64 results about "Computation history" patented technology

The invention provides a large language model long-term memory method based on human cognition inspiration, which comprises the following steps of: calculating semantic similarity distribution of historical interaction text vectors according to a sliding window, and calculating local information entropy of the window; determining an event boundary based on the local information entropy difference of the adjacent windows; taking each discrete event as the memory of the model, and performing partition management on the memory; when the event memory partition is full, the memory event with the low retention rate is transferred to a long-term memory area; when the interactive content relates to historical information, retrieving related historical events in the event memory partition by using a contextualized memory retrieval method; and the large language model generates content for the user in combination with the current context and the retrieved related historical events. According to the method, memory coding and storage, multi-level dynamic memory management and event-level situational memory retrieval methods based on the event cutting theory are adopted, and therefore the problem that an existing large language model long-term memory method lacks dynamic memory and situational memory is solved.

The invention relates to an electric power internet of things operation and maintenance management method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining sensor data, carrying out the cleaning and time alignment of the sensor data, and obtaining standard operation data; based on a historical data window corresponding to the standard operation data, calculating a mean value and a standard deviation of the historical data, and further generating a dynamic detection threshold value; comparing the standard operation data of the current time node with a dynamic detection threshold value, and screening out potential abnormal data; calculating an abnormal trend factor corresponding to the potential abnormal data, and determining an abnormal development trend of the potential abnormal data according to the abnormal trend factor; and determining a corresponding causal influence weight based on a preset causal relationship network, and calculating a fault score according to the causal influence weight and the abnormal trend factor, so as to generate an early warning report according to the fault score. The method has the effect of improving the management efficiency of the power equipment.

The invention discloses a multi-round dialogue method and device based on a large model, equipment and a storage medium, and relates to the field of natural language processing, and the method comprises the steps: screening out a target round dialogue corresponding to a current round dialogue from historical dialogues; extracting triple information of the current round of dialogue and the target round of dialogue, and constructing a target cue word based on the domain data and the triple information corresponding to the current round of dialogue; the triple information comprises intention information, entity information and slot position information; inputting the target cue word into a preset dialogue large model, and determining a target key value cache of the target cue word and a historical key value cache corresponding to the target round of dialogue; and generating a dialogue reply of the current round of dialogue based on the target key value cache and the historical key value cache. By calculating the semantic similarity of the historical dialogue and the current round, combining entity recognition and only retaining part of content, the memory used by the dialogue is effectively reduced, logic breakage is avoided, and the use efficiency of the video memory is effectively improved through key value cache multiplexing.

The invention provides an electric power information multi-modal data dynamic integration method and system, relates to the technical field of electric power information data integration, and realizes multi-modal data fusion through a technical chain of asynchronous receiving, feature extraction, elastic alignment, dynamic verification and intelligent compensation. The method comprises the following steps: acquiring a sensor numerical value, an infrared image and an environment audio heterogeneous data stream in real time; extracting each modal feature and calculating a historical deviation coefficient; sensor data is used as a reference axis, and a multi-modal maximum feature coincidence interval is determined through sliding window matching; performing cross-modal feature cross validation in the extended dynamic interaction interval, and marking credible fusion features; redundant feature compensation is triggered for the modals which do not pass verification, and an integrated feature vector is generated; and if continuous failure occurs, marking the data source as a low-confidence data source. Through elastic time alignment and a dynamic verification mechanism, the problem of native asynchronization of power multi-modal data is solved, the feature fusion reliability is improved, and accurate fault information can be provided.

The invention relates to the technical field of hydrological forecasting, and discloses a medium-and-long-term runoff intelligent forecasting method, which comprises the following steps: collecting drainage basin data, preprocessing, extracting features, generating a feature vector sequence, and dividing a training set; constructing a physical constraint recurrent neural network module, and training based on a total loss function of physical constraint loss; constructing a sliding window online self-adaptive correction module, and initializing a sliding window to store actually measured runoff values and initial runoff forecast values at the latest O moments; inputting the feature vector at the current moment into a trained physical constraint recurrent neural network module to generate a preliminary runoff forecast value, calculating a historical average deviation according to a window state and correcting the historical average deviation to obtain a final forecast result; and finally, forming a new sample pair by the actually measured runoff value and the initial runoff forecast value, adding the new sample pair into the window, and removing the oldest sample to realize dynamic updating of the window. According to the invention, medium and long term runoff intelligent forecasting with high precision, high physical consistency and on-line adaptive capability is realized.

A computing system for filtering indirect tax legislation includes processing circuitry that implements an indirect tax legislation filtering program. The processing circuitry imports customer sales tax calculation history data from a customer sales tax calculation history database, imports customer configuration data from a customer configuration database, and generates a product category report indicating taxable products sold by a customer. New and / or changed tax rules from the monthly data update are imported, and, in response to receiving an instruction to identify products affected by a monthly data update of indirect tax rule changes, the product category report is filtered with the monthly data update of the tax rule changes. A legislative change report indicating product categories and associated products affected by the tax rule changes is generated and output.

The invention provides a causal reasoning and confidence-driven gas overrun risk judgment method and device, and the method comprises the following steps: firstly constructing a scenario semantic context based on underground monitoring data and an equipment state, and inputting the scenario semantic context into a fine-tuned large language model to generate a risk evolution causal chain containing hidden intermediate nodes; then, respectively calculating a historical case support degree, a real-time data goodness of fit and a knowledge fragment matching degree, and obtaining a comprehensive confidence coefficient based on arithmetic average logic; and the system automatically executes a hierarchical response strategy of full-automatic confirmation, man-machine collaborative research and judgment or low-confidence suppression according to a comparison result of the comprehensive confidence and a preset threshold value, and performs model closed-loop optimization by using feedback data. According to the method, by introducing a recessive logic completion and three-dimensional confidence verification mechanism, interpretability analysis and reliability quantification of the gas over-limit risk are achieved, and the problems that a traditional method is high in false alarm rate and lacks physical basis are effectively solved.

The invention discloses a load curve decomposition method and system based on a curvature equipartition rational B-spline basis function, and the method comprises the steps: calculating the curvature of a historical load sequence, generating a continuous curvature function, determining a group of B-spline node sequences meeting the cumulative absolute curvature equipartition condition, and enabling the node distribution to be adaptive to the load fluctuation intensity; a B-spline primary function is generated based on the node sequence, and a rational B-spline primary function is constructed through weight optimization so as to finely describe a continuous time fluctuation mode; a fitting model is established in combination with time period virtual variables, and a load curve is decomposed into two parts, namely a time period characteristic load level representing different time period reference power consumption intensity and a time characteristic load curve representing a continuous change fluctuation form. The defect that a traditional uniform spline is poor in adaptability to a non-uniform load form is overcome, self-adaptive decomposition with definite physical significance is achieved, and a better-quality data basis is provided for load analysis and prediction.

The invention provides a channel prediction method and system based on a multi-cycle feature convolutional network. According to the method, context sensing weighted input and a dynamic multi-model pool mechanism are introduced. A terminal device obtains environmental context features in real time, dynamically calculates the weight of each time step in a historical channel state information sequence based on the features, and generates a weighted optimized input sequence. Meanwhile, the network equipment pre-trains and maintains a plurality of multi-period feature convolutional network models specific to different environments, and the network can effectively capture the complex time-varying rules in the period and during the weeks in the channel, so that the accuracy of channel state information prediction is improved. And the terminal equipment immediately selects and activates the optimal sub-model in the model pool for prediction according to the real-time context features, and locally performs increment fine adjustment. According to the invention, through management of the data input end and rapid and adaptive switching of the prediction model end, the precision of channel prediction and the adaptive capacity to environmental sudden change are effectively improved.

The invention relates to the technical field of software operation and maintenance simulation, and discloses a software operation and maintenance full-process simulation system based on digital twinning. According to the system, a data acquisition module is used for acquiring a historical operation data set and a real-time operation data set of a software system; the feature parameter extraction module performs feature extraction on the historical data, generates a historical feature parameter set and establishes a dynamic change feature matrix; the digital twin state modeling module constructs a multi-dimensional virtual state space according to the matrix, and determines a key early warning index set by calculating the state distribution density of historical fault events; the real-time risk analysis module extracts real-time characteristic parameters to form a state vector, and performs spatial correlation calculation on the state vector and the key early warning index set to generate a real-time risk correlation value; and the full-process simulation engine module generates an operation and maintenance risk prediction model based on the data, simulates a fault probability and outputs operation and maintenance optimization configuration, so that precise simulation and optimization of the software operation and maintenance full process are realized.

This invention discloses a method for recommending points of interest (POIs) based on user spatiotemporal behavior and social information. The method includes constructing a spatiotemporal knowledge graph based on user POI visit sequences and social relationships; learning POI representations in the spatiotemporal knowledge graph and constructing a POI transfer matrix; updating POI features through graph convolution using the POI transfer matrix as weights; calculating a user preference representation based on the user POI visit sequence and POI features; calculating the weights of historically visited POIs based on the spatiotemporal distances between POIs and the distances between POIs and user preferences; inputting the POI sequence into a recurrent neural network and updating a hidden representation using the weights of historically visited POIs; and connecting the hidden representation with the user preference representation, inputting the result into a recommendation model to generate the next POI that matches the user's preferences. This method can effectively model user spatiotemporal behavior, capturing behavioral patterns and user preferences, and making POI recommendations more accurate.

The present invention relates to a method, device, terminal and medium for avoiding data loss in a peering process. Before an OSD in a down state in a placement group starts up and enters the peering process, a historical epoch sequence is calculated, and by judging whether the historical epoch sequence contains an abandoned OSDmap version, it is determined whether data writing occurs in the placement group during the period of the historical epoch sequence, thereby avoiding the situation where data writing cannot be judged when data is written to the abandoned OSDmap version. The analyzed historical epoch sequence of data writing is not less than the actual situation, thereby avoiding missing data writing of the placement group. When an OSD in the placement group that has data writing during the down state starts up and enters the peering process, the status of the remaining OSDs in the placement group is detected. If the remaining OSDs are all in the down state, the peering process of the OSD is suspended, and the peering process is completed only after the remaining OSDs are started up, thereby avoiding data loss caused by the OSD completing the peering process alone when data is written to the placement group during the down state.

The invention relates to the technical field of fuel contract templates, and discloses an AI-based fuel contract template collaborative editing recommendation method and system, and the method comprises the steps: obtaining historical contract editing data of each contract editor in a contract editing team, and determining a plurality of historical contract editing data sequences; dividing the historical contract editing data sequence into a score-adding and score-reducing historical contract editing data sequence, and calculating a score-adding contract editing factor and a score-reducing contract editing factor; calculating a historical contract editing data recommendation factor of the contract editing team according to the score adding contract editing factor and the score reducing contract editing factor; and selecting the contract editing team corresponding to the maximum historical contract editing data recommendation factor, carrying out collaborative editing on the fuel contract template, calculating the historical contract editing data recommendation factors to improve the recommendation precision and recommendation efficiency of the contract editing team, and editing and recommending a low-risk and high-efficiency contract editing team for the fuel contract template. And the comprehensiveness and safety of contract collaborative editing are ensured.

The invention discloses an electric power prediction method and device, equipment, a storage medium and a product, and relates to the technical field of data processing, and the method comprises the steps: extracting time sequence information from a historical electric power feature sequence obtained in advance, and calculating the time sequence features of the historical electric power feature sequence according to the time sequence information; time coding is carried out according to the sampling date of the time sequence feature and historical power feature sequence and the sampling date of a preset prediction sequence, and an input sequence time code and a prediction sequence time code are obtained; a correlation matrix is constructed based on input sequence time coding and prediction sequence time coding, power prediction is carried out based on the correlation matrix, a power prediction result is obtained, and the problems that a traditional prediction model is insensitive to a time sequence, position coding cannot represent complex periodic characteristics and the like are solved. Therefore, precise modeling of the power data time sequence and the complex period is realized, and the prediction precision and practicability are remarkably improved.

The present application relates to the technical field of communication detection, and discloses a training method and device of a border gateway protocol anomaly detection model and computer equipment, wherein the historical route statistical features and historical graph topology features corresponding to historical BGP data are extracted, and the features are trained by using the border gateway protocol anomaly detection model, the drift detection score between the historical anomaly detection result and the historical BGP data is calculated in the training process, and the drift detection threshold is used as a reference benchmark to adaptively update the multiple parameter values of the border gateway protocol anomaly detection model in combination with the drift detection score. Therefore, even if the BGP abnormal condition caused by the dynamic change of the network environment is faced, the BGP abnormal data can be accurately detected, and the high efficiency of the BGP abnormal data anomaly detection can be ensured.

The application discloses a kind of based on multi-source data's electric quantity abnormality monitoring method and system, method includes: calculating historical load fluctuation coefficient and carrying out first level division, obtain user terminal set;Real-time load data sequence is obtained and real-time load feature is extracted;Combining static attribute data and real-time load feature, user terminal set is carried out second level division, and obtain user terminal sub-set;Based on historical load fluctuation coefficient interval, target user terminal is selected in sub-set and electric quantity abnormality detection is carried out, and obtain abnormality degree set;Calculate dispersion and adopt screening strategy to optimize abnormality degree set, and obtain target abnormality degree set;According to target abnormality degree, the electric quantity abnormality degree of each user terminal in sub-set is determined.The application realizes the accuracy and high efficiency of electric quantity abnormality monitoring by the synergistic mechanism of double-layer classification, representative selection, dynamic detection, dispersion optimization and result inference.

The invention relates to the technical field of legal artificial intelligence, and discloses a legal contract term AI examination method, which comprises the following steps: acquiring a contract to be examined, historical contract data and user game parameters, analyzing a contract text into a structured graph containing legal functional units, generating vectorized representation, and calculating the evolution distance between historical contracts to obtain the legal contract term AI. The method comprises the following steps: constructing a historical contract evolutionary tree structure, combining an evolutionary tree, a commercial risk scene and user game parameters, performing simulation deduction by applying a genetic algorithm, evaluating advantages and disadvantages of various strategy combinations through a payment function, guiding a mutation direction by utilizing evolutionary fitness scores of historical contracts, and searching an optimal strategy combination. And by comparing the difference between the optimal strategy combination and the original contract, an accurate modification suggestion or a complete contract draft is generated. According to the method, evolutionary intelligence of historical contracts is combined with game theory simulation, the limitation of static and template review in the prior art is overcome, and the strategic and personalized level of review is improved.

The application discloses a media creation task distribution method and system based on multi-dimensional feature matching, belonging to the field of data processing. The method first receives a multi-modal task requirement containing a text report, a reference material set and a structured constraint, and generates a multi-dimensional structured task feature vector group through joint semantic understanding and feature decomposition. Relying on the extraction of features from the author's historical works and the calculation of the historical originality index, a static capability portrait is constructed. Through the release of style transfer innovation challenge tasks, the creation output is evaluated and the innovation index is calculated to generate a dynamic capability vector. Combined with the task feature vector, the author's static portrait and dynamic vector, the original adaptation score is obtained through the prediction network, and the risk is corrected according to the two types of original innovation indexes to generate an author ranking recommendation list. In the task execution, the similarity of the creation content, the reference material and the copyright library features is compared in real time, and a hierarchical copyright guidance prompt is triggered, which can improve the efficiency of media creation task distribution and control the copyright risk.

The invention relates to an intelligent question and answer interaction method and device for psychological health education, equipment and a medium. According to the method, historical dialogue data of a user is collected and subjected to text preprocessing and serialization, and an instant emotion vector is extracted based on a pre-training language model to capture current emotion expression of the user; a potential emotional state is decoupled from a dialogue history by using a time sequence model constructed by a variational auto-encoder and a recurrent neural network, a cross-dialogue evolution rule of the potential emotional state is tracked to generate a state vector sequence, a historical emotional baseline is calculated on the basis, and a risk detection threshold and a response generation strategy are dynamically adjusted through statistical deviation analysis; finally, a personalized response is generated in combination with real-time user input and the emotional state vector, continuous tracking of the long-term emotional state of the user and personalized security strategy self-adaption in a pure dialogue environment independent of biological signals are achieved, and the accuracy of emotion recognition and the security protection capability of a mental health interaction system are effectively improved.

The present application relates to the field of cable temperature monitoring, more particularly, the present application relates to a new energy load cable temperature monitoring method and system. The method comprises: obtaining a temperature sequence of a preset period; calculating a first-order difference sequence of the history sequence, iteratively segmenting the first-order difference sequence, constructing a segmentation evaluation function, obtaining a plurality of sub-difference sequences, and constructing a network model for any sub-difference sequence; inputting the to-be-predicted sequence into any network model to output the corresponding temperature prediction value, taking the importance of the calculated network model as the weight, calculating the product of the weight and the temperature prediction value, taking the sum of the products of each network model as the final temperature prediction value, and completing the temperature monitoring. Through the technical scheme of the present application, the accuracy of temperature prediction can be improved, and the response speed and stability of temperature monitoring can be improved.

The present invention provides a method, program, and apparatus for efficiently compressing historical data, which is data recording the past price movements of financial instruments, as well as a method, program, and apparatus for decompressing compressed historical data. [Solution] Various differences are calculated from the preceding and succeeding records of the historical data (however, to ensure that data close to 0 appears as frequently as possible, the difference between the opening price and the closing price of the previous data is calculated for the opening price, and the difference between the high price, low price, and closing price and the opening price of the current data is calculated for the high price and the closing price). The numerical values ​​related to price are converted to integers, and for integers that may be negative, both positive and negative numbers close to 0 are mapped to non-negative integers close to 0. By representing the non-negative integers with the fewest possible octets, the historical data can be represented with less data than before.

The application provides a simulation method and system of a dual active bridge converter with an invariant circuit admittance matrix, wherein the method comprises the following steps: equivalent the dual active bridge converter to an equivalent circuit in which each branch is a parallel branch of a resistor and a historical current source; calculating a system admittance matrix according to each circuit parameter; checking a switching change condition, and calculating the historical current source according to circuit information of a previous time step; and solving a new node voltage according to a system node voltage equation, and then solving a branch voltage and a branch current until iteration is completed. The application calculates the system admittance matrix which does not change with a system state, and reflects the switching change in the equivalent current source of the switching branch. In the simulation process, only the equivalent current source of the switching branch needs to be updated, the overhead of recalculating the system admittance matrix is avoided, and a large memory required for precalculating a large number of system admittance matrices is also avoided.

The invention belongs to the technical field of optimization scheduling, and provides a source network load storage cooperative control method and system based on learning type adaptive model prediction control, and the method employs a learning type prediction method to continuously calculate a historical prediction error, and is used for optimizing an actual value. Obtaining instantaneous power surplus according to the optimized actual value; obtaining a dynamic electricity price signal according to the instantaneous power surplus and the reference electricity price; according to the dynamic electricity price signal, establishing a mathematical model comprising a source network load storage unit, and solving the mathematical model to carry out optimization control by taking the maximum operator income as an objective function and taking the power balance constraint, the energy storage SOC dynamic constraint and the abandoned electricity quantity constraint as constraint conditions; according to learning type prediction correction, source network load storage collaborative optimization is realized, users are guided to adjust loads through dynamic electricity prices, renewable energy sources and energy storage utilization rates are improved, and the operator income of source network load storage element-containing projects is maximized.

An information processing device (3) is provided with: a storage unit (31) that stores input data (D10) to be optimized; and a control unit (30) that, on the basis of the input data, executes optimization calculation in which candidates for a solution optimized for one or more management indexes are repeatedly calculated. The input data includes constraint conditions for limiting the solution optimized in the optimization calculation. The control unit stores, into a storage unit, a calculation history including a solution violating the constraint conditions among a plurality of solutions calculated as the candidates in the optimization calculation (S12), and generates a plan in which at least a part of the management indexes is improved from the result of the optimization calculation by referring to the stored calculation history (S14).

A deep reinforcement learning optimization compensation method for nonlinear intermittent processes, which expands a three-dimensional input data matrix into a two-dimensional matrix; performs standardization processing; constructs a JY-KPLS model; constructs a JY-KPLS model optimization problem; solves the optimization problem; calculates the similarity of historical and query data; according to the similarity, m old data and n new data are selected from the old and new process data sets respectively, and the difference is calculated with the current query data; the deviation sample is used as the data set to establish a JITL-JYKPLS local model to solve the mismatch problem; the compensated model and the optimization system are interacted and trial-and-error trained; if the total reward value of the current batch training exceeds the total reward value of the previous batch training, the optimized system of the current training is used for batch-to-batch optimization; otherwise, the optimization system of the previous batch is used for batch-to-batch optimization; the final product quality is output. The method can significantly improve the quality of the final product.

The intelligent power grid operation and maintenance monitoring method and system based on artificial intelligence provided by the present invention relate to the technical field of data processing. In the present invention, data extraction processing can be first performed on a target power grid component to obtain historical power grid operation data corresponding to the target power grid component; then, the data correlation between the historical power grid operation data and each reference power grid operation data included in a pre-configured reference power grid operation data set is calculated respectively; finally, based on the data correlation between the historical power grid operation data and each reference power grid operation data, and in combination with the abnormality degree reference value pre-configured for each reference power grid operation data, an abnormality degree characterization value corresponding to the historical power grid operation data is analyzed and output, and then the target operation safety degree corresponding to the target power grid component is determined based on the abnormality degree characterization value. Based on the above content, the reliability of operation safety analysis can be improved to a certain extent.

The application discloses a load curve decomposition method and system based on curvature equalization rational B-spline basis functions, and the method comprises the following steps: calculating the curvature of a historical load sequence, generating a continuous curvature function, and determining a set of B-spline node sequences meeting the cumulative absolute curvature equalization condition according to the continuous curvature function, so that the node distribution is adaptive to the load fluctuation intensity; generating B-spline basis functions based on the node sequences, and constructing rational B-spline basis functions by optimizing the weights, so as to finely depict the continuous time fluctuation mode; and combining time period virtual variables to establish a fitting model, and decomposing the load curve into two parts, i.e., a 'time period characteristic load level' representing the reference electricity intensity in different time periods and a 'time characteristic load curve' representing the continuous change fluctuation mode. The method overcomes the defect that the traditional uniform spline has poor adaptability to non-uniform load forms, realizes adaptive and clear physical meaning decomposition, and provides a higher-quality data basis for load analysis and prediction.

The invention relates to the technical field of human resources, in particular to a human resource demand prediction method and system, and the method comprises the steps: collecting internal data and external environment data of an enterprise, and obtaining standardized feature data after processing; a dual-channel weight calculation module constructed based on an attention mechanism respectively calculates a dependency weight of historical time sequence data and an influence weight of external environment data, and fuses the dependency weight and the influence weight to obtain a total weight of each data; based on the standardized feature data and the total weight of each data, a preset basic time sequence model is adopted to capture a periodic trend, an enhanced neural network is combined to learn a long-term and short-term dependency relationship, and meanwhile, a causal inference engine is constructed to quantify causal association between variables to obtain an initial demand prediction result; the dynamic error between the initial prediction result and the actual demand is calculated, when the error exceeds a preset adjustment threshold value, incremental learning is triggered to update the model parameters, and the calibrated demand prediction result is obtained, and the method has the advantages of being high in dynamic adaptability and high in prediction precision.

A processing device obtains an input (302), where the input specifies a set of devices to be placed and routed for a circuit design. In response to the input, the processing device executes a machine learning model (304) to calculate a probability distribution function on a historical device placement library, the probability distribution function estimates the suitability of each historical device placement in the library of historical device placement for placing and routing the set of devices specified in the input. The processing device presents (306) a graphical representation of a defined number of historical device placement from the library of historical device placement, the historical device placement being estimated to be suitable for placement and routing of the set of devices based on the probability distribution function.

The application provides a shared energy storage resource configuration method and device, computer equipment and medium, and belongs to the field of energy configuration. The method comprises the following steps: obtaining a power demand prediction value sequence and a capacity demand prediction value sequence in a continuous unit time length according to a multiple relationship between a unit time length of a load side resource regulation system participating in operation and a minimum selling time length of a shared energy storage right, and then determining a power configuration amount and a capacity configuration amount; calculating a tail risk quantitative value of a yield distribution in a historical market scenario, and taking a weighted combination of the tail risk quantitative value and an expected yield as a target function to be optimized, so as to maximize the target function to obtain a power calling price and a capacity calling price of the load side resource regulation system in a time period corresponding to the minimum selling time length. Thus, the peak power demand in the entire service period can be covered, the operation risk caused by insufficient power configuration can be avoided, and the stability and economy of system operation are enhanced.