7528 results about "Nerve network" patented technology

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for controlling an agent interacting with an environment. In one aspect, a method comprises: receiving a user text sequence that characterizes a robotic task to be performed by the agent in the environment; converting the user text sequence into an input text sequence in a standardized format for a visual language understanding task; obtaining an observation image characterizing a state of the environment; processing (i) the input text sequence in the standardized format and (ii) the observation image using a policy neural network to generate a policy output that defines an action to be performed by the agent in response to the observation image; selecting an action to be performed by the agent using the policy output; and causing the agent to perform the selected action.

The invention discloses an intelligent driving multi-sensor fusion data processing system, which belongs to the technical field of fusion data processing and comprises a data preprocessing unit, a data fusion unit and a data post-processing unit. The data preprocessing unit normalizes the original data of the laser radar, the camera, the millimeter wave radar and the ultrasonic sensor through a standardization, cleaning, noise reduction and calibration synchronization module; the feature level fusion module deeply digs the characteristics of each sensor, and fuses geometric, visual, motion and close-range obstacle features by using a deep neural network; the decision-making level fusion module generates a driving instruction through weighted voting and fuzzy logic reasoning through classification and situation evaluation; and the data post-processing unit combines the vehicle and road condition optimization instruction, evaluates the risk, and stores the data feedback optimization system. The system can improve the sensing precision and decision reliability, enhances the expansion adaptability of the system, and provides guarantee for safe and efficient operation of intelligent driving.

The invention relates to the technical field of welding defect identification, and discloses a welding defect identification method and system based on a molten pool image. According to the method, a multispectral high-speed camera is used for collecting a molten pool dynamic image sequence in the welding process, after multi-scale morphological filtering preprocessing is conducted, a fusion feature vector is extracted through a depth separable convolutional network, and then the fusion feature vector is input into a defect classification model adopting a heterogeneous graph neural network architecture to obtain a defect probability distribution matrix. Then, constructing a dynamic sparse optimization model to position defects, generating a defect space coordinate set, and finally, outputting welding defect types and position information through hierarchical verification framework processing. According to the method, the problems of welding image noise interference, complex defect characteristics and the like are effectively solved, the accuracy and reliability of welding defect identification are improved, and powerful technical support is provided for welding quality control.

The invention provides a vehicle collision risk early warning method, system and device and a medium, and belongs to the technical field of vehicle safety control. The method comprises the steps that surrounding environment information of a vehicle is collected in real time through a vehicle built-in sensor and a GPS module; a multi-sensor fusion device technology is used to carry out comprehensive processing on surrounding environment information of a vehicle, and an environment perception model is generated after data integration; based on the environment perception model, target identification is carried out by using a computer vision technology based on a convolutional neural network so as to identify travelers, vehicles, traffic signs and obstacles; performing behavior prediction based on a deep learning model to predict a behavior track of a front vehicle or pedestrian; evaluating a potential collision risk in the current scene in combination with the target recognition result and the behavior prediction result; when a potential collision risk is detected, an early warning signal is sent to a driver, and when the driver does not take corresponding measures, active safety measures are automatically triggered to avoid collision or reduce accident damage.

The invention relates to the technical field of power battery packs, and discloses a power battery pack state evaluation method and system, and the method comprises the steps: collecting an operation characteristic data set of a power battery pack under various working conditions, and constructing a hybrid evaluation model; performing parameter configuration on the double-extended Kalman filter to obtain an improved filter; acquiring real-time voltage, current and temperature data, and inputting the real-time voltage, current and temperature data into the improved filter for state calculation to obtain a charge state value and an internal parameter value; inputting the charge state value and the internal parameter value to perform multi-branch adversarial neural network analysis to obtain a fault mode type and a fault severity; according to the method, the multi-factor health index is calculated, particle swarm optimization fitting is carried out on the degradation curve of the power battery pack, the residual service life prediction value is obtained, more accurate residual service life prediction is further achieved, the precision in a wide temperature range is remarkably improved, and the difference between monomers in the power battery pack can be effectively processed.

The invention relates to the technical field of agricultural intelligent detection, in particular to a fruit tree disease and pest detection method based on machine vision, which comprises the following steps: acquiring visible light, infrared light and ultraviolet light images of a fruit tree at multiple angles through a multispectral camera to obtain complete fruit tree surface information, and preprocessing the acquired images to obtain a fruit tree disease and pest detection result; comprising the steps of denoising, illumination correction and image alignment, extracting feature vectors, including the color, shape and size of scabs, pest distribution and other features, of pest and disease damage from preprocessed images through a convolutional neural network, fusing the pest and disease damage features with environmental data by using a multi-modal data fusion algorithm, and enhancing the accuracy of pest and disease damage classification. Classifying and identifying diseases and pests based on a support vector machine classification algorithm; according to the dynamically adjusted threshold value and suggestions provided by the expert system, a disease and pest diagnosis report is generated, and prevention and control measure suggestions are provided; according to the invention, efficient and accurate detection and management of fruit tree diseases and insect pests can be realized, and the intelligent level of agricultural management is improved.

The invention discloses a concrete building outer wall detection method and system, computer equipment and a storage medium, and relates to the technical field of prefabricated building detection, and the method comprises the steps: obtaining multi-angle image data of a concrete building outer wall; preprocessing the image data, and extracting a target area of the outer wall of the concrete building; extracting feature information representing outer wall defects from the preprocessed image data; inputting the feature information into an artificial intelligence detection model, and judging whether an outer wall defect exists or not; and according to an output result of the detection model, determining distribution positions of the defects on the surface of the outer wall, and classifying the detected defects. According to the concrete building outer wall detection method provided by the invention, the unmanned aerial vehicle avoids low efficiency and safety risks of high-altitude operation in manual detection. Meanwhile, the convolutional neural network can automatically extract multi-level features from the image, various defects are accurately recognized, and the accuracy problem of traditional infrared detection under the change of environmental factors is solved.

The invention relates to the technical field of data processing, in particular to a distributed multi-source heterogeneous sensor data processing method and system. The method comprises the following steps: collecting environmental parameters of a leakage area in real time through a distributed sensor array; performing coordinate system unification and timestamp alignment on the environmental parameters of the leakage area to generate a standardized leakage situation data set; extracting gas diffusion characteristics and thermal imaging characteristics of the standardized leakage situation data set to calculate a gas cloud cluster diffusion path probability; performing risk decision instruction generation on the gas cloud cluster diffusion path probability based on a preset leakage level classification neural network to obtain a risk decision instruction; calling a matched emergency plan based on the risk decision instruction; and analyzing the implementation steps of the emergency plan and performing instruction conversion to generate an emergency plan instruction. According to the invention, through real-time data acquisition, intelligent risk assessment and automatic emergency response, the timeliness responsiveness of data processing is improved.

The invention discloses a partial discharge detection method and device, and the method comprises the steps: outputting a multi-mode signal matrix after noise reduction through employing a self-adaptive noise reduction algorithm according to a multi-mode sensing signal during the operation of a generator; based on the multi-modal signal matrix, a space-time convolutional neural network is adopted to extract discharge features, meanwhile, a topological relation between signal time-frequency features and sensor space is captured, and a multi-modal feature fusion tensor is output; according to the multi-modal feature fusion tensor, a model is generated through a dynamic map, manifold learning and a particle swarm optimization algorithm are combined, and a dynamic fault map containing discharge intensity, phase and frequency point distribution characteristics is output; and based on the dynamic fault map, calculating a real-time discharge danger coefficient by using a risk prediction model, and outputting a discharge grading early warning instruction and a maintenance priority sequence. According to the embodiment of the invention, high-precision and traceable discharge detection and graded early warning can be realized.

The invention relates to a control method and device for wind load resistance of a mooring unmanned aerial vehicle and a storage medium. The method comprises the following steps: collecting and modeling a mooring unmanned aerial vehicle to obtain a system dynamics model; state estimation is carried out to obtain system wind load disturbance data and expected attitude data; performing optimal control solution on the plurality of mooring points to obtain tension distribution control data of each mooring point; inputting into a discrete time neural network controller for attitude compensation calculation to obtain thrust vector compensation control data; performing adaptive robust control on the attitude of the mooring unmanned aerial vehicle to obtain a rotating speed control instruction of each motor of the four rotors; and performing safety threshold judgment, and when any monitoring data exceeds a preset threshold, outputting a working mode switching instruction and executing safe landing control. According to the invention, a multi-stage safety protection mechanism is established, and reliable operation of the mooring unmanned aerial vehicle under extreme conditions is ensured through real-time monitoring and mode switching.

The invention relates to the technical field of intelligent control, in particular to a digital control method and system applied to on-line operation of multiple punching machines, and aims to acquire historical data of a production line, acquire operation state data of the punching machines in real time, realize product appearance defect detection through a YOLOv7 model in combination with product appearance images and quality inspection parameters, and improve product quality. Generating an offset quality inspection parameter in combination with a quality inspection standard; and generating a process abnormality positioning result by using the offset quality inspection parameters and the appearance defect detection result. And fusing the abnormal process positioning result with the operation state data of the punching machine to form multi-mode comprehensive operation characteristic data which is used for identifying the abnormal process. And after an abnormal process is detected, analyzing an abnormal propagation path through a dynamic working condition adaptive graph neural network model, extracting current working condition features, performing similarity matching with historical data, and extracting optimal historical reference parameters. And generating a production process adjustment parameter by dynamically correcting the historical reference parameter. And the adaptive capacity and the flexible response capacity of the production line to complex working conditions are improved.

The invention relates to the technical field of signal devices, and discloses a traffic signal control method and system based on vehicle-road cloud multi-modal data fusion, and the method comprises the steps: collecting multi-modal traffic data synchronously in real time through a vehicle-end sensor, road-side sensing equipment and a cloud Internet platform; fusing the heterogeneous data by adopting a space-time alignment algorithm, and constructing a standardized space-time feature matrix; traffic flow prediction is carried out based on a multi-layer space-time diagram neural network trained by a federated learning mechanism, and a signal control instruction is generated through reinforcement learning and a multi-objective optimization model; and issuing the green wave parameter, the dynamic timing scheme and the cross-domain coordination strategy to a roadside signal machine through the cloud edge coordination architecture to execute control. The problems that in the prior art, low-delay private network communication cannot be achieved, the data fusion efficiency is low, unmanned driving is not supported, and the deployment cost is high are solved, and the purposes of low-delay communication, high reliability and low risk are achieved.

The invention relates to the technical field of mobile energy storage vehicles, and discloses an AI-based mobile energy storage vehicle running state monitoring and analysis system which comprises a main controller, an AI monitoring coprocessor and a state analysis coprocessor. The main controller calls a monitoring analysis instruction and sends the monitoring analysis instruction to the state analysis coprocessor; an AI monitoring coprocessor dispatches a state analysis instruction and processes multi-modal data; and the state analysis coprocessor analyzes the parameters to generate a reference monitoring signal, cooperatively monitors the battery temperature, the charging and discharging efficiency and the load fluctuation in real time, and generates an abnormal correction signal and a state control instruction in combination with an analysis optimization mode. The system realizes accurate analysis of multi-source data and dynamic control of the energy storage unit through data preprocessing, feature dimension reduction and deep neural network prediction, improves monitoring accuracy, dynamic adaptability and intelligent level, and ensures safe and efficient operation of the mobile energy storage vehicle.

The invention relates to the technical field of health monitoring, in particular to a cardiovascular disease data fusion and multi-dimensional health assessment method, which comprises the following steps of: acquiring multi-source data of a patient through a sensor and wearable equipment, preprocessing the acquired data, ensuring data integrity, extracting physiological features by adopting a convolutional neural network, and evaluating the health of the patient. The method comprises the steps of extracting non-physiological data features through principal component analysis, constructing a multi-dimensional feature set, identifying cardiovascular disease related risk factors through a random forest algorithm, distributing weights through a logistic regression algorithm to generate individualized risk scores, analyzing time sequence trends of the risk factors based on a long and short-term memory network model, and generating dynamic risk scores in real time. When the score exceeds a preset threshold value, an early warning mechanism is automatically triggered, and a patient and medical staff are notified through an application program; according to the method, accurate assessment and personalized intervention of cardiovascular health are realized through multi-source data fusion and dynamic analysis, and the method has high efficiency and timeliness.

An agricultural vehicle includes a propulsion system configured to move the agricultural vehicle, a steering system configured to orient the agricultural, a navigation controller operably coupled to the propulsion system and the steering system, and a camera and a radar operably coupled to the agricultural vehicle. A computing device is operably coupled to the radar and the camera and includes instructions that, when executed by at least one processor, cause the computing device to receive image data from the camera, receive radar data from the radar and generate a radar point cloud based on the radar data, identify one or more objects in the image data using a trained neural network to generate labeled image, fuse the labeled image data with the radar point cloud to obtain fused data, and perform one or more navigation operations based on the fused data. Related agricultural vehicles and methods are also disclosed.

The present invention discloses a dyskinesia rehabilitation training method and system based on electroencephalogram signal recognition; the method includes: the convolutional neural network is used to recognize the electroencephalogram signal, analyze the patient's motor intention, convert the motor intention into a control instruction for the exoskeleton device, drive the patient's limb movement by controlling the movement of the exoskeleton device, and assist the patient to complete the movement disorder rehabilitation training. In this method, the self-attention mechanism and self-distillation training were used to establish an electroencephalogram signal recognition model, analyzing the patient's motor intention to help motor nerve remodeling, which broke the passive and single problem of traditional rehabilitation methods, realized active rehabilitation of patients, and significantly improved the rehabilitation efficacy.

The invention belongs to the technical field of power distribution network voltage regulation and control, and discloses a distributed photovoltaic complex power prediction-cluster-cooperation-based power distribution network voltage regulation and control method, which integrates photovoltaic historical data, inputs an improved back propagation neural network model and outputs predicted photovoltaic active power output. Estimating the reactive capacity boundary of each node in real time based on the running state of the network-following inverter; dividing a distributed photovoltaic cluster by establishing a two-dimensional modularity function of a net load index and an equivalent electrical distance; a multi-device differential cooperative control strategy is provided for the voltage out-of-limit risk in the cluster; and constructing an optimization function with minimum network loss and voltage offset as a target, and optimizing and solving the function by using an improved multi-organization particle swarm optimization algorithm to obtain a multi-device adjustment sequence and a device action amount. According to the method, the renewable energy consumption capacity is improved and the network loss is reduced while the voltage stability of the power distribution network is ensured, and the comprehensive adjustment cost is optimized.

The invention relates to the technical field of image processing, and discloses a device and a method for recognizing longitudinal tearing of a mining conveying belt, which are used for improving the accuracy of recognizing the longitudinal tearing of the mining conveying belt. The device comprises: an acquisition module performing structured light projection and image acquisition to obtain original image data; the extraction module performs preprocessing and contour line extraction on the original image data to obtain geometric feature information; the generation module performs feature point analysis and stress distribution analysis on the geometric feature information to generate dynamic change features and a stress distribution diagram; the fusion module carries out fusion analysis on the sensing data set to obtain conveyor belt tearing risk evaluation data; the analysis module performs tearing probability distribution analysis based on a multi-modal neural network on the conveyor belt tearing risk evaluation data to obtain tearing identification probability distribution data; and the early warning module performs early warning information analysis according to the tear identification probability distribution data to obtain target early warning information, and performs real-time early warning according to the target early warning information.

The invention relates to the field of power equipment, and discloses a power equipment fault early warning system, which comprises a data acquisition module, a data fusion module, a state evaluation module, a trend prediction module, an early warning judgment module and an information interaction module. Key operation parameters are cooperatively acquired through multiple types of sensors, time series data are uniformly calibrated by adopting a timestamp mechanism, the problems of fragmentation of operation state information of power equipment and superposition of acquisition errors are effectively solved, and then feature fusion and dimension reduction compression are performed on high-dimensional heterogeneous data by introducing a principal component analysis and auto-encoder neural network, so that the operation state information of the power equipment is acquired. According to the method, redundant information is eliminated, meanwhile, key discrimination features are reserved, the sensing dimension of the system for the equipment operation state is more comprehensive, the representation capacity is higher, the Bayesian network and the support vector machine are adopted to jointly evaluate the equipment state health level, higher state recognition accuracy is achieved in a dynamic scene, and the method is suitable for popularization and application. And the model generalization ability is enhanced through historical samples, so that the equipment state can be judged more stably.

The invention provides an optical storage and charging cooperative control method and system based on multi-energy complementation. The method comprises the following sub-steps: respectively establishing a photovoltaic power generation model, an energy storage system model and a charging load model; acquiring historical illumination information, charging information and electricity price information, constructing a prediction model based on a neural network, and predicting and outputting illumination intensity, charging load demand power and electricity price in a future time period; constructing a target optimization function by taking the annual net cost and the power deviation rate as optimization targets; according to the method, the cooperative optimization control of the photovoltaic power generation, energy storage and charging system is realized, the annual net cost is reduced, the power deviation is reduced, the energy storage and charging cooperative control strategy is established, the target optimization function is solved by adopting the improved whale optimization algorithm, and the charging and discharging power sequence is generated according to the optimal solution obtained by solving and is input to the system for execution. And the operation efficiency and reliability of the whole system are improved.

The invention discloses an intelligent water and fertilizer integrated precise irrigation control method. The method comprises the following steps: step 1, establishing a multi-modal data acquisition network; step 2, inputting the data obtained in the step 1 into a trained LSTM neural network, and predicting future crop transpiration water demand; 3, according to nutrient concentration distribution fed back by a soil conductivity sensor in real time, combining with a water and fertilizer coupling migration model, dynamically calculating a nitrogen-phosphorus-potassium ratio of each irrigation zone, realizing precise mixing of a fertilizer mother solution and irrigation water through a three-channel proportioning pump, and implementing a variable irrigation decision; 4, dynamically adjusting the pulse irrigation duration of each branch through an MPC algorithm; and step 5, recovering the residual liquid at the tail end of the pipe network to a buffer tank, detecting the EC value of the recovered liquid through an electrochemical sensor, and taking the EC value as a reference parameter automatic correction proportioning algorithm of the next irrigation period. Accurate control over the irrigation process can be achieved, so that the water and fertilizer utilization efficiency is improved, and resource waste and environmental pollution are reduced.

The invention relates to the technical field of building appearance defect detection, in particular to a building appearance defect detection method and system based on an unmanned aerial vehicle, and the method comprises the steps: obtaining the building information of a target building, and generating a hierarchical scanning path and a three-dimensional obstacle avoidance flight path, a visible light image, an infrared thermodynamic diagram and laser radar point cloud information are collected for space-time alignment processing, and an attention mechanism neural network is used for extracting multi-scale features to generate a detection report containing defect three-dimensional coordinates, damage levels and safety risk assessment. The method achieves the purpose of efficiently and accurately detecting the building appearance defects, can adapt to complex building structures and environmental conditions, supports defect trend prediction and maintenance decision, and remarkably improves the building safety management efficiency.

The invention discloses a hanging basket transportation safety management method and system, and relates to the field of damage detection. Collecting weight data of the hanging basket; establishing a steel wire rope health state prediction model by using a multi-layer perceptron (MLP) algorithm, predicting the damage degree of the steel wire rope, and obtaining a health state prediction result of the steel wire rope; according to the health state prediction result of the steel wire rope, a bidirectional long-short-term memory neural network BiLSTM is used for prediction, and a damage trend prediction result of the steel wire rope in a future period of time is obtained; comparing the collected weight data of the hanging basket with a preset safety weight range to obtain a judgment result of whether the weight of the hanging basket exceeds the limit or not; according to the health state prediction result of the steel wire rope, the damage trend prediction result and the judgment result of whether the weight of the hanging basket exceeds the limit or not, a decision tree algorithm is adopted to generate a hanging basket operation control scheme; aiming at low safety of a hanging basket in the prior art, the safety of the hanging basket is improved.

The invention discloses a multi-sensing and data physical fusion frozen soil hot water force response test system and method, and the system integrates an ultra-weak fiber grating, an active heating fiber, a miniature dielectric constant sensor and an optical frequency domain reflection technology, and constructs a freezing process-oriented temperature, moisture, ice content and strain synchronous monitoring network; the method comprises the steps of constructing a multi-field fusion data set based on thermal disturbance lag correction, temperature-strain decoupling and a moisture-strain residual term compensation mechanism, and extracting key criterion characteristics of an ice lens growth rate, frost heaving force evolution and a shear deformation change rate; and in combination with a physical information neural network (PINN) embedded into a hot water force control equation, risk identification and grade judgment of the freezing abnormal behavior are realized. The method supports indoor heating and water pressure linkage adjustment, realizes a complete closed loop of multi-source sensing-risk identification-response verification under model driving, and breaks through the bottlenecks of low multi-physics field coupling identification precision, large parameter cross interference and insufficient risk identification real-time performance of the existing monitoring technology.

A forest point cloud reconstruction method and system based on combination of a handheld radar and an unmanned aerial vehicle, relating to the technical field of remote sensing. The method comprises: on the basis of an unmanned aerial vehicle and a handheld radar device, using a multi-band radar sensor to perform forest terrain and vegetation information acquisition to generate original radar data. The use of the multi-band radar sensor improves the data acquisition accuracy and the coverage range, so that forest terrain and vegetation information is recorded more comprehensively and accurately; a generative adversarial network improves the data quality, and reduces the noise, thereby ensuring the accuracy of subsequent analysis; a convolutional neural network and a random forest algorithm are combined to improve the accuracy of vegetation classification and terrain feature recognition; the generation of comprehensive environmental feature data integrates multidimensional environmental parameters, providing richer information for in-depth analysis; a three-dimensional point cloud model visually displays a forest structure; and time series analysis and a long short-term memory network improve the accuracy of forest change trend prediction, and provide decision support for forest management strategies.

Apparatuses, systems, and techniques are presented to predict gaze of an observer. In at least one embodiment, a network is trained to predict a gaze of one or more users based, at least in part, on one or more gazes corresponding to objects not always visible to the one or more users.

The invention provides a new energy photovoltaic dynamic inspection method and system based on artificial intelligence, and relates to the technical field of photovoltaic power station intelligent inspection. Inspection is triggered according to weather early warning, performance warning or timed tasks; initial path planning is carried out by combining terrain, weather and historical data, and the path is updated by dynamic obstacle avoidance through an RRT * algorithm; multi-modal data, including visible light images, infrared thermal imaging, EL detection data and positioning data, are acquired during inspection of the unmanned aerial vehicle; the unmanned aerial vehicle data and the ground sensor data are integrated to generate a unified fault feature matrix; positioning a defect area in real time by using a deep neural network, judging a defect type and dividing a fault level; and finally, the health degree of the photovoltaic system is scored according to the fault level, and the safe operation trend is analyzed. The multi-modal data real-time fusion and dynamic path planning are realized, the fault identification precision and the inspection efficiency are improved, the manual inspection cost and risk are reduced, and powerful support is provided for intelligent operation and maintenance of a photovoltaic system.

A system configured to improve audio processing by performing dereverberation and noise reduction during a communication session. In some examples, the system may include a deep neural network (DNN) configured to perform speech enhancement, which is located after an Acoustic Echo Cancellation (AEC) component. For example, the DNN may process isolated audio data output by the AEC component to jointly mitigate additive noise and reverberation. In other examples, the system may include a DNN configured to perform acoustic interference cancellation, which may jointly mitigate additive noise, reverberation, and residual echo, removing the need to perform residual echo suppression processing. The DNN is configured to process complex-valued spectrograms corresponding to the isolated audio data and / or estimated echo data generated by the AEC component.

The invention provides a road slope data monitoring method under a complex geological condition, and relates to the technical field of road slope monitoring, and the method comprises the steps: carrying out the temperature compensation of a change rate, analyzing the thermal expansion characteristic of a material, calculating the deformation caused by the temperature change, correcting the monitored side length, and generating a temperature compensation side length; according to the temperature compensation side length and the surrounding rock rheological equation, correcting the two adjacent side lengths through the strain difference of the adjacent sides so as to obtain the final side length; combining the final side length with the reference interior angle to generate a topological structure; based on the topological structure, constructing a three-dimensional deformation index matrix containing a diagonal displacement difference, an adjacent side curvature ratio and an interior angle variation coefficient; and inputting the index matrix, the support structure strain space-time matrix and the rock mass fragmentation tensor into a space-time attention graph neural network together, and outputting a deformation evolution trend including a diagonal displacement difference, an adjacent side curvature ratio and an interior angle variation coefficient. According to the invention, high-precision real-time monitoring can be realized.

The invention relates to the technical field of air compressors, and discloses a centrifugal air compressor surge suppression method and device.The method comprises the steps that pressure pulsation feature extraction is conducted on a pressure signal of a centrifugal air compressor, and pressure pulsation features are obtained; extracting a flow statistical feature and a frequency band energy feature of the flow signal to obtain a surge early warning feature space; carrying out characteristic decomposition on the bearing vibration signal, and extracting a characteristic frequency to obtain a surge early warning characteristic set; inputting the surge early warning feature set into a three-channel neural network for surge early warning prediction to obtain a surge early warning level; according to the surge early warning grade, control parameters of a guide vane angle controller, an anti-surge valve opening controller and a return valve opening controller are generated, the guide vane angle, the anti-surge valve opening and the return valve opening are synchronously adjusted, and a surge margin value is calculated. According to the method, accurate control over the guide vane angle, the anti-surge valve opening and the return valve opening is achieved, the surge margin value is calculated, and the control accuracy of the guide vane angle, the anti-surge valve opening and the return valve opening is improved. And real-time optimization and adjustment of the control strategy are ensured.