlinear burst, social situation factors lack time dynamics modeling, and a cross-scene cold start period is long. Network scene characteristics are quantized through a scene heterogeneity index (SHI), a double-branch parallel architecture is constructed, Bi-GRU is used for modeling nonlinear time sequence dependence, ARIMAX is used for depicting linear trends and exogenous variables, a noise adaptive fusion mechanism is introduced to dynamically adjust model weights, and 7-14-day rapid deployment is realized in combination with a transfer learning strategy of parameter freezing. On a BAI data set, the model MSE is 0.078, the MAPE is 6.26%, and the MSE and the MAPE are respectively reduced by 69.4% and 58.7% compared with those of ARIMAX and LSTM; and the error amplification is only 56% under the condition of 20% noise. In cross-scene application, the cold start period is shortened from 30-60 days to 7-14 days. After the method is applied to a university library scene, the network operation and maintenance cost is reduced by 35%, the service response time is shortened by 57%, and the resource utilization rate is increased from 47% to 83%.

The invention discloses a time sequence prediction method and system based on endogenous-exogenous cooperation. The method comprises the following steps: acquiring endogenous input and exogenous input of a prediction target; integrating the endogenous input into an endogenous embedded vector; integrating the exogenous input into an exogenous embedded vector; splicing the endogenous embedded vector and the exogenous embedded vector to obtain a joint feature; endogenous characterization is optimized through a multi-stage time attention mechanism; endogenous embedding and exogenous embedding are fused through a dynamic interaction attention mechanism, and cross-modal interaction features are obtained; fusing the optimized endogenous representation with the cross-modal interaction feature to obtain an intermediate dynamic representation; and performing time series prediction based on the intermediate dynamic representation. Unified characterization of multi-source heterogeneous features is achieved, local short-term correlation and global long-term dependence of endogenous variables are integrated through self-adaptive weight distribution, and collaborative correlation, dynamically evolved along with time, between input of the internal source and input of the external source can be accurately described.

The invention discloses an industrial equipment cooperative task anomaly detection method and system, and belongs to the technical field of industrial equipment detection. The method comprises the steps of preprocessing time sequence data of multiple mechanical arms, determining an initial window based on a task complexity index, dynamically determining an optimal time window through Bayesian optimization, and adopting the window to train a VAE-LSTM model to extract time dynamic features; a mechanical arm is modeled into graph nodes, a collaborative relation topological graph is constructed, local interaction features and global topological features are extracted through a multi-scale attention mechanism, and collaborative spatial features are extracted through graph convolution after self-adaptive fusion is conducted through a gating mechanism; and finally fusing the spatio-temporal features and carrying out anomaly detection. According to the method, through a dynamic window and a multi-scale attention mechanism, the space-time interaction relation in the mechanical arm cooperative task is captured in a self-adaptive mode, the accuracy, the real-time performance and the adaptability of anomaly detection are remarkably improved, and the defects of an existing method in the aspects of dynamic task adaptation and multi-modal data fusion are effectively overcome.

A technique is disclosed to use time series phasor data to perform real-time dynamic line rating of electric power transmission lines. A variety of techniques are used to generate well-poised solutions to the determination of transmission line parameters from phasor data. Line health information can also be determined from changes to transmission line parameters, such as galloping, icing, vegetation encroachment, imperfect splicing, and conductor corrosion.