353 results about "Network deployment" patented technology

The invention relates to the technical field of coal mining equipment, and discloses a mining roadway stability real-time monitoring and early warning system which comprises the following steps: S1, multi-source sensor network deployment, S2, data transmission and preprocessing, S3, stability index dynamic calculation, S4, fusion early warning model construction and S5, graded early warning triggering. Roadway surrounding rock deformation, stress, vibration and environmental parameters are collected in real time through a multi-source sensing network, efficient data transmission and preprocessing are achieved in combination with an industrial looped network, surrounding rock strain energy density, displacement convergence rate, support failure coefficient and other key indexes are dynamically calculated, a fusion early warning model is constructed based on machine learning, and the early warning accuracy is improved. According to the method, accurate evaluation of the stability level is achieved, grading alarm and emergency control are automatically triggered when a threshold value is reached, a'perception-analysis-decision-response 'closed loop is formed through the design, the real-time performance of roadway stability monitoring and the early warning reliability are remarkably improved, and accidents such as roof fall and wall caving are effectively prevented.

The invention provides a network congestion control method and system based on multi-agent reinforcement learning, computer equipment and a storage medium, and relates to the technical field of network congestion control, a communication network comprising a plurality of communication nodes is constructed, the communication nodes are mapped into independent agents, and the independent agents are used for controlling the network congestion. Designing a weighted composite reward function containing indexes such as throughput and delay by taking a flow-level index as a local state and a data sending rate adjustment coefficient as an action; the intelligent agent selects action execution according to a current state based on a strategy network, calculates an environment reward based on the current state, the action, a next state after the action is executed and a weighted composite reward function, and stores the environment reward in a shared experience playback pool; training a value network and a strategy network by using the data in the pool; the trained strategy network is deployed at each node, distributed congestion control is realized, the stability of a congestion control decision in a complex network environment is improved, and bandwidth resources are distributed fairly while the throughput is ensured.

The invention relates to the technical field of airport safety monitoring, in particular to an airport runway intrusion identification method based on an ESNB algorithm, and the method comprises the steps: constructing an airport runway intrusion data set; the method comprises the following steps: constructing an improved Inception U-Net network, and obtaining a teacher network based on the improved Inception U-Net network; based on the teacher network, obtaining an optimal student network through a multi-objective evolutionary algorithm; based on the teacher network and the airport runway intrusion data set, performing knowledge distillation on the optimal student network to obtain an ESNB network; according to the scheme, the ESNB network is deployed in an airport and is used for airport runway intrusion recognition, and the problems that existing airport runway intrusion recognition is low in efficiency and high in computing power requirement can be effectively solved.

The invention discloses a disaster area robot operation path and energy consumption optimization method and system, and the method comprises the steps: constructing a multi-factor coupled unit path energy consumption prediction model, fusing the terrain, real-time wind resistance and battery attenuation factors of a disaster area, and generating a precise energy consumption map; a hierarchical reinforcement learning path scheduling framework is established, an upper-layer strategy network generates a global path by taking energy consumption and time optimization as double targets, a lower-layer execution network processes sensor data in real time based on MPC, and path stability and low energy consumption are considered during obstacle avoidance; designing task plug-in mechanism optimization scheduling triggered by an event; and compressing an upper-layer strategy network by adopting knowledge distillation, and deploying to a robot edge end. According to the method, the endurance and task completion rate are improved through the precise energy consumption model, global-local control disjunction is cracked through the layered architecture, the high-real-time requirement is met through the plug-in mechanism, the edge deployment adapts to the weak network / broken network environment, and the method is suitable for high-real-time scenes such as post-disaster inspection and material transportation.

The invention relates to the field of low-altitude communication network planning methods, in particular to a 5G three-dimensional coverage and capacity planning design method for low-altitude economy, which comprises the following steps: performing airspace range limitation and application scene division on a target airspace, establishing a mobile model of a low-altitude terminal, and providing dynamic input for simulation; the method comprises the following steps: establishing a dual-frequency heterogeneous network, and determining base station addresses covering targets at different heights and meeting different signal station spacing; performing three-dimensional coverage and quality simulation on the established dual-frequency heterogeneous network, performing network KPI evaluation on the simulation until the standard is reached, and outputting a network deployment planning scheme; deploying according to a network deployment planning scheme, predicting low-altitude service flow, and adjusting a beam direction and a PRB reservation ratio to realize on-demand allocation of resources; and carrying out actual measurement on the deployed network, collecting an air measurement index, comparing a comparison deviation between the air measurement index and a simulation result, and carrying out fine adjustment on the network according to the comparison deviation. According to the invention, diversified communication requirements are met, and the success rate of terminal access is improved.

The invention discloses a photovoltaic-energy storage-charging multi-stage scheduling and market bidding optimization method and a photovoltaic-energy storage-charging multi-stage scheduling and market bidding optimization device. The method comprises the following steps: constructing a data-driven random environment model reflecting photovoltaic output, electricity price fluctuation and charging load uncertainty by adopting a mode of combining time sequence clustering and a non-homogeneous Markov chain based on historical operation data; modeling a scheduling and bidding problem of the optical storage and charging integrated station into a multi-stage Markov decision process model which comprises day-ahead decision and joint optimization of multiple intra-day rolling adjustment; a deep reinforcement learning algorithm is utilized to train the network, and a strategy regulation and control network which can adapt to various uncertain scenes and meet equipment physical constraints is obtained; and deploying the trained strategy regulation and control network in an energy management system to realize global coordinated scheduling and bidding of the optical storage and charging integrated station. According to the method, the economic benefit is remarkably improved, the robustness is greatly enhanced, the decision is globally coordinated and optimized, the real-time decision capability is strong, and the expandability and portability are good.

The invention discloses a micro-grid parallel-separation operation switching method, which relates to the technical field of micro-grids and comprises the following steps: S1, collecting real-time state data of a micro-grid, calculating system instability and generating a stability state label; s2, generating an ultra-short-term prediction result, and calculating an energy buffer potential well parameter set and a virtual inertia coefficient in combination with a stability label; s3, triggering a mode switching instruction based on the stability state label, executing the energy buffer potential well parameter set and the virtual inertia coefficient to control the output of the energy storage system, and completing grid-connected and off-grid mode switching; s4, the harmonic content of the power grid is collected, and a protection constant value and reactive power output are dynamically adjusted according to the instability degree of the system; and S5, constructing a system support framework by constructing a hierarchical communication network, deploying a redundancy mechanism and adopting a staged implementation strategy. And through real-time stability evaluation, virtual inertia adjustment and ultra-short-term prediction optimization, the stability, the electric energy quality and the adaptive capacity in the grid-connected and off-grid switching process of the micro-grid are improved, and high reliability and flexibility of the system are ensured.

A method comprises receiving a request to predict a deployment configuration for at least one application, analyzing code of the at least one application to identify one or more additional applications on which the at least one application will depend, identifying a plurality of network paths between the at least one application and the one or more additional applications, and using one or more machine learning algorithms to predict execution times for the at least one application over the plurality of network paths. The predicted execution times for the at least one application over the plurality of network paths are inputted to a network graph model. The network graph model predicts the deployment configuration for the at least one application based at least in part on the predicted execution times, wherein the deployment configuration comprises a subset of the plurality of network paths.

The invention belongs to the technical field of bridge construction, and particularly relates to a suspension bridge prefabricated cable strand method construction traction system control method based on reinforcement learning. The method comprises the specific steps that modeling is conducted on the PPWS method cable strand traction process through Simulink; a reinforcement learning environment is built, a state space and an action space are defined, the state space comprises the position, the speed and the acceleration of a puller, and the action space comprises PID controller parameters; a DDPG reinforcement learning model is adopted for training; the trained DDPG strategy network is deployed in hardware control, dynamic optimization of PID parameters is achieved according to state data collected by a sensor, and intelligent control over cable strand traction is achieved. According to the method, the multi-target reinforcement learning framework including the track smoothness reward, the speed stability reward and the speed sudden change punishment is constructed, PID control parameters are dynamically optimized in combination with the DDPG algorithm, accurate regulation and control of the operation speed of the puller are achieved, the control precision and the construction efficiency of the main cable traction process are remarkably improved, and the human intervention requirement is reduced.

The invention discloses a dam safety detection method and system based on multi-source detection data, and relates to the technical field of dam safety detection, and the method comprises the steps: cooperatively collecting dam multi-source data through multiple platforms and multiple sensors; the collected multi-source data are preprocessed; constructing a multi-scale teacher network, and performing high-precision feature learning and risk quantification by using labeled multi-source data; teacher network knowledge migration is carried out through a knowledge distillation technology, and a lightweight student network is trained in combination with cross-domain pseudo data; and deploying the trained lightweight student network to an unmanned aerial vehicle or a robot dog, and carrying out dam safety real-time detection. Through integration of multi-source data acquisition, cross-domain mutual training of teachers and students, lightweight model deployment and automatic early warning decision, pain points of single data, difficulty in cross-domain adaptation, insufficient precision, deployment limitation and low efficiency are solved, the system can be directly deployed on an unmanned aerial vehicle or a robot dog, dependence on cloud computing power is not needed, data acquisition and analysis time delay is reduced, and the system can be widely applied to unmanned aerial vehicles or robot dogs. And emergency scenes are quickly responded.

The invention discloses an adaptive controller tuning method based on reinforcement learning. The method comprises the following steps: S1, constructing a photovoltaic control task environment; s2, inputting the state sequence into a long short-term memory network, and outputting an embedded state vector; s3, constructing a hierarchical policy network comprising a basic policy module and a fast adaptation module; s4, training the hierarchical strategy network, inputting the embedded state vector into the hierarchical strategy network to generate a controller parameter action, interacting with a photovoltaic control task environment by using a multi-layer feedforward neural network to obtain response and reward feedback, and executing target value calculation and strategy optimization; s5, obtaining a general strategy initial parameter by using a meta-learning optimization method; and S6, deploying the optimized hierarchical strategy network to a target photovoltaic control task. The method is suitable for photovoltaic control and other multi-working-condition dynamic environments, and has the advantages of being high in strategy migration capacity, high in robustness and excellent in self-adaptive performance.

A topology-reprogrammable test environment is provided that can support the needs of CI / CD / CV in the field. The system disclosed provides a highly scalable network architecture to simplify the implementation of network slicing, TaaS and network CI / CD, and solves problems related to the complexity of cloud-native network (CNN) deployments. A Network Cell (NC), comprises or consists of a Containerized Network Function (CNF), a Containerized Digital Twin (CDT), and a Containerized Test Agent (CTA). The CDT has at least two personalities, e.g., an emulator of the CNF in the same NC or a nodal of the CNF. The choice of personality of the CDT is controlled by the CTA of the NC. A number of NCs use a 3D IP address to interconnect and form a new kind of CNN over the infrastructure of VRs.

The invention discloses a boiler furnace three-dimensional combustion temperature field reconstruction method and system, and the method comprises the steps: S1, multi-mode sensing network deployment, S2, time-space synchronization data collection, S3, sound wave flying time extraction and acoustic modeling, S4, flame image processing and radiation temperature conversion, S5, three-dimensional temperature initial field generation, and S6, multi-mode joint objective function construction. S7, carrying out regularization constraint iteration solution; and S8, carrying out three-dimensional temperature field visualization output. According to the method, high penetrability and sensitivity of sound waves to medium temperature and rich radiation information contained in flame images are fully utilized, and the precision and spatial resolution of temperature field reconstruction are remarkably improved by establishing a joint objective function and performing collaborative optimization. Meanwhile, a regularization method and an adaptive filtering technology are introduced, interference caused by measurement noise and uncertain problems is effectively suppressed, and the stability and robustness of the reconstruction process are enhanced.

The invention relates to the technical field of network deployment, in particular to a method and system for interactively configuring parameters through an LLM and a router gateway. The method comprises the following steps: arranging corresponding data of a natural language, a router command and a command execution result, and forming Quest / Answer training data required by a large model for providing a semantic mapping basis of a router configuration field for the large model; and training the general model through a large model fine tuning technology, so that the general model translates a router configuration related natural language into a specific router command and translates a command execution result into a natural language. According to the method, corresponding data pairs of a natural language, a router command and an execution result are constructed, and targeted fine adjustment is performed on a large model, so that the model has a bidirectional semantic translation capability in the field of router configuration, and the technical bottleneck that a traditional interaction scheme depends on independent interface development is broken through.

The invention discloses a leakage monitoring method based on an LNG fuel gas system. The leakage monitoring method comprises the steps that a self-adaptive frequency modulation continuous wave active acoustic scanning network is established; blind source separation is carried out on mixed signals in acoustic scanning network abnormal events by adopting self-adaptive kernel independent component analysis; constructing a leakage feature mapping model of the physical information neural network; leakage source accurate positioning and quantification based on acoustic tomography and Bayesian reasoning are carried out; multi-modal decision fusion is carried out based on the multi-dimensional data sources received in parallel, a false alarm suppression mechanism is set, and time continuity verification and space consistency verification are carried out; establishing a reinforcement learning model for autonomously optimizing a monitoring strategy according to environment change and system state, and realizing adaptive optimization; the strategy network after self-adaptive optimization is deployed at the cloud, actions are generated regularly according to the current state, the actions are issued to the regional gateway and the edge node for execution, and iterative updating is carried out, so that the monitoring accuracy in a complex environment is improved, and the false alarm rate is reduced.

According to the wireless network intelligent optimization deployment method and system disclosed by the invention, the network deployment strategy is issued to the edge node, so that the calculation amount of the integrated controller is reduced, and the optimization efficiency of the network deployment strategy is improved; besides, strategy chromosomes are evolved through a double-variation mechanism, and a dynamically changing gene mutation probability and an environment evolution range are introduced, so that the algorithm can quickly jump out of a local optimal solution, dynamically track and adapt to the change of a network environment, and the convergence speed of the evolutionary algorithm is remarkably improved; and finally, analyzing the strategy chromosome according to the network state data, and determining an optimized deployment strategy. According to the method, the overall overhead and delay of a network deployment strategy are remarkably reduced, and the efficiency is improved.

The invention is suitable for the field of fire extinguishing and automatic control, and provides a digital twin-driven dynamic topology network intelligent fire-fighting cooperation method and system, and the method comprises the steps: building a digital twin model of a fire-fighting monitoring site, carrying out the real-time monitoring of data through various devices, and enabling the model to be dynamically updated; establishing a dynamic topology network, deploying communication nodes and optimizing topology; and then analyzing a monitoring scene, defining task priorities in combination with resource and environment information, allocating tasks and scheduling communication nodes, and finally visually outputting a result. According to the system, task allocation, resource scheduling and communication optimization are incorporated into a unified mathematical framework, the defect that a traditional fire-fighting decision is split is overcome, and the rescue efficiency is improved.

The invention relates to the technical field of Internet of Things communication, in particular to a LoRa network dynamic link self-adaptive control method and system based on environment awareness, and the method comprises the steps that the system firstly presets initial communication parameters for a terminal and collects the data and link information of the terminal; based on the building and traffic data on the transmission path, an environmental perception model is constructed to quantify environmental impact factors, and a dynamic evaluation matrix is established in combination with performance indexes. Calculating the comprehensive performance index of the current link based on the matrix, and predicting the change of the signal-to-noise ratio to obtain a relative difference value; the system generates optimization suggestions of communication parameters by analyzing a performance index change trend and applying an adaptive algorithm, and predicts brought performance improvement. By balancing energy consumption and time delay cost required by performance improvement and adjustment, parameter adjustment is executed only when the income is greater than the overhead. And finally, completing the optimization of all terminals within the fixed network distribution time, and evaluating the overall performance of the network. According to the invention, the accuracy of dynamic link adaptive control of the LoRa network is improved.

The invention provides a heterogeneous perception adaptive low-bit neural network deployment method, and relates to the technical field of artificial intelligence and heterogeneous computing, and the method comprises the steps: firstly obtaining the hierarchical computing feature information of each layer of a neural network and the dynamic feature parameter information of heterogeneous hardware, and forming a multi-level basic information set; and then a hierarchical efficiency association model is constructed to describe the association relationship among the calculation precision, the hardware dynamic characteristics and the layer calculation efficiency. During online operation, a hardware real-time load state and an energy efficiency constraint condition are tracked to generate a dynamic state monitoring result, and the dynamic state monitoring result and the dynamic state monitoring result are combined to generate a hierarchical deployment configuration scheme by adopting a multi-objective optimization algorithm. And finally, a lightweight runtime scheduler is called to allocate calculation tasks according to the scheme, and interlayer dependency data is loaded and executed, so that dynamic neural network deployment across hardware equipment is realized, and the deployment effect and the operation efficiency are improved.

The invention belongs to the technical field of bridge construction, and particularly relates to a design method for a temporary pushing buttress based on reinforcement learning. The method comprises the following steps that reinforcement learning modeling is carried out, and a temporary buttress design agent is constructed; constructing a finite element simulation environment for simulating the structural response of the buttress scheme; training a temporary buttress design agent by adopting a depth deterministic strategy gradient algorithm; and strategy network deployment: deploying the trained temporary buttress design agent model to an engineering design platform, and realizing automatic generation of a pushing temporary buttress design scheme in an actual scene. According to the method, the intelligent and dynamic temporary buttress design optimization is realized by constructing the multi-target reinforcement learning framework including buttress stability, material consumption and girder deformation and combining with the depth deterministic strategy gradient algorithm to train the intelligent body capable of automatically carrying out temporary buttress design under different environment constraints; and the design efficiency of the temporary buttress is obviously improved.

Visible light communication (VLC) enabled unmanned aerial vehicles (UAVs) have evolved as a promising technology for the fifth generation and beyond communications. The proposed invention describes a system and methodology for deploying VLC-enabled UAVs, which serve as flying base stations (V-FBSs) over the target area: disaster regions, concert and fest areas, and search and rescue operations zones. The V-FBSs provide both communication and illumination. The proposed system follows CRAN architecture, which helps reduce the CAPEX and OPEX significantly. Further, present disclosure delivers a complete method to deploy the V-FBS network along with the detailed synchronization process required for autonomous deployment. It offers a complete 3-D deployment of the V-FBSs, which ensures minimum interference while satisfying the promised QoS and providing an energy-efficient position. The proposed network is scalable and can significantly reduce the outage. Since VLC is used for the present invention, the network does not interfere with the neighboring RF networks.

The invention relates to the technical field of pulse neural network computing hardware, in particular to a neural network computing circuit of a pulse self-attention mechanism. According to the method, invalid or inefficient pulse events are dynamically screened through the hardware mask module, the operation number is remarkably reduced, and calculation path delay and logic resource occupation are reduced; meanwhile, in order to adapt to novel networks with binary architecture such as QKFormer, calculation and storage of a V matrix are eliminated, and storage resources and calculation resources are further reduced; in addition, the event coding module only generates active neuron events, and input sparsity is achieved. A configurable IF neuron model is adopted, exponential operation is avoided, hardware implementation is facilitated, and the method is suitable for binary network deployment. The modular architecture can support function extension, assembly line and parallel work; and the parallelism degree of the design can be determined according to the actual data pulse distribution rate. Event driving and mask pruning are combined, so that the overall computing resources of the circuit are greatly reduced, and the power consumption is reduced.

The invention discloses a flight control end intelligent algorithm deployment method based on adaptive pruning, and the method comprises the following steps: S1, collecting flight control task state data and resource state data, and generating a flight control task resource state data set; s2, constructing a task characteristic vector based on the flight control task resource state data set; s3, inputting the task characteristic vector into a pruning strategy generation module to generate a dynamic grouping pruning strategy; s4, executing a dynamic grouping pruning operation on the improved Shuffle Net structure, and generating a task adaptive pruning network; s5, deploying the task adaptive pruning network to flight control end embedded hardware, and generating an intelligent reasoning control result; s6, monitoring the execution effect of the flight control task corresponding to the intelligent reasoning control result in real time; and S7, dynamically adjusting the task characteristic vector and the dynamic grouping pruning strategy to form an updated task adaptive pruning network. According to the method, the improved Shuffle Net and adaptive pruning are adopted, and real-time optimization and efficient deployment of a flight control end intelligent algorithm are realized.

The invention discloses a dynamic trapping network deployment method and system fusing attack behavior preferences, and relates to the technical field of network security. The method comprises the following steps: modeling a network topology into a directed acyclic graph, and constructing a dynamic trapping network topology structure fused with attack behavior preference based on attacker behavior preference analysis in combination with situation awareness; modeling an attack and defense confrontation process as a Markov multi-stage dynamic game process, and solving Nash equilibrium to obtain an optimal strategy of each stage; the behavior preference of an attacker is updated in real time by using Bayesian learning and a sliding window mechanism, and the adaptability of a dynamic trapping network to attack and defense situations is enhanced. According to the method, the problems of configuration stiffness and poor attack and defense situation change adaptability in dynamic trap network deployment are solved, the active defense capability aiming at attacker behaviors can be improved, and an effective solution is provided for dynamic deployment of a trap network in a dynamic network environment.

The invention discloses an optimal control method for an intermittent leaching process, which comprises the following steps of: solving a state-action track of the intermittent leaching process by adopting an OCP strategy, and performing behavior cloning on a strategy network of a PPO intelligent agent by taking the state-action track as initial expert data to obtain an initial strategy; iteratively training a strategy network in the PPO intelligent agent through interaction with the intermittent leaching process based on the initial strategy; each iterative training comprises the following steps: collecting a state-action trajectory of the intermittent leaching process under the current strategy network, and calculating a dominant function estimated value of each step of the intermittent leaching process by adopting a dynamic enhanced generalized dominant function; optimizing the strategy network and the value network based on the loss function value; and deploying the optimal strategy network obtained by training into a control system of the intermittent leaching process so as to output an optimal control action according to the real-time state vector. The method can get rid of dependence on an accurate mechanism model, and can solve the problems of sample efficiency, reward design, training stability and the like.

The invention belongs to the field of communication networks, and particularly discloses a robot system based on high-speed LVDS networking and a control method of the robot system. According to the invention, a PCS layer cache forwarding scheme of the LVDS interface is adopted, so that the signal quality of network communication among a plurality of joint driving cards can be improved, and the stability of data transmission of a complex robot system is effectively ensured; a high-speed LVDS networking communication mode is adopted, high-bandwidth transmission and reliable transmission of data are effectively guaranteed, the anti-interference capacity is improved through a differential signal interface mode, FPGA pin IO resources are saved, the network deployment cost is greatly reduced, and the real-time performance and stability of network communication of a robot system can be effectively guaranteed; each board card adopts an FPGA + MCU architecture, the FPGA is responsible for data analysis, transmission and motor drive control, the MCU is responsible for FOC algorithm calculation, and the architecture can give full play to respective advantages of the FPGA and the MCU.

The utility model provides a multi-axis advanced control and digital twinning training platform, including: a housing, a controller assembly arranged on the housing, a multi-axis drive control assembly and a motor application system, the controller assembly includes a small PLC, a medium PLC and a digital twinning display and control all-in-one machine, the small PLC, the medium PLC and the multi-axis drive control assembly are arranged in the middle of the housing, and the motor application system is connected with the multi-axis drive control assembly and the digital twinning display and control all-in-one machine. The digital twin display and control all-in-one machine is arranged below the medium-sized PLC, and the motor application system is arranged above the small-sized PLC, the medium-sized PLC and the multi-axis driving control assembly. According to the utility model, network deployment is carried out on multi-axis control communication, mainstream axis control on the market is realized, multi-axis synchronous control and a real-time feedback mechanism are realized, the comprehensive ability of students is improved, and the students can more intuitively understand the principle and application of typical data acquisition and analysis of axis control.

The invention relates to a network discovery method, a first device, a second device, a chip, a computer readable storage medium, a computer program product, a computer program and a communication system. The network discovery method comprises the following steps: a first device sends a discovery request on a first channel; wherein the discovery request is used for indicating or requesting a second device on the first channel to send discovery request reply information, and the discovery request reply information comprises network deployment information. According to the embodiment of the invention, energy consumed by network discovery can be saved.

The application provides a low-altitude airborne terminal heterogeneous network vertical handover method and system, the method comprises the following steps: constructing and dynamically updating a three-dimensional network coverage information library; predicting the future route of the airborne terminal based on its real-time motion state, and obtaining a candidate network set from the three-dimensional network coverage information library according to the prediction result; generating a handover decision weight for each network in the candidate network set based on a preset evaluation model; determining whether to trigger vertical handover through fuzzy logic decision based on the handover decision weight, the current network state and the operating parameters of the airborne terminal; if the vertical handover is triggered, performing a vertical handover operation based on the candidate network set and the handover decision weight. The application is based on the pre-collected network deployment information, combined with the actual flight information for vertical handover, the switching execution logic is simple and efficient, which can effectively avoid the communication interruption problem caused by switching failure or frequent switching, and significantly reduce the power consumption in the communication process.

The application discloses a space micro-motion target recognition method based on multi-band data features and text information fusion, comprising: dividing narrowband radar scattering cross-section area data, wideband high-resolution range image data and text data of space micro-motion targets under multiple frequency bands to generate a training set and a test set; constructing a target recognition network comprising a feature extraction module, a multi-band graph attention module and a classifier; inputting the training set into the target recognition network, training through a back propagation algorithm, testing by using the test set, and obtaining a target recognition network meeting the conditions; and deploying the network on a device to recognize space micro-motion targets in real time. The application fully excavates complementary information of different frequency bands and different modal data by extracting multi-band multi-modal features and dynamically fusing by using a graph attention mechanism, significantly improves the recognition accuracy of space micro-motion targets, and can be widely applied to ground-based wideband and narrowband radar systems.