42 results about "Algorithmics" patented technology

The present invention relates to the field of pipeline leakage monitoring. Disclosed is an oil and gas pipeline leakage wave identification and monitoring system. In the present invention, an mCNN is combined with LFLBs for performing feature extraction on an acoustic wave signal collected by a DFB, and the collected data improves information completeness; a three-way parallel one-dimensional CNN used in the present invention exhibits good temporal resolution and sensitivity to high-frequency feature transformations in signals; and the present invention integrates advantages of different scales, enabling the algorithm to learn more features, and incorporating the LFLBs to further extract high-level local features. An mCNN-LFLBs network model of the present invention exhibits significant innovation and advancement on the technical level, and also demonstrates extremely high value in actual application. The network model not only provides a novel and efficient technical means for critical fields such as natural gas pipeline inspection, but also introduces new ideas and methods to research fields related to deep learning and signal processing.

The invention discloses a power transmission line engineering quality defect acceptance method based on generative adversarial and reinforcement learning, and relates to the technical field of power engineering quality detection and intelligent image recognition, and the method comprises the steps: carrying out the sample amplification of original image data of a power transmission line tower through a generative adversarial network module, and obtaining an amplified training data set; constructing a PaFPN feature extraction network according to the amplification training data set, extracting multi-scale defect feature vectors and generating a coding feature matrix; establishing a reinforcement learning agent module, taking the coding feature matrix as state input, learning an optimal defect detection strategy through a Q-learning algorithm, and outputting a defect detection parameter combination; and performing feature fusion on the coding feature matrix to generate a fusion feature vector, inputting the fusion feature vector into a classifier network, and outputting a defect category label and a confidence score of the power transmission line project quality. According to the invention, the automation level and the detection precision of power transmission line project quality defect acceptance are improved.

The invention belongs to the technical field of data processing, and discloses an active intelligent operation and maintenance monitoring method for a data center, which comprises the following steps: step 1, component modeling and topology configuration; 2, carrying out distributed health detection and data acquisition; 3, performing real-time health assessment and anomaly detection; 4, performing intelligent alarm and root cause analysis; 5, unified operation and maintenance and closed-loop control are carried out; and step 6, dynamically optimizing the intelligent operation and maintenance strategy. According to the method, a monitoring object and a dependency relationship are clarified through component modeling and topological configuration, and specific scenes, such as multi-mode acquisition, active detection, index pulling, log analysis, coverage message queue theme accumulation and database connection pool exhaustion, of distributed health detection are combined. The quantitative health score is calculated based on the preset scoring model, and by combining with the dynamic baseline learned by the ARIMA or LSTM algorithm, the module abnormity can be actively detected in different periods and weekly updating, and the problems of fault discovery lagging and incomplete monitoring coverage are solved.

The invention discloses a novel power distribution network real-time topology tracing method and system, and the method comprises the steps: carrying out the comprehensive collection of the voltage, current and power of a key node based on a power distribution network SCADA system, and constructing a historical measurement data set containing a topological structure label; designing and training a Transform deep learning model suitable for the characteristics of the power distribution network; the feature space of the Transform model is optimized on the basis of the maximum margin principle; constructing a topological graph model of the power distribution network, carrying out weight assignment, and rapidly tracing and positioning a problem region when a fault or an abnormal condition occurs by using an improved depth-first search algorithm; and constructing a multi-scene topology traceability collaborative decision-making system, and generating a visual traceability report including a fault area topology structure, an influence range and a key equipment state. According to the method, the problems of incomplete data collection, limited learning ability of an identification algorithm, lack of an efficient local tracing mechanism and the like in novel power distribution network topology tracing are solved, and the fault processing efficiency and the operation reliability of the power distribution network are improved.

The invention discloses a wafer defect detection method and a computer program product, and relates to the technical field of semiconductor measurement. The wafer defect detection method comprises the following steps: acquiring a defect image sample with a real defect label or a false positive defect label; the method comprises the following steps: firstly, respectively performing morphological structure analysis on areas to be detected in a defect image sample to determine defect morphological characteristics, performing frequency domain transformation and energy analysis to determine defect texture characteristics and performing boundary line gray gradient analysis to determine defect boundary characteristics, and then integrating the characteristics of the three types of areas to be detected into multi-dimensional characteristics; and learning a mapping relationship between the multi-dimensional features and the defect tags through a decision tree generation algorithm in combination with the defect tags so as to construct a target decision tree model. By means of the mode that the multi-dimensional features cooperatively describe the physical and structural essence of the defects and the decision tree model is combined for classified learning, the real defects and the false positive defects can be effectively distinguished, and the accuracy of wafer defect detection is improved.

The present application discloses a method and device for constructing a digital power system based on multifunctional intelligent agents. The method comprises: respectively extracting feature production element data and control production element data from power data of a power plant side and power data of a power grid side; constructing a target data vector on the basis of a power function requirement, and determining key control production element data associated with the power function requirement; using an artificial intelligence algorithm to learn a mapping function between the target data vector and the key control production element data, constructing a functional operator using the mapping function as a core, and constructing an intelligent agent on the basis of the functional operator; and adding intelligent agents corresponding to a plurality of different power function requirements into a digital power system as power function implementation units. The system can realize data-driven diversified and digital power functions, and implement more accurate capturing of the relationship between data, thereby forming high-value data assets in the power system.

1. The name of the design product: artificial intelligence comprehensive experiment box. 2. The use of the design product: for teaching simulation test algorithm learning and practice, visual detection, etc. Teaching product. 3. The design points of the design product: in shape. 4. The picture or photo that best indicates the design points: perspective view.

The application discloses an intelligent pushing method and system based on intention recognition, relates to the technical field of data processing, and implements the following contents: S10, language information, text information and historical information of a user are collected, and the collected information is mapped to a high-dimensional space; S20, a LightGBM algorithm is used to learn high-dimensional features of the information, and then according to feature importance, unimportant features are removed, and the data dimension is reduced; S30, based on a BiGRU-Attention model, output data of step S20 is processed, and key information reflecting a user intention is output; S40, a demand scheme similar to the key information is found from a database, and the demand scheme with the highest similarity is intelligently pushed to the user. The application can effectively improve pushing efficiency and pushing accuracy.

The present application relates to the technical field of data processing, and especially relates to a box-type transformer substation intelligent integrated design method based on cloud data, which comprises: integrated storage of historical system graph data, component data, cabinet type data and heat dissipation scheme data; through a machine learning algorithm, the correlation between each data after classification and arrangement is learned, and the machine learning algorithm is applied to a prediction model; the prediction model is trained, and according to new input system graph data, corresponding component data, cabinet type data and heat dissipation scheme are output; in the data transmission process, encryption processing is carried out, and the encryption processing adopts a transmission layer security protocol. The present application simplifies the design process, improves the accuracy and speed of the design, optimizes the matching of each component and the heat dissipation scheme, solves the parameter complex correlation problem through systematic data analysis, and can ensure the safety of data in the transmission and storage process, so as to comprehensively improve the design efficiency and quality of the box-type transformer substation.

The invention discloses a navigation method, device and equipment based on a GRPO algorithm and a medium, and relates to the technical field of reinforcement learning, and the method comprises the steps: calculating the average similarity between a current strategy and a plurality of previous iteration strategies based on KL divergence; updating a step length factor through an average reward change rate and an average similarity determined based on a plurality of iterated rewards; determining a gradient estimation correction item based on the gradient estimation of the sampling trajectory, determining target gradient estimation according to the gradient estimation correction item and the original gradient estimation, and updating the current strategy through the target gradient estimation and the updated step length factor; when the current strategy is updated, the importance weight is cut, the target function of the GRPO algorithm is corrected according to the cut weight, the GRPO algorithm is trained based on the corrected function and the updated strategy, so that the intelligent agent learns the optimal strategy based on the trained GRPO algorithm, and the outlet of the labyrinth is determined according to the optimal strategy. Therefore, the stability of the algorithm is improved.

This invention discloses a device type identification method, terminal device, and storage medium based on traffic sampling. It eliminates the need to collect complete device traffic data; device identification is achieved simply by sampling traffic at the gateway. To address the issue of missing traffic features, traditional tensor imputation algorithms learn the embeddings corresponding to tensor rows, columns, and depths, but cannot generalize to unknown rows, columns, or depths. This results in repeated retraining, time-consuming and costly feature imputation. This invention proposes an inductive tensor imputation method that utilizes historical information to learn and generate embedding functions, enabling fast and effective device traffic feature imputation.

Disclosed are artificial intelligence (AI)-based security risk prediction system and method for targets to be protected in a cloud environment. The method includes: collecting cloud logs and system logs for the targets to be protected in real time; learning all activity logs included in the cloud logs and the system logs for the targets to be protected of a corresponding member company through an AI algorithm; identifying a new activity among activities for the targets to be protected based on a learning process through the AI algorithm, and in response to the identified new activity being a new activity related to security, identifying a first activity pattern comprising the corresponding new activity; identifying an order of an preparatory activity for the new activity in the first activity pattern; identifying a risk score corresponding to the order of the preparatory activity for the new activity; and calculating a risk score of each target to be protected by summing identified risk scores of all new activities.

The invention discloses an energy management method and system based on big data, and belongs to the technical field of energy management, and the method comprises the steps: building a digital twin model of an enterprise energy system through a digital twin modeling module, and building a physical and virtual real-time mapping relation; the method comprises the following steps: integrating multi-dimensional data such as energy consumption, business operation, environmental parameters and real-time electricity price through a multi-source data fusion module, and generating a unified energy state feature vector based on an adaptive fusion weight; learning an optimal energy configuration strategy by adopting a deep Q network algorithm through a reinforcement learning optimization module; the energy configuration effect is evaluated in real time through the closed-loop feedback adjustment module, feedback adjustment parameters are generated and transmitted to the front module, the four modules form a deep coupling closed-loop cooperative system, and real-time modeling, self-adaptive optimization and closed-loop feedback of an energy system can be achieved.

Provide a system. 【Solution means】 Means for collecting information from a past contract database and learning the characteristics of fraudulent contracts using a machine learning algorithm; Means for analyzing real-time information transmitted from an information processing device during the contract procedure and interpreting this information using natural language processing technology; Means for obtaining image data of personal identification materials and evaluating their authenticity using image recognition technology; Means for comparing the analyzed data with learned abnormal patterns to detect abnormalities; Means for performing a risk assessment on the detected abnormalities and calculating a risk score; Means for sending a warning to the person in charge based on the risk score and proposing additional confirmation procedures; Means for continuously improving the machine learning model upon receiving feedback; Means for analyzing personal information and identification information in real time when opening an e-commerce transaction, and collating with past patterns of unauthorized use to detect abnormalities; Means for instructing additional personal verification when an abnormality is detected; A system including the above.

This invention discloses an autonomous driving reward learning and control method based on ensemble maximum entropy deep inverse reinforcement learning, comprising the following steps: Step 1: Learning the reward function and control operation in a highway autonomous vehicle driving environment and constructing it into a Markov decision process model; Step 2: Constructing a soft Q-learning model based on value pruning to obtain expert demonstrations, and dividing the inverse reinforcement learning task into sub-tasks according to expert preferences; Step 3: Establishing a strong learner ensemble model and recovering the reward function through maximum entropy deep inverse reinforcement learning; Step 4: Achieving the fusion of reward functions of each sub-task through linear combination, thereby improving the learning accuracy of the reward function. This invention considers the problems of gradient explosion, gradient vanishing, and data overflow in soft Q-learning. It learns expert demonstrations through an improved soft Q-learning algorithm and proposes an ensemble maximum entropy deep inverse reinforcement learning algorithm based on the learned expert demonstrations, which can better achieve decision control.

The application provides a manufacturing network maintenance-detection joint optimization method based on deep reinforcement learning, and the steps are as follows: firstly, for the machine level, a machine reliability model considering the influence of feed quality and a processing quality model considering the influence of machine reliability are constructed under the condition that the dynamic production speed caused by machine failure shutdown is considered; secondly, the system evaluation of the manufacturing network state and performance is carried out based on the reliability model and the quality model; and a manufacturing network maintenance and quality detection joint optimization model is built; finally, at the system level, the economic operation of the manufacturing network is taken as the standard of strategy evaluation, and a deep deterministic policy gradient algorithm is designed to learn the optimal strategy of quality detection and maintenance under the given manufacturing network state. The application can well balance the contradiction between the economic benefits and the operation risks of the manufacturing network, and has better adaptability to dynamic and diversified manufacturing scenes.

The invention relates to the technical field of mixed questions and answers in the financial field, in particular to a text-table complex questions and answers method integrating enhanced example selection and inference enhancement, which comprises the following steps: S1, acquiring a training set; s2, learning an optimal strategy through a Markov decision process in combination with a Q-learning algorithm, and extracting a plurality of candidate examples from the training set to construct an example I; s3, dynamically screening a plurality of candidate examples related to the test example of the current question and answer from the training set through multi-level similarity calculation and integer linear programming optimization to construct an example II; s4, constructing a Prompt template of a test example of the current question and answer based on the example I and the example II; and S5, inputting the test example of the current question and answer and the Prompt template corresponding to the test example into the large language model, and outputting the answer of the question to be answered in the test example through the large language model. According to the method, the robustness of the model on a complex numerical reasoning task can be enhanced, so that the accuracy of text-table complex questions and answers is improved.

A low-orbit satellite network based on DDRL aircraft task offloading method, the steps of which include: (1) constructing a LEO low-orbit satellite network system into a LEO satellite mobile edge computing network system; (2) modeling the LEO low-orbit satellite network system utility maximization problem as a joint decision problem of optimal task offloading and resource allocation; (3) converting the joint decision problem into a Markov decision process MDP, and using a double deep Q network algorithm DDQN to learn the optimal task offloading decision under a double deep reinforcement learning DDRL framework, and using a time difference triple policy gradient algorithm TD3PG to obtain the optimal resource allocation strategy. Simulation experiments show that compared with the benchmark algorithm, the scheme can effectively access and coordinate concurrent tasks, and has better convergence and superiority under different environmental variables.

This invention relates to the field of autonomous driving technology and provides an intelligent quadtree decomposition path planning method for complex dynamic scenarios. The method includes: modeling the quadtree decomposition depth selection problem as a Markov decision process and extracting environmental features as state input; employing an improved Q-learning algorithm to learn the optimal depth selection strategy and constructing a composite reward function that balances planning success rate, path quality, computational efficiency, and depth adaptability; based on the learned strategy, adaptively selecting the quadtree decomposition depth according to environmental features, constructing the quadtree, and performing path planning, while simultaneously combining global planning and local replanning to avoid dynamic obstacles. This invention achieves adaptive matching between quadtree decomposition depth and environmental complexity, solving the problem that a fixed depth cannot adapt to dynamic environmental changes, and significantly improving computational efficiency and robustness in dynamic scenarios while ensuring planning accuracy.

The invention discloses a contract risk retrieval system and method based on intra-group advantages and multi-target distillation, and belongs to the technical field of contract risk retrieval, and the contract risk retrieval system comprises a GRPO contract retrieval strategy optimization module, a large model distillation module and a batch contract processing module. According to the method, corresponding terms of contract signing can be automatically matched, the response speed of law and regulation revision is greatly improved, the lag problem that a traditional rule engine needs long-term reconstruction is avoided, the cross-regional compliance adaptation capability is remarkably enhanced, and the compliance empty window period is effectively eliminated; according to semantic level risk identification, large model semantic understanding and GRPO strategy optimization are cooperated, hidden risks, depending on contexts, of trial period agreed economic compensation calculation and the like can be accurately analyzed, the missed judgment problem caused by lack of semantic understanding of a traditional rule engine is solved, the high-frequency high-risk point identification accuracy is greatly improved, and the labor dispute probability is remarkably reduced.

An artificial intelligence device according to an embodiment of the present disclosure may comprise: a memory for storing an energy prediction model trained through deep learning or a machine learning algorithm; and one or more processors for receiving an energy prediction request including a prediction request time point, and in response to the receiving of the energy prediction request, acquiring, through the energy prediction model, an energy prediction result from the prediction request time point to a time point after a preset time interval.

The application discloses an energy management method and system based on big data, and belongs to the technical field of energy management. The method constructs a digital twin model of an enterprise energy system through a digital twin modeling module, and establishes a real-time mapping relationship between the physical and virtual. Through a multi-source data fusion module, multi-dimensional data such as energy consumption, business operation, environmental parameters and real-time electricity prices are integrated, a unified energy state feature vector is generated based on adaptive fusion weights. Through a reinforcement learning optimization module, a deep Q network algorithm is used to learn the optimal energy configuration strategy. Through a closed-loop feedback adjustment module, the energy configuration effect is evaluated in real time, feedback adjustment parameters are generated and transmitted to the front-end module, and the four modules form a deep-coupled closed-loop collaborative system. The application can realize real-time modeling, adaptive optimization and closed-loop feedback of the energy system.

The application discloses a target detection method based on multi-path component reconstruction residual error, comprising the following steps: step 1: based on the double auto-encoder training algorithm of environmental clutter, learning the clutter structure features in the multi-path channel and reconstructing the multi-path echo; step 2: synthesizing the reconstructed multi-path echo obtained in step 1 into a reconstructed image, based on the target pre-detection algorithm of the reconstruction residual error, comparing the difference between the original image and the reconstructed image, and used for distinguishing whether the to-be-detected region contains target echo. The application fully explores the multi-path characteristics of the clutter, and realizes the weak and small target detection under the condition of no target sample through the multi-path echo reconstruction residual error.

The invention relates to the field of space-time prediction, and discloses an intelligent egg distribution method based on a space-time double-attention model, which comprises the following steps: acquiring historical time sequence data and historical space data of eggs in a store in a target area, and integrating and preprocessing the two data to obtain a multi-dimensional feature vector; constructing an algorithm learning library based on the multi-dimensional feature vectors; constructing a time attention model and a space attention model based on an algorithm learning library, sequentially inputting the sample matrix into the two models for calculation, and outputting a prediction matrix; constructing a loss function based on the sample label matrix and the prediction matrix, and optimizing dual-model parameters through iterative training to obtain a trained poultry egg sales prediction model; and using the trained model to carry out sales volume prediction, generating a distribution scheme according to a prediction result, and carrying out intelligent egg distribution.

Disclosed are artificial intelligence (AI)-based security risk prediction system and method for targets to be protected in a cloud environment. The method includes: collecting cloud logs and system logs for the targets to be protected in real time; learning all activity logs included in the cloud logs and the system logs for the targets to be protected of a corresponding member company through an AI algorithm; identifying a new activity among activities for the targets to be protected based on a learning process through the AI algorithm, and in response to the identified new activity being a new activity related to security, identifying a first activity pattern comprising the corresponding new activity; identifying an order of an preparatory activity for the new activity in the first activity pattern; identifying a risk score corresponding to the order of the preparatory activity for the new activity; and calculating a risk score of each target to be protected by summing identified risk scores of all new activities.

This application relates to the field of data processing and discloses a process card optimization method, device, equipment and storage medium for intelligent cloud simulation. The method involves obtaining equipment operation data in the production workshop; using big data processing technology in the cloud platform to clean the production workshop data and remove abnormal data in the production workshop data; establishing a multi-level simulation model for the process card based on the FEA finite element analysis model, setting production scenarios and process parameters based on the historical data of the production workshop, and using the multi-level simulation model for the process card to simulate and calculate the production scenarios and process parameters; learning the initial production workshop data and the initial process card production plan based on the improved GAN generative adversarial network; using the reinforcement algorithm Q learning to define the state, action and reward in the learning process of the GAN generative adversarial network; and adaptively adjusting the process card based on the production data to solve the problems of insufficient real-time performance, flexibility and intelligence in traditional methods.

The application discloses a bearing intelligent diagnosis method based on a generalized domain data fusion strategy and kernel sparse representation, designs a generalized domain data fusion strategy for dictionary learning, specifically uses an improved Kalman filter fusion framework to project time domain and frequency domain signals to a generalized domain state space and realizes signal adaptive fusion, and secondly, in order to avoid the influence of time shift characteristics on a dictionary learning model, develops a kernel discriminative sub-dictionary learning method, specifically uses a Gaussian kernel function to map the fused generalized domain signals to a high-dimensional feature space, then learns a specific category kernel discriminative sub-dictionary in a data-driven manner through a kernel K-SVD algorithm, then uses the learned specific category kernel discriminative sub-dictionary to realize sparse representation of unknown bearing signals in a high-dimensional space, and finally realizes intelligent identification of the bearing health state according to a minimum reconstruction error criterion. The application enhances the sparse representation ability and discriminative feature mining ability of the dictionary model for nonlinear data.

The invention discloses a deep learning-based microscopic core pore network equivalent model generation method. Comprising the following steps: S1, based on casting body slices and core analysis data, learning the form and distribution characteristics of pore throats by using a machine learning algorithm, generating a plurality of similar sub-graphs, performing local feature optimization on the plurality of similar sub-graphs, and splicing the plurality of optimized similar sub-graphs into a machine learning equivalent model; s2, performing pore structure characteristic parameter similarity evaluation on the machine learning equivalent model; and S3, performing model seepage performance evaluation on the machine learning equivalent model. The method has the beneficial effects that the machine learning model is constructed based on the casting body slice and rock core analysis data, the pore structure characteristic parameter similarity evaluation and the model seepage performance evaluation are combined to ensure that the generated model is highly matched with the actual pore structure, and a reliable equivalent model is provided for accurate characterization of reservoir microscopic characteristics.

The invention provides a dynamic occlusion target complementing and labeling method and system based on BEV time sequence fusion, and is applied to the technical field of data processing. According to the method, BEV time sequence fusion is taken as a core, multi-dimensional sensing data such as a time sequence image and a three-dimensional coordinate are collected firstly according to dynamic shielding target complementation and labeling requirements, and key parameters such as a complementation confidence threshold value are set in combination with detection precision and labeling specifications; the data quality is optimized through integrated processing, and a standardized training data set is constructed and sorted according to completion contribution degrees. A dynamic completion labeling strategy is determined based on scene complexity, hardware computing power and the like, parameters such as sliding window size and the like are adapted, and data sets are split and then imported into training in parallel. A BEV time sequence fusion algorithm is utilized to learn a mapping relation between shielding features and a target form and a motion rule, multi-modal features are dynamically weighted and fused, and accurate completion and labeling are realized by combining adaptive core parameters such as shielding types and target scales.

The invention relates to the field of cloud computing, and particularly provides a method and a device for improving the management capability of a key application in a cloud environment, firstly, collecting basic information and service information of different life cycle nodes of the application, generating an application portrait, and secondly, forming a key application with the application as a core by utilizing numerous association relationships among resources, an application portrait topological graph is generated, meanwhile, total-branch detection nodes are constructed for distributed cloud centers, all the cloud centers collect and detect applications needing service escorting, the applications are reported to a center end for data analysis, and a distributed dial test node network is constructed; and finally, the central end learns and analyzes the nodes generating influence in the application portrait topology through an algorithm, generates a first diagnosis condition of service escort, and forms a service escort report or gives an alarm in time. Compared with the prior art, the method has the advantages that the problem finding capability can be improved on the whole, the customer active fault reporting rate is reduced, the monitoring granularity is improved, the resource utilization rate is improved, and the fault response time is shortened.