34 results about "Time series generation" patented technology

The invention discloses a time sequence generation method and system based on uncertainty quantization and frequency domain constraint, and belongs to the technical field of artificial intelligence and data mining. Outputting a mean value of the completion value and a variance representing the uncertainty of the completion result; an entropy weight adaptive mask generation mechanism is constructed, a soft mask matrix of continuous values is generated according to variance and local observation density, and high uncertainty corresponds to low mask weight; constructing a mixed loss function containing time domain weighted loss and frequency domain consistency loss; and training a diffusion model based on the soft mask matrix and a mixed loss function to generate a regularized high-fidelity time sequence. According to the method, the Bayesian uncertainty quantization and soft mask mechanism is introduced, so that the problems that a binary mask mechanism in the prior art is rigid and cannot sense completion errors are solved, and meanwhile, the periodic characteristics of generated data are ensured in combination with frequency domain constraints.

A system for Time Series Generation (TSG) model selection. The system is configured to perform a method including: receiving a user prompt, the user prompt comprising a user input and a user-provided TSG dataset; providing the user input as input to a first machine learning model, to determine the user prompt as a TSG query; providing the user input and the user- provided TSG dataset as input to a second machine learning model, to select at least one shortlisted TSG model from a TSG database; and using the first machine learning model, providing the at least one shortlisted TSG model as a response to the TSG query.

The invention discloses an actuarial calculation engine code compiler, an operation method and a related product. The actuarial calculation engine code compiler comprises an editing layer, an analysis layer and a mapping layer. The editing layer is configured to compile an actuarial process about the target actuarial business by adopting actuarial business terms and actuarial logic in response to the operation of the client; the analysis layer is configured to analyze the actuarial process and identify variables, formulas and time sequences in the actuarial process; generating a main syntax tree comprising target actuarial business semantics and a sub syntax tree comprising operation symbols and functions; and a mapping layer configured to parse the main syntax tree and the sub-syntax tree into machine executable code. In the embodiment of the invention, the actuarial process is edited through the actuarial business terms and the actuarial logic, and the actuarial process is converted into the machine executable code, so that the actuarial process adapts to high professionality and high suitability of professional evaluation of actuarial professionals.

The present disclosure relates to a user learning data prediction method, device, medium, equipment and program product, comprising: obtaining a learning data time sequence, the learning data time sequence comprising actual data of a target analysis dimension at a plurality of historical time nodes; generating an interval grey number sequence according to the learning data time sequence, the interval grey number sequence comprising interval grey numbers of the target analysis dimension at the plurality of historical time nodes, the interval grey numbers being determined according to corresponding actual data; and predicting the target analysis dimension of a future time node according to the interval grey number sequence to obtain a prediction interval of the target analysis dimension. The present disclosure represents the historical learning data of a user as a learning data time sequence, represents the uncertainty variable of the target analysis dimension by an interval grey number to obtain an interval grey number sequence, and thus predicts the target analysis dimension of a future time node, so that the prediction interval of the target analysis dimension is more in line with the small sample and uncertainty characteristics of individual learning data.

The invention relates to the field of time synchronization information security, and provides a time synchronization information security examination method and device based on dual-scale feature collaboration, equipment and a medium, and the method comprises the steps: generating a large-scale sample set based on a historical time sequence; utilizing the large-scale sample set to train and generate a large-scale branch model; generating a small-scale sample set based on a predicted value output by the large-scale branch model; using the small-scale sample set to train and generate a small-scale branch model; generating a final predicted value which comprises predicted values of the large-scale branch model and the small-scale branch model; and performing security review on the real-time time synchronization information based on the final predicted value. According to the method, the dual-scale branch neural network model with a cooperative training mechanism and a dual optimization mechanism are utilized to realize complementary enhancement of long and short scale feature extraction of time synchronization information, and cooperative consideration of the model on long-term security situation stable monitoring and short-term abnormal feature sensitive identification is completed; and the accuracy and the reliability of time synchronization information security examination are obviously improved.

A method and system for generating synthetic data is provided, in which longitudinal time series data is retrieved and a neural network is trained to generate synthetic time series data that meets a privacy metric based on the longitudinal time series data, where the longitudinal time series data is unlabeled and univariate.

The application provides a high-reliability anti-interference engine rotating speed measurement data processing method, including the following steps: generating a square wave interval time sequence based on a square wave time sequence, generating an interval time sequence multiset based on the square wave interval time sequence, and obtaining a calculated rotating speed through the interval time sequence multiset; performing abnormal value checking and replacement on the calculated rotating speed to obtain an optimized rotating speed; and performing smoothing filtering processing on the optimized rotating speed to obtain an output rotating speed. The method can reasonably divide and classify the time intervals collected by the rotating speed sensor, reduces effective data loss, improves rotating speed measurement accuracy, suppresses rotating speed jump, and calculates a more reasonable and reliable real-time rotating speed under a large amount of high-frequency noise, random interference, intermittent or continuous time sampling abnormality and the like.

This invention discloses a method and apparatus for generating long-term series data of photovoltaic power generation from a photovoltaic power plant. The method includes: inputting pre-prepared time-series data of power generation from a target photovoltaic power plant under different weather conditions and random time-series data of the same dimension into a pre-constructed generative adversarial network model for generating typical daily power curves of the photovoltaic power plant under different weather conditions, thereby generating typical daily power time-series data of the target photovoltaic power plant under different weather conditions; constructing a daily weather type time-series generation model based on implicit Markov and Monte Carlo simulations; and sorting the typical daily power time-series data of the target photovoltaic power plant under different weather conditions using the daily weather type time-series generation model to generate long-term series data of power generation from the target photovoltaic power plant.

Adaptive Forecasting System This description concerns a control system configured to supply and / or receive, in a buffer tank, a quantity of a resource destined for or received from a target system (102). The control system comprises: - at least one sensor (106) configured to periodically measure, within a first time interval, quantities of the resource supplied and / or received; - a processing device (110) configured to perform a processing operation on the measurements taken by the at least one sensor. This processing operation corresponds to the generation of a first time series whose values ​​are sums of the quantities measured at several time instants. The processing device further comprises a neural network (202) to generate at least one quantity estimate based on the first time series and to control an actuator (108). Figure for the abstract: Fig. 1A

The invention discloses a TTAO-LSTM less-data container cloud resource load prediction method, and relates to the technical field of deep learning. The method comprises the following steps: firstly, carrying out normalization processing on original container cloud resource load data, and dividing time sequence data into a plurality of fixed-length sample sequences through a sliding window mechanism; carrying out data enhancement on small sample load data by utilizing an improved TimeGAN structure so as to improve the training data volume and the model generalization ability; a TTAO algorithm is adopted to search and optimize key hyper-parameters of the LSTM network, and an optimal load prediction model is constructed; and finally, inputting the sample after data enhancement into the trained TAO-LSTM model to obtain a prediction result. According to the TTAO-LSTM less-data container cloud resource load prediction method provided by the invention, data enhancement is performed on small sample data through time sequence generation, and the generated data is sent to the TTAO optimized LSTM model, so that prediction of a future resource load trend is completed, a high-precision result is output, and accurate prediction of a less-sample resource load task is realized.

Provided is a technique that makes it possible, as evaluation of a control plan with respect to a control target, to evaluate a plan sequence while taking into consideration long-term influence. An information processing apparatus includes: an acquisition section that acquires state information and a plan sequence, the state information indicating a state of at least one of a control target and an environment, and the plan sequence being a time series of control plans with respect to the control target; a generation section that generates a state sequence using output obtained by inputting the state information and the plan sequence into a learned model, the state sequence being a time series of pieces of state information each indicating a predicted future state; and a calculation section that calculates, using the state sequence, a success probability of the plan sequence which has been acquired by the acquisition section.

The invention relates to the technical field of intelligent decision making, can be applied to business system platforms such as financial science and technology and medical health, and discloses a time sequence generation method, device and equipment and a medium, and the method comprises the steps: obtaining sequence description information inputted by a target user, recognizing the field category of the sequence description information, and obtaining a field matrix corresponding to the field category; semantic features of the sequence description information are extracted, feature mapping is conducted on the semantic features through the domain matrix, and semantic vectors of domain categories are obtained; calculating the similarity between the semantic vector and each semantic tag vector in a semantic prototype library, and calculating a weighting coefficient of the semantic tag vector according to the similarity; based on the semantic vector and the weighting coefficient, utilizing a conditional diffusion model to perform reverse denoising to generate an initial time sequence of the sequence description information; and correcting the initial time sequence according to a constraint strategy corresponding to the field category to obtain a time sequence of the sequence description information. Through the method, the time sequence generation accuracy can be improved.

The invention belongs to the field of photovoltaic power prediction, and discloses a TimeGAN-based photovoltaic power prediction method in consideration of turning weather, and the method comprises the steps: detecting an abnormal value through a reconstruction error, carrying out the filling of missing data through a bidirectional K-nearest neighbor algorithm of meteorological similarity weighting, and selecting high-correlation meteorological features; aiming at the condition of insufficient turning weather samples, expanding a turning weather data set through a time sequence generative adversarial network; and outputting a photovoltaic output feature sequence through a CEEMDAN-VMD double decomposition algorithm, and inputting the photovoltaic output feature sequence and the high-correlation meteorological feature sequence into an SSA-LSSVM model for photovoltaic output prediction. According to the method, the photovoltaic output and meteorological characteristic sequence is fully extracted through correlation analysis and double decomposition, the prediction model is jointly input, parameter selection is optimized, the influence of meteorological driving force on the system uncertainty in the turning weather is enhanced, and the accuracy and stability of photovoltaic output prediction in the turning weather are improved.

The application discloses a kind of wetland water quality monitoring pollution source intelligent tracking method based on deep learning, to solve the problem that wetland hydrodynamic boundary is complex, parameter is difficult to obtain, leading to the position of pollution source is difficult to be stably deduced from monitoring result, large positioning error, the application is by obtaining wetland water system spatial data and water quality monitoring data and preprocessing to construct observation water quality time series and working condition time series, construct the initial map of wetland water system including monitoring point and candidate water inlet and candidate water inlet set;Time-varying adjacency matrix sequence, direction constraint matrix sequence and condition vector sequence are generated from working condition time series;Train the forward prediction model including the space-time graph Transformer network limited by direction-constrained attention and the conditional neural operator network modulated by condition vector;In candidate water inlet set, the pollution source parameter is iteratively updated with differentiable inversion optimization, and the maximum weight corresponding water inlet number and coordinate are output, realize the stable tracking positioning of pollution source water inlet and coordinate under the condition of dam scheduling, tide and other working conditions, improve inversion stability and reduce positioning error.

The invention discloses a high-resolution photosynthetically active radiation absorption ratio time sequence generation method based on random forest iterative interpolation, and the method comprises the steps: building a Gaussian process regression model with Sentinel-2 surface reflectance as input and GLASS FAPAR as output, carrying out the downscaling of the GLASS FAPAR, and obtaining a FAPAR result of a 10-meter clear sky pixel; carrying out pre-filling on the missing value by utilizing GLASS FAPAR, establishing a random forest model, and carrying out second filling on the missing FAPAR; updating the pre-filled FAPAR value by using the result of the second filling, re-establishing the random forest model, and performing third filling on the missing FAPAR; and applying the final model to the tinel-2 reflection data of the target area until the precision reaches a target threshold, and generating a high-resolution space-time complete FAPAR product of the target area. According to the method, the FAPAR time sequence with the resolution ratio of 10 meters can be accurately generated, and important data support is provided for applications such as vegetation monitoring, ecological system evaluation and carbon cycle research.

The invention relates to a method for processing time series of noisy images of a same area, the method comprising: generating a set of time series of images from an input image time series by combining by first linear combinations each pixel of each image of the input image time series with selected neighboring pixels in the image and in an adjacent image of the input image time series; applying filtering operations in cascade to the set, each filtering operation combining each pixel of each image of each time series of the set by second linear combinations with selected neighboring pixels in the image and in an adjacent image in each time series of the set; performing an image combination operation to reduce each time series of the set to a single image; introducing a model image of the area as a filtered image in the set; and combining each image in the set into an output image, by third linear combinations.

This invention proposes a vector data processing method for drainage systems oriented towards SWMM modeling, relating to the technical fields of urban flood simulation and vector data processing. The method involves reading and unifying the basic vector data of the modeling area, and reading time series files. Based on the pipe network topology, outlet nodes are identified through set operations, and outlet attribute vector data conforming to SWMM model requirements is constructed. Based on the outlet name mapping relationship, the attributes and topology of manholes, pipes, and sub-catchments are updated. Rain gauges are deployed and their attributes configured. Two types of time series are processed to generate standardized time series files conforming to SWMM format requirements. After data preparation, the data is converted into an SWMM model file. This invention achieves standardized conversion from raw input data to model input files, offering advantages such as clear logic, simple calling, and strong operability. It provides an efficient means for preliminary data preparation in urban flood simulation and improves the reliability of flood simulation results.

The application discloses a kind of renewable energy annual output time series generation method, belong to power system control technical field, the method, using Ito stochastic process to the current residual component and the current trend component in historical output time series by stationary calibration are modeled, obtain initial simulation trend component and initial simulation residual component;Further respectively to the inverse difference processing of both obtain target simulation trend component and target simulation residual component, finally reconstruct target output time series using the target simulation trend component, the target simulation residual component and the periodic component;The scheme can generate massive simulation curves by Ito stochastic process to the mathematical modeling of non-periodic component, while accurately reflecting its random characteristics and time series correlation, can accurately quantify the key features of renewable energy under long time scale, according to the above key features, the new energy consumption capacity of power grid can be accurately grasped.

The invention discloses a machine learning-based low-altitude air route time slot distribution method and system. The method comprises the steps of constructing a four-dimensional space-time grid, dividing voxel units, collecting an airspace state and forming a feature sequence; based on the state feature sequence, constructing a voxel space-time diagram containing space and time edges; fusing intrusion time coding, inputting the improved STSGCN model, and generating a risk and time sequence; constructing a reversible reasoning link, and executing reverse time sequence reasoning to obtain a reverse time sequence; generating oscillation parameters by using the time sequence, and forming a dynamically adjusted time slot adjustment sequence; and constructing an airspace elastic field, executing grid deformation, generating an elastic grid and outputting a scheduling result. According to the invention, through combination of improved STSGCN prediction and reversible predicted intrusion time reasoning with time slot self-adjustment and airspace elastic grid updating, dynamic safe distribution of low-altitude airway time slots and airspace capacity improvement are realized.

The invention discloses an interpretable text semantic driving time sequence generation method based on a Diffusion Transform model, and relates to the technical field of artificial intelligence and time sequence generation methods. The method comprises the following steps: converting a natural language condition into a control signal generated for a time sequence; dividing a time sequence characteristic explicit into three parts, namely a trend part, a season part and a residual error part, and capturing a long-term time dependence and periodic mode in combination with a time sequence Transform; a time sequence conforming to text semantics is obtained through an iterative sampling process of DDPM, and interpretable trend, season and residual component decomposition results of the time sequence are output at the same time; text condition coding and time embedding are fused into a joint condition vector, intermediate features of a denoising network are modulated through self-adaptive normalization, so that the generation process is dynamically guided by text semantics, and an interpretable text time sequence conforming to an instruction is generated. According to the method, the stability and diversity of the generation process can be improved, and the interpretability and analyzability of the generation result are improved.

The application provides a time series generation method based on a frequency domain double learning mechanism, including: acquiring noisy "future one-year temperature" time series data, performing Fourier transform on the time series data to the frequency domain to obtain a frequency spectrum; using a linear layer neural network to analyze the frequency spectrum, calculating the importance weight of each frequency column on the frequency spectrum, and obtaining a weighted frequency spectrum based on the importance weight multiplied by the corresponding frequency column; dividing the frequency spectrum into multiple blocks; sending all the blocks into an attention network for attention calculation to obtain an attention output; performing residual connection on the attention output and the weighted frequency spectrum to obtain a fused and optimized frequency signal; converting the fused and optimized frequency signal back to the time domain through inverse Fourier transform to obtain a seasonal component; extracting a trend component from the noisy "future one-year temperature" time series data; and adding the seasonal component and the trend component to obtain predicted "future one-year temperature" initial data.

In some implementations, sequences of time series values determined from machine data are obtained. Each sequence corresponds to a respective time series. A plurality of predictive models is generated for a first time series from the sequences of time series values. Each predictive model is to generate predicted values associated with the first time series using values of a second time series. For each of the plurality of predictive models, an error is determined between the corresponding predicted values and values associated with the first time series. A predictive model is selected for anomaly detection based on the determined error of the predictive model. Transmission is caused of an indication of an anomaly detected using the selected predictive model.

The invention discloses a method for predicting the remaining service life of a power transmission line. The method comprises the following steps: firstly, extracting fault probability features from historical operation data by using a random forest; then, determining a cable life degradation point and a degradation speed through k-means clustering based on the load data, and constructing a segmented residual service life model to generate a training label; then, a time sequence generative adversarial network is adopted to enhance fault data, and the problem of data imbalance is solved; and finally, constructing a CNN-LSTM prediction model, and carrying out training by taking the fault probability characteristics as input and taking the segmented residual life as a label. When the method is applied, real-time data is processed by the same process and input into the trained model, so that an accurate residual service life prediction value can be output. Therefore, the dependence of a traditional method on a physical model and balanced data is effectively overcome, and the prediction precision is remarkably improved.

The application discloses a power system load data super-resolution reconstruction method and system, relates to the technical field of power system data quality improvement, and comprises the following steps: weighting weather influencing factors under different weather conditions, calculating the correlation between load data and weather influencing factor data, obtaining key influencing factor data, and performing dynamic range normalization processing by using a sliding window; high-resolution load data is obtained by using a trained time series generation model to perform super-resolution reconstruction on the normalized data; during training, the time series generation model adjusts the loss function weight according to the change of the current training loss; after training is completed, the combination features of the normalized load data and the key influencing factor data are extracted to perform super-resolution reconstruction. The defects of insufficient resolution of low-quality sampling data are overcome, the quality of reconstructed data is improved through meteorological factor constraint, and the problem that low-resolution sampling data cannot meet the training or use requirements is solved.

The invention discloses a safety monitoring system and method for a construction period of constructional engineering. The method comprises the following steps: collecting multi-modal data, and standardizing to generate a monitoring sequence; constructing a space-time diagram, and learning causal priori and lag indexes by using an attention network to obtain a causal parent set and a condition set; topological analysis is carried out on the monitoring sequence and the causal parent set, a topological scale index is established, and a time sequence component is reconstructed; inputting the topological scale time sequence component into a time sequence to generate an adversarial network, generating an extreme working condition sample and forming an extension set; screening invariant causal features based on the extension set and the condition set to obtain a robust causal feature set; calculating an improved Liang-Kleeman information flow rate, an output scale level and a global information flow rate under the robust causal feature set and the topological scale index; identifying a risk source, and performing intervention, comparison and evaluation on the key driving variables; and executing graded early warning. According to the invention, the real-time performance, the interpretability and the early warning reliability are improved.

A method for automatically determining a composite index for evaluating an electric power grid includes (a) receiving a time series of sensor measurements from one or more sensors within an area served by the electric power grid, (b) generating, from the time series of sensor measurements, a plurality of indexes for evaluating the electric power grid, and (c) generating the composite index based on the plurality of indexes for evaluating the electric power grid.

This invention relates to the field of time series generation technology, and more particularly to a method and system for generating time series data with extended similarity upper bound constraints. The method includes: normalizing slices of multi-domain time series data, dividing the data into sliding windows and performing discrete Fourier transforms, extracting the amplitude and phase features of the windows, and fusing domain information to obtain low-dimensional embedding vectors; calculating frequency domain similarity and constraining its upper bound using learnable two-dimensional convolutional kernels to obtain a constraint metric; then, attention encoding the embedding vectors to output period-aligned time series feature codes; finally, introducing a domain condition vector and using the attention mechanism of the metric constraint for iterative denoising to generate cross-domain time series data with period matching to the target domain. This invention solves the problems of period misalignment and weak generalization in cross-domain time series generation through upper bound constraints, frequency domain window fusion, and domain condition injection, generating high-quality sequences with period alignment and consistent distribution.

This invention discloses a digital twin-based method for estimating the wear of semi-autogenous grinding mill (SAG) liners, belonging to the technical field of mineral processing equipment. The invention includes: constructing a digital twin model for estimating the wear of SAG liners; setting the initial ore filling rate, steel ball filling rate, application conditions, and application scope of the digital twin model; generating an equal amount of ore and steel ball particles based on historical drum speed, feed rate, and ball feed data of the SAG and adding them to the drum of the digital twin model; performing numerical simulation to calculate the wear of the SAG liners; enhancing and expanding the time-series generative adversarial network according to a 1:30 ratio of input data to generated data; calculating errors and correcting the model to obtain the final result. Compared with the discrete element method (DEM), this invention effectively solves the problems of long simulation time and delayed simulation results, contributing to the development of liner wear estimation methods in the control loop of intelligent semi-autogenous grinding mill systems.

The application discloses a kind of based on array dynamic scheduling's metallurgical electric furnace feeding control method and system, it is related to industrial automation control technical field, method includes receiving the feeding request of target furnace top stock bin, corresponding formula array is called based on array scheduling algorithm;According to the production unit of target furnace top stock bin belongs to automatically determine corresponding transport path;Determine the set of discharge port participating in batching and select target discharge port as reference zero point, the spatial difference of each discharge port in discharge port set relative to reference zero point is mapped as dynamic time series by the running speed of belt based on the synchronous trigger mechanism of space-time mapping, generate the start time interval of each discharge port corresponding;According to dynamic time series, discharge port and corresponding quantitative belt are started in turn, quantitative belt is controlled to run using subtraction discharge strategy, through the combination of collaborative control of space-time mapping and adaptive discharge strategy, realize the intelligentization, high accuracy and high reliability operation of whole process of feeding.

The invention relates to a non-intrusive electric bicycle load monitoring method based on a lightweight neural network. The non-intrusive electric bicycle load monitoring method comprises the following steps: collecting total electric energy consumption data, segmenting a time sequence by adopting a sliding window, and generating a power segment; fusing time-frequency domain features; aligning the sequence by adopting a dynamic time warping algorithm; at 5W, event monitoring is triggered, current sequences before and after an event are extracted, the sequence distance is calculated through dynamic planning, and the phase deviation problem is solved; the aligned fragments are input into a MobileNetV3-Small model, and the power change rate is calculated; and constructing a neural network model by taking the feature vector [Ft, Pt] as input and taking the state of the electric appliance and the energy consumption value as output: deploying the model at STM32H745 edge equipment, and performing real-time reasoning. According to the method, a MobileNetV3 lightweight architecture and a time-frequency domain DTW algorithm are combined, the deployment bottleneck of a traditional NILM in a resource-limited MCU is broken through, and an efficient solution is provided for edge computing energy management.