1636 results about "Cluster based" patented technology

The invention discloses an intelligent power grid optimal scheduling method and system based on multivariate energy storage cooperative scheduling, and relates to the technical field of power grid optimal scheduling, and the method comprises the following steps: building a prediction model based on first data, generating prediction data, coupling energy storage characteristic parameters of different types of energy storage equipment with the prediction data, and obtaining a prediction model; establishing a multi-energy collaborative scheduling model; dynamically screening the energy storage scheduling strategy set based on a preset real-time performance evaluation index to generate an optimal strategy subset; according to the optimal strategy subset, performing differentiated charging and discharging control instructions on the energy storage equipment cluster; and collecting second data in the charge and discharge control process, calculating a deviation value between the second data and the prediction data, converting the deviation value into a feature vector, inputting the feature vector into a preset incremental learning algorithm, and optimizing parameters of the multi-energy collaborative scheduling model. Layered screening is implemented in combination with real-time performance evaluation indexes, and it is ensured that the optimal scheduling scheme can be rapidly selected in different time periods and under the uncertain disturbance condition.

Job clusters take time to instantiate. A computing system is provided comprising a plurality of non-interactive job clusters, a control database storing a task queue, and a controller. The controller instantiates one or more clusters of the plurality of non-interactive job clusters based on a size of the task queue and monitoring if the one or more clusters are successfully instantiated. Each of the one or more clusters, after successfully being instantiated by the controller, executes a dispatcher process that includes: querying the control database to identify an available task from the task queue; obtaining and processing the available task; and, after completion of the available task, further querying the control database prior to terminating.

The invention relates to the technical field of equipment fault prediction, in particular to a cup labeling equipment fault prediction method and system based on machine learning. According to the method, equipment operation state parameters are converted into a multi-mode pulse sequence with a timestamp synchronization characteristic; loading the purified pulse flow to a quantum bit array for entanglement state evolution, and extracting a three-mode entanglement association tensor; carrying out dimensionality reduction projection on the three-mode correlation tensor to an equipment degradation manifold space, and determining quantum tunneling probability density distribution; constructing a time-varying Hamiltonian of an equipment degradation state based on quantum tunneling probability density distribution, and generating a degradation track cluster according to the time-varying Hamiltonian; and performing time sequence convolution processing on the degradation track cluster, performing probability amplitude amplification on a fault critical point in the track cluster by using an energy level splitting characteristic of a time-varying Hamiltonian, and generating a space-time probability cloud picture. The fault evolution law can be visually presented, the accuracy and timeliness of early fault early warning are improved, and a reliable basis is provided for predictive maintenance.

A method, system, and device for wind speed prediction and layout optimization in wind power generation are provided. The method includes: obtaining a basic wind resource dataset of a target region; constructing a physics-informed neural network model based on the basic wind resource dataset; obtaining wind speeds data at a specific location in a velocity field based on the physics-informed neural networks and constructing a training dataset; training the physics-informed neural network model based on the training dataset; reconstructing a wind speed distribution within the velocity field and predicting wind speeds for a next time period with a wind farm using the trained physics-informed neural network model; and optimizing a layout of a wind turbine cluster based on a reconstructed wind speed distribution within the velocity field. The present application reconstructs a two-dimensional velocity field of the wind farm by training the PINN and enables accurate ultra-short-term wind speed prediction.

The invention provides a cluster network flow prediction method based on multi-scale time feature fusion, and belongs to the technical field of computer network flow prediction. The method comprises the following steps: determining a multi-index prediction sequence based on traffic load characteristics of cluster IP instances, and constructing a high-quality time sequence data set; fourier transform and discrete wavelet transform are used for time-frequency feature analysis, and noise filtering and data dimension reduction are completed; projecting sequences of different time granularities to a unified model dimension, performing one-dimensional channel convolution merging, inputting the merged sequences into a time encoder and a cross-channel encoder, and capturing cross-scale long-term time dependence and a coupling relationship between variables; in the loss function design, time domain and frequency domain loss are fused, double-domain error calculation is carried out on a prediction result and a label through Fourier transform, and the robustness of a model to non-stationary fluctuation is enhanced; and through linear layer decoding and reverse normalization processing, the abstract feature is converted into an actual flow prediction value. According to the invention, the precision and reliability of cluster network flow prediction are significantly improved.

The system and methods for determining the representative samples from a large imperfectly labeled dataset to support data processing and inferences for machine-learning applications. The method includes accessing data samples that may be processed to generate embedded vectors along with a set of reference labels. For each label, clustering is performed to group at least some of the embedded vectors together into clusters based on the associated inherent patterns, followed by a refinement process to select relevant clusters from the clustered patterns. One or more embedded vectors from the selected clusters are passed to a statistical technique to generate representative embedded vectors for each label. The statistical technique is configured such that the weights of selected embedded vectors within each of the cluster are the same. These representative embedded vectors may be further fed into a machine-learning model to predict a label from the set of reference labels for a given prompt.

The invention relates to the technical field of deep learning, in particular to an AI chip test parameter adaptive optimization method based on deep learning, which comprises the following steps: acquiring historical test data of an AI chip, and calculating correlation strength among different failure modes based on the historical test data; identifying a failure coupling matrix according to the edge weight, and converting a preset static detection parameter constraint boundary into a dynamic constraint space changing along with a failure detection state; a multi-level optimization framework is constructed, the upper layer executes failure type correlation analysis and generates constraint propagation information, the middle layer optimizes a parameter cluster based on the constraint propagation information, and the lower layer adjusts a single detection parameter and outputs a parameter optimization result; establishing a neural network mapping model, and obtaining a nonlinear mapping relationship between the detection parameters and the failure types; based on the physical state parameters, the nonlinear mapping relation is adjusted, the dynamic constraint space is updated, parameter optimization is executed again, a parameter optimization result is output, and an optimal test parameter combination is output.

The invention discloses a federated learning client selection method based on reinforcement learning and a federated learning system. The method comprises the following steps: a client performs local detection training and collects loss, delay data and other related data; the central server divides the clients by using mean shift clustering based on data distribution of the clients, and regards each cluster as an independent agent; the intelligent agent dynamically selects a client based on the multi-dimensional state information, and optimization selection is carried out by adopting an exploration strategy; the client uploads model update after local training; and the central server carries out aggregation updating, and the intelligent agent optimizes client selection through a multi-target reward function according to a feedback adjustment strategy, maximizes a model convergence speed and balances communication and calculation overhead. By using the method and the system of the invention, under challenged environments of data heterogeneity, computing resource limitation, communication delay and the like, the client can be intelligently selected, the training efficiency of federated learning is optimized, and the performance and generalization ability of a global model are improved.

The invention discloses a liquid cooling resource intelligent distribution regulation and control method and system, and the method comprises the steps: collecting first operation data of a target computing power cluster, and carrying out the first preprocessing of the first operation data, and obtaining second operation data; establishing a first model, and evaluating the operation state of the target computing power cluster based on the second operation data in combination with the first model; performing regulation and control according to the operation state in combination with a first regulation and control strategy, wherein the first regulation and control strategy comprises a first target function and a constraint condition; and solving the first objective function to obtain an optimal regulation and control scheme. By continuously optimizing parameters such as the flow speed, the flow and the pressure of the cooling liquid and dynamically adjusting the temperature of a cooling liquid outlet of a heat exchanger and the rotating speed of a fan, intelligent and dynamic management of the computing power cluster cooling resources can be achieved, and efficient and safe operation is ensured.

The invention relates to the technical field of intelligent building energy management, and particularly discloses an air conditioner load cluster optimization regulation and control method and system based on building thermal inertia modeling, and the method comprises the steps: collecting building structure parameters, building material thermophysical parameters, indoor and outdoor temperature and humidity historical data, air conditioner operation data and real-time electricity price data of a building; building a building thermal inertia model based on the collected data; inputting the collected air conditioner operation data, indoor and outdoor temperature and humidity historical data into a building thermal inertia model, and predicting air conditioner loads under different working conditions; constructing an optimized objective function, and solving the optimized objective function by adopting a genetic algorithm to obtain an optimal air conditioner load regulation and control strategy; and according to the optimal air conditioner load regulation and control strategy, the air conditioner load cluster is regulated and controlled in real time. The energy utilization efficiency can be effectively improved, the operation cost is reduced, and intelligent and refined regulation and control of the air conditioner load cluster are achieved.

A method, system, and readable storage medium for fault line selection in distribution networks is provided. The method includes: obtaining the zero-sequence current of each feeder and the zero-sequence voltage of the busbar within a preset time window after a fault occurs; using these to process the feeder's short-time window zero-sequence instantaneous power curve cluster in the distribution network through KPCA (Kernel Principal Component Analysis) for dimensionality reduction, determining the principal component scores; and performing BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) clustering based on these scores to identify whether a feeder is faulted. This clustering process allows for precise and rapid identification of the faulted feeder, even when the current is small, improving detection accuracy. This solves the problem of quickly identifying the faulted feeder in a small current grounding distribution network during single-phase grounding faults.

The invention relates to the technical field of transformer substation intelligent patrol, and provides a transformer substation three-dimensional fusion patrol method and system based on digital twinborn and autonomous identification. According to the method, a fused three-dimensional model is constructed through multi-source data acquisition and a three-dimensional Gaussian splash algorithm, and in combination with deep learning-based point cloud semantic segmentation and clustering, an equipment-patrol means coverage relationship is generated. Creating a virtual inspection proxy object based on a three-dimensional virtual environment, and controlling terminals such as an unmanned aerial vehicle to collect real-time video image data; the system carries out automatic identification on pictures, automatically completes equipment level alignment and standard point location identification, generates fine control of camera zooming, horizontal rotation, pitching and the like, and realizes standardized view finding and acquisition. By combining an enhanced recognition algorithm, traditional image processing and a deep learning model are fused, model self-evolution is realized through incremental learning, flexible expansion and collaboration of various patrol terminals are supported through a unified interface, and refined, real-time and intelligent patrol operation and maintenance requirements of an intelligent substation are met.

The invention discloses an image feature matching optimization method based on intra-class space consistency in the technical field of computer vision and image processing. The method comprises the following steps: feature point extraction and preliminary matching: extracting feature points from a query image and a reference image and performing preliminary matching; initialization and transformation model estimation: initializing a matching point set and a residual error, and calculating an initial transformation model; error calculation and matching point set updating: calculating the error of the matching point pair, and updating the matching point set by adopting a dynamic screening method; performing residual optimization: judging whether the optimal condition is reached or not based on the residual, and deciding whether to continue iteration or not; performing intra-class space consistency clustering and isolated cluster elimination: performing clustering analysis after the optimal residual error is obtained, and eliminating isolated clusters based on an intra-class space consistency separation ion structure; and outputting a result: outputting a matching point set after the isolated clusters are removed. The method solves the problem that a traditional feature matching method is difficult to completely remove mismatching in a complex scene and is sensitive to noise.

The invention provides an intelligent inspection and diagnosis system, method and device for a photovoltaic power station unmanned aerial vehicle and a medium, and the system comprises a heterogeneous perception fusion module which is used for constructing digital mirror image mapping of a power station global physical field through multi-mode sensor space-time coding and electromagnetic fingerprint matching; the group intelligent decision module is used for realizing autonomous task negotiation and anti-fragile path evolution of a multi-unmanned aerial vehicle cluster based on a dynamic Bayesian game model; the embedded diagnostic kernel module is used for running a quantization feature extraction algorithm at an edge computing node to realize component-level defect subsurface-level diagnostic reasoning; the superbody self-evolution module is used for continuously reconstructing a diagnostic knowledge graph by means of a genetic programming framework to realize Darwin asymptotic optimization of a system cognitive architecture; the problems that in unmanned aerial vehicle inspection, fault type judgment and accurate geographic positioning cannot be conducted online, photovoltaic array arrangement changes and sudden shielding scenes cannot be self-adapted, and transmission bandwidth limitation and analysis lag are likely to happen in a 5G weak coverage area are solved.

The system and method for processing a digital pathology image using a machine learning model that includes a self-supervised hierarchical Vision Transformer (ViT) configured to perform unsupervised clustering with multiple classification tokens. The method includes receiving a digital pathology image that depicts a tissue slice stained with histological dyes. The digital pathology image may be processed to generate a result comprising multiple predicted classifications of individual patches of the digital pathology image. The result is generated by a machine-learning model using a self-supervised hierarchical Vision Transformer (ViT) that may further comprise a multi-head self-attention module configured to predict a crosspatch relevance metric using an attention mechanism for each individual patch in the digital pathology image thereby assigning the individual patches to a cluster based on the crosspatch relevance metrics.

The invention provides a cluster reliability test method based on fault simulation, which comprises the following steps: acquiring injection parameter information, and performing fault simulation in a cluster based on the injection parameter information and a fault transfer model to obtain a complex fault scene; key performance index records are obtained in real time based on the complex fault scene; executing a reliability detection step based on a preset reliability detection model and the key performance index record, and obtaining a data analysis report; and obtaining a reliability score, an influence analysis result and a risk prediction result based on a data analysis report, a weighted scoring algorithm, an anomaly detection algorithm and linear regression analysis, and completing reliability detection of the cluster. According to the cluster reliability test method and system based on fault simulation provided by the invention, the test efficiency and the reliability detection level are greatly improved, the fault simulation is carried out in the cluster and the reliability detection model is combined, so that the accurate quantification of the reliability detection result is realized, and the reliability detection efficiency and the detection capability are improved.

The invention discloses a data processing method and system for micro-motion detection array station distribution, and relates to the technical field of micro-motion detection, and the method comprises the steps: according to the actually measured starting point coordinates and the azimuth angle of a measuring line, setting the measuring point distance, and the radius and number of nested triangle circumcircles, generating field station coordinates in batches; collecting and preprocessing station waveform data; calculating a spatial autocorrelation coefficient of any center station and a non-center station; dBSCAN clustering is carried out based on the station distance; performing spatial averaging on the spatial autocorrelation coefficient according to a clustering result; fitting the averaged spatial autocorrelation coefficient and a zero-order Bessel function to obtain a frequency dispersion spectrum, picking up a surface wave phase velocity in the frequency dispersion spectrum, and calculating an apparent shear wave velocity; inverting the shear wave velocity of the waveform data by adopting a self-adaptive gradient inversion strategy; and generating a velocity comprehensive profile of a depth domain and a frequency domain according to the surface wave phase velocity, the apparent shear wave velocity and the inverted shear wave velocity. The invention provides a new method for station distribution to data processing for micro-motion detection.

The invention provides an intelligent decision-making method based on multi-dimensional risk coupling modeling of new energy ships and related equipment. The method comprises the following steps: fusing multi-modal data of each new energy ship to obtain a target multi-modal feature vector; mining risk implicit association strength by using a large language model, and constructing a risk knowledge graph; performing risk time sequence evolution prediction by adopting a dynamic Bayesian network based on the atlas to obtain a single-ship risk prediction result; constructing a graph structure according to the single ship risk and the operation parameters, and carrying out space coupling modeling by adopting a graph convolutional network to obtain a cluster risk prediction result; and generating an optimal operation and maintenance strategy by adopting reinforcement learning based on a cluster risk prediction result and a reward function by taking a high-fidelity digital twinborn body as a virtual environment. Therefore, according to the method, the problem of closed-loop adaptive control from risk deduction to decision response in a complex navigation scene is effectively solved by constructing a unified modeling mechanism of multi-dimensional risk coupling and linking real-time control strategy generation.

The invention discloses a layered energy management method of a fuel cell hybrid tramcar with self-adaptive working condition perception. In the recognition layer, a sliding window mechanism is adopted to extract time domain and frequency domain features of load conditions, feature data are clustered based on a spectral clustering algorithm driven by a deep auto-encoder, a data set with category labels is obtained, and a deep dynamic learning vector quantization neural network classifier is trained; in the strategy layer, a double-delay depth deterministic strategy gradient reinforcement learning algorithm is adopted, a reward function is constructed, and lithium battery SOC fluctuation penalty term limit parameters in the reward function are adaptively adjusted according to the real-time load working condition category output by the recognition layer; training the reinforcement learning agent to obtain an optimal power distribution scheme between the multi-stack fuel cell power generation system and the lithium battery; and according to the performance degradation degrees of different fuel cell stacks, a distributed cooperative control strategy considering performance difference is adopted to distribute the output power of each stack, so that the coordinated control of the running state of the multi-stack fuel cell power generation system is realized.

The embodiment of the invention discloses a multi-microgrid regulation and control method and device based on federal hierarchical reinforcement learning, and a medium, belongs to the technical field of smart grids, and solves the problem of low microgrid dispatching precision. Each microgrid uploads a model gradient parameter corresponding to the local scheduling model to a central server; the central server determines an aggregation weight according to the similarity between the micro-grid data distribution characteristics so as to update a global model; the central server dynamically divides each micro-grid into a plurality of micro-grid clusters according to model gradient similarity, and constructs a leader-follower game model in each micro-grid cluster; the central server outputs a first scheduling strategy corresponding to each micro-grid based on the updated global model, the plurality of micro-grid clusters and the leader-follower game model; and each micro-grid periodically performs secondary optimization on the first scheduling strategy based on the local feature data to obtain a second scheduling strategy corresponding to the current micro-grid.

The invention discloses a complex search scene-oriented unmanned cluster distributed collaborative decision-making method, which comprises the following steps of: modeling a search area into a three-dimensional space-time grid, establishing an unmanned aerial vehicle cluster comprehensive hit rate model and a cost function, and solving approximate optimal path distribution; establishing a target consistency judgment model based on multi-view space overlapping, and judging whether observation targets are the same target or not; target encircle collaborative decision-making is realized by improving reinforcement learning; forming a consensus unmanned aerial vehicle cluster based on group consensus, and performing position and speed estimation on a target in combination with an unmanned cluster observation error and target steering; cooperative network graph relation construction is carried out on non-consensus unmanned aerial vehicles through a dynamic graph theory, so that space-time scheduling is carried out, and surrounding point distribution and conflict resolution are optimized. According to the method, through path optimization, target consistency judgment, reinforcement learning of a hunting strategy, consensus target state collaborative prediction and a dynamic graph scheduling mechanism, efficient collaborative hunting and task allocation of multiple unmanned aerial vehicles in a complex environment are realized.

Various embodiments include a wireless communication network that comprises resource allocation circuitry. The resource allocation circuitry hosts a traffic forecasting machine learning model, a resource forecasting machine learning model, and a resource allocation machine learning model. The resource allocation circuitry obtains traffic data for a provisioning engine cluster and provides the traffic data to the traffic forecasting model. The resource allocation circuitry obtains an output that comprises a traffic prediction for the provisioning engine cluster and provides the prediction to the resource forecasting model. The resource allocation circuitry obtains an output that comprises a hardware requirement prediction for the provisioning engine cluster and provides the hardware requirement prediction to the resource allocation model. The resource allocation circuitry obtains an output that comprises a hardware allocation recommendation for the network provisioning engine cluster. The resource allocation circuitry allocates hardware resources to the cluster based on the hardware allocation recommendation.

This disclosure relates to a privacy preserving machine learning platform. In one aspect, a method includes identifying a request for processing an input feature vector by a machine learning model using a multiple multi-party computation (MPC) cluster including a plurality of MPC computing systems. Each feature of the input feature vector is encoded to generate an encoded weight vector. A respective share of the encoded weight vectors is generated for each computing system and provided to a corresponding computing system to generate a partial prediction for the respective share. The MPC cluster collects modified partial predictions for the input feature vector from the rest of the multiple MPC computing systems, where each of the modified partial predications is generated based on the partial prediction by a corresponding MPC computing system. A final prediction is generated by the MPC cluster based on the respective partial predictions.

The invention provides a task scheduling method and system for big data, and relates to the technical field of big data, and the method comprises the steps: obtaining task parameters and a dependency relationship between tasks, and calculating a task priority score; according to the data locality coefficient of the task, the cluster parameter and the cluster relevance correction coefficient, determining the fitness of the task and the cluster; determining a target cluster of the tasks and a task sequence according to the task priority score and the fitness of the tasks and the cluster; when the number of failure times of the task exceeds a time threshold value, fusing is triggered, if cross-cluster dependence fails, the dependent task is migrated to the current cluster, the fitness is recalculated, and the priority of the task is recovered after continuous success; by means of dynamically adjusting task priorities, optimizing cluster resource allocation, establishing fusing and migration mechanisms and the like, the core problems of thread waste, priority disability, stability risk and the like in the traditional technology are solved, so that the service stability and the processing efficiency of a data intermediate station in a complex scene are improved.

The invention provides a multi-source heterogeneous data dynamic updating and agent knowledge base construction method, and relates to the technical field of data processing, and the method comprises the steps: organizing heterogeneous data through a three-dimensional feature matrix, constructing feature mapping through singular value decomposition and cross decomposition, and executing recursive tensor completion to generate a fusion feature space; extracting multi-scale features and determining a stable knowledge entity based on comprehensive measurement; constructing a network structure and dividing knowledge clusters; and performing differentiation fusion of the knowledge clusters based on the life cycle parameters. According to the method, efficient integration of heterogeneous data, accurate extraction of knowledge entities and dynamic optimization of knowledge structures are realized, and the intelligent level of knowledge management is improved.

The invention provides a virtual power plant group resource scene adaptive scheduling method and system, and a storage medium, and the method comprises the steps: building a typical external feature model of a virtual power plant based on the resource characteristics and core parameters of different types of distributed resources; generating a feasible region of the single equipment based on power constraint, electric quantity constraint and climbing constraint of the single equipment in the virtual power plant, and aggregating the feasible region of the single equipment to form an aggregated feasible region of the virtual power plant; based on a typical external feature model of the virtual power plant and different service scene requirements, dynamically adjusting response capability index weights in different service scenes, and based on an aggregation feasible region of the virtual power plant, constructing a virtual power plant dynamic aggregation model adapted to multiple scenes; and solving the dynamic aggregation model of the virtual power plant by taking minimization of the power generation cost of the virtual power plant as a target to obtain an optimal scheduling scheme of the virtual power plant.

The invention belongs to the technical field of computer software, and particularly relates to a heterogeneous database unified deployment management method based on a cluster management platform, which comprises the following steps: predefining metadata of various database services; the method comprises the following steps: pre-constructing a multi-version program package warehouse, and establishing a structured warehouse directory system; obtaining a chip architecture and an operating system type of the target node, and generating environment fingerprint information; dynamically selecting a database installation package matched with the current environment and a dependency item set of the database installation package by mapping a warehouse directory system; through a unified database adaptation layer, a standardized operation interface is called, and installation, configuration and start-stop operation of the database are executed; a visual deployment and arrangement function based on component roles is provided by expanding a Web service interface of the management platform; and according to a deployment arrangement result and the environment fingerprint information, calling a deployment script to complete automatic deployment and configuration of the database cluster. According to the invention, unified adaptation, deployment and full-life-cycle management of various domestic databases are realized.

The invention relates to a cluster-oriented large model parallelization method and device and an electronic device.The method is applied to the field of large models.The method comprises the steps that operator information of a cluster-oriented large model in a preset microprocessing batch and a preset parallelization mode is obtained, the operator information comprises operator time information and operator memory information of operators in the cluster-oriented large model; the cluster comprises one or more types of accelerators; determining an initial operator parallel configuration strategy of a plurality of assembly lines in the large model based on the operator information, model memory information required by the large model and a memory extreme value of an accelerator; and performing recursion processing on the initial operator parallel configuration strategy according to a preset load balancing mode to obtain a target operator parallel configuration strategy, and running the cluster-oriented large model based on the target operator parallel configuration strategy. According to the method and the device, the utilization efficiency of chip calculation performance during large model parallel configuration is improved, and high efficiency and wide application range of large model training are realized.

The invention discloses a spinning machine cluster pressure cooperative control method based on multi-source evidence fusion, and the method comprises the steps: obtaining the real-time state parameters of each spinning machine, and calculating the derived state parameters of each spinning machine and a connection region thereof through the real-time state parameters; forming an abnormal evidence according to the abnormal state parameters of each spinning machine, and fusing the abnormal evidences of all the spinning machines through a DS evidence theory to obtain a global abnormal confidence coefficient; dynamically generating a cluster security envelope and a dynamic security threshold of the real-time state parameters according to the global abnormal confidence and the state parameters of each section; and real-time state judgment is carried out through the dynamic safety threshold and the global abnormal confidence coefficient, and the spinning machine cluster is controlled to work.

The invention discloses a server non-perception resource scheduling method and system based on cluster telescopic self-adaptive reinforcement learning. The method comprises the following steps: step 1, using a monitoring assembly based on a time sequence database to collect various operation load indexes of a cluster server; and step 2, constructing a function resource scheduler by using a Kubernetes scheduling framework. And step 3, constructing a scheduling decision agent by using a cluster telescopic adaptive reinforcement learning algorithm, inputting a variable-length observation sequence, and outputting a node deployment action. And step 4, using a historical cloud service load data set to construct a load generator to simulate a production environment to train and strengthen an intelligent agent deep neural network. And 5, using the trained scheduling decision agent to make a production environment scheduling decision. According to the method, the existing reinforcement learning algorithm is subjected to elastic cluster-oriented algorithm optimization, the expansibility and robustness of the scheduling algorithm when the cluster executes node scaling are improved, and the method has practical application value.