Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

123 results about "Problem transformation" patented technology

Data processing method and equipment

The invention relates to the technical field of data processing, in particular to a data processing method and equipment. In the embodiment of the invention, a target problem input by a user is solved through a multi-agent mode, a task planning agent can independently think and understand the requirements of the user and split service analysis logic according to target information and the target problem, and the target problem of a natural language is converted into each to-be-executed sub-task; and then the code generation agent is called to generate a code corresponding to each subtask, and each code is executed to obtain a target processing result for the target problem, so that the efficiency and accuracy of data processing are improved.
Owner:新奥新智科技有限公司

Content generation, model construction, and data processing

PCT designated stage expiredWO2025045093A1Natural language data processingData packDialog system
Provided are a content generation method and apparatus, a model construction method and apparatus, and a data processing method and apparatus, a device, and a medium. The method comprises: in response to a dialogue initiated by a user, obtaining dialogue data (202); when this round of dialogue is the first round of dialogue, the dialogue data being a current question of the user in this round of dialogue; when this round of dialogue is not the first round of dialogue, the dialogue data comprising a current question of the user and previous historical questions of the user; inputting the dialogue data to an independent question transformation model, so that the independent question transformation model transforms the dialogue data into an independent question that represents a user intention of the user in this round of dialogue (204), wherein training samples used by the model comprise a plurality of rounds of dialogue samples and an independent question sample, and the independent question sample indicates a user intention of a latest round of dialogue sample among the plurality of rounds of dialogue samples; and inputting the transformed independent question to a multi-round dialogue system, so that the multi-round dialogue system generates reply content for this round of dialogue on the basis of the independent question (206). The reply accuracy of the multi-round dialogue system can be improved.
Owner:RAJAX NETWORK &TECHNOLOGY (SHANGHAI) CO LTD

Intelligent question and answer inference system based on knowledge graph

The invention belongs to the technical field of intelligent question-answering systems, and particularly relates to an intelligent question-answering inference system based on a knowledge graph, which is characterized in that firstly, a knowledge graph construction module fuses multi-source data to generate a structured graph, and after a user inputs a natural language question, a question-answering analysis module completes intention classification and entity disambiguation and converts the question into structured query; an inference engine module fuses symbol rules and graph neural network inference through a hybrid inference sub-module, and a dynamic weight adjustment sub-module optimizes weights according to errors and attenuation factors to generate an inference result; the knowledge updating module incrementally updates the atlas in real time and detects conflicts, the interactive interface module visually presents a result, and the evaluation optimization module iteratively optimizes parameters in combination with offline evaluation and online feedback. The whole process is from user question asking to result output, accurate reasoning and continuous performance improvement are achieved, and multi-field question and answer requirements are met.
Owner:SICHUAN JOYOU DIGITAL TECH CO LTD

Method and device for extracting overlapping peaks based on mode decomposition

InactiveUS20240393177A1Image analysisSpectrum investigationAugmented lagrange multiplier methodOptimization problem
The present disclosure discloses a method and a device for extracting overlapping peaks based on mode decomposition, which includes: obtaining a problem formula for finding peak values according to an between-class variance function representation, adding a penalty term and a Lagrange multiplier to the function by using an augmented Lagrange multiplier method, and transforming an optimization problem formula containing two variables and one constraint condition into an unconstrained extreme value problem containing three variables; performing quadratic optimization on the unconstrained extreme value problem and transforming the unconstrained extreme value problem into a minimization problem formula; setting a convergence condition according to the minimization problem formula, updating three of the variables in the minimization problem formula, and stopping iteration until the preset convergence condition is met; wherein a final calculation result is peak values of two spectral axial response signals when stopping iteration.
Owner:OPTOSKY (XIAMEN) PHOTONICS INC

Distributed elastic consensus optimal control method under denial of service attack of multi-agent system based on zero-sum game

The invention provides a zero-sum game-based distributed elastic consensus optimal control method under denial of service attack of a multi-agent system, and the method specifically comprises the steps: firstly, constructing a multi-agent formation model in which a leader and a plurality of followers cooperatively move through graph theory knowledge and a multi-agent second-order state equation; secondly, in order to reduce the influence of denial of service attack on communication topology, a time-varying weight distributed elastic observer is provided to estimate the state of a leader, and the attacked condition of the leader is considered; then, by constructing an augmentation system, a distributed consistency tracking problem with a leader is converted into a local tracking problem between each follower and a virtual leader thereof; and finally, in order to solve the zero-sum game problem, introducing a Hamiltonian-Jacobian-Ansaxophone equation to realize optimal control input under maximum external disturbance, and realizing algorithm design by using single-evaluation reinforcement learning with experience playback and combining a gradient descent method.
Owner:WUHAN TEXTILE UNIV

Intelligent curriculum schedule arrangement method and system

The invention relates to the technical field of artificial intelligence, in particular to an intelligent curriculum schedule arrangement method and system, and the method comprises the following steps: defining teacher nodes, student nodes, curriculum nodes and classroom nodes based on the information of teachers, students, curriculums and classrooms, building the relation between the nodes, analyzing the time conflict, burden and classroom capacity, endowing each edge with a corresponding weight, and carrying out the arrangement of the curriculum schedule. And obtaining graph structure data. According to the method, the information node relation network is constructed, the course arrangement problem is converted into graph structure data processing, teacher arrangement, student demands, course capacity, time conflicts and other constraint conditions are quantified, node states are updated layer by layer, the course arrangement influence is accurately evaluated, the optimization efficiency is improved, constraint weight distribution is dynamically adjusted, key contradictions are preferentially solved, and the course arrangement efficiency is improved. According to the method, balance and rationality are considered, a curriculum schedule is adjusted in combination with real-time data, student course selection and teacher time change are rapidly adapted, resource waste and insufficient coordination of static arrangement are avoided, and an efficient and flexible curriculum arrangement scheme is realized.
Owner:JIANGSU VOCATIONAL COLLEGE OF BUSINESS

Quantum-classical multi-stage optimization method for unit combination and transmission line switching

The invention provides a quantum-classical multi-stage optimization method for unit commitment and transmission line switching, and the method comprises the steps: decomposing an original model into a plurality of modules, introducing consensus variables between nodes into the modules, further decomposing the modules into a plurality of sub-problem models, and coordinating the unit power generation and transmission line switching through the external circulation between the modules and the internal circulation of the modules, and the optimal unit scheduling scheme of the power system and the corresponding optimal power grid topology can be obtained. According to the method, a main problem for optimizing discrete variables is converted into an equivalent Hamiltonian in a secondary unconstrained binary optimization form suitable for quantum calculation, the Hamiltonian is efficiently solved through a quantum computer, and the solving speed is increased; the modules of the method are subjected to efficient iterative interaction through cutting plane Hamiltonian, so that the global search capability is improved; according to the method, a sensitivity analysis theory is applied to the construction of the cutting plane, and the effectiveness of the cutting plane and the convergence of the method are ensured; the optimal unit scheduling and the power grid topology can be quickly searched, and the power generation energy consumption is reduced.
Owner:GUANGXI UNIV

Method for identifying key node set under cascade failure of marine transportation network based on multi-population genetic algorithm

The invention belongs to the field of combinatorial optimization, and relates to a heuristic algorithm combined with global marine transportation network priori knowledge, which is used for solving a key port set problem in a global marine transportation network, and comprises the following steps: modeling a marine transportation network, modeling a marine transportation network cascade failure phenomenon, and carrying out optimization on the marine transportation network cascade failure phenomenon; a key port gathering problem of the marine transportation network is converted into a single-target optimization problem; a multi-population genetic algorithm framework is adopted, priori knowledge of the marine transportation network is fused into initial population generation of all sub-populations, and therefore initial sub-populations with high quality are formed; a simulated annealing mechanism is introduced in the algorithm iteration process to improve the algorithm performance and enhance the search capability. The method is superior to a current mainstream heuristic method and an index method on a sea transportation data set, and an effective reference is provided for stable operation of a sea transportation network.
Owner:DALIAN UNIV OF TECH

Knowledge Distillation Via Learning to Predict Principal Components Coefficients

Provided is an approach for knowledge distillation based on exporting Principal Components approximations (e.g., Bregman representations) of one or more layer-wise representations of the teacher model. In particular, the present disclosure provides an extension to the original Bregman PCA formulation by incorporating a mean vector and orthonormalizing the principal directions with respect to the geometry of the local convex function around the mean. This extended formulation allows viewing the learned representation as a dense layer, thus casting the problem as learning the linear coefficients of the compressed examples, as the input to this layer, by the student network. Example empirical data indicates that example implementations of the approach improve performance when compared to typical teacher-student training using soft labels.
Owner:GOOGLE LLC

Earth-rock multi-target intelligent allocation method, equipment, medium and program product

The invention relates to the field of earth-rock engineering scheduling, in particular to an earth-rock multi-target intelligent allocation method and device, a medium and a program product. The method specifically comprises the following steps: S1, collecting and sorting basic data; s2, constructing an earthwork allocation mathematical model; s3, using a sparrow search algorithm to optimize the earth-rock allocation scheme; and S4, scheme implementation and dynamic adjustment. The method comprises the following steps: firstly, constructing a mathematical model of earth-rock deployment, taking the shortest transportation time and the shortest transportation path as objective functions, comprehensively considering constraint conditions such as earth-rock supply and demand balance, transportation vehicle load limitation and construction progress requirements, and converting an actual deployment problem into a mathematical planning problem; and then, a sparrow search algorithm is introduced to carry out intelligent deployment, an optimal deployment scheme is obtained, efficient deployment of earthwork in time and path dimensions is realized, and a scientific and reasonable earthwork deployment decision scheme is provided for engineering construction.
Owner:CCCC FOURTH HARBOR ENG CO LTD +1

Knowledge base-oriented interactive causal inference support method and system

According to the knowledge base-oriented interactive causal inference support method and system provided by the invention, a user causal analysis intention is precisely limited through interactive dialogue, a multi-dimensional intention is precisely captured, a fuzzy causal problem is converted into a computable analysis parameter, and ambiguity is remarkably reduced; multiple strategies such as an explicit causal declaration mode, time sequence correlation, statistics of significant correlation and comparative difference analysis are integrated to collect causal evidence, the four strategies are executed in parallel, multiple evidence sources of causal inference are covered, and limitation of a single method is avoided; and the query efficiency is optimized in combination with a special index and a graph algorithm, and finally evidence is presented in a structured and interactive mode and confidence is clearly distinguished, so that the response efficiency is greatly improved. The system solves the problems of fuzzy intention, single evidence and low efficiency in traditional causal inference, is suitable for the fields of industrial fault analysis, medical diagnosis, social phenomenon research and the like, and provides scientific and transparent causal hypothesis generation support for users.
Owner:SHANGHAI JEINTAI INFORMATION TECHNOLOGY CO LTD

Incremental cooperative mechanical arm multi-target trajectory optimization method based on trust region and corresponding electronic equipment

The invention discloses an incremental cooperative mechanical arm multi-target trajectory optimization method based on a trust region and corresponding electronic equipment. The incremental cooperative mechanical arm multi-target trajectory optimization method comprises the steps that S1, data are acquired; s2, establishing a model; S3, converting and solving a problem; S4, dividing a track and performing optimization execution; according to the method, the high efficiency and practicability of the cooperative mechanical arm in the complex environment are ensured through the multi-dimensional optimization model, the improved trajectory optimization method based on the trust region is adopted, and improvement is achieved in the aspects of reducing the calculation time, optimizing the trajectory length of the joint space, improving the trajectory smoothness and the like; according to the method, an incremental trajectory optimization framework is used, the environment information can be sensed in real time, and the motion information of the human body is fused into the trajectory optimization process in real time, so that the man-machine safety can be guaranteed, and the trajectory optimization quality and the calculation efficiency are effectively improved; the trajectory optimization method disclosed by the invention is superior to the prior art in processing man-machine cooperation tasks in complex scenes, and the application of the cooperation mechanical arm is perfected.
Owner:SHANGHAI UNIV

Multi-agent formation obstacle avoidance method based on distributed model predictive control

The invention discloses a multi-agent formation obstacle avoidance method based on distributed model predictive control, which obtains an optimal control sequence of a system by solving an optimization problem in a limited predictive time domain so as to realize online closed-loop control of the system in the whole control time domain. A leader agent moves according to a preset track, other agents consider consistency performance indexes between the leader agent and other agents in a neighborhood, a control target is constructed, then a target function is constructed based on the control target, and the position and speed consistency performance indexes serve as the target function. The collision avoidance requirements and the obstacle avoidance requirements between the intelligent agents are constructed into constraint conditions; according to the method, an obstacle avoidance problem is converted into an optimization problem of a target function, and then a distributed model prediction control method is used for solving the optimization problem to obtain an optimal control sequence, so that an expected formation obstacle avoidance function is realized, the system control performance is improved, the calculation complexity is reduced, and the method has relatively good practicability.
Owner:CHENGDU AIRCRAFT INDUSTRY GROUP

Subgraph-based temporal knowledge graph question answering method

The present invention discloses a subgraph-based temporal knowledge graph question-answering method, including: extracting and parsing the time information hidden in the question through the knowledge in the temporal knowledge graph, replacing the question in natural language format with a simplified question by using a regular expression; converting the question with time constraints into a vector representation, using the temporal knowledge graph scoring function and the time-sensitive function to obtain candidate entities and their semantic scores; constructing a temporal neighbor subgraph for each question; trimming the temporal neighbor subgraph by time constraints; using a time activation function to quantitatively score each entity in the temporal neighbor subgraph to obtain a subgraph reasoning score; fusing the score semantic score and the subgraph reasoning score to obtain the final answer. The present invention improves the ability to identify the time information contained in the question; and obtains a reliable answer to the question by reasoning the answer that meets the time constraint through the temporal subgraph.
Owner:NAT UNIV OF DEFENSE TECH

Metasurface multi-beam rapid generation method based on semi-supervision

The invention discloses a metasurface multi-beam rapid generation method based on semi-supervision, and the operation process of the method comprises the following steps: 1, introducing an attention mechanism based on a convolutional neural network, and designing a cross entropy loss function; step 2, training the model by using a large number of samples obtained by caliber field superposition, migrating trained model parameters to a new model, and training based on the samples optimized by a traditional algorithm; step 3, semi-supervised learning is introduced, a model obtained through initial training is utilized to label unlabeled data samples, labeled sample data is added to label data samples, then the network model is retrained, the process is continuously repeated, and the number of the samples is iteratively increased; and 4, inputting required directional diagram data to obtain a corresponding coding matrix. Based on the method, a multi-beam coding problem is converted into a multi-label classification problem, the condition of repeated traversal in a codebook set in a traditional algorithm is avoided, and the time complexity is reduced.
Owner:BEIJING UNIV OF POSTS & TELECOMM

Mechanical arm model parameter identification method and system based on deep learning

The invention relates to a mechanical arm model parameter identification method and system based on deep learning, and the method comprises the following steps: 1, converting a parameter identification problem of a space mechanical arm kinetic equation into a deep learning parameter identification problem, and designing a feedforward network approximate mass matrix; step 2, constructing a loss function to optimize a model parameter deep learning process of a mechanical arm kinetic equation; 3, realizing structure description of a network layer in deep learning; 4, performing fractional order optimization on the triangular matrix obtained in the step 1, and designing an activation function for the deep learning part; and 5, evaluating the performance of the deep learning algorithm on the mechanical arm parameter identification problem. According to the method, a system identification parameter set is simplified, and the number of identification parameters is reduced, so that the convergence speed of deep learning training is improved, the learning efficiency is enhanced, the high efficiency and precision of an algorithm are ensured, and the identification precision and generalization ability are improved.
Owner:HARBIN INST OF TECH +1

Data analysis method and device, storage medium and processor

The invention discloses a data analysis method and device, a storage medium and a processor. In the scheme, description information of an insurance service problem is obtained, and the description information is converted into a data query vector; retrieving data matched with the data query vector in a corresponding knowledge base based on a plurality of preset recall channels to generate a retrieval result; converting the insurance business problem into a standard insurance business semantic expression by the large language model according to the data caliber to obtain target problem description information; generating a corresponding target SQL query statement by the large language model according to the target problem description information, the SQL template and the database table structure, and querying matched insurance service data based on the target SQL query statement; and performing data analysis on the insurance service data by the large language model according to the target problem description information and the data analysis template to obtain an analysis result. According to the technical scheme, the accuracy of insurance service data analysis is remarkably improved through deep collaboration of a multi-path recall mechanism and a large language model.
Owner:ABC FINANCIAL TECH CO LTD

DIKWP semantic modeling method for complex problems of enterprises

The invention provides a DIKWP semantic modeling method. Complex problems of an enterprise are converted into a five-layer linkage semantic map. The method comprises the following steps: 1, cleaning multi-source heterogeneous data to form nodes by a data layer; (2) information layer construction domain ontology clarification concepts and constraints; (3) the knowledge layer aligns with the knowledge base to infer and complement the relationship to generate a knowledge graph; (4) the wisdom layer outputs scheme elements, benefits and risks based on decision rules; and (5) modeling each intention layer refining strategic target constraint layer. Full-link semantic mapping from data to intention is achieved, consistency and accuracy of problem definition are improved, cross-department implicit causality is mined, it is ensured that a scheme is accurately aligned with a strategic target, an interpretable semantic basis is provided for automatic scheme generation, scene simulation and optimization decision, and the method has remarkable commercial application value.
Owner:HAINAN UNIV

Depth reinforcement learning-based reasoning task scheduling and resource allocation method in vehicle edge intelligent system

The invention discloses a reasoning task scheduling and resource management method in a vehicle edge intelligent cooperative network. The method comprises the following steps: constructing a reasoning task system model and a deep reinforcement learning decision model under a vehicle-RSU-edge cooperative network; constructing a task processing time delay model, an energy consumption model and a reasoning error rate model, and further establishing a long-term system performance optimization problem for comprehensively optimizing the total cost; based on a problem decomposition technology, a complex optimization problem of a task processing decision in the multi-layer heterogeneous network is converted into decomposable sub-problems; constructing a Markov decision process, and converting an original optimization problem into a deep reinforcement learning optimization problem; based on an improved SD3 deep reinforcement learning algorithm, a task reasoning position decision and resource allocation strategy is trained and applied. According to the method, dynamic optimization of task reasoning decision, transmission power distribution, computing resource distribution and transmission rate distribution is realized, task processing delay, energy consumption and reasoning precision are effectively balanced, and the efficiency of a vehicle edge intelligent system is improved.
Owner:BEIJING UNIV OF POSTS & TELECOMM

Micro-grid edge control method and system based on multi-objective collaborative optimization

The invention discloses a micro-grid edge control method and system based on multi-objective collaborative optimization, and belongs to the technical field of micro-grid control. The method comprises the following steps: acquiring micro-grid operation parameters in real time through a multi-source sensor network; performing data preprocessing and feature extraction at the edge computing node; a parameter space is pre-divided based on an optimal segmentation theory to generate a critical region (CR) and a linear decision mapping table thereof, and a millisecond-level initial decision is realized in a table look-up mode; constructing a dynamic weighted optimization function aiming at economy, safety and environmental protection to perform collaborative optimization; a lightweight deep Q network (DQN) is adopted to carry out on-line fine tuning on the mapping relation; implementing a hierarchical response strategy according to a comprehensive performance index (DPI); and updating the strategy through a cloud edge cooperation mechanism and realizing network disconnection fault tolerance. According to the method, the complex optimization problem is converted into low-delay table lookup operation on the edge side, the problems of high centralized control delay, local strategy stiffness, uneven resource allocation and the like are effectively solved, and efficient, real-time and stable operation of the microgrid is achieved.
Owner:GUIZHOU XIANGBIN NEW ENERGY TECHNOLOGY CO LTD

Interactive question answering method and system based on artificial intelligence

The invention discloses an artificial intelligence-based interactive question-answering method and system, and relates to the technical field of interactive question-answering, and the method comprises the steps: converting a user input text question into vector representation, constructing a relation graph by combining a dependency relation tree of a text entity, capturing a dependency relation in the question as an edge weight, forming an adjacent matrix, and constructing a node degree matrix; and determining a standardized graph Laplacian matrix, performing feature decomposition, determining an optimal clustering number, clustering language phrases of the text question, and determining task fragment sets of different clusters. According to the method, semantic association between language phrases is achieved by constructing the dependency relationship graph, meanwhile, syntactic relationships are fused, the integrity of text structure information is ensured, the degree of nodes serves as a quantitative index of local strength in the graph, the information directly supports subsequent Laplacian matrix normalization processing, a unified scale is provided, and the method is suitable for being applied to the field of text processing. And the imbalance problem of the data range is avoided.
Owner:HANGZHOU LUXIANG TECH CO LTD

Data-driven enhanced optimization solving method for security constraint unit commitment problem

The invention discloses a data-driven enhanced optimization solving method for a security constraint unit commitment problem. The method comprises the following steps: constructing a security constraint unit commitment problem model of a power system; historical operation load data of a power system is collected, a confidence data set is constructed by using a Turing statistical method, and the confidence data set is divided into a training set and a verification set according to a proportion; constructing a neural network model, and inputting the training set and the verification set for network training; new load data of a power system is used as input, a multi-layer neural network is used as a classifier, and a logic strategy of unit operation is predicted online; by eliminating integer variables, an SCUC problem is converted into a continuous convex optimization problem, and by sorting a plurality of candidate integer solutions, feasibility of the model is simplified and optimized, and rapid online solution of the method is realized. According to the method, a high-quality approximate optimal solution can be provided, the solving speed is high, and the expandability is high. Not only can efficient power system scheduling optimization be realized, but also the peak regulation capability and the emergency response capability of the power grid can be enhanced.
Owner:STATE GRID LIAONING ECONOMIC TECHN INST

Multi-frame joint detection method based on graph neural network link prediction task

The invention provides a multi-frame joint detection method based on a graph neural network link prediction task, and the method comprises the steps: carrying out the correlation test of observation data passing a first threshold through the priori knowledge of target physical motion, constructing a trace point correlation graph, and carrying out the detection of a multi-feature link prediction network based on the graph neural network design. The multi-feature link prediction network integrates the echo structure of single-frame observation, Doppler, space-time coupling of multi-frame observation and other features, and completes a multi-frame detection task. According to the method, a multi-frame detection problem is converted into link prediction in a graph, track search and track judgment are integrally completed, the target track can be directly observed and output from multiple frames to simplify processing logic, and track judgment can be more accurately carried out while track search performance loss is reduced. Compared with a traditional single-frame and multi-frame detection method, the method has the advantages that false trace points can be effectively suppressed, and meanwhile, the detection performance of the target is improved.
Owner:XIDIAN UNIV

DeepSeek-R1-based production scheduling optimization method

The invention relates to a production scheduling optimization method based on DeepSeek-R1. The method is used for improving modeling and coding efficiency when a production scheduling optimization problem is solved. According to the method, natural language processing and code generation capabilities of DeepSeek-R1 are introduced into a production scheduling task, and a scheduling optimization process including three stages of cue word design, modeling solution and iterative optimization is constructed. In the cue word design stage, a production scheduling problem is converted into structured cue word input including problem hypothesis, constraint conditions, optimization targets and the like, and a model is guided to automatically extract problem elements; in the modeling solving stage, DeepSeek-R1 automatically generates solving codes according to cue words to obtain an initial scheduling scheme; and in the iterative optimization stage, the user obtains a final scheme conforming to the problem hypothesis and constraint through check-feedback circulation based on the initial scheduling scheme. The method provided by the invention is finally applied to an island type assembly line scheduling optimization case, and the effectiveness of the method is verified.
Owner:BEIJING INST OF TECH

SAR (Synthetic Aperture Radar) ship target detection system and method for efficient neural architecture search

The invention discloses an SAR ship target detection system and method for efficient neural architecture search, and belongs to the technical field of SAR ship target detection, and the system comprises a super-network architecture module, a multi-path search module, a PANet module and a model construction module. The super-network architecture module is used for constructing an efficient super-network framework based on weight sharing and convolution kernel compounding by reconstructing a convolution kernel; the multi-path search module is used for constructing a multi-path search structure and converting a discrete multi-path selection problem into a continuous differentiable optimization problem through a Gumbel-Softmax function; the PANet module constructs a searchable PANet structure by unifying a backbone network and a PANet architecture search space to a joint optimization framework; the model construction module constructs a target detection model based on an efficient super-network architecture, a multi-path search structure and a searchable PANet structure, and performs SAR ship target detection by using the target detection model.
Owner:ANHUI UNIV +1

Unit commitment optimization method based on solution space expansion and space-time diagram modeling

The invention provides a unit commitment optimization method based on solution space expansion and space-time diagram modeling. A space enhancement strategy and a diagram neural network modeling technology are innovatively combined. According to the method, disturbance optimization and diversity constraints are introduced, a training data set containing rich approximate optimal solutions is constructed, and the generalization ability of the model is remarkably improved. Meanwhile, the unit commitment problem is converted into a space-time dependency graph structure, the sequential relation and structural constraint between variables are effectively extracted through a graph neural network, and deep modeling and prediction of the optimization problem are achieved. Furthermore, by designing a double-order constraint repair mechanism, feasibility correction and fine adjustment are carried out on a prediction result, and the feasibility and scheduling precision of a solution are guaranteed. The finally constructed scheduling framework has good real-time performance and deployability, and can provide efficient and accurate combined scheduling decision support for a large-scale power system.
Owner:STATE GRID JIANGSU ELECTRIC POWER CO LTD NANJING POWER SUPPLY COMPANY

Multi-UAV cooperative task allocation method, system and medium

The present application relates to the field of drone control technology and discloses a method, system and medium for cooperative task allocation among multiple drones. The method comprises: S1, defining the problem of cooperative task allocation among multiple drones; S2, solving the optimal control strategy for the problem of cooperative task allocation among multiple drones; S3, solving the dual optimization problem based on the primal-dual theory; S4, solving the convex dual optimization problem using the Q-function and Schur Complementation theory transforms the convex dual optimization problem into a semidefinite programming problem. S5, based on the properties of matrix congruence, transforms the semidefinite programming problem into a model-free semidefinite programming problem. S6, using a solver, obtains the optimal control strategy. S7, by varying the weight coefficients and repeating S2-S6, obtains the optimal energy loss function bound. This application, based on the Q-learning method, requires only a small amount of data collection, not a precise dynamics model or extensive data, to obtain the optimal strategy for multi-UAV cooperative task allocation.
Owner:QINGDAO UNIV OF TECH

Spectral element tearing and splicing region decomposition method for embedding multi-pole perfect matching layer technology

The invention discloses a spectral element tearing and splicing region decomposition method embedded with a multi-pole perfect matching layer technology, and the efficiency and precision of seismic wave forward modeling simulation are improved through the fusion of the multi-pole perfect matching layer technology and a region decomposition strategy. The time domain convolution of the multi-pole perfect matching layer is eliminated through a recursion auxiliary equation, the boundary truncation calculation complexity is reduced, glancing waves and long-distance propagation waves are efficiently absorbed, and false reflection is restrained. According to the regional decomposition strategy, a global three-dimensional problem is converted into a two-dimensional interface problem through implicit interface constraint and explicit angular point continuity execution, meanwhile, the explicit calculation advantage of a spectral element method on an angular mass matrix is kept, and rapid solution is achieved without matrix inversion. A load balancing mechanism dynamically allocates tasks of a computational domain and a perfect matching layer, and parallel computing efficiency is guaranteed. The system supports high-order interpolation and large-scale grid division of a complex geologic model, appropriately increases memory consumption, improves engineering practicability, and provides an efficient and reliable simulation tool for deep ground resource exploration.
Owner:ANHUI UNIV

Robust approximation method, device and system of Koopman operator and medium

The invention discloses a robust approximation method, device and system of a Koopman operator and a medium, and the method comprises the steps: constructing a Hankel matrix through time delay embedding for a nonlinear system in a noise environment, introducing a correction matrix to carry out the dynamic adjustment of the Hankel matrix, and obtaining a corrected Hankel matrix delta H; the Koopman operator approximation problem containing noise data is converted into a robust optimization problem, and a robust optimization objective function J is designed; and when a real-time data stream arrives, dynamically correcting delta H and re-solving J by utilizing the estimated value KN of the Koopman operator in the previous step and combining an incremental Hankel matrix updating strategy, and iteratively generating a Koopman operator KN + 1 at the current moment. According to the method, the influence of noise on the system is fully considered, high-precision data modeling is performed on the noisy nonlinear system, and modeling, analysis and prediction of the nonlinear system are realized at the modeling cost of the linear system.
Owner:HUBEI CENT CHINA TECH DEV OF ELECTRIC POWER +2

A method and system for event extraction based on graph parsing

The present invention discloses an event extraction method and system based on graph parsing. By treating multiple events included in an input sentence as a whole and linking the multiple events to form an event graph, the problem of extracting events from the input sentence is transformed into a graph parsing method for analyzing and generating an event graph from the input sentence. This method no longer relies on event trigger words, explicitly models the correlation between multiple events, solves the argument sharing phenomenon, and alleviates the long-tail problem. At the same time, an effective decoding algorithm is designed based on the Transformer-based generation model to improve the performance of event extraction. In addition, a pre-trained sequence-to-sequence model is adopted to improve the data sparsity problem. In the generation model based on the event graph, dependency syntactic information is utilized, and a graph attention neural network is used to encode the dependency information, and the dual attention mechanisms of the dependency graph encoding layer and the sentence encoding layer are fused to improve the performance of event extraction.
Owner:NANJING NORMAL UNIVERSITY