72 results about "Medoid" patented technology

The invention belongs to the technical field of power system operation control, and discloses a virtual power plant aggregation operation optimization method considering flexible and adjustable resource hierarchical and partitioned regulation, which is characterized in that dynamic clustering and clustering are performed on a distributed power supply based on an improved K-medoids algorithm, and response characteristic analysis is performed on an electric vehicle, an energy storage load and a temperature control load; by establishing a self-adaptive layering and partitioning mechanism, dynamic layer region boundary adjustment is realized. An optimization model with the lowest cost as the target is constructed, the optimal resource scheduling scheme is solved by comprehensively considering various cost factors and operation constraint conditions, and efficient configuration of resources and effective control of the cost are achieved; and the adjustable load and the energy storage resource are guided to participate in system regulation and control, so that the peak-valley difference of the system is reduced with relatively low regulation and control cost. According to the method, the resource scheduling efficiency is effectively improved, high-quality peak regulation auxiliary service is provided for the power grid, and efficient utilization of distributed energy is assisted.

The invention belongs to the technical field of intrusion detection, and discloses an industrial Internet of Things intrusion detection method based on time sequence clustering. An intrusion detection model based on TCN-GRU-Temporal Attention is provided, and the model combines the advantages of long and short term memory processing, dynamic feature capture and key information focusing, and is especially suitable for processing complex time sequence tasks. A multi-dimensional time sequence clustering algorithm based on an evaluation index is provided, k-medoids clustering edge nodes based on DTW are used, and the intrusion detection accuracy and the system response speed in a federated learning environment are improved. The industrial network intrusion detection method provided by the invention also shows good performance under non-independent identically distributed data, and effectively improves the recognition capability for complex attack modes.

The invention discloses a method for predicting the charging load of an electric vehicle, and the method comprises the steps: collecting the historical charging load data of the electric vehicle, and carrying out the one-time decomposition of the time series data of the charging load through employing an ICEEMDAN model, and obtaining 10 components; performing sample entropy calculation on components obtained by primary decomposition, performing signal classification by adopting K-medoids clustering according to a sample entropy result to obtain a high-frequency component, an intermediate-frequency component and a low-frequency component respectively, optimizing parameters in an MTS-Mixers model, a Crossform model and a DeepESN model by using an improved AO optimization algorithm, and performing signal classification by adopting K-medoids clustering to obtain a high-frequency component, an intermediate-frequency component and a low-frequency component; and predicting the high-frequency component, the intermediate-frequency component and the low-frequency component by using an MTS-Mixers model, a Crossform model and a DeepESN model, and finally reconstructing prediction results of the three models to obtain a prediction result of the charging load.

The invention relates to the technical field of photovoltaic power generation system monitoring, in particular to a photovoltaic power generation power medium and short term prediction method and system. Preprocessing the photovoltaic power generation power data, and performing feature selection; carrying out similarity evaluation by adopting a DTW algorithm, and carrying out modeling on a relationship between weather conditions and generated power; similar day clustering is carried out through K-Medoids in combination with a DTW algorithm; and training the prediction model and evaluating the model effect. According to the method, firstly, the correlation degree of meteorological factors and photovoltaic power generation power is analyzed through K-Medoids-DTW, and similar day clustering analysis is carried out according to weather conditions; a TimeXer model is used for performing eco-eco variable and eco-eco variable joint time sequence prediction modeling, and accurate prediction of photovoltaic power generation power under three weather conditions of sunny days, cloudy days and cloudy and rainy days is realized. Medium-short term photovoltaic power generation power prediction can be carried out under three common weather conditions, and the accuracy and efficiency of online monitoring and energy management of a photovoltaic system are improved.

The invention discloses a transformer fault intelligent diagnosis system and method based on a K-Medoid and SMOTE optimization support vector machine, and the system comprises a data collection module which is used for collecting dissolved gas and operation parameters in transformer oil in real time; the data preprocessing module is connected with the data acquisition module and is used for completing data standardization and abnormal value detection; the sample balancing module is connected with the data preprocessing module and used for reducing majority class redundancy through K-Medoid clustering and expanding minority classes through an SMOTE algorithm; the intelligent classification module is connected with the sample balance module and adopts a weighted SVM classifier to train and predict; and the result output and visualization module is connected with the intelligent classification module and is used for displaying the classification result, the confidence coefficient and the key performance indexes. According to the method, the problem of classification performance degradation caused by sample imbalance in the prior art is solved, and the recognition capability of the SVM model on minority class faults and the overall diagnosis precision are improved.

The invention discloses a rural energy system multi-energy load scene construction method, and relates to the field of comprehensive energy, and the method comprises the steps: obtaining a historical multi-energy load time sequence of a rural energy system; according to the historical multi-energy load time sequence, a scene generation model is adopted to generate a multi-energy load scene set in a future time period; the scene generation model is constructed based on a Transform model and a regularization relative loss generative adversarial network, and a cosine annealing algorithm and a hot restart mechanism are adopted to dynamically adjust the learning rate when the scene generation model is trained; and reducing scenes in the multi-energy load scene set by adopting a dynamic time warping distance method and a K-Medoids clustering method to obtain a typical multi-energy load representative scene set. According to the method, on the basis that the generated scene keeps a certain diversity, the historical data evolution law and the coupling characteristic between the source and the load are restored to the maximum extent, and extraction and reduction of the representative scene are achieved.

The invention discloses a flood classification method and system based on dynamic time warping and multi-cluster coupling. The method comprises the following steps: firstly, collecting and processing flood flow time sequence data to obtain a data set composed of standardized floods; secondly, calculating the similarity between every two of the K floods by adopting a dynamic time warping method to form a similar matrix; and flood classification is carried out based on DTW coupling hierarchical clustering and K-medoids multi-clustering. The system comprises a data acquisition module, a data preprocessing module, a comparison module and a classification output module. According to the method, hierarchical clustering is optimized again, and a multi-clustering method is combined, so that the reasonability of flood classification is improved, and the accuracy of flood classification is improved. The method can be applied to the fields of flood early warning, flood control dispatching and the like, and has important practical value and social significance.

The invention relates to the technical field of electric digital data processing, in particular to an ophthalmology department clinical nursing data preprocessing method and system. The method comprises the following steps: realizing stage clustering of pathological images through a K-Medoids algorithm based on predefined typical images; carrying out deformable image registration on the images in the same stage, and extracting key focus features through group difference analysis and structural pattern recognition; generating a weighted mask by using the lesion features, and realizing local enhancement processing of the original image; further extracting non-image modal data corresponding to the image and carrying out structured coding; and finally, constructing a time axis taking the image acquisition time as a reference, and aligning and fusing the image modality and the non-image modality to generate multi-modality comprehensive information. The system supports co-processing and feature enhancement of multi-source data, provides more accurate input for subsequent intelligent analysis and nursing intervention, and improves clinical aid decision-making efficiency and reliability.

A wind storage joint optimization regulation and control method considering conditional value-at-risk comprises the following steps: acquiring prediction data of wind power plant output and day-ahead electric energy market clearing price in each transaction period, and generating a large number of uncertain scenes through a Monte Carlo simulation method; then scene reduction is carried out through a K-medoids clustering algorithm to generate a reasonable number of typical day-ahead spot electricity price and new energy output scenes, the two typical scenes are permutated and combined to obtain final scene data, and the final scene data is input into a pre-constructed wind storage joint participation day-ahead electricity market output regulation and control optimization model; taking the day-ahead expected net income maximization as the target, introducing CVaR to measure the income risk, setting a corresponding target function and constraint conditions, and solving to obtain an output regulation and control plan and a bidding decision of the wind power plant and the novel energy storage in each simulation transaction scene. According to the method, the complementary advantages of novel energy storage and new energy are utilized, the new energy output fluctuation and deviation punishment risks caused by prediction deviation are reduced, the market competitiveness of clean energy can be improved, and new energy consumption is promoted.

The invention relates to the technical field of big data management and database optimization, in particular to an index tuning method and system for medical multi-modal data mixed query, and the method comprises the following steps: accessing structured and unstructured medical data into a data lake; performing expansion processing on a nested structure in the structured data according to field types, extracting semantic feature vectors from unstructured data, packaging the unstructured data into a standardized JSON format, performing data fusion, and loading the fused data to a PostgreSQL database; analyzing a mixed query SQL statement submitted by a user, and constructing high-dimensional mixed feature representation; carrying out clustering compression on the query request set by adopting a K-Medoids clustering algorithm to generate a representative query set; performing virtual index simulation evaluation on the structure field by utilizing HypoPG, creating a real index for the vector field in combination with a pgvector plug-in, and collecting a query performance index; and based on IBST multiplexing historical evaluation data, in combination with a structure change prediction model based on a Transform architecture, outputting an optimal index combination. According to the method, the query performance and the index recommendation efficiency are remarkably improved.

The invention discloses a periodic power consumption mode optimization method based on an encryption technology and a K-Medoids algorithm, and relates to the technical field of intelligent power grids, and the method comprises the steps: obtaining the historical load data of periodic power consumption of a consumer and a cost saving rate expected value; selecting and updating each evaluation factor based on the historical load data of different periods, and iteratively distributing the historical load data of each period of non-evaluation factors to the evaluation factor class with the highest similarity level, so as to determine each optimal evaluation factor and obtain the corresponding power consumption mode of each optimal period; and obtaining each power consumption cost saving rate of each optimal period power consumption mode compared with the recent period power consumption, determining the power consumption cost saving rate closest to the cost saving rate expected value, and determining the corresponding optimal period power consumption mode as the expected periodic power consumption mode. According to the invention, it is ensured that the power consumption cost saving rate reaches the expectation, the periodic power consumption mode with the minimum total power consumption cost is found, and the encryption technology is used to guarantee the communication security.

The invention relates to an ocean data extraction method of k-medoids based on neighborhood variance optimization. The method comprises the following steps: calculating a plurality of first neighborhood data points of a first data point in a target ocean environment data set; determining a variance between the first data point and a plurality of first neighborhood data points; selecting K first data points with variances smaller than a second threshold as K initial clustering center points; clustering the target marine environment data set according to the K initial clustering center points; constructing a retrieval database according to a clustering result; and extracting target data points according to the retrieval database. According to the method, multiple first neighborhood data points corresponding to each data point are found for each data point, the variance of each data point in the neighborhood of the data point is calculated based on the multiple first neighborhood data points, the compactness of data distribution in the neighborhood is quantized through the variance, and part of data points with relatively small variance are selected as initial clustering center points; the method can select an initial clustering center point with a scientific basis, and improves the extraction efficiency and accuracy of the marine environment data.

The invention relates to a hydrogen energy dynamic scheduling method and system considering source load uncertainty in an industrial park, and belongs to the technical field of energy scheduling, and the method comprises the steps: obtaining wind-solar power generation and hydrogen load historical data, and constructing a data set; obtaining an intrinsic mode function component through variational mode decomposition; constructing a modal feature matrix based on the components, and performing K-medoids clustering by adopting a dynamic time warping distance to generate a typical scene and probability; establishing a day-ahead scheduling optimization model, and making a hydrogen production and energy storage plan by taking total cost minimization as a target; and in actual operation, a real-time scheduling instruction is generated by minimizing the plan deviation and the equipment variation through the intra-day rolling optimization model. Through multi-scale optimization and scene generation, the scheduling precision, economical efficiency and real-time adaptability are improved, and the method is suitable for an electric-hydrogen coupling system in an industrial park.

The application discloses a differential privacy data desensitization method based on time sequence random mapping and related devices, comprising: obtaining various target time sequence data that needs to be desensitized in a power transaction center, and clustering the target time sequence data by using a k-medoids clustering algorithm based on dynamic time warping distance measurement to obtain time sequence data sets of different clusters; based on the maximum compression ratio of the target time sequence data, the window value size corresponding to different target time sequence data in the same cluster is calculated, and the average compression ratio of all target time sequence data is determined, and the average compression ratio is taken as the target window size set during desensitization; based on the pre-determined security protection requirement level of the target time sequence data, the privacy budget required by differential privacy is determined, and the target time sequence data is processed according to the target window value size and the differential privacy desensitization mode of time sequence random mapping to obtain power market subject data after time sequence desensitization.

The invention discloses a cross-border ecological economy zoning method. The method comprises the following steps: S1, constructing a cross-border ecological economy zoning index system; s2, a PartitioningAround Medoid clustering algorithm, namely a PAM clustering algorithm, is divided around the central point, and the PartitioningAround Medoid clustering algorithm is divided around the central point; and S3, selecting the optimal clustering number. According to the method, multidisciplinary theories and methods of geography, resource science, ecology, environmental science, economics, sociology and the like are crossly fused, a national boundary system is broken through, an ecological economy division scheme of a cross-administrative region boundary based on a grid scale is completed, and a certain method and theoretical reference is provided for cross-border region ecological economy division and collaborative development mode research.

The application discloses a transformer winding looseness fault diagnosis method based on an improved MPE and K-medoids algorithm, which is used for forming a transformer winding looseness MPE value criterion and realizing winding looseness fault diagnosis. The method steps are as follows: 1, measuring points are arranged on a transformer box body, vibration signals of each measuring point are acquired, and a vibration amplitude maximum measuring point D is selected as an optimal measuring point; 2, vibration signals of the transformer winding in different states are measured; 3, a particle swarm optimization algorithm is used to optimize parameters in a traditional MPE algorithm, so that the overall change of the MPE value is stable; 4, the MPE value of the measuring point D is calculated by using the optimized MPE algorithm; 5, the MPE values under two adjacent scale factors are selected as horizontal and vertical coordinates of a clustering coordinate system; 6, the K-medoids algorithm is used in the coordinate system to realize accurate classification of the transformer winding fault type; and 7, the MPE value criterion is summarized and formed, and a database is established. The method reduces the cross aliasing phenomenon of the traditional MPE algorithm, and realizes accurate judgment of the transformer fault type.

The application discloses a kind of based on trust value improvement K-medoids's blockchain trusted sensor anomaly detection method, comprising the following steps: S1, trust value initialization, and the trust value of the BTS node participating in this consensus is updated by node trust value update formula;S2, n BTS node is clustered into k groups, and the optimal BTS node is determined as center point;S3, partition for the rest BTS node;S4, when there is client to provide new transaction request, BTS layer according to partition voting result, the same partition vote different from other BTS node is regarded as abnormal target set A, the node set of credit value below 60 points is recorded as B, the node in the intersection of A intersection B is judged as abnormal node, the beneficial effects of the application are: improved K-medoids clustering algorithm, more in line with the demand of BTS practical application, can more accurately and efficiently exclude BTS abnormal node;While clustering result has strong stability, abnormal BTS node under different environments can maintain higher detection rate.

The invention belongs to the field of Internet of Things technology and reinforcement learning, and discloses a method for optimizing information age of data acquisition under a wireless rechargeable sensor network, which comprises the following steps of: firstly, determining initial cluster division of hovering points and nodes of an unmanned aerial vehicle according to geographic positions, electric quantity states and data priorities of sensor nodes by adopting a weighted K-Medoids clustering algorithm; and then, a dual depth Q network (DDQN) algorithm in deep reinforcement learning is used for performing unmanned aerial vehicle path planning and dynamic adjustment of a hovering strategy, minimization of intra-cluster average information age (AoI) and overall network average AoI is taken as an optimization target, and node weights in a clustering algorithm are dynamically adjusted through a feedback mechanism of the DDQN algorithm, so that continuous optimization of the AoI is realized. The data acquisition delay is effectively reduced, the information acquisition real-time performance and the network overall performance are improved, and the method has the advantages of being efficient, flexible, high in adaptability and high in expandability.

The invention belongs to the field of computer software, and provides an unmanned aerial vehicle intelligent body task execution and optimal trajectory generation method based on deep reinforcement learning. The unmanned aerial vehicle intelligent body task execution model constructed through the TD3 algorithm can realize more accurate unmanned aerial vehicle action value evaluation according to the unmanned aerial vehicle observation value. During iterative training of an unmanned aerial vehicle agent task execution model, task execution actions of an unmanned aerial vehicle are constrained by using an observation space and an action space, and it can be ensured that each decision in task execution of the unmanned aerial vehicle is beneficial for the unmanned aerial vehicle to complete a task execution target; the unmanned aerial vehicle agent task execution model is trained through the unmanned aerial vehicle agent task execution function, and trajectory prediction of task execution of the unmanned aerial vehicle is improved. Clustering analysis is carried out on trajectory information of all execution tasks of the unmanned aerial vehicle intelligent body through a K-Medoids center clustering algorithm, the limitation of single reasoning decision can be effectively avoided, and the task completion reliability is improved.

The present invention discloses a method for generating a distributed photovoltaic random output scenario, and the steps are as follows: S1. Collect historical output data of each photovoltaic site; S2. Input the processed historical output data into a graph convolutional generative adversarial network to train the graph convolutional generative adversarial network; S3. Input the output data of each photovoltaic site collected into the trained graph convolutional generative adversarial network to generate a photovoltaic output scenario and complete data augmentation of the output scenario; S4. Use the ST-DBSCAN clustering algorithm to cluster the power data of the photovoltaic power station, select typical photovoltaic sites, and splice the outputs of each typical photovoltaic site according to the time dimension; S5. Use the K-medoids clustering method to select typical output dates and select generated output scenarios to obtain typical output scenarios with multiple photovoltaic sites. The present invention makes the distributed photovoltaic output scenario more comprehensive and diverse, and at the same time improves the accuracy of the distributed photovoltaic output scenario.

The invention relates to the technologies of the Internet of Vehicles, block chains, consensus algorithms and the like, and discloses a dynamic clustering reputation collaborative hierarchical consensus architecture for the Internet of Vehicles. In the method, an architecture of'cluster grouping optimization topology + reputation dynamic management and control + hierarchical consensus efficiency improvement 'is constructed: grouping is carried out according to a communication distance between a vehicle and a clustering center, equipment computing power and a reputation value by using an enhanced K-Medoids algorithm, and a low-delay topology foundation is laid; based on an improved EigenTrust algorithm, node reputation is evaluated in combination with data authenticity, consensus participation degree, resource contribution degree and historical impact factors, and node permission is dynamically controlled; layered efficiency improvement is performed according to the number of vehicle nodes, a multi-layer architecture is established, local efficient processing is adapted, global consensus is realized, malicious interference and communication consumption are reduced, and the security of a consensus network is improved.

This invention discloses a method for evaluating the quality of marine controlled-source electromagnetic data, belonging to the field of data quality evaluation technology. The method includes: projecting the acquired marine controlled-source electromagnetic data onto a two-dimensional plane, whereby the real and imaginary parts of the data signal are projected onto the horizontal and vertical coordinates to obtain the data distribution; performing cluster statistics on the data distribution using the K-medoids clustering method to obtain cluster centers; and using the cohesion (CP) calculation formula to perform aggregation degree analysis on the data distributed according to the cluster centers to obtain the data quality assessment result. This invention can directly quantify and evaluate data based on the cohesion of the electromagnetic data distribution characteristics, without defining empirical functions, requiring no specific data volume, and is unaffected by fly-spot data, thus improving the applicability and accuracy of the evaluation method.

The application discloses a rural energy system multi-energy load scene construction method, relates to the field of comprehensive energy, and comprises the following steps: acquiring a historical multi-energy load time sequence of a rural energy system; generating a multi-energy load scene set of a future time period by using a scene generation model according to the historical multi-energy load time sequence; the scene generation model is constructed based on a Transformer model and a regularized relative loss generation adversarial network, and the scene generation model dynamically adjusts a learning rate by using a cosine annealing algorithm and a hot restart mechanism during training; and the scenes in the multi-energy load scene set are reduced by using a dynamic time warping distance method and a K-Medoids clustering method, so that a typical multi-energy load representative scene set is obtained. According to the application, the generated scenes can restore the coupling characteristics between the historical data evolution law and the source load to the maximum extent on the basis of maintaining a certain diversity, and the representative scenes can be extracted and reduced.

The present invention relates to an active contour model image segmentation method based on K-medoids clustering, comprising: using a K-medoids clustering algorithm to perform binarization processing on an image to be segmented to obtain a fitting result, and acquiring a foreground image and a background image; constructing a pre-segment fitting function based on the binarization fitting result; using a zero level set that meets the Lipschitz condition to replace the contour curve of the pre-segment fitting function to obtain a KM pre-segment fitting function; using a gradient descent method to minimize the energy functional of the KM pre-segment fitting function to obtain a gradient flow equation; using an adaptive regularization function to regularize a data-driven term of the gradient flow equation; using a rulesig function to regularize the zero level set function, using a kernel function to smooth and shorten the curve of the regularized zero level set function, and outputting a level set function as a segmentation result of the image to be segmented.

The invention relates to a topological clustering method for a power grid operation scene fusing a graph editing distance and K-medoids, and the method comprises the steps: carrying out the construction and preprocessing of a power grid topological graph, abstracting a power system into a graph structure, carrying out the unified coding of nodes and edges, and guaranteeing the consistency of the graph structure; counting dissimilarity of the topological graph by adopting a heuristic algorithm based on A *; k topologies are randomly selected from the topologies to serve as initial clustering centers, each non-central topology is distributed to a cluster where a central point closest to the non-central topology is located, for each cluster, the central point is tried to be replaced, the topology with the minimum GED sum from all points in the cluster to the central point is selected to serve as a new central point, and the distribution and updating steps are repeated; the central point does not change any more or reaches the maximum number of iterations; k clustering clusters are output, topologies far away from all the clusters are screened out and judged to be of an abnormal or non-standard structure, each cluster represents the center topology of the cluster, and each cluster center corresponds to a typical power grid topology mode.

The application discloses a kind of combination clustering scene reduction method and system based on potential decision result, the uncertainty factor of each node in power system is expressed as multidimensional curve input domain in the form of time sequence curve;Based on multidimensional curve input domain, the power supply planning problem solving model under different time scale scene is constructed, considering operation checking, multidimensional curve input domain is converted into potential decision result domain for clustering analysis;SOM neural network clustering is carried out to potential decision result, and preliminary clustering result is obtained;SOM and k-medoids combination clustering method is used to finally classify preliminary clustering result, and finally the typical scene used in operation checking is obtained.The application is used to cluster the most representative scene set in power supply planning typical scene operation checking, to reduce the complexity of solving model to the greatest extent under the condition of guaranteeing the reliability of checking result, improve calculation efficiency.

The application discloses a response potential evaluation method of adjustable resource cluster based on an AE-LSTM algorithm, and comprises the following steps: 1, collecting original power data and excitation intensity data; 2, decomposing the power consumption power sequence of the adjustable resource individual by using a discrete wavelet transform; 3, classifying the decomposed sequence of the adjustable resource cluster by using K-medoids clustering based on a DTW algorithm; 4, extracting the characteristic value of the center point data of each sequence of the clustered adjustable resource cluster by using the AE-LSTM algorithm, obtaining the parameters of the potential evaluation probability distribution, and calculating the response potential probability distribution of the adjustable resource individual. The application can effectively improve the characteristic value extraction efficiency of the AE-LSTM neural network, solve the problem that individual differences are neglected in adjustable resource potential evaluation, and thus can realize effective evaluation on the scheduling potential of the adjustable object in the adjustable cluster.

The application relates to the technical field of motion detection, in particular to a real-time motion mode switching identification system and method based on a single inertial sensor, which comprises a motion mode conversion data stream automatic acquisition module, a space-time gait parameter feature processing module and an improved K-medoids clustering algorithm model which are sequentially connected; the motion mode conversion data stream automatic acquisition module comprises an inertial motion sensing unit and a probability distribution modeling unit; the space-time gait parameter feature processing module comprises a feature extraction unit, a sorting unit and a cross-validation unit; and the improved K-medoids clustering algorithm model is used for classifying and training and predicting different switching motion states according to selected gait features, so as to realize accurate identification of a complex human motion switching process. The real-time motion mode switching identification system based on a single inertial sensor provided by the application mainly solves the problems of low accuracy and large complexity when a wearable device switches motion modes.

The invention relates to the technical field of power systems, in particular to a wind power output analysis method and device based on integrated clustering, equipment and a medium. According to the method, a K-medoids algorithm is adopted to perform preliminary clustering to obtain a clustering center matrix, and on this basis, a hierarchical clustering algorithm is adopted to perform secondary clustering to obtain a clustering center. And finally, dividing the original data according to the final clustering center, and using the center point as typical output for subsequent analysis. On the basis that original data information is reserved as far as possible, the high efficiency of the K-medoids algorithm and the accuracy of hierarchical clustering are exerted, the clustering speed is increased, and the clustering effect is improved.

The invention requests to protect a maximum security and rate resource allocation method for a multi-relay unmanned aerial vehicle (RMA) assisted rate division multiple access (RSMA) communication system. Under the minimum transmission rate of secondary users, the level of unmanned aerial vehicle flight area limitation, the limitation of unmanned aerial vehicle collision avoidance, the limitation of the number of unmanned aerial vehicle service users and the pre-coding matrix power constraint, the maximum safety and rate of the multi-relay unmanned aerial vehicle assisted RSMA communication system are maximized. The method is characterized in that a multi-relay unmanned aerial vehicle is used to assist an RSMA communication system, a K-Medoids algorithm is used to carry out user association on the unmanned aerial vehicle, deep reinforcement learning is used to optimize unmanned aerial vehicle relays to carry out system resource allocation, and a QMIX multi-agent deep reinforcement learning algorithm is used to solve combination of unmanned aerial vehicle trajectory planning and resource scheduling. A solution for the flight path of the unmanned aerial vehicle is provided, and the method is more practical and easier to transplant to process complex scenes.