435 results about "System deployment" patented technology

The invention belongs to the field of power systems, and discloses a power distribution network fault autonomous diagnosis and self-healing control method and system based on deep reinforcement learning, and the method comprises the steps: carrying out the collection and preprocessing of the multi-source heterogeneous data of a power distribution network; key feature vectors are extracted from the multi-source heterogeneous data based on a graph convolutional network, and a state representation model is constructed; based on an asynchronous dominant actor-commentator algorithm, constructing a fault diagnosis agent capable of quickly diagnosing faults; a meta-reinforcement learning algorithm is adopted, a dynamic reconstruction strategy library is generated through pre-training, and a self-healing control agent capable of rapidly adapting to various different fault scenes to generate an optimal reconstruction strategy is constructed; constructing a self-healing control module based on virtual impedance matching, wherein the self-healing control module is used for intelligent reconstruction and self-healing control of the power distribution network; a risk-sensitive reward function and a game equilibrium strategy optimization method are introduced to improve performance and robustness; and finally carrying out system deployment and engineering verification.

The embodiment of the invention provides a BIM-based hydraulic engineering digital management method and system. The method comprises the following steps: performing multi-dimensional monitoring system deployment on a to-be-monitored area to obtain multi-dimensional hydrological data; constructing a first BIM based on the topographic data of the to-be-monitored area, the real-time work area image, the hydraulic engineering construction information of the design stage and the equipment deployment information; fusing the multi-dimensional hydrological data, the water area change data, the construction progress data and the operation and maintenance monitoring data into the first BIM, and constructing a second BIM including a full life cycle; and through the equipment characteristic curve and the water conservancy project physical model, in combination with the multi-dimensional hydrological data and the water area change data, predicting a water conservancy project structure change trend and potential risk factors in the second BIM, fusing the water conservancy project structure change trend and the potential risk factors into the second BIM, and displaying an implementation effect and improvement suggestions in real time in the second BIM. A user is assisted to realize hydraulic engineering digital management, and the engineering management efficiency is improved.

The invention discloses a collaborative scheduling method based on a security agent, and relates to the technical field of network security design. The specific operation of the security agent collaborative scheduling method comprises the steps of system deployment and initialization, security operation task execution process, agent self-learning and capability evolution implementation, cross-domain security agent collaborative adaptation implementation and security operation visualization and traceability implementation. Connection paths between the security agent and business data, a private knowledge base, a security tool and multiple models are broken through, the problem that all elements in a traditional mode lack efficient communication and collaboration is solved, and by means of agent routing dynamic scheduling, A2A protocol interaction, MCP protocol tool calling, RAG business interface calling and multi-model combination, the security of the security agent is improved. Integration and intellectualization of the safety operation process are achieved, the continuity and the response speed of the operation process are improved, safety operation is more efficient and collaborative, and complex and variable safety requirements are met.

The invention relates to the field of electronic communication, and provides a multi-protocol dynamic switching method and system based on an FPGA. The method comprises the following steps: carrying out regional division on an FPGA chip through a layout planning tool to obtain a static logic region and a plurality of reconfigurable logic partitions; performing hanging processing on the current protocol module based on the protocol switching trigger signal, and writing running state data of the current protocol module into a state cache unit to obtain protocol context snapshot data; according to the target protocol type, bit stream loading is carried out on the corresponding reconfigurable logic partition through configuration of an access port, and a reconfiguration protocol module is obtained; and reading protocol context snapshot data, and performing state recovery on the reconstruction protocol module to obtain a target protocol data path. According to the invention, rapid protocol switching is realized, the utilization rate of hardware resources is improved, and the system deployment cost is reduced.

The invention relates to the technical field of crack detection, in particular to a building crack intelligent detection and prediction method based on multi-modal data fusion, which comprises the following steps: constructing a multi-source heterogeneous data set including an image data set, a thermal imaging data set and a vibration data set; simulating complex environment interference through a GAN (Generative Adversarial Network), and dynamically enhancing image training data; designing a double-branch multi-mode fusion network, and extracting the edge and texture features of the crack; designing a multi-modal fusion branch, and carrying out feature alignment and weighted fusion on vision, thermal imaging and vibration data; and based on a long short-term memory network LSTM and a structural mechanics model, a crack trend prediction module is developed, and early warning is provided for building structure safety. According to the method, intelligent monitoring and early warning of building structure safety are realized by constructing a multi-source heterogeneous data set, enhancing image training data through a GAN, extracting features by using a double-branch multi-mode fusion network, predicting a crack trend based on an LSTM and a structural mechanical model, and finally integrating system deployment.

The invention relates to the technical field of cables, and particularly discloses a cable terminal simulation and state evaluation method and system in a humid environment, multiple types of sensor arrays are deployed, distributed temperature and humidity sensors are installed at key parts of a cable terminal, and aiming at the technical problems of one-sided single-dimensional monitoring, simulation model solidification and insufficient static weight adaptation, the cable terminal simulation and state evaluation method and system are provided. Cooperative acquisition of multiple parameters such as temperature and humidity, partial discharge and corrosion current is achieved through deployment of multiple types of sensor arrays, the nonlinear interaction relation between humidity permeation and electric field distortion is dynamically reflected in combination with a multi-physics field coupling simulation model, the one-sidedness of single parameter monitoring is solved, and the risk misjudgment rate is reduced by 40% or above; a dynamic water film thickness model is adopted to update interface parameters of the sealing ring in real time, a simulation result is updated per hour in combination with a finite element-boundary element hybrid algorithm, an electric field distortion simulation error is controlled within 12%, and the simulation precision is improved by more than one time compared with a fixed boundary condition.

The invention belongs to the technical field of speech translation, and particularly relates to a video speech automatic translation method based on a pre-training model, and the method comprises the following steps: 1, preprocessing video and audio data; step 2, voice recognition and language detection; step 3, machine translation and text post-processing; step 4, speech synthesis and audio mixing; step 5, synchronizing video processing and subtitles; step 6, quality control and multi-dimensional evaluation; 7, carrying out model iteration and data closed loop; and step 8, system deployment and engineering implementation. Through deep fusion of efficient transfer learning of the pre-training model and the multi-modal technology, a high-precision, low-cost and easy-to-expand video speech translation solution is constructed, the time and labor cost of globalized content production is greatly reduced, the cross-language communication efficiency is improved, immersive multi-language experience is provided, and the method is suitable for popularization and application. And a data-driven continuous optimization mechanism is established, so that the system performance is improved along with the increase of the use scale.

A robotic fleet resource provisioning system includes a computer-readable storage system storing a fleet resources data store and resource provisioning rules. The fleet resources data store maintains a fleet resource inventory indicating fleet resources, each with features, configuration requirements, and a status. The resource provisioning rules are accessible to an intelligence layer to ensure that provisioned resources comply with the resource provisioning rules. The system receives a request for a robotic fleet to perform a job and determine a job definition data structure. The definition data structure defines a set of tasks that are to be performed in performance of the job. The system determines a robotic fleet configuration data structure corresponding to the job based on the set of tasks and the fleet resource inventory. The system determines a respective provisioning configuration for each respective fleet resource. The system deploys the robotic fleet to perform the job.

Techniques for host fleet management in a cloud provider network are described. Forecast data including a forecasted demand for virtual machines in each capacity pool of a set of capacity pools is obtained. A mathematical optimizer application is executed to generate a first optimal fleet plan, the mathematical optimizer application having an objective function to minimize a number of new host computer systems to add to the set of host computer systems to satisfy the forecasted demand for each capacity pool, the first optimal fleet plan includes an identification of a set of hardware types and, for each hardware type in the set, a quantity of new host computer systems of the hardware type. A plurality of new host computer systems is deployed, based on the first optimal fleet plan, for a hardware type in the set of hardware types into the set of host computer systems.

The invention discloses a heterogeneous computing power resource dynamic scheduling system and method based on multi-objective optimization, and the system is deployed in a digital ecological platform complex based on multivariate consensus and embedded intelligent management. Comprising a platform access module, a data acquisition and perception module, a multi-target modeling and optimization module, a dynamic scheduling and execution module and a fault-tolerant mechanism module. The system registers as a computing power scheduling service node through an intelligent contract interface, obtains and verifies the compliance of a computing task, collects heterogeneous computing power resource node state data, and generates a scheduling decision by dynamically adjusting the weight of a target function; evaluating task suitability by using an AMCU suitability scoring model, and executing a scheduling AMCU strategy; according to the method, the technical problems of low resource scheduling efficiency and poor fault-tolerant capability in a heterogeneous computing power environment are solved, and the system throughput and the resource utilization rate are improved.

In the implementation of techniques for deploying machine learning models with automated resource management, a system receives logic corresponding to a machine learning model and computing resource data corresponding to a plurality of computing resources available. Based on the logic and the computing resource data, the system generates the machine learning model and an allocation of one or more computing resources of the plurality of computing resources available for the machine learning model, in which the machine learning model conforms to the logic. Upon generation of the machine learning model and the allocation of the one or more computing resources, the system deploys the machine learning model and the allocation of the one or more computing resources of the plurality of computing resources available for the machine learning model.

The invention discloses a system and a method for automatically measuring pipe network parameters based on three-dimensional scanning of a wellhead of an inspection well, and belongs to the technical field of municipal engineering measurement. The system comprises a wellhead fixing assembly, a three-dimensional laser scanning assembly, a posture sensing assembly, a distance measuring assembly, a coordinate input assembly, a control processing assembly and a telescopic rod assembly. According to the method, through system deployment leveling, coordinate reference establishment, working condition judgment, point cloud acquisition, data processing and result output, automatic measurement of inspection well depth, pipe network pipe diameter, pipeline trend and pipe orifice center coordinates is realized. The system is fixed at a wellhead, does not need to descend, and adapts to scenes such as shallow wells / deep wells and good / weak signal areas; based on a high-density point cloud and an automatic algorithm, the measurement precision is high, the efficiency is improved, and output data can be directly used for pipe network modeling. The method solves the problems of low efficiency, strong manual dependence and poor safety of traditional measurement, and is suitable for parameter measurement of various inspection wells and pipe networks.

The invention discloses an anti-countercurrent system based on V2G equipment and a control method, and the system comprises at least two charging and discharging modules which are connected to one side of a power line, and the other side of the power line is connected to a power grid; the charging and discharging module comprises a first bidirectional electric meter and a V2G charging pile. One end of the V2G charging pile is connected with the power line through the first bidirectional ammeter, and the other end is connected with the charging terminal. The bidirectional main electricity meter is connected between the power line and the power grid; the contactor is connected between the bidirectional main ammeter and the power line. The anti-countercurrent control module is in communication connection with the bidirectional main electric meter, the contactor, the first bidirectional electric meter and the V2G charging pile; according to the invention, a virtualized anti-countercurrent control architecture is adopted, any V2G charging pile or cloud server is used as an anti-countercurrent control module, a traditional hardware controller is replaced by existing equipment computing power, global power regulation and control are realized based on data provided by the bidirectional main ammeter and the first bidirectional ammeter, and the system deployment cost and complexity are significantly reduced.

The invention provides a small and medium-sized enterprise digital transformation evaluation method, which comprises the following steps of: S11, analyzing and identifying regularly acquired small and medium-sized enterprise digital transformation evaluation files to obtain quantifiable multi-dimensional evaluation indexes; s12, constructing an evaluation index matching model based on the multi-dimensional evaluation indexes, and further constructing a smart contract architecture; s13, regularly collecting and analyzing the successful cases of the digital transformation of the small and medium-sized enterprises, performing scoring guidance for the quantifiable indexes corresponding to each successful case, and constructing a grade evaluation model; s14, determining quantifiable indexes corresponding to the small and medium-sized enterprises in transformation, deploying corresponding smart contracts in internal business systems of the small and medium-sized enterprises, and collecting key transformation data; and S15, analyzing the key transformation data through the grade evaluation model, and outputting a visual evaluation report. According to the invention, comprehensive and accurate quantitative evaluation of the digital transformation degree of small and medium-sized enterprises is realized, the objectivity, efficiency and credibility of evaluation are improved, and a user is helped to know the current situation of enterprise transformation.

The invention discloses an architecture system and method for large-scale task allocation, and belongs to the field of spaceflight measurement and control, and the system comprises a large-scale hardware device which is used for deploying processing units with different performances and various heterogeneous physical communication link resources; the hardware resource dynamic modeling and visualization framework is used for realizing a dynamic modeling system based on multi-source heterogeneous hardware topology perception; the resource topology awareness database captures and synchronizes system deployment changes by maintaining the dynamic topological relation of hardware resources, presents the life cycle state of the hardware resources, and supports system-level resource situation visualization and decision making; a versioning management engine of component arrangement is applied, a multi-version executable unit is packaged based on an atomization component, and an application instance is completed through a combined modeling component; and deploying a scheduling decision system for deploying a scheduling decision. The method can adapt to task deployment requirements of a current large-scale spaceflight measurement and control system.

The embodiment of the invention provides an energy storage system deployment optimization method and device based on a starfish optimization algorithm. The method is applied to the energy storage system deployment field. The method comprises the steps of establishing an energy storage system multi-objective optimization function and a constraint condition according to an optimization objective; under a constraint condition, solving a multi-objective optimization function of the energy storage system by utilizing a starfish optimization algorithm, and obtaining the position of a starfish individual with the lowest fitness in all starfish individual fitness when a preset iteration termination condition is met by iterating the positions and fitness of starfish individuals; and taking the position of the starfish individual with the lowest fitness as a final deployment scheme of the energy storage system. In this way, the deployment performance and optimization efficiency of the energy storage system are remarkably improved, the service life of energy storage system equipment is prolonged, cost minimization, system reliability and energy utilization rate maximization are achieved, and the optimization process is more flexible, more practical and more applicable.

Linear unretimed interfaces lack clock and data recovery and retiming circuits implemented on a sophisticated digital signal processor. When the interface is not a retimed interface, it can be especially useful to extract some information about the link to allow for debugging and optimization of system deployment. An efficient solution can be implemented in unretimed interfaces to compute metrics such as the impulse response and signal histogram. Having the metrics allows for the solution to check and monitor the quality and the status of the link. Based on the link quality and status information, it is possible to address non-idealities and optimize performance of the unretimed interfaces and the link.

The invention discloses a subway transport capacity matching and passenger flow collaborative management and control method based on multi-agent reinforcement learning, and relates to the technical field of urban rail transit operation management. The method comprises the following components: S1, a subway system digital twin model construction step, S2, a multi-agent reinforcement learning training environment construction step, S3, an extreme scene simulation and agent training step, S4, a virtual-real data interaction and model iteration step and S5, an actual system deployment and decision execution step. According to the method, a high-fidelity digital twin model of a subway system is constructed, extreme scenes including passenger flow retention caused by rainstorm and instantaneous large passenger flow of a large competition field are simulated in the model, and a multi-stage sampling method based on scene clustering is combined to perform training for ten millions of times; and the decision-making capability and the cooperation efficiency of the multiple agents in the face of complex and extreme conditions are obviously improved.

The invention discloses an unmanned aerial vehicle subsystem health assessment method and system, and relates to the technical field of unmanned aerial vehicle health management, and the method comprises the steps: obtaining the data of a multi-source sensor of an unmanned aerial vehicle, screening the features strongly correlated with a fault state through a Spearman rank correlation coefficient, and obtaining key monitoring parameters; according to the self-adaptive mutation mechanism, the virtual evolution strategy and the double-objective optimization strategy, an improved multi-objective genetic algorithm optimization model is constructed, then key monitoring parameters are input, and a visual report of a health state assessment result and confidence rating is obtained. According to the method, the problems that traditional single-target optimization neglects the misjudgment rate, the parameter search efficiency is low and the evaluation credibility is insufficient are solved, accurate evaluation is achieved, the evaluation accuracy is improved to 84.86%, the key misjudgment rate is reduced to 0.009%, the convergence speed is improved by 40% compared with a traditional method, the safety and real-time performance of unmanned aerial vehicle health state recognition are remarkably optimized, and the method is suitable for popularization and application. The method is suitable for embedded system deployment.

The invention discloses a relay unmanned aerial vehicle track optimization method, which comprises the following steps of: firstly, constructing a lossy relay unmanned aerial vehicle communication system model under finite length coding, and covering communication time slot division, unmanned aerial vehicle flight constraint and a channel model; deducing outage probability expressions of two communication strategies under finite length coding, wherein the outage probability expressions are an independent decoding strategy and a joint decoding strategy respectively; then, a trajectory optimization problem is converted into a Markov decision process (MDP), and a state space, an action space, a reward function and a state transition mechanism are clarified; solving by adopting a joint optimization algorithm based on depth deterministic strategy gradient (DDPG) to realize joint optimization of the unmanned aerial vehicle track and the communication strategy; and finally, outputting an optimal unmanned aerial vehicle track and strategy configuration to provide reference for actual system deployment. The method can effectively reduce the outage probability of the system and improve the communication reliability under the condition of limited length coding, and is suitable for various scenes in 6G and future air-ground cooperative communication networks.

The invention relates to the technical field of engineering structure health monitoring and artificial intelligence, and discloses an offshore wind turbine tower drum base three-dimensional stress distribution prediction and visualization method, which comprises the following steps: generating batch working conditions by presetting a bending moment range and an interval, and calling a finite element software automation script to execute static analysis; node stress values are extracted and integrated into a structured training data set; customizing neural network training: constructing a full-connection network, and training a model by adopting a standardized preprocessing and early stop mechanism; generating an interactive cloud picture: generating a 3D stress cloud picture HTML (Hypertext Markup Language) file supporting rotation / zooming / data query based on a spatial interpolation algorithm and a Plotly tool, and automatically marking the coordinate and the numerical value of a maximum stress point; and end-to-end system deployment: packaging as a Flask Web application, and inputting a bending moment value by a user to generate an interactive cloud picture in real time. The stress analysis period is greatly shortened, the engineering application threshold is reduced, and rapid evaluation and remote cooperation of the offshore wind turbine tower drum base can be efficiently supported.

The invention provides a robot operating system and method based on hardware isolation and a dynamic task engine, belongs to the technical field of artificial intelligence, and aims to solve the problem that the real-time performance and the intelligence are bidirectionally opposite for a long time in an architecture level in the field of robots. The system comprises a heterogeneous multi-core SoC, a universal operating system deployed in a non-real-time domain, a kernel real-time operating system deployed in a real-time domain, an inter-core communication bus based on shared physical memory configuration and an integrated development tool chain based on a VSCode plug-in. The method comprises the steps that a robot operating system based on hardware isolation and a dynamic task engine is built; task generation, routing and deployment are carried out; and when the real-time domain executes the task, the high-frequency state data is reported to the non-real-time domain through the inter-core communication bus to form closed-loop cooperative control.

The invention discloses a park large-screen centralized management and control method and system of an intelligent large-screen management platform, and relates to the technical field of intelligent large-screen management, and the method comprises the steps: building the network connection between a platform and a large screen through the configuration of communication parameters and the input of large-screen equipment identification information, and completing the equipment access authentication through the network connection relation; collecting parameter information and an equipment operation state of the accessed equipment; based on the parameter information of the accessed device, setting a content distribution strategy of a display area, and executing format adaptation and split screen template configuration on the multi-format content; and receiving a voice instruction, completing user identity recognition, and executing alarm grading and display content adjustment according to the collected equipment operation state. According to the method, an access mechanism based on automatic matching of communication parameter configuration and equipment identification information is constructed, so that the access efficiency of heterogeneous large-screen equipment and the automation level of system deployment are effectively improved.

The invention discloses an enterprise overseas marketing automation service method based on multi-mode AI cooperation, and relates to the technical field of artificial intelligence and digital marketing. According to the method, enterprise multi-modal data is aggregated through an intelligent data center, an AI model is driven to automatically generate an independent station page and a marketing video adaptive to a target market, and the independent station page and the marketing video are synchronously published to a social media platform. A cross-platform intelligent customer service module is deployed in the system, user consultation is analyzed in real time, a preset notification rule is triggered, and response efficiency and compliance are guaranteed. And dynamically optimizing the display content of the master station based on a multi-modal AI scoring mechanism, and realizing accurate recommendation in combination with user behavior analysis. Compared with a traditional method, the system solves the problems of multi-platform operation splitting, manual response delay, poor content adaptability and the like through multi-AI model collaboration and automatic process integration, and remarkably improves the overseas marketing efficiency and precision.

The invention discloses a park electric energy intelligent management regulation and control method based on the Internet of Things, and belongs to the technical field of community power grid building platforms, and the park electric energy management regulation and control through a management regulation and control platform comprises the steps of firstly installing an intelligent electric meter for data acquisition and monitoring system deployment, and then training a load prediction model by using historical load data, and finally, according to the trained load prediction model, predicting time-sharing electric power of the proprietors in the future. And meanwhile, according to future weather forecast data, a reference electricity price model is established in a linear regression mode, so that the future time-sharing reference electricity price is predicted. After various basic data are prepared, a park power supply and demand model is constructed in a linear programming mode, the total power consumption cost is minimized, in subsequent power supply management, a manual intervention step is added, the power consumption ratio, cost analysis and energy storage state of each owner are displayed through a visual component, a threshold value is set, and when the detected deviation exceeds the threshold value, the power supply management is completed. And prompting an administrator for intervention.

The invention discloses a chirp interval modulation backscatter communication method and a chirp interval modulation backscatter communication system based on variable-length scrambling codes, which mainly solve the problems of high power consumption of backscatter labels and high system deployment cost in the prior art. According to the scheme, a sending end sends an environment LoRa signal; the backscattering label modulates information to be sent to an environment LoRa signal in a time interval modulation mode, and modulates a variable-length scrambling code to an anchor symbol; a receiving end performs frequency shift compensation on the backscattering signal by adopting a frequency shift compensation method, demodulates the backscattering signal by combining a sliding window with a variable-length scrambling code, intercepts a backscattering signal sample through the sliding window, and performs de-spreading and fast Fourier transform (FFT) on the signal in the window; the position of an anchor symbol is determined by detecting the maximum value of an FFT peak value, information sent by the backscattering label is recovered from the position of the anchor symbol, and data demodulation is completed. On the premise of keeping demodulation reliability, the power consumption of the tag is effectively reduced, the deployment cost of a backscatter communication system is reduced, and the method can be used for ultra-low power consumption Internet of Things communication.

The invention discloses an environment deployment method and system of a cloud native electric power auxiliary service market system, and belongs to the technical field of electric power system informatization. The method comprises the following steps: acquiring test environment configuration information and a system running state; constructing an infrastructure and deploying a system service component according to the test environment configuration information, and generating a test environment consistent with the production environment; testing the test case in the test environment to obtain a test result; verifying the system operation state and the test result according to a preset verification rule to obtain a test report; the test report comprises a test success report and a test failure report; when the test success report exists and the test coverage is in accordance with the preset range, allowing the system to be deployed to the production environment; and when the test report does not exist, the test failure report exists or the test coverage does not meet the preset requirement, the system is prohibited from being deployed to the production environment. The problems that the test environment generation efficiency is low and deployment verification is not strict can be solved.

The invention relates to the technical field of non-electrical variable control or adjustment systems, in particular to an unmanned aerial vehicle (UAV) positioning method and system, which comprises an airborne positioning module integrated with an ultra wide band (UWB) transceiver, an inertial measurement unit, a visual sensor module and a global navigation satellite system receiving module, and cooperates with an external UWB positioning beacon. High-precision, continuous and reliable positioning of the unmanned aerial vehicle is realized through core function modules such as multi-modal data real-time quality evaluation, dynamic weight fusion positioning, environment perception and situation recognition, robustness and integrity monitoring and the like. According to the method, the positioning precision and the system robustness can be remarkably improved, the complex environment adaptability and the anti-interference capability are enhanced, and the system deployment and maintenance cost is effectively reduced.

The invention is suitable for the technical field of railway safety, and provides an intelligent and efficient railway signal centralized monitoring and maintenance system, a maintenance platform comprises a three-level networking architecture, and the three-level networking architecture comprises a national railway group level monitoring operation and maintenance subsystem, a bureau group company level monitoring operation and maintenance subsystem and a station level monitoring operation and maintenance subsystem. The national railway group level monitoring operation and maintenance subsystem deploys an active-active data center; the bureau group company-level monitoring operation and maintenance subsystem is configured with a dual-computer hot standby server cluster and comprises an acquisition and analysis unit, an application server and a database server; the station-level monitoring operation and maintenance subsystem is integrated with an edge computing node; the three stages are interconnected through a data communication network; according to the comprehensive maintenance platform of the railway signal centralized monitoring system, an integrated, intelligent and efficient operation and maintenance management system is constructed, comprehensive monitoring and precise maintenance of the railway signal system are achieved through the functions of real-time monitoring, fault early warning, intelligent diagnosis, decision support and the like, safe and stable operation of the railway signal system is ensured, and the service life of the railway signal system is prolonged. And the overall efficiency and service quality of railway operation are improved.

The invention discloses a generator set defect detection method based on deep learning, and belongs to the technical field of generator sets, and the method comprises the following steps: S1, collecting and marking multi-source heterogeneous data; s2, data preprocessing and feature engineering; s3, constructing a deep neural network model: designing and constructing a composite deep learning model architecture fused with multi-modal data processing capability; s4, performing model training and verification; s5, optimizing and lightening the model; s6, edge computing system deployment and real-time detection; s7, carrying out online feedback and model iteration; and S8, generating a visual diagnosis interface and a report. By integrating various heterogeneous data sources such as the vibration sensor, the acoustic sensor and the thermal infrared imager, multi-angle real-time monitoring of the running state of the generator set is realized, and in the acquisition stage, the synchronous acquisition card is utilized to ensure the time precision.