285results about How to "Achieve co-optimization" patented technology

The invention relates to the technical field of multi-axis linkage numerical control machining, and discloses a multi-axis linkage numerical control machining state monitoring method and system.The method comprises the steps that vibration, acoustic emission and cutting force signals in the multi-axis linkage numerical control machining process are collected and preprocessed, and first multi-source sensor data are obtained; performing feature extraction and compressed sensing to obtain a low-dimensional feature subset; performing fusion judgment to obtain an abnormal state detection result; training a deep learning model comprising a backbone network and a plurality of branch networks to obtain a state evaluation model; and inputting second multi-source sensor data collected in real time into the state evaluation model for state evaluation to obtain a processing state classification result and a parameter estimation value. According to the method, cooperative state evaluation and prediction simulation of multiple machine tools are realized.

The invention relates to the technical field of image classification, and discloses a breast image classification evaluation method and system based on deep learning, and the method comprises the steps: carrying out the preprocessing of sample breast image data, and obtaining comprehensive feature data; performing feature enhancement based on the comprehensive feature data to obtain enhanced feature data; performing multi-modal bidirectional feature fusion on the enhanced feature data to obtain multi-modal fusion feature data; based on the multi-modal fusion feature data, constructing and training a semi-supervised local feature reservation dense connection context-aware graph convolutional network to obtain a graph volume integral class model; according to the multi-modal fusion feature data, constructing and training a multi-branch residual attention reconstruction network to obtain a multi-scale feature fusion model; through the image volume integral class model and the multi-scale feature fusion model, classification evaluation and prediction result integration are carried out on the real-time mammary gland image data to obtain a target mammary gland image classification result, and the method improves the accuracy and reliability of mammary gland image classification.

The invention discloses a tunnel segmented exhaust method and system based on train working conditions, a medium and a product, and relates to the technical field of ventilation. The method comprises the steps that a target tunnel is divided into a plurality of control sections according to fan installation configuration of the target tunnel; acquiring train passing data and train configuration information of the control section, and calculating a flue gas growth value of the control section in a preset time period; collecting flue gas concentration data of the control section, and calculating a ventilation demand index of the control section in combination with the flue gas growth value; collecting wind direction parameters of the control sections, determining a ventilation control relationship between the adjacent control sections based on the ventilation demand indexes and the wind direction parameters, and generating a ventilation cooperation scheme; according to the ventilation cooperation scheme, operation parameters of the tunnel fans in all the control sections are calculated, and control instructions corresponding to the operation parameters are sent to the tunnel fans. By implementing the application, the fan operation can be adjusted according to the actual train operation condition, and the ventilation effect is optimized.

The invention relates to the technical field of solid state disks, and discloses a load collaborative scheduling method and system for a solid state disk, and the method comprises the steps: carrying out the data collection of a main controller of the solid state disk, and obtaining the original feature data; a Kolmogorov-Arnold neural network is input to carry out nonlinear feature transformation and linear combination operation, and a load prediction result is obtained; performing hierarchical processing to obtain a hierarchical scheduling model; constructing a multi-objective optimization function and carrying out constraint setting to obtain a constraint optimization equation set; performing convex difference function decomposition and augmented Lagrange function construction on the constraint optimization equation set, and solving through an alternating direction iteration method to obtain a scheduling decision scheme; according to the scheduling decision scheme, load balancing control is carried out on data write-in physical address allocation, data read cache scheduling and multi-channel parallel transmission, a real-time performance monitoring index is obtained, and the method realizes load collaborative scheduling of the solid state disk.

The invention provides a task allocation and optimal scheduling method and system for a multi-robot collaborative industrial assembly line, and relates to the technical field of robots, comprising the steps of constructing a task allocation strategy network through deep reinforcement learning, encoding workpiece types and attribute information into a state vector input network, a multi-objective optimization function including total assembly time minimization, load balance degree maximization and energy consumption minimization is designed, a near-end strategy optimization algorithm is adopted to train network parameters, a robot task mapping matrix is output as a distribution scheme, and the network parameters are dynamically optimized and updated in an online learning mode. The task allocation efficiency of the multi-robot collaborative assembly production line can be remarkably improved, the dynamic optimization scheduling of the production line is realized, the system energy consumption is reduced, and the assembly efficiency is improved.

The invention relates to the technical field of panoramic video processing, in particular to a multi-modal fusion panoramic video intelligent content analysis method, which comprises the following steps of: acquiring a panoramic video frame and a synchronous audio thereof; performing spherical projection transformation on the panoramic image data, and converting the equidistant projection image into a spherical image; generating a plurality of plane images based on the spherical image; performing candidate target position detection based on the plurality of plane images; based on a candidate target position detection result, an image target detection model is utilized to realize instance segmentation of a target, and target features are extracted to generate a target embedded description vector sequence; based on the synchronous audio, an audio recognition processing module is used for carrying out time sequence information extraction, and an acoustic attention sequence is formed; a spectrum sensor is adopted to realize association fusion of the target embedded description vector sequence and the acoustic attention sequence, and an audio and video collaborative attention sequence is generated; and based on the audio and video collaborative attention sequence, establishing an audio and visual event identification model, and obtaining event content described by the panoramic content.

The invention provides a wireless communication and computing resource collaborative optimization method and system. The method provided by the invention comprises the following steps: establishing an initialized global model matched with a business demand for each service provider, and matching service resources of each service provider according to the business demand; establishing a service agent corresponding to each service provider; each service agent generates an optimal strategy based on the corresponding service resources and the state information of other service providers; sending the initialized global model to a distribution client according to an optimal strategy; each distribution client trains the received to-be-trained model based on the distributed service resources, and sends the trained model to the corresponding service provider; and aggregating the received trained models by each service agent to obtain a global model of the corresponding service. The method and the system provided by the invention are used for realizing collaborative optimization of wireless communication and computing resources and improving the overall performance and the resource utilization efficiency of federated learning.

The invention provides an intelligent water-saving control method and system for an agricultural irrigation device, and the method comprises the following steps: generating a farmland state prediction index based on soil comprehensive information, weather information and crop information, and generating a farmland irrigation strategy; calculating an irrigation emergency degree based on the farmland state prediction index; if the irrigation emergency degree reaches the highest emergency degree, the filtering irrigation module is controlled to execute alternate irrigation operation; if the irrigation emergency degree does not reach the highest emergency degree, calculating to obtain the backwashing emergency degree of the filtering irrigation module; if the backwash emergency degree is greater than or equal to the irrigation emergency degree, generating a backwash strategy, and sequentially executing the backwash strategy and the farmland irrigation strategy; if the backwash emergency degree is smaller than the irrigation emergency degree, the filtering irrigation module is controlled to execute alternate irrigation operation on the target farmland. According to the invention, intelligent irrigation and backwashing control can be realized, and a better water-saving effect is achieved.

The invention belongs to the technical field of human body posture estimation, and discloses an ultra-low light human body posture estimation method based on image frequency division feature enhancement, and the method comprises the following steps: collecting ultra-low light human body images to construct a training data set; constructing an ultra-low light human body posture estimation model based on image frequency division feature enhancement; in the training stage, training data is enhanced by using a frequency sensing image enhancement module, and then the training data is input to a network trunk for training to obtain a trained model; and a test stage: inputting a test data set into the trained network model to obtain a human body posture estimation result. The network model provided by the invention can guide the network to generate richer and clearer image features by using the frequency perception image enhancement module, and collaborative optimization of image enhancement and human body posture estimation is realized, so that the accuracy of extremely low light human body posture estimation is improved.

The invention discloses a Chinese named entity recognition method based on stacked grid structure information enhancement, and belongs to the technical field of natural language processing and named entity recognition. A stacked grid structure and a hierarchical compressed sensing module are used for introduction and lossless compression of word vectors, a long input sequence and information loss are avoided, and the recognition efficiency is improved. Meanwhile, the vocabulary fusion process and the context feature coding process are decoupled to reduce unnecessary attention calculation, and the calculation efficiency is improved; font structure information of Chinese characters is introduced as a supplement, the relevance of entities on multiple granularities of fonts, characters and words is mined through multiple cross attention, and the context representation ability of a text is enhanced; an entity recognition method based on duplex collaboration is used, collaborative learning between span extraction and label classification is achieved through soft label constraint and a multi-task joint scoring function, label classification training is completed through full samples, error propagation caused by span extraction errors is reduced, and the recognition accuracy is improved.

The invention provides a federated learning-based cloud edge size model collaboration method and system, and electronic equipment, and the method makes full use of cloud edge resources, releases the size model capability: a large-scale model is trained at a cloud end, a small model suitable for a local scene is trained by an edge server and terminal equipment, and through a federated learning mechanism, a large-scale model is trained at the cloud end; and model collaborative optimization among different levels of nodes is realized. Data are available and invisible, data privacy and safety are guaranteed, private data of clients on the edge side do not need to be out of the domain, the cloud side shares the model capacity, the data are available and invisible, and data privacy and safety are guaranteed. The invention provides an efficient, flexible and extensible model training and optimizing method, computing power resources of cloud edges can be fully utilized, user privacy is protected, and a global large model is continuously optimized.

The invention relates to the technical field of sound console, and discloses a balance adjustment method, system and device for a sound console volume controller, and a medium. The method comprises the following steps: processing an audio signal of an input end of the sound console to obtain an initial system parameter matrix; performing probability density function calculation on the initial system parameter matrix to obtain an audio state data matrix and a noise feature data matrix; performing first-order volume control calculation based on the audio state data matrix to obtain a volume control data matrix; performing second-order frequency equalization calculation according to the volume control data matrix to obtain a frequency equalization control data matrix; performing iterative optimization calculation on the control parameters to obtain an optimal control parameter matrix; and performing dual-channel parallel calculation on the transfer function of the volume control channel and the transfer function of the frequency equalization channel to obtain an audio signal of the output end of the sound console, thereby realizing independent adjustment of volume control and frequency equalization, and ensuring high quality of audio output.

The invention provides an edge computing resource allocation method and system based on multi-objective optimization, and relates to the field of data processing, and the method comprises the steps: determining the resource demand feature of each unmanned retail store and the resource demand association feature between any two unmanned retail stores according to the historical computing resource demand information of a plurality of unmanned retail stores, determining an edge calculation layout scheme, and setting a plurality of edge calculation devices; predicting future computing resource demands of the plurality of unmanned retail stores according to the resource demand feature of each unmanned retail store, the resource demand association feature between any two unmanned retail stores and the demand prediction auxiliary information; acquiring resource allocation auxiliary information; according to the method, the edge computing resource allocation scheme is determined according to the future computing resource demands and the resource allocation auxiliary information of the unmanned retail stores, the edge computing resources are allocated to the unmanned retail stores, and the method has the advantage of improving the flexibility of edge computing resource allocation in an unmanned retail scene.

The invention discloses an energy storage system control method based on multi-target cooperative operation. The method comprises the following steps: acquiring state data of an energy storage system in a current control period and state data of the energy storage system connected with a power grid; based on the state data of the energy storage system and the state data of the power grid, calculating an operation state comprehensive evaluation value of the energy storage system in the current control period; and determining a priority regulation and control target based on the comprehensive evaluation value of the operation state in combination with the state data of the energy storage system and the state data of the power grid, and controlling the energy storage system to regulate and control the priority regulation and control target in the next control period. A plurality of regulation targets of the energy storage system are comprehensively considered, and the comprehensive evaluation value of the operation state is calculated according to the real-time state data of the system to determine the priority regulation target, so that the energy storage system is promoted to regulate the corresponding targets in a targeted manner in the next period, the relationship among different targets is effectively coordinated, the one-sidedness caused by single target regulation is avoided, and the regulation efficiency is improved. And overall collaborative optimization is realized.

The invention discloses an optimization system for intelligent operation and maintenance of a new energy generator set. The optimization system comprises a data acquisition module for acquiring operation state data and external environment data of the generator set through a sensor network; the data processing and storage module is used for carrying out preliminary data processing by utilizing edge computing equipment; the algorithm model module comprises a maintenance optimization model and a dynamic energy efficiency optimization model based on reinforcement learning; the invention discloses an optimization method for intelligent operation and maintenance of a new energy generator set. The optimization method comprises the following steps: S1, data acquisition and preprocessing; s2, constructing a maintenance optimization model based on reinforcement learning; s3, constructing a dynamic energy efficiency optimization model; s4, fault prediction and maintenance strategy optimization; and S5, system integration and feedback execution. According to the method, the operation parameters of the generator set are adjusted in real time, it is guaranteed that the generator set is always in the optimal operation state under different external environment conditions, and the dual purposes of maximum utilization of resources and minimum energy consumption are achieved.

The invention provides a small sample image classification method fusing global and adaptive local information, and relates to the technical field of image processing, and the method comprises the steps: S1, obtaining small sample images to establish a data set, and dividing the data set into a support set and a query set; s2, inputting the data set into a global sensing network DCCP-Net for pre-training; s3, inputting the support set and the query set into a task aware network QSA-TAN for meta-training; s4, inputting a pre-training result and a meta-training result into a global-local knowledge collaborative network GLKC for joint training to output probability distribution differences of the small sample images, and classifying the small sample images according to the probability distribution differences to obtain a classification result; according to the method, through collaborative optimization of knowledge in the global data space and the local task space, image characteristics and category information to which the images belong can be better highlighted, generalization of a small sample image classification model is remarkably improved, and therefore the intra-class change problem of the images is effectively solved.

The invention discloses an electric power-traffic-information network joint planning method and system, and the method comprises the steps: firstly, building a communication network model considering the traffic flow in consideration of the demands of a large-scale growth intelligent electric vehicle for communication; the method comprises the following steps: firstly, simulating the influence of node traffic flow on communication requirements under a user equilibrium condition by coupling a traffic network and an information network, secondly, effectively coupling a power-traffic-information network through the power consumption of infrastructures in the traffic network and the information network, establishing a joint planning model in order to minimize the total construction cost and the operation cost, and finally, establishing a joint planning model. And converting the model into a mixed integer linear programming model by utilizing piecewise linearization and a large M method, and solving to obtain an optimal programming strategy of the power-traffic-information network. According to the method, a joint programming model is simplified into a mixed integer linear programming model, the solving difficulty is remarkably reduced, and three-network collaborative optimization is achieved.

According to the method for intelligently realizing online control and optimization of the incinerator based on the AI, the technical means of deep learning, reinforcement learning, a multi-mode integration model and the like are utilized, deep analysis and mining are carried out on operation data of the incinerator, an intelligent control model is established, all the operation data of the incinerator are analyzed in real time, combustion parameters are dynamically adjusted, and the control accuracy of the incinerator is improved. The online intelligent control and optimization of the incinerator are realized, so that the efficient, stable and safe operation of the incinerator is realized.

The invention discloses a power distribution network fault repairing method and system based on an information physical system. The method comprises the following steps: acquiring basic parameters of a power layer and a communication layer in a power distribution network; constructing a double-layer optimization concurrent fault recovery model according to the basic parameters; the upper layer of the double-layer optimization concurrent fault recovery model is a communication layer fault recovery model, and the lower layer of the double-layer optimization concurrent fault recovery model is a power layer fault recovery model; solving the communication layer fault recovery model to obtain a recovery link of the communication layer, reconstructing the power layer fault recovery model based on the recovery link to obtain a power layer fault recovery reconstruction model, and solving the power layer fault recovery reconstruction model to obtain a fault recovery scheme to control the power distribution network to be repaired; a collaborative recovery model with recovery of a communication layer prior to a physical layer as a strategy is constructed based on the CPS, ordered recovery of the power distribution network during CPS concurrent faults is realized, and the fault recovery efficiency of the power distribution network is also greatly improved.

The invention discloses a logistics scheduling optimization method based on deep reinforcement learning and related equipment, and the scheme provided by the invention is based on a deep reinforcement learning algorithm, and through constructing a unified deep reinforcement learning optimization model, three key elements of facility site selection, personnel scheduling and path planning are integrated, so that the logistics scheduling optimization method based on deep reinforcement learning is realized. Reinforcement learning and a multi-objective optimization algorithm are utilized to reinforce learning, through self-learning of an intelligent agent in a dynamic environment, a scheduling strategy is continuously optimized, through multi-objective optimization, collaborative optimization of all parts of logistics scheduling is achieved through balance of cost and efficiency, and through continuous learning and strategy adjustment in the dynamic environment, the logistics scheduling efficiency is improved. The efficient scheduling of the logistics network is realized, and the accuracy of the scheduling decision is ensured.

The invention discloses a tundish cover castable, and relates to the technical field of tundish refractory materials. The tundish cover castable is prepared from the following raw material components in percentage by mass: 7.8 to 9.6 percent of alumina of 8 to 5mm, 2.6 to 4.3 percent of alumina of 5 to 3mm, 2.6 to 4.3 percent of alumina of 1 to 0mm, 17.4 to 20.8 percent of alumina of 200 meshes, 9.6 to 13 percent of mullite of 5 to 3mm, 22.6 to 26 percent of mullite of 3 to 1mm, 9.6 to 13 percent of mullite of 1 to 0mm, 7 to 8.7 percent of mullite of 200 meshes, 2.6 to 3.5 percent of composite binding agent, 0.8 to 1.7 percent of aluminum silicate, 0.05 to 0.07 percent of SiC nanowire, 0.2 to 0.3 percent of aluminum titanate and the balance of water. The invention discloses a tundish cover castable which comprises the following components in percentage by weight: 0.4-0.7% of alumina micro powder, 0.08-0.11% of sodium tripolyphosphate, 2.6-3.5% of silica fume, 0.87% of steel fiber and 1.3-1.7% of high-alumina cement. Compared with a traditional process, the tundish cover castable has the advantages that the overall performance of the tundish cover castable is remarkably improved, the breaking strength and compression strength of the castable are effectively enhanced, the service life is remarkably prolonged, and the production cost is reduced. And the production cost is greatly saved, and the economical efficiency and the operation rate of continuous casting operation are improved.

The invention provides a process parameter optimization method for liquid addition type plant function fiber production, and belongs to the technical field of plant function fiber production, and the process parameter optimization method comprises the following steps: 1) collecting process parameters and fiber performance index historical data; 2) analyzing the correlation between the process parameters and the fiber performance, and screening out key influence parameters; 3) establishing a multi-target prediction model based on the key parameters, and generating a simulation data set by using Latin hypercube sampling; 4) performing clustering screening on the simulation data to obtain a Pareto optimal process parameter combination; 5) verifying the accuracy of the prediction model through experiments, and carrying out iterative optimization on the model; and 6) finally performing refined optimization by adopting a response surface method to obtain optimal production process parameters. According to the scheme, the technical problem that better optimized parameters cannot be obtained due to the fact that technological parameters of liquid addition type plant function fiber production are adjusted by means of experience of engineers in the prior art is solved.

The invention belongs to the technical field of the Internet of Things and energy management, and particularly relates to an energy consumption data collector based on the Internet of Things and a collection method, the collector comprises a data processing module, a network state evaluation module, a data value evaluation module and a communication and control module, real-time protocol analysis and data conversion are carried out on original data, and the energy consumption data is obtained. Generating a standard energy consumption data packet with a unified timestamp and a device identifier, storing the standard energy consumption data packet according to a receiving sequence, and eliminating a time reference deviation caused by an instrument clock difference; communication signal strength and round-trip time with a remote Internet of Things platform are collected in real time, through multiple times of detection and normalization processing, a maximum value is selected as a network state factor, a dynamic change reference and a self-adaptive fluctuation interval are calculated based on a historical power data sequence, an instantaneous deviation rate and a continuous deviation duration are determined, and the network state factor is determined. A data value factor is generated through processing, and the network state factor and the data value factor are matched with a preset period adjustment mode to realize collaborative optimization.

The invention discloses an integrated energy system planning and scheduling method and device considering demand side response. The method comprises the following steps: establishing load models of price type demand response, excitation type demand response and alternative type demand response; establishing an initial planning scheduling model by taking the lowest annual operation cost and the lowest demand side response cost of the full-life-cycle integrated energy system as a first optimization target; solving the initial planning scheduling model in combination with operation constraint conditions and user satisfaction constraints to obtain load transfer quantity and reduction quantity after price-type demand response and excitation-type demand response, and a real-time pricing mechanism; carrying out alternative demand response based on a real-time pricing mechanism to obtain an optimized price curve and load substitution quantities of various energy sources; and solving the planning scheduling optimization model taking the lowest annual operation cost of the full-life-cycle integrated energy system as a second optimization target to obtain an optimal capacity configuration result and an operation scheduling plan of each energy device in the integrated energy system.

The invention belongs to the technical field of berth distribution, and discloses a dry bulk cargo wharf berth distribution and ship loader combined dispatching system based on self-adaptive MOEAD. The invention provides a dry bulk cargo wharf berth allocation and ship loader combined dispatching system based on self-adaptive MOEAD. The system comprises an input module, an algorithm module, a berth allocation module, a loading and unloading dispatching module, a notification module, a control module, a loading and unloading operation module and a port leaving module. The input module is connected with the scheduling module and is used for inputting a ship plan table; the scheduling module is connected with the berth distribution module and the loading and unloading scheduling module and is used for scheduling through a scheduling algorithm; the berth distribution module is connected with the notification module and is used for carrying out berth distribution on the ship; and the loading and unloading scheduling module is connected with the control module and used for scheduling loading and unloading equipment. The operation efficiency of port loading and unloading is improved. The energy consumption of port loading and unloading equipment is reduced; the algorithm is high in convergence speed and has good convergence and distribution.

The invention discloses a wireless communication dynamic anti-interference optimization method and platform based on edge computing, and relates to the technical field of wireless communication anti-interference, and the method comprises the steps: collecting wireless channel state information and external interference source features in real time at an edge node, and constructing a real-time interference feature vector according to a collected data set; generating an interference prediction matrix; constructing a dynamic interference topological graph among network nodes, taking the dynamic interference topological graph as environment input, and utilizing reinforcement learning to generate an adaptive anti-interference strategy; and executing self-adaptive anti-interference strategy compensation, issuing a compensation result to the edge node, and executing anti-interference control. The technical problems that in the prior art, wireless communication anti-interference real-time response lags behind, global adaptability is insufficient, and local immediate response and overall reliable anti-interference requirements are difficult to consider are solved, and edge real-time anti-interference and cloud global collaborative optimization are achieved. And the technical effects of real-time performance, reliability and adaptability of wireless communication anti-interference are improved.

The invention relates to the technical field of power grid energy control, in particular to a micro-grid and main power grid energy interaction control method and system based on a neural network, and provides the following scheme: constructing a regional power grid model comprising an upper layer model and a lower layer model; a regional power grid model is trained through a neural network model, and energy interaction strategies in different operation states are generated; dynamically adjusting an energy interaction strategy and updating the power grid model based on real-time monitoring data; a convolutional neural network is adopted to optimize the power flow of the main power grid, and energy loss is minimized; micro-grid energy management is optimized through a reinforcement learning model, and the operation cost is minimized; the main power grid and the micro power grid are subjected to collaborative optimization through a genetic algorithm, and an optimal energy interaction strategy is generated; the method can effectively improve the economical efficiency and energy efficiency of power grid operation, and is suitable for a complex and changeable power grid operation environment.

The present invention discloses an intelligent fault diagnosis method and system for explosion-proof motors in the natural gas industry, the method comprising the following steps: S10: acquiring multi-source heterogeneous data of explosion-proof motors; S20: performing enhancement processing on the multi-source heterogeneous data of explosion-proof motors to obtain an enhanced multimodal data set of explosion-proof motors; S30: performing feature extraction on the enhanced multimodal data set of explosion-proof motors using a deep learning model to obtain a high-dimensional feature vector; S40: based on the high-dimensional feature vector, a multi-task learning network is used to perform fault classification and fault severity evaluation to obtain a fault type probability distribution and a fault severity index; S50: generating an intelligent diagnosis report of the explosion-proof motor including maintenance suggestions according to the fault type probability distribution and the fault severity index. The method can make full use of multi-source heterogeneous data and realize an explosion-proof motor fault diagnosis method of intelligent diagnosis, so as to improve the diagnostic efficiency and accuracy and realize preventive maintenance.

The invention discloses an AI marking closed-loop method based on active learning and difficult case mining, and the method comprises the steps: carrying out the parallel calculation of uncertainty, representativeness and diversity three-dimensional value indexes after training an initial model, and carrying out the dynamic weighting screening of a high-value sample through a closed-loop feedback controller; difficult cases are recognized, DBSCAN clustering is adopted, similar sample expansion is retrieved, and the samples are stored in a dynamic pool; annotations are distributed, intelligent judgment is triggered for dispute samples after consistency verification, and labels are determined by fusing node reputation, model confidence and feature similarity; a mixed loss function is adopted for incremental training, and if the performance does not reach the standard, sampling, difficult examples and training parameters are adjusted in a linkage mode and then retry is conducted; the effectiveness of the difficult cases is evaluated after each round of iteration, the invalid difficult cases are eliminated, redundant clusters are combined, and the pool capacity is adaptively maintained; and the performance of the monitoring model is iteratively optimized until the standard is reached. According to the method, the labeling cost is reduced, and the boundary sample and long tail category recognition capability is improved.

The invention discloses a multi-model collaborative knowledge distillation optimization method based on edge computing. The method comprises the following steps that a plurality of edge devices construct local models respectively and train the local models through collected local data, and the edge devices upload the trained models to a server; real-time performance evaluation is carried out on local models on the edge devices, whether the performance evaluation result meets preset conditions or not is judged, if not, the local models are updated, and the model updating comprises the following steps that 1, a multi-model collaborative distillation strategy is adopted, and knowledge exchange is carried out on the local models on the different edge devices; 2) carrying out data heterogeneity and imbalance processing; 3) carrying out resource adaptive scheduling; (4) the server carries out local model aggregation and issues model updating content in the iteration process to the edge device; if yes, the local model does not need to be updated in the iteration. According to the invention, the performance of multiple models on the edge device can be improved, the communication overhead is reduced, and the data privacy is ensured.