203 results about "Adaptive capacity" patented technology

Embodiments of the present application disclose a model training method and device. After obtaining a pre-training model, the embodiments of the present application additionally set corresponding routing modules before each calculation module of the pre-training model to obtain an initial training model, and then perform multi-stage reinforcement learning training on the initial training model by using preset samples, to update the basic model parameters of the initial training model and the module parameters of each routing module based on the model output result and the routing information output by the routing module for non-key token representation in the first stage training, and to update the basic model parameters of the initial training model and the module parameters of each routing module based on the model output result in the second stage training, thereby obtaining a target model. Thus, the embodiments of the present application can make the model have token self-adaptive capability, to improve the reasoning efficiency and throughput of the model and reduce the calculation cost of the model.

The application belongs to the technical field of mechanical arm motion planning, and provides a kind of adaptive sampling and multi-channel expansion mechanical arm path planning method, first, the environment obstacle set is constructed, the point to obstacle distance, the clearance configuration and the segment level minimum clearance are established, and the unified safety criterion is constructed;Subsequently, according to the approximation state of bidirectional search tree, the sampling window is adaptively updated, and the random reference sample is generated in combination with the stagnation detection mechanism;On this basis, the hierarchical multi-channel expansion mechanism of direct pointing expansion, bypass expansion, potential field wall expansion and global bottom expansion is used to generate candidate new nodes;When a feasible new node is generated, it is connected to the current search tree and real meeting detection is carried out, if the double trees are successfully connected, path backtracking, splicing, simplification and full path safety review are further executed, and the final executable path is output.The application can improve the safety, search efficiency and environmental adaptability of the path planning of the mechanical arm in the complex limited environment.

The application provides a joint resource allocation and trajectory design method and system based on deep reinforcement learning, which is suitable for a multi-cell integrated sensing and communication system assisted by a UAV. The method comprises the following steps: initializing a starting position of the UAV, communication and sensing system parameters, and a deep reinforcement learning model; scheduling users of multi-cell base stations under the condition of a given UAV position; allocating transmission power of each communication node under the current user scheduling result to meet the communication performance constraint and the sensing accuracy constraint; outputting a motion control strategy of the UAV by the deep reinforcement learning model based on the power allocation result and system state information, updating the flight trajectory of the UAV; and judging whether the joint optimization process converges or not, and outputting the optimization result if the joint optimization process converges, and then continuing the iterative optimization. The application alleviates the coupling problem between resource allocation and trajectory design in a multi-cell ISAC scenario, and improves the adaptive ability and comprehensive performance of the system in a dynamic environment.

This invention discloses a service-oriented edge processing method and system for satellite networks using S-UPF (Service-Oriented Components), relating to the field of satellite communication technology. The method includes: decoupling and encapsulating user plane containers to a ground repository; clustering S-UPFs; instantiating containers from the repository to the platform on demand; identifying service characteristics after receiving data streams to obtain forwarding intentions and target container groups; planning inter-satellite forwarding paths based on a comprehensive network state; and calling the target containers for on-orbit processing followed by inter-satellite transmission. This invention solves the technical problems in existing satellite networks where user plane functions lack service-oriented encapsulation and dynamic on-demand deployment capabilities, leading to rigid on-board processing, uneven resource utilization, and low inter-satellite collaboration efficiency. It achieves the technical effect of accurately clustering and dynamically instantiating service capabilities through multi-dimensional satellite state profiling, constructing efficient inter-satellite service-oriented forwarding paths, and improving the adaptive capability, resource utilization efficiency, and inter-satellite forwarding collaboration performance of satellite network on-orbit edge processing.

The utility model relates to hose connecting technical field especially relates to an anti -shock flexible hose connecting assembly, including flexible connecting piece, both ends of flexible connecting piece are provided with flexible joint, the outside of flexible connecting piece is equipped with shock insulation cover, the surface of shock insulation cover is evenly spaced and is provided with several shock absorbers along the circumference. Through setting up flexible joint, utilize its elastic deformation ability to realize quick plug -in connection with hose, simple operation, high installation efficiency, in addition, set up shock insulation cover outside flexible connecting piece, and arrange multiple shock absorbers on its surface, constitute double damping structure, shock absorber absorbs external vibration energy by its high damping characteristic, play the primary buffering effect, shock insulation cover is as the secondary damping structure, effectively isolates the vibration transmission path, thereby significantly reduce the dynamic stress suffered by the connecting part, improve the adaptive capacity of assembly under the complex vibration environment.

This invention discloses a production data-driven intelligent scheduling method and system for mechanical equipment, relating to the technical field of equipment scheduling. The method includes: constructing a set of equipment operating characteristics; comprehensively evaluating the current load rate, processing capacity, and failure risk of each piece of mechanical equipment; constructing a scheduling objective function using equipment survivability scores as scheduling decision weight parameters; performing task allocation and processing path optimization calculations to generate an intelligent scheduling scheme for the mechanical equipment; and monitoring the scheduling execution process in real time to adaptively adjust the production scheduling scheme. This invention solves the technical problems in existing technologies where static scheduling cannot adapt to real-time changes in equipment status and lacks a dynamic risk feedback mechanism, leading to equipment overload, unplanned downtime, and cascading production delays during scheduling scheme execution. It achieves the technical effects of improving the production scheduling system's adaptability to dynamic equipment operating conditions, reducing the risk of unplanned downtime, and optimizing equipment utilization and production delivery reliability.

The application discloses a kind of hearing-impaired person intelligent terminal adaptive feedback method, intelligent terminal and system, it is related to smart home and barrier-free interaction technical field.The present application is aimed at the defects of single feedback mode, lack of intelligent scheduling and no adaptive capacity, constructs an end-to-end intelligent feedback scheme.First, the user situation characteristics are collected in real time by multidimensional sensor, and the situation characteristic vector is constructed;At the same time, the importance score of visitor is calculated by extracting the multi-modal features of visitor.Secondly, the situation characteristics and visitor characteristics are input into the deep Q network, and the optimal multi-modal feedback device combination and feedback intensity level are intelligently decided.Then, the graded response strategy is executed, and the reminder intensity is dynamically adjusted based on the user response.Finally, the decision model is continuously optimized through online learning mechanism.The present application realizes the leap from passive fixed reminder to active intelligent perception feedback, significantly improves the use experience and efficiency of hearing-impaired person intelligent terminal.

The application relates to the technical field of lithium battery material preparation, and discloses an LFP positive electrode high-density homogenate process optimization method, a system and a storage medium. The method comprises the following steps: collecting slurry state and process parameter data in real time, analyzing a correlation strength index, identifying an abnormal slurry state mode, extracting a quantitative feature index, obtaining historical upstream and downstream data if the quantitative feature index is not in a preset range, combining an initial slurry state to evaluate the influence degree of upstream fluctuation, generating a risk level, formulating a compensation scheme according to the risk level, verifying the effectiveness of the scheme through virtual simulation, generating an adjustment instruction and delivering the adjustment instruction to a production control system, updating process parameters and storing the process parameters in a historical database, and the application improves the self-adaptive capacity of the homogenate process to upstream fluctuation.

This invention belongs to the interdisciplinary field of artificial intelligence and geological resource assessment, specifically relating to a method and system for predicting the energy storage potential of salt caverns based on multi-source geological information. It aims to address the problems of low prediction accuracy and difficulty in balancing breadth and precision in existing technologies due to single-source data, static models, and lack of adaptive capabilities. The method includes: acquiring multi-source data such as seismic, well logging, core, and geostress data; constructing a geological feature fusion vector containing seven core parameters; establishing a high-dimensional nonlinear mapping model driven by a deep neural network; dynamically adjusting the prediction range based on data coverage density and confidence level; and achieving continuous model iteration through online incremental learning. The system integrates data quality assessment, anomaly handling, and distributed computing modules. By adopting the above technical solutions, this application can achieve a significant improvement in prediction accuracy and adaptive intelligent adjustment of the prediction range.

The application discloses a deep learning-based time-space fusion early warning method for deep rock burst, and constructs a dual-branch Transformer-CNN time-space feature fusion early warning model. The model captures local mutation features in a microseismic sequence through a parallel multi-scale CNN network branch, combines a Transformer network branch to extract global evolution features, realizes comprehensive mining of rock burst precursor information and identification of key precursor information, and further enhances the dynamic adaptability of the early warning model to complex microseismic sequences by using a self-adaptive gating fusion mechanism, thereby enhancing the adaptability of the early warning model in a complex monitoring environment. In addition, a learnable position coding is introduced to replace a traditional fixed coding mode, so that the early warning model can more flexibly capture time sequence dependence in mining stress evolution, and the problems of gradient vanishing and low calculation efficiency of a traditional cyclic network in long sequence modeling are overcome. Finally, the classification accuracy of the rock burst danger level and the early warning reliability are effectively improved.

This invention relates to the field of hydrological forecasting technology and discloses a medium- and long-term intelligent runoff forecasting method, comprising: collecting watershed data, preprocessing and extracting features, generating a feature vector sequence and dividing it into training sets; constructing a physical constraint recurrent neural network module and training it based on the total loss function of physical constraint loss; constructing a sliding window online adaptive correction module, initializing the sliding window to store the measured runoff values ​​and preliminary runoff forecast values ​​of the most recent O times; inputting the feature vector of the current time into the trained physical constraint recurrent neural network module to generate preliminary runoff forecast values, calculating the historical average deviation according to the window state and correcting it to obtain the final forecast result; finally, forming a new sample pair of measured runoff values ​​and preliminary runoff forecast values ​​and adding it to the window while removing the oldest sample to achieve dynamic window updates; this invention achieves high-precision, high-physical-consistency, and online adaptive intelligent forecasting of medium- and long-term runoff.

The present application relates to a microgrid adaptive scheduling algorithm automatic discovery method based on meta-learning; first, a microgrid meta-learning framework is constructed, a scheduling rule search space is defined, including an agent network output form and a meta-network structure, and a self-healing mechanism is designed to realize self-learning ability; second, a multi-microgrid parallel training mechanism is designed, multiple agents interact with their respective environments simultaneously, and a unified scheduling rule is used for learning; third, a meta-gradient optimization method is used, the agent update process is penetrated through back propagation, and the meta-network parameters are optimized to maximize the cumulative income of the agent group; finally, the discovered scheduling algorithm is evaluated in microgrid environments of different scales and different equipment configurations to verify its generalization performance and adaptive ability; the present application can automatically discover the scheduling algorithm, breaks through the limitations of traditional methods relying on artificial design, has strong generalization ability and the ability to quickly adapt to new environments, and can be applied to microgrid systems of different scales and configurations.

Embodiments of the present application disclose a kind of group adaptive integrated control method, device, electronic equipment and storage medium, involve embodied intelligent technical field, wherein, the method includes: according to target scene characteristic, utilize physical modeling and domain randomization technology to generate diversified environment task set, and pre-processing is carried out to ensure data quality, train reinforcement learning coach based on these tasks, combine decoupling and backtracking policy distillation technology to synthesize training sequence, enhance the generalization and adaptive ability of model, adopt long-time memory model structure design base model, integrate multi-agent observation module to capture time sequence and spatial correlation, dynamic noise injection and asynchronous optimization are carried out in training process, the trained field base model is deployed to target scene, form collaborative network, and set fault-tolerant mechanism to deal with agent offline and other sudden situations, continuously collect actual data to optimize model parameters.The present application effectively solves the problems of high customization, weak migration and high maintenance cost in the prior art.

The application relates to a control method for dynamic walking of a biped robot and the biped robot, and belongs to the technical field of robot control, which comprises the following steps: 1, a self-recurrent cerebellar model neural network is used to establish a dynamic model of the biped robot with a disturbance term, and dynamic robust walking of the biped robot is converted into a problem of realizing stability of a multi-input multi-output nonlinear system with a bounded uncertain term; 2, an adaptive self-recurrent cerebellar model neural network error observer is designed to estimate an error upper limit; 3, an adaptive law of network weight is designed to realize real-time updating of the network weight space and to adjust parameters of each learning factor; and 4, a boundary value estimation algorithm is used to compensate for an estimation error and feedback to a robot walking system, so that the biped robot can realize asymptotic stable walking. The application enables the control system to adapt to time-varying characteristics of the biped walking system on line, and has continuous learning and adaptive capacity for unknown dynamics.

This invention discloses a method and system for adaptive optimization control of thermal power units under all operating conditions, mainly relating to the field of automatic control technology for thermal power generation. It includes: collecting historical and current operating data of the thermal power unit; constructing a dynamic operating condition partitioning model based on unit load, matching steady-state benchmark parameters to each benchmark operating condition subspace; determining the target subspace of the current operating condition and extracting the corresponding benchmark parameters; constructing an adaptive optimization controller, using the benchmark parameters as the target and current data as feedback, generating superimposed control quantities through rolling time-domain solution and outputting them to the distributed control system; collecting unit response data to update the operating condition partitioning model, realizing dynamic subspace correction and online self-learning of benchmark parameters. The beneficial effects of this invention are: achieving adaptive optimization control under all operating conditions, improving control accuracy and stability under varying load conditions, possessing online self-learning capabilities, and enhancing the unit's adaptability to coal quality fluctuations and equipment characteristic changes.

The invention relates to the technical field of artificial intelligence, and provides a mountain wind energy resource dynamic evaluation method and device, equipment and a medium, which can perform attention fusion on multi-dimensional initial data, so that fusion features consider feature importance difference and comprehensiveness. A mountain wind energy evaluation data set with a wind energy power density label is constructed, and a basis is provided for supervised training of the model; the state space completely covers core variables influencing mountain wind energy evaluation, and the problem that a traditional model is insufficient in state space dimension and cannot represent a complex coupling relation is solved. The action space is in non-static output, so that the adaptive capacity of the model is improved; the reward function solves the technical problem that precision and efficiency cannot be considered at the same time; a dual experience playback mechanism improves the evaluation precision of the model on a complex wind field scene; the target network delay updating strategy enables the target network to maintain parameter stability in an iteration period, and the problems of training oscillation and slow convergence caused by real-time fluctuation of target return are solved.

This invention discloses a physical twin modeling method for cross-condition industrial control systems. It constructs a data-physical fusion physical twin model applied to an anomaly detection framework for industrial control systems. This framework uses the physical twin model as its core, expressing industrial physical processes through a system of differential equations and introducing physical constraints using PINN. Simultaneously, it continuously calibrates the model using real-time operational data to achieve dynamic updates. Furthermore, it achieves adaptive capabilities across conditions, stages, and scenarios through parameter sharing and transfer learning. The trained physical twin model is compared with real-time data, and the existence of anomalies is determined by analyzing the deviation between predicted and observed values. This framework organically combines the high fitting properties of data-driven approaches with the interpretability of physical-driven approaches, providing industrial control systems with high-precision, interpretable, and transferable anomaly detection capabilities.

This invention relates to the field of industrial intelligent control technology, specifically to an intelligent decision-making and collaborative management system for the entire industrial process based on digital twins. It includes: a data acquisition and modeling unit; a digital twin real-time synchronous simulation unit; an intelligent decision-making unit; and a cross-process collaborative management and control execution unit. This invention deeply binds equipment health status with process control decisions, and performs adaptive compensation based on the equipment's lifecycle degradation characteristics, ensuring that the industrial process decisions are adapted to the actual operating conditions of the equipment. By constructing a dynamic causal graph and real-time adjusting node weights, it transforms the results into a constraint mask matrix to shield against illegal decision-making actions that harm equipment health. Simultaneously, based on Weibull distribution fitting of the equipment aging trajectory to determine the degradation rate, it adjusts the safety compensation margin of control commands, and integrates these to generate an intelligent decision-making scheme with adaptive equipment degradation capabilities, ensuring that decisions match the equipment's health status and aging degradation characteristics.

The application relates to an intelligent linkage control method and system of automatic fire-fighting facilities. The method comprises the following steps: acquiring a fire alarm voice signal and on-site context information, analyzing and screening to obtain a confirmed keyword, combining the context information to generate a text feature sequence; acquiring an on-site image stream, removing irrelevant backgrounds to obtain to-be-fused visual features, fusing the text and image features, and reducing dimensions to generate a structured control alarm data stream; solving a safety boundary optimization problem based on the data stream, generating and broadcasting a legal safety action sequence; fitting a distribution of action execution feedback data, calculating a parameter matrix decay offset, and when the offset exceeds a threshold, reconstructing parameters to generate an updated linkage control parameter matrix. The method can realize intelligent linkage control of fire-fighting facilities and improve the safety and self-adaptive capacity of fire-fighting linkage.

The application relates to the technical field of artificial intelligence, and discloses a multi-modal large language model intelligent decision-making method and system fusing cross-domain information. The method comprises the following steps: performing hierarchical coding processing on cross-domain multi-modal input data corresponding to a target decision event, generating observation evidence data and conditional constraint data, and performing modal confidence estimation processing, cross-domain conflict relationship construction, gating fusion processing and historical decision feedback updating on the basis of the observation evidence data and the conditional constraint data. The method can distinguish the category tendency of the corresponding mode from the reliability of the evidence, identify real conflicts and pseudo conflicts, reduce the interference of low-quality modes and invalid differences on the fusion decision result, and simultaneously perform closed-loop correction on the modal confidence estimation parameters and the gating fusion parameters based on the historical decision feedback data, thereby improving the accuracy, stability and self-adaptive ability of the multi-modal large language model intelligent decision-making fusing cross-domain information in a time-varying environment.

The present application relates to a kind of cluster collaborative control method based on continuous mean shift theory, belong to involve multi-agent distributed collaborative control field, including three parts: desired configuration figure input and contour fitting, target shape physical parameter determination, distributed configuration control algorithm design based on continuous mean shift theory.The present application adopts Bezier curve to express desired shape, has strong representation ability to complex non-convex shape, target shape data storage is less, reduces memory requirement, and continuous control law can avoid the oscillation possibly caused by grid-based discrete control law.The present application uses distributed computing mode, and multi-agent system can enter desired shape area according to local perception information and actively explore unknown shape area, effectively reduce the requirement to agent communication ability, increase engineering feasibility, have strong robustness and scale expandability.The present application has strong adaptive capacity, and does not need global label and initial position configuration.

The application discloses an AI calculation intelligent scheduling method for multi-device intelligent management, and relates to the technical field of artificial intelligence, which comprises the following steps: collecting device operation parameters and environmental parameters and preprocessing them into state feature vectors; extracting time sequence correlation by using a space-time correlation analysis model, and constructing a resource load feature space through principal component analysis; performing multi-dimensional parallel calculation on task overhead based on a deep residual network and a bidirectional long short-term memory network fusion model, and outputting an execution performance index; constructing a scheduling augmented matrix according to the performance index and a constraint condition, and generating an optimal scheduling instruction sequence by using an improved heuristic search algorithm; and issuing the instruction and performing closed-loop dynamic correction. The application aims to solve the problems of low scheduling accuracy and uneven load in multi-device intelligent management, improve the accuracy and response speed of resource scheduling, optimize the load balancing and energy efficiency performance among heterogeneous devices, and enhance the dynamic adaptability and stability of the system.