683 results about "Single node" patented technology

The invention discloses an adjustable load security access method for a virtual power plant, and relates to the technical field of power system automation, and the method comprises the following steps: in the operation process of a virtual power plant scheduling system, after an upper-layer scheduling center issues a unified adjustment instruction according to an operation demand, all access nodes are monitored in real time, and the adjustment instruction is sent to the upper-layer scheduling center; the method comprises the following steps: receiving a scheduling instruction from an access node, collecting response behavior data of the access node to the scheduling instruction, preprocessing the collected original response data of a single node, and carrying out unified coding and formatted storage on historical and current response behaviors according to node identification and scheduling time sequence information to construct a structured data set. According to the method, accurate identification and dynamic correction of the adjustable load frequency response offset risk in the virtual power plant are realized, the heterogeneous load perception and control capability is improved, frequency staggering resonance and scheduling failure caused by inconsistent response are avoided, the system cooperative adjustment stability and the load safety access reliability are enhanced, and the method has a good application prospect.

The invention relates to the technical field of virtual power plant operation monitoring, and discloses a virtual power plant distributed resource cluster anomaly monitoring system, method and equipment, and the system comprises a resource node data collection module, an anomaly trend score calculation module, an association graph model construction module, a single node influence calculation module and a comprehensive anomaly early warning output module. In the prior art, a single-point alarm mode depends on a fixed threshold value or a rule base, and especially under the conditions that distributed resources in a virtual power plant are complex in type and space coupling and controller sharing exist between nodes, accurate identification and dynamic early warning of trend anomalies, cooperative faults and potential propagation paths are difficult to realize. According to the method and the device, the multi-dimensional parameter modeling and the graph neural network anomaly propagation analysis are fused, so that the anomaly in the distributed resource cluster is identified and dynamically evaluated, and the early warning accuracy and the intelligent operation and maintenance efficiency of the virtual power plant are improved.

The invention relates to a federated learning system based on multi-key homomorphic encryption and adaptive differential privacy, and belongs to the technical field of privacy computing. According to the system, on the premise that no trusted third party exists, a multi-client collaborative key generation and threshold decryption mechanism is achieved, it is ensured that model parameters are always in an encrypted state in the aggregation process, and leakage of a single node is prevented. By introducing a parameter sensitivity analysis and selective encryption strategy, the system only encrypts high-risk parameters, and the encryption burden is effectively reduced. Meanwhile, in combination with an adaptive privacy budget allocation mechanism, the system dynamically adjusts noise intensity according to a model training state, and model performance and convergence speed are maintained while privacy protection capability is improved. According to the method, high robustness and collusion resistance are realized, stable operation under the condition that part of clients are offline is supported, and the method is suitable for application scenes such as medical treatment and finance with high data sensitivity and strict performance requirements.

The embodiment of the invention discloses a collaborative optimization method and device for a distributed micro-grid and a medium, and relates to the technical field of energy management, and the method comprises the steps: carrying out the real-time data collection through a multi-element sensor network disposed at each distributed energy node in advance, so as to obtain the real-time energy data corresponding to each distributed energy node; according to the edge node corresponding to each distributed energy node, carbon emission of the distributed energy nodes is predicted by using the real-time energy data, and node-level predicted carbon emission parameters are determined, so that single-node optimization reference information corresponding to each distributed energy node is determined; through a preset global center node, collecting single node optimization information corresponding to each distributed energy node for aggregation, and determining global model gradient information; and based on the global model gradient information and the single-node model gradient information, carrying out collaborative optimization on the single-node optimization strategy to determine a collaborative optimization strategy corresponding to each edge node.

Embodiments described herein relate to computer systems and methods for digital asset custodian that seamlessly integrates off-chain multi-party computation (MPC) signing with on-chain party identity authentication. The system comprises a service provider node, multiple computation nodes equipped with robust secure hardware, and a Blockchain accessible to all participating nodes. The nodes deliver a threshold signing service with utmost security—ensuring that no single node can recover the private key, while simultaneously recording and mutually authenticating each node's identity on-chain. The system boasts intrinsic support for hierarchical address creation and signing, facilitated by an enhanced protocol. Its scalability, particularly in the quantity of computation nodes, is complemented by efficient communication tailored to the demands of the MPC protocol.

The invention discloses a power distribution network fault positioning method and equipment based on a graph convolutional neural network, belongs to the technical field of power distribution network fault positioning, and is used for solving the technical problems that a traditional fault positioning method depends on feature information of a single node, neglects the overall structure of a power distribution network and is difficult to deal with a complex and changeable power distribution network. The method comprises the steps of performing data preprocessing on a collected data set to obtain a target data set; constructing an optimization matrix based on the target data set and the topological structure diagram of the to-be-tested power distribution network; performing node feature extraction on the topological structure diagram to obtain a node feature vector; training a graph convolutional neural network through the optimization matrix and the node feature vectors, and performing sample balance adjustment to obtain a power distribution network fault positioning model; and inputting the to-be-detected data into the power distribution network fault positioning model for identification to obtain state parameters of each node, and positioning a fault node according to the node state parameters. According to the invention, the accuracy and robustness of power distribution network fault positioning are improved.

The invention provides a digital twin hydraulic engineering large language model reasoning method based on tensor parallel processing. The method sequentially comprises the following steps: S1, gradient calculation optimization; s2, carrying out self-attention tensor dimension reduction blocking; s3, normalizing and parallelizing a low-dimensional layer; s4, inter-layer overall weak extension resource scheduling processing is carried out on the low-dimensional layer after middle-layer normalization processing in the step S3; and S5, converting an intermediate tensor result of a decoder into a specific predicted value. According to the method, the parameters and the input data of the large-scale language model are subjected to tensor segmentation and are distributed to the computing nodes for parallel processing, so that the reasoning efficiency of the large-scale model is remarkably improved. By using the method, the calculation efficiency and the real-time performance of the hydraulic engineering large language model can be greatly improved, the calculation nodes can work simultaneously by processing the reasoning task of the large-scale model in parallel, the calculation burden of a single node is remarkably reduced, larger-scale hydraulic engineering data can be processed, and the calculation efficiency and the real-time performance of the hydraulic engineering large language model are improved. And efficient and accurate risk prediction capability can be maintained.

The invention provides a new energy automobile supply chain network risk node identification method, system and device, and the method comprises the steps: taking a new energy automobile supply chain enterprise as a node, constructing an adjacent matrix, and further constructing a new energy automobile supply chain network; calculating a network index of each node in the network; calculating an immune propagation evaluation index of each node in the network, wherein the immune propagation evaluation index is used for reflecting the influence intensity of infection propagation caused by a single node; simulating attacks by traversing nodes, and calculating survivability indexes of the nodes; and sorting the nodes based on the network index, the immune propagation evaluation index and the survivability index, and determining key nodes in the nodes according to a sorting result. According to the scheme, through comprehensive and multi-dimensional fusion risk propagation key node identification, potential hidden key nodes in the new energy automobile supply chain network can be found, and risk prevention of the new energy automobile supply chain network is optimized.

The invention discloses a link state analysis method and device for a broadband ad hoc network, equipment and a medium. The method comprises the following steps: S1, acquiring coordinates and a signal intensity sequence of each node in a monitoring time period; s2, acquiring a link instability coefficient of the single node in the monitoring time period according to the difference of the nearest distances of the single node at different moments and the dispersion degree of the signal intensity sequence of the node; s3, according to the difference of the link instability coefficients of all the detection nodes in each emergency group, the correlation degree of the signal intensity sequences of any two detection nodes, and the dispersion degree of the coordinate distance of any two detection nodes at all moments, obtaining the regional co-variability of each emergency group; s4, the link global stability of the broadband ad hoc network system is obtained according to the mean value of all the link stability indexes; and S5, evaluating the link state according to the global stability of the link. According to the method and the device, the dynamic environment interference is integrally analyzed, so that the link state analysis accuracy is improved.

The invention provides a lower limb joint angle prediction method and system based on space-time attention in the technical field of exoskeleton robot control, and the method comprises the steps: S1, collecting a large number of I MU historical signals through a single-node IMU sensor disposed at a shank, carrying out the preprocessing of each IMU historical signal, and constructing a data set; s2, creating a joint angle prediction model based on a multi-scale convolution module, a space-time attention module and a long-term memory module; s3, dividing the data set into a training set, a verification set and a test set based on a preset proportion, training the joint angle prediction model through the training set, verifying the trained joint angle prediction model through the verification set, and testing the joint angle prediction model passing verification through the test set; and S4, performing lower limb joint angle prediction on the joint angle prediction model passing the test. The lower limb joint angle prediction method has the advantages that the lower limb joint angle prediction precision and applicability are greatly improved.

The invention relates to a heterogeneous cluster-oriented resource allocation method and device and a storage medium, and the method comprises the steps: executing a plurality of iterations of an initial model on a single node of each heterogeneous cluster, and obtaining the performance characteristics of the model; generating a plurality of parallel strategy combinations comprising the first parallelism degree and the second parallelism degree; generating a plurality of assembly line parallel combinations based on the first degree of parallelism, calculating the load imbalance rate of each assembly line parallel combination, and screening out N combinations with the minimum load imbalance rate from the assembly line parallel combinations; traversing the second parallelism degree and the micro-processing batch in each stage corresponding to each parallel combination to generate a plurality of stage strategy combinations; and calculating video memory occupation and execution time under each stage strategy combination, screening out K stage strategy combinations of which the video memories are smaller than or equal to a video memory threshold value and the execution time is shortest, and carrying out resource allocation on the heterogeneous cluster based on the stage strategy combinations. According to the invention, the problems of low training efficiency and low resource utilization rate are solved.

The invention discloses a data packet analysis method and device based on eBPF and electronic equipment. Relates to the field of cloud computing, and the method comprises the steps that a to-be-analyzed data packet captured by an eBPF kernel mode program from an operating system kernel is acquired based on an extended Berkley data packet filter eBPF user mode program, the eBPF user mode program is a program running in a user space of an operating system, and the eBPF kernel mode program is a program running in a kernel space of the operating system; determining a target analytic function; a target analytic function on a function calculation platform is called through the flow entry function, and the function calculation platform is a service system for operating the function according to needs; and based on the target analysis function, performing analysis processing on the to-be-analyzed data packet to obtain an analysis result of the to-be-analyzed data packet. According to the method and the device, the problem of limited processing capability of an eBPF flow analysis system on a single node in related technologies is solved.

The invention provides a traffic flow prediction method based on space-time decomposition and an attention mechanism. The method comprises the following steps: firstly, decomposing original traffic flow data into a low-frequency trend item and a high-frequency event item through Fourier transform and inverse transform thereof, and introducing time embedded information to enhance the expression ability of time sequence characteristics; secondly, for a trend term, constructing a trend processing branch by using time convolution, a time attention module and a trend category learning module, and emphatically extracting trend information of a single node; for an event item, a time convolution and multi-time slice space attention gating module is adopted, an event processing branch is constructed, and event influence and spatial dependence between different nodes are emphasized to be captured. And finally, performing weighted fusion on the features of the trend item and the event item through an adaptive fusion module to obtain a more accurate and robust traffic flow prediction result. The method can improve the prediction precision while effectively processing the spatio-temporal information, and is suitable for flow prediction in a complex traffic scene.

The invention provides a low-altitude pipeline route planning method and device, and the method comprises the steps: building a three-dimensional rectangular map of route planning based on a three-dimensional digital airspace model; taking each three-dimensional grid in the three-dimensional rectangular map as a single node, and determining a low-altitude pipeline route planning optimization scheme through an A * algorithm based on a cost function; wherein the cost function is determined according to the moving cost of the starting point grid reaching the end point grid through any path and the route cost of the starting point grid reaching the end point grid through any path; the route cost comprises distance cost and threat cost. Therefore, aircraft pipeline route planning in a low-altitude environment is realized, effective help is provided for low-altitude airspace management, and low-altitude operation safety is guaranteed.

The invention provides a distributed high-performance resolving method for a large-scale GNSS (Global Navigation Satellite System) network. The distributed high-performance resolving method comprises the following steps of S1, setting project resolving parameters; s2, according to the total number of single-day baseline resolving tasks, selecting to adopt a distributed-single node parallel mode or a distributed-cluster parallel mode; s3, uploading the data to be solved; s4, resolving by adopting a corresponding mode; s5, performing quality inspection on the single-day baseline solution; s6, analyzing a time sequence; s7, carrying out net adjustment analysis; and S8, obtaining a calculation result. According to the distributed high-performance resolving method for the large-scale GNSS network, the computing power of a computer is deeply excavated, the advantages of high efficiency of a shared memory system and high flexibility of a distributed system are fully played, centralized parallel and distributed parallel collaborative operation is realized, and the production efficiency and the computing resource utilization rate are greatly improved.

According to the user sensitive data protection system based on the cloud service and the block chain, an SM3 algorithm is adopted to fuse biological characteristics and dynamic tokens to generate an encryption credential, high-security identity authentication is achieved, differential fragmentation encryption is carried out on data based on semantic classification, Rabin threshold fragmentation and AES + ECC double-layer encryption is adopted for data of the highest security level, and the security level of the data is improved. A dynamic node optimization model is designed; an optimization coefficient is calculated according to the capacity, the transmission rate and the fault rate; the storage efficiency is improved by more than 40% in combination with a time period regulation factor; the node temperature and the vibration frequency are monitored in real time through a sensor; and verifying the validity period of the dynamic token and the fragment permission mark by using an intelligent contract to realize GDPR compliance fine-grained access control. The system remarkably improves the anti-attack capability and storage reliability of data, and is suitable for high-sensitivity scenes such as highway toll collection, finance and medical treatment.

The invention relates to a distributed LoRA unloading operation method and system. The method comprises the following steps: initializing a Kademlia DHT and a communication protocol which are respectively used for managing fragmented storage and inter-node communication; fragmenting the weight matrix W of the pre-training model and the LoRA matrixes F and G, storing fragments into corresponding storage managers, and managing fragment information through DHT; generating a calculation task context according to a calculation formula; the calculation manager dynamically queries required storage information according to the context of the calculation task, pulls required data fragments from the storage manager through a communication protocol, and executes the calculation task on the calculation manager; and summarizing the obtained task calculation results through an aggregate communication protocol, and returning the final calculation result to a user or storing the final calculation result in a specified storage position. According to the method, the memory overhead pressure of a single node is unloaded by using remote resources, and PEFT operation of nodes in a resource-constrained environment is supported.

A multi-node artificial intelligence topology adapts to service many different overall purposes. Support processing nodes, discriminative AI elements, generative AI elements along with input, output and communication circuitry along with other outside interactions provide the nodal basis for the overall topology. Therewithin, outputs of several nodes drive a single node which uses influence balancing to optimize its own output. Influence is delivered in feed forward and feed back manner. Segmented processing is provided where sections of an overall output goal is processed through the topology in segments, e.g., chapter by chapter of a novel, episode by episode, a full topology processing using internal cross node influence followed by a second full topology processing using both internal cross node and cross segment influence. Pseudo random templating providing constraints used to progress through segments to control an output flow. AI elements can be fully software, use acceleration circuitry, and employ neural network circuitry such as analog and digital versions thereof. Topologies also adapt between local and remote processing locations on a node by node basis, where, for example, some AI elements or nodes operate in the cloud, while other AI elements operate on a particular user's device or other user devices located remotely. Topologies adapt in real time to move nodes to away from a user's device to a cloud counterpart and vice versa as circumstances change.

The invention discloses a method and a system for improving data interaction efficiency of a hyper-converged architecture, relates to the technical field of hyper-converged architectures, and aims to solve the problems of network bottleneck and storage delay of data transmission among different virtual machines or containers in the conventional hyper-converged architecture. According to the scheme, a data transmission optimization module needs to be introduced into a hyper-converged architecture, so that the data access speed and the network efficiency are improved through data prefetching and an intelligent routing technology; the distributed cache module is used for relieving the pressure of a single node and ensuring accurate synchronization of data through data locality optimization and cache consistency management; the parallel data processing module is used for improving the data processing efficiency through task parallelization and data fragmentation technologies; and the self-adaptive load balancing module is used for dynamically adjusting resource allocation according to the system load and the resource use condition so as to realize the balance of calculation, storage and network load. The method reduces the delay, improves the throughput, and improves the overall performance and resource utilization rate of the system.

The invention discloses a computer self-check cycle optimization method based on the Internet of Things in the technical field of server maintenance, and relates to the technical field of server maintenance. Server node operation state data is collected in real time, a dynamic load evaluation model is constructed, and a comprehensive load index is calculated; and dynamically dividing performance states based on the threshold values L1 and L2, and adjusting a self-checking period and a module priority. Calculating a single-node risk index Ri by adopting a fault risk model, generating a fault generalization risk factor R1 through a probability graph model in combination with network topology, and triggering a linkage mechanism; dividing and dynamically adjusting a self-checking period according to the early warning level; during advanced early warning, redundant resources are distributed to preferentially process key faults, edge calculation is combined to carry out data cleaning and standardization, distributed management is realized through a consistent Hash algorithm, and finally, a visual interface is used for displaying an optimized scheduling result, so that a closed-loop optimization system of data acquisition, risk prediction and resource scheduling is formed.

The invention relates to the technical field of slope reinforcement, and discloses an anchoring device for preventing landslide geological disasters, the anchoring device is arranged on a slope surface, distributed in a grid shape and formed by splicing a plurality of anchoring units, each anchoring unit comprises an anchor rod support and four frames, the four frames are evenly distributed on the periphery of the anchor rod support, and the anchor rod support is arranged on the periphery of the anchor rod support. The anchor rod support is detachably connected with the frame; a center hole is formed in the anchor rod support, and an anchor rod inserted into a slope body is installed at the center hole. Every two adjacent frames are connected through a set of reinforcing mechanisms. The modular structural design is adopted, specifically, the detachable anchor rod supports and the frame are combined, the weight of a single node is small, carrying is convenient, manual carrying and rapid splicing are facilitated, a latticed splicing system formed by the scheme can adapt to terrains with different slopes, and the construction efficiency can be effectively improved.

The invention relates to the technical field of keyboard pairing, and discloses a keyboard multimode pairing method and device, equipment and a storage medium. The method comprises the following steps: establishing a switching time calculation model of a keyboard according to the reference switching time of keyboard hardware, a protocol node cooperation coefficient, an equipment response weight coefficient and a pre-connection optimization coefficient; calculating the theoretical optimal switching time of the current switching scene; selecting a working mode of single-node direct switching, double-node pre-connection or three-node parallel pre-connection of a keyboard Bluetooth node, a 2.4 GHz node and a USB node; activating a corresponding communication node, sending a connection handshake signal to the target equipment and establishing a communication link pre-connection state; and switching keyboard input and output from the current equipment to the target equipment to finish quick connection switching between the equipment. According to the invention, a pre-connection mechanism is adopted, the target equipment connection is established in advance before the user completes the switching instruction, passive response switching is converted into active prediction switching, and instantaneous equipment switching is realized.

The embodiment of the invention provides a data processing method and device based on a block chain network and a storage medium. The method relates to the technical field of data processing in the technical field of block chains. According to the method, a target standby node calls a system contract of a block chain network to generate first transaction information when detecting that a target master node has a fault, and under the condition that a consensus result of a first proposal block corresponding to the first transaction information indicates that the main node of the target node cluster is sponsored to be switched from the target main node to the target standby node, switching the main node in the target node cluster from the target main node to the target standby node. And the first proposal block is added to the block chain of the target standby node, so that the availability and fault-tolerant capability of the single node of the block chain can be improved, and the block confirmation delay is reduced, thereby improving the stability and performance of the whole block chain system.

A honeycomb-shaped active power distribution network topological structure selection and optimization method based on a graph theory comprises the following steps: taking line impedance and rated apparent power as constraints, constructing a multi-target weight factor, adopting an improved spectral clustering algorithm to cluster and distribute power grid nodes, and transiting a micro-grid group after cluster division into a honeycomb-shaped active power distribution network; abstracting each micro-grid as a honeycomb distribution network single node, constructing a connection relation mathematical representation of the HSPH and the micro-grid, and realizing sparse matrix method modeling of the honeycomb active power distribution network; optimizing a site selection strategy of the HSPH by taking a maximized HSPH return on investment index and a system stability index as a target function; and optimizing the energy storage capacity of the HSPH based on the net load time sequence data of the micro-grid group, and finally generating an optimal topological structure of the honeycomb-shaped active power distribution network. And through honeycomb topology and HSPH addressing and sizing optimization, the new energy consumption capability of the power distribution network is improved, and the economy, reliability and flexibility of the power distribution network are improved.

The invention relates to the technical field of automatic feeding, in particular to special automatic feeding equipment for Plasma, which comprises a fixing frame, a first plate feeding device, a second plate feeding device, a plugging and unplugging device, a transmission device and a distributed cooperative control system are arranged on the fixing frame, and the devices are electrically connected with a central controller; the first plate feeding device and the second plate feeding device are arranged side by side, the plugging and unplugging device is located on one side of the second plate feeding device, and the conveying device passes through the lower portion of the plugging and unplugging device. The distributed cooperative control system comprises a distributed control node triggered by a dynamic event, a digital twin cooperative optimization platform and a fault-tolerant self-healing module. And the distributed control node optimizes an actuator action strategy through a reinforcement learning algorithm. Plugging positioning precision of a clamping plate frame reaches + / -0.02 mm, the dislocation rate of a PCB is smaller than or equal to 0.1%, efficiency is optimized, production changing time is shortened by 83%, productivity is improved by 25%, reliability is enhanced, single-node fault recovery time is smaller than or equal to 5 s, and system availability is larger than or equal to 99.99%.

The invention provides a model training method, device and system based on federated learning and safety protection, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the current training of a target model based on local training data, and obtaining a current model parameter; the local training data is obtained by performing quantum encryption on the original training data; performing homomorphic encryption on the current model parameter to obtain a current encrypted model parameter; uploading the current encryption model parameter to an aggregation node; and receiving global model parameters distributed by the aggregation nodes, and carrying out next training on the target model based on the global model parameters. According to the method, the device and the system provided by the invention, quantum encryption is adopted in an acquisition stage, and homomorphic encryption is adopted in an interaction stage, so that the privacy of an aggregation process is ensured, and a potential attacker cannot obtain model updating details of any single node; in the collaborative training process, it is ensured that data privacy is strictly protected, data are not attacked or tampered, and the efficiency of multi-party collaborative training is improved.

The present disclosure relates to a liquid cooling server cabinet, including a cabinet body, a liquid cooling module, and a pluggable assembly. The cabinet body is provided with a receiving cavity, and the pluggable assembly is arranged in the receiving cavity. The pluggable assembly includes at least one pluggable portion. The liquid cooling module includes at least one liquid cooling unit corresponding to the at least one pluggable portion. The pluggable portion is configured for a plugging and mounting of a single-node server that defines a cooling cavity. When the pluggable portion is mounted with the single-node server, the liquid cooling unit is communicated with the cooling cavity of the corresponding single-node server.

The invention relates to the technical field of data processing and storage, in particular to an asynchronous verification and batch storage system and method for high-concurrency voucher data. Certificate verification is executed based on the memory snapshot, and verification time consumption is improved from the millisecond level of database IO to the microsecond level of memory calculation; a batch persistence mechanism enables the throughput of the system to be improved by dozens of times, and a concurrent storage scene of ten thousand-level TPS can be easily supported; the voucher storage response time is reduced from a hundred millisecond level to a millisecond level; the response is immediately returned to the user after the verification is passed, the user almost does not need to wait, and the operation experience is smooth; the introduction of the message queue enables the system to have peak clipping and valley filling capabilities, and can deal with burst traffic leisurely; the multi-node cluster deployment supports horizontal capacity expansion, the fault of a single node does not influence the overall service availability, and the system reliability is improved.

The invention relates to a configuration implementation method and system for a large-scale FPGA (Field Programmable Gate Array) interconnection system-on-chip. The system comprises 32 data interconnection FPGA core particles and a control FPGA core particle, the control FPGA is connected with the data FPGA through 32 groups of independent Slave Serial interfaces to realize parallel configuration, and state monitoring is carried out through 32 groups of UART (Universal Asynchronous Receiver / Transmitter) interfaces; configuration data are received externally by adopting an RGMII interface, and an upper computer is connected through a system UART. The method comprises the following steps that: after the FPGA is controlled to be self-started from the Flash, an upper computer issues a configuration data packet through the RGMII; the FPGA is controlled to synchronously drive 32 groups of Slave Serial time sequences to complete parallel configuration; the state is fed back in real time through a dual-mode monitoring mechanism, and timeout fault-tolerant logic is implanted. According to the configuration implementation method and system, the configuration efficiency is improved, interface signals are optimized and reduced, single-node fault isolation is supported, and the system availability is improved.

The invention discloses a multi-node MPI set communication performance modeling method and system, and the method comprises the steps: firstly determining model parameters needing to be measured according to the information of a multi-core processor in a node and the hierarchical information of a platform network; then measuring intra-node communication parameters and inter-node communication parameters on different network levels on the platform; analyzing a set communication source code and simulating a communication algorithm behavior to obtain all communication characteristics of each round of communication of the algorithm; and finally, corresponding parameters are selected in combination with communication characteristics, and the communication time of each round is represented by a mathematical formula of the corresponding parameters. Compared with an existing model, supercomputing network hierarchical structure information and a point-to-point communication primitive type used by a communication algorithm bottom layer are introduced, and the application range of the model is expanded from a single-node shared memory model to a hybrid model of a multi-node shared memory and an interconnection network; the model can more accurately predict the communication time of the multi-node set communication operation.