A cloud platform operation and maintenance method and system

Abstract

The present invention provides a cloud platform operation and maintenance method and system, which relates to the field of data processing technology. The method comprises obtaining multi-dimensional data of a cloud platform, integrating the multi-dimensional data of the cloud platform into a high-order tensor based on the Tucker decomposition method, and performing non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform; generating a spatiotemporal multi-scale topological map of the cloud platform based on the high-order tensor structure of the cloud platform; constructing a dynamic prediction model of the cloud platform based on the spatiotemporal multi-scale topological map of the cloud platform; preprocessing the multi-dimensional data of the cloud platform, and sending the preprocessing results to the dynamic prediction model for prediction to obtain a state prediction result of the cloud platform in a preset time period; and obtaining a preset operation and maintenance plan based on the state prediction result of the cloud platform in the preset time period for operation and maintenance. The present invention combines the characteristics of long-range dependence and sudden fluctuations to automatically perform resource scheduling and anomaly detection.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a cloud platform operation and maintenance method and system. Background Art

[0002] With the widespread adoption of cloud computing, cloud platforms have become critical infrastructure for data storage, computing resources, and application hosting for many businesses and individuals. Cloud platforms, by providing virtualization technology, on-demand scalable computing power, and elastic storage, have significantly facilitated the rapid development of the information age. However, while offering efficient and reliable services, cloud platforms also face increasingly complex operations and maintenance challenges. Existing cloud platform operations and maintenance technologies primarily rely on simple monitoring and statistical analysis methods, such as load balancing, capacity planning, and resource monitoring, relying on manual operations or static algorithms for optimization and anomaly detection. These traditional methods often rely on fixed rules and thresholds and are difficult to adapt to the dynamic changes and load fluctuations of cloud platforms. In particular, when faced with complex network environments, dynamic resource consumption, and sudden load fluctuations, they are prone to resource waste, performance bottlenecks, or system crashes. Consequently, existing technologies are unable to cope with the complexity, dynamism, and multi-factor interactions of cloud platform operations and maintenance, lacking real-time, intelligent optimization methods.

[0003] Therefore, there is an urgent need for a cloud platform operation and maintenance method and system to solve the above technical problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a cloud platform operation and maintenance method and system to solve the above problems. To achieve the above objectives, the technical solutions adopted by the present invention are as follows:

[0005] On the one hand, this application provides a cloud platform operation and maintenance method, including:

[0006] Acquire multi-dimensional data from the cloud platform, including energy consumption data, environmental change data, computing task data, storage data, and abnormal event data;

[0007] Integrating the multi-dimensional data of the cloud platform into a high-order tensor based on the Tucker decomposition method, and performing non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform;

[0008] Performing multi-scale topological graph construction and geometric mapping processing in hyperbolic space on the high-order tensor structure of the cloud platform to generate a spatiotemporal multi-scale topological graph of the cloud platform;

[0009] A nonlinear delay differential equation model is constructed based on the spatiotemporal multiscale topological graph of the cloud platform, and combined with a singular spectrum analysis method to obtain a dynamic prediction model of the cloud platform;

[0010] Preprocess the multidimensional data based on the cloud platform, and input the preprocessing results as input data into the dynamic prediction model of the cloud platform for prediction, and obtain the status prediction results of the cloud platform in the preset time period;

[0011] Based on the status prediction results of the cloud platform in a preset time period, a preset operation and maintenance plan is obtained for operation and maintenance.

[0012] On the other hand, this application also provides a cloud platform operation and maintenance system, including:

[0013] An acquisition unit, configured to acquire multi-dimensional data from the cloud platform, the multi-dimensional data including energy consumption data, environmental change data, computing task data, storage data, and abnormal event data;

[0014] An integration unit, configured to integrate the multi-dimensional data of the cloud platform into a high-order tensor based on a Tucker decomposition method, and perform non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform;

[0015] A mapping unit, configured to construct a multi-scale topological map of the high-order tensor structure of the cloud platform and perform geometric mapping processing in a hyperbolic space to generate a spatiotemporal multi-scale topological map of the cloud platform;

[0016] A construction unit is used to construct a nonlinear delay differential equation model based on the spatiotemporal multiscale topological graph of the cloud platform, and combine it with a singular spectrum analysis method to obtain a dynamic prediction model of the cloud platform;

[0017] The prediction unit is used to pre-process the multi-dimensional data of the cloud platform and input the pre-processing results as input data into the dynamic prediction model of the cloud platform for prediction, thereby obtaining the state prediction results of the cloud platform in a preset time period;

[0018] The operation and maintenance unit is used to obtain a preset operation and maintenance plan for operation and maintenance based on the status prediction result of the cloud platform in a preset time period.

[0019] The beneficial effects of the present invention are:

[0020] By combining advanced algorithms such as fractional Brownian motion, nonlinear delay differential equations, and multifractal detrended fluctuation analysis, this invention can achieve accurate modeling and real-time analysis of multidimensional data on cloud platforms. By inputting multidimensional data such as the cloud platform's load, resource consumption, and network bandwidth into a dynamic prediction model, combined with the characteristics of long-range dependencies and sudden fluctuations, the future state of the cloud platform can be accurately predicted, and resource scheduling and anomaly detection can be automatically performed. This method can dynamically adjust anomaly detection thresholds and optimization strategies based on the actual operating conditions of the cloud platform, greatly improving the automation and intelligence level of cloud platform operation and maintenance, avoiding the lag and static decision-making problems in traditional technologies, and providing more efficient, accurate, and flexible operation and maintenance support for the cloud platform.

[0021] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or be understood by practicing the embodiments of the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 This is a flow chart of the cloud platform operation and maintenance method of the present invention;

[0024] Figure 2 Schematic diagram of the cloud platform operation and maintenance system of the present invention.

[0025] Markings in the figure:

[0026] 701, acquisition unit; 702, integration unit; 703, mapping unit; 704, construction unit; 705, prediction unit; 706, operation and maintenance unit. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first" and "second" are used only to distinguish the description and should not be understood as indicating or implying relative importance. Example 1:

[0029] like Figure 1 As shown, this embodiment provides a cloud platform operation and maintenance method, including step S1, step S2, step S3, step S4, step S5 and step S6.

[0030] Step S1: Acquire multi-dimensional data from a cloud platform, where the multi-dimensional data includes energy consumption data, environmental change data, computing task data, storage data, and abnormal event data;

[0031] It is understandable that this step not only involves real-time monitoring of various indicators of the cloud platform, but also needs to ensure the diversity and comprehensiveness of the data, covering multiple dimensions including energy consumption, environmental changes, computing tasks, storage and abnormal events. Specifically, energy consumption data reflects the energy consumption of the cloud platform when performing computing tasks, which is crucial for evaluating resource utilization efficiency and cost control; environmental change data focuses on external environmental factors, such as climate change, temperature and humidity, which affect the cooling efficiency, equipment performance and stability of the cloud platform's data center; computing task data involves the status, scheduling information and load conditions of various computing tasks on the platform, reflecting the platform's operating load and resource usage; storage data is related to the utilization of storage devices in the cloud platform, storage bandwidth, data access frequency, etc., helping to analyze data liquidity and storage bottlenecks; abnormal event data records various anomalies that occur during the operation of the platform, such as failures, downtime, resource allocation errors, etc., and is an important basis for anomaly detection and fault diagnosis.

[0032] Acquiring this multi-dimensional data involves more than just acquiring static data through basic sensors or monitoring tools. It also involves the continuous monitoring and real-time updating of dynamic data streams, ensuring sufficient understanding of the cloud platform's status and the ability to respond to rapidly changing workloads and environmental changes. Technically, real-time monitoring and big data analysis technologies, combined with the cloud platform architecture, provide comprehensive platform awareness. The richness and real-time nature of this information provides powerful data support for subsequent anomaly detection, dynamic prediction, and adaptive optimization. The multi-dimensional data obtained through this step not only provides the foundational data for subsequent data analysis and model building, but also provides a real-time, comprehensive data source for subsequent accurate predictions, optimization strategies, and adaptive operations and maintenance.

[0033] Step S2: Integrate the multi-dimensional data of the cloud platform into a high-order tensor based on the Tucker decomposition method, and perform non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform;

[0034] It can be understood that this step effectively integrates and optimizes the sampling of the cloud platform data through Tucker decomposition and non-uniform sampling, providing a precise high-dimensional data representation for subsequent analysis and model training. This also improves data processing efficiency and avoids the interference of redundant information on subsequent analysis. This method can effectively capture the complex interactions between various dimensions in the cloud platform, revealing potential patterns and dependencies, and providing important data support for accurate prediction and optimization. In this step, step S2 includes steps S21, S22, S23, and S24.

[0035] Step S21: Based on the Tucker decomposition method, the multi-dimensional data of the cloud platform is represented as the product of a set of core tensors and multiple factor matrices, wherein each factor matrix represents the feature space of data of different dimensions, thereby obtaining the multi-dimensional interaction features of the cloud platform;

[0036] It can be understood that this step integrates the multidimensional data of the cloud platform into a high-order tensor. Each dimension of data (such as time, node, task type, resource usage, etc.) corresponds to an axis of the tensor. Since the multidimensional data of the cloud platform is usually a highly complex matrix with very large dimensions and scale, directly using the original data for analysis is extremely complex and computationally expensive. In order to effectively extract the interactive features in the data and reduce the computational complexity, this step uses Tucker decomposition to simplify the high-dimensional data into multiple factor matrices and a core tensor. Tucker decomposition decomposes the original multidimensional data into multiple factor matrices, where each factor matrix represents the feature space of data of different dimensions. For example, energy consumption data is in one factor matrix, computing task data is in another factor matrix, and storage data is in another factor matrix. The core tensor reveals the interactive relationship between these dimensions and represents the interaction between different feature dimensions.

[0037] Each factor matrix represents the data feature space for a dimension. These matrices can be viewed as "compressed representations" of each dimension, reducing the complexity of the original data by extracting the important features along that dimension. For example, for energy consumption data on a cloud platform, the factor matrix contains energy consumption patterns over different time periods. For compute task data, the factor matrix represents the resource requirements for different task types or load levels. Each factor matrix extracts the key features of the data in that dimension using matrix decomposition techniques (such as singular value decomposition and principal component analysis). During the Tucker decomposition process, these factor matrices are multiplied with the core tensor to produce a high-order tensor. The resulting data representation reveals the complex interactions between various dimensions. The core tensor is typically obtained by optimizing the product of factor matrices. It captures the complex relationships between various dimensions of the cloud platform (such as compute resources, storage, and network). For example, the core tensor reveals energy consumption trends under high compute loads or the interplay between task scheduling and storage bottlenecks.

[0038] The combination of factor matrices and core tensors obtained through Tucker decomposition not only effectively reduces the dimensionality and simplifies the complexity of the data, but also retains the key interactive features of the cloud platform while reducing computing costs.

[0039] Step S22: performing non-uniform sampling on the multi-dimensional interaction features of the cloud platform, and filtering the multi-dimensional interaction features to obtain a data set that meets preset conditions;

[0040] It's understandable that this step sets an appropriate sampling strategy based on the volatility and change patterns of data in each dimension of the cloud platform. For example, when the cloud platform load increases dramatically, storage pressure increases, or network bandwidth fluctuates, the system will increase the sampling frequency to capture key changes. When the load is low and the system is running smoothly, the sampling frequency can be appropriately reduced to avoid processing redundant data. Through non-uniform sampling, the cloud platform can dynamically adjust the sampling frequency based on real-time status, ensuring that sampled data fully covers important time periods.

[0041] After non-uniform sampling, this step further involves data filtering, selecting datasets that meet pre-defined criteria from the resulting multi-dimensional interactive features. These criteria are typically based on the cloud platform's operational objectives, anomaly detection requirements, and resource optimization strategies. These criteria can include threshold requirements for key metrics (load, energy consumption, etc.) or specific focus on system performance bottlenecks. These criteria can be defined based on historical data, system health, and specific task requirements, allowing the non-uniform sampling results to be filtered out for the most important components for subsequent analysis.

[0042] Step S23: construct a high-resolution spatiotemporal graph based on the data set that meets the preset conditions, where each functional node represents a specific function of the cloud platform, and each functional edge represents the interaction between different functional nodes;

[0043] It's understandable that this step associates information related to various cloud platform functions within the dataset that meets pre-defined criteria. Each data point (such as load and energy consumption) is mapped to a specific functional node. For example, compute task data is associated with the "compute node" function, while storage data is associated with the "storage node" function. Each functional node is assigned corresponding metrics (such as resource consumption and performance) based on its actual role and status within the cloud platform. These nodes serve as the basic building blocks of the graph, representing the core functional modules of the cloud platform.

[0044] In the spatiotemporal graph, each functional edge represents the interaction between different functional nodes. These edges reveal the dependencies or interactions between functional modules. For example, fluctuations in computing resource load can affect the efficiency of storage resource utilization, and the execution of computing tasks can cause fluctuations in bandwidth demand. This step defines interaction edges between nodes based on multidimensional cloud platform data (such as task scheduling, resource allocation, and data flow). The weight of each edge reflects the strength of the interaction and dependency between different functional nodes. Edge weights can be determined by calculating the similarity or dependency between two nodes, for example, based on metrics such as load fluctuation, data transmission volume, and resource consumption.

[0045] Next, based on the high-frequency, real-time data collected, the nodes and edges of the spatiotemporal graph are updated and adjusted over time based on the actual system operating status. For example, during peak load periods, the interaction edges between compute nodes and storage nodes are weighted and adjusted to reflect the resource constraints of the cloud platform. A high-resolution spatiotemporal graph continuously captures changes in the interactions between nodes in the cloud platform and updates the graph structure in real time. By constructing a high-resolution spatiotemporal graph, the functional modules of the cloud platform and their interactions can be accurately represented. This graph not only reflects the usage of internal cloud platform resources but also dynamically adapts to changes over time.

[0046] Step S24: Based on the tensor construction technology, the data sets of different dimensions in the high-resolution spatiotemporal graph that meet the preset conditions are merged together to form a high-order tensor structure.

[0047] It can be understood that this step extracts the functional modules and their interactions represented by the nodes and edges in the graph. This data is then used as different dimensions of a tensor. Each dimension can represent a specific function, such as compute nodes, storage nodes, or network nodes; each edge represents the interaction between different nodes. Nodes and edges that meet pre-defined criteria (for example, key nodes such as those experiencing periods of significant load fluctuations or storage bottlenecks) are screened for subsequent processing. Then, using tensor construction techniques, the feature spaces of the data from each dimension are integrated to form a high-order tensor. Each dimension represents a functional node of the cloud platform or its state, and each tensor element stores specific interaction or status information between the functional modules of the cloud platform. For example, a tensor might contain a "compute node" dimension, a "storage node" dimension, and a "bandwidth usage" dimension, with each tensor element corresponding to the interaction or status characteristics between these functional nodes at a specific moment in time.

[0048] Next, this step standardizes the data of each dimension in the high-order tensor so that the data of each dimension can be compared under the same dimension. By removing outliers, normalizing the data range, etc., it is ensured that each element of the tensor can truly reflect the changes and interactions in the state of the cloud platform. After standardization, the multi-dimensional structure of the data set will be more in line with the requirements of subsequent analysis and modeling. By fusing the data in the high-resolution spatiotemporal graph into a high-order tensor structure based on tensor construction technology, the multi-dimensional data of the cloud platform can be effectively integrated into a compact and information-rich data representation. This structure not only effectively captures the interactive relationship between the various functional modules of the cloud platform, but also can further explore the potential laws and patterns of the cloud platform through methods such as tensor decomposition.

[0049] Step S3: constructing a multi-scale topological map and performing geometric mapping processing in a hyperbolic space on the high-order tensor structure of the cloud platform to generate a spatiotemporal multi-scale topological map of the cloud platform;

[0050] It can be understood that this step, through multi-scale topological graph construction and hyperbolic space mapping, enables the cloud platform's spatiotemporal multi-scale topological graph to accurately represent each functional module and their interactions. Through hierarchical clustering and adaptive algorithm optimization, the multi-scale topological graph refines the hierarchical structure of the functional modules in the cloud platform and reveals the deep interactions between nodes. The geometric mapping of hyperbolic space further enhances the representation of the graph structure, revealing nonlinear and unbalanced interaction patterns in the cloud platform. In this step, step S3 includes steps S31, S32, S33, and S34.

[0051] Step S31: construct a multi-scale topological graph of the high-order tensor structure of the cloud platform, wherein the data of the high-order tensor is decomposed into a graph structure of multiple scales, wherein at each scale, each scale node represents a different physical module of the cloud platform, and each scale edge represents the interaction relationship between nodes, thereby obtaining a preliminary topological structure of the cloud platform;

[0052] As can be understood, this step first performs data decomposition. At each scale, the high-order tensor data is assigned to a graph structure. At each scale, nodes represent physical modules (such as virtual machines and physical devices) within the cloud platform, while each edge represents the interaction or dependency relationship between these nodes. This decomposition allows for the observation of interactions between cloud platform modules at different levels, allowing for subsequent analysis and modeling based on the data characteristics at different scales. After constructing the node and edge relationships at each scale, the graph structures at all scales are integrated to form a preliminary topological structure for the cloud platform. Combining the graph structures at each scale yields a hierarchical topological graph. Each topological graph at each scale represents a different perspective on the cloud platform. For example, a global perspective provides a macroscopic view of the interactions between various functional modules of the cloud platform, while a local perspective allows for in-depth analysis of the detailed interactions between specific modules (such as virtual machines and physical devices). This hierarchical approach effectively simplifies the complex structure of the cloud platform, facilitating multi-dimensional, cross-level analysis. By decomposing the cloud platform's high-order tensor structure into topological graphs at multiple scales, the multi-level interactions between specific modules of the cloud platform can be effectively revealed. The graph structure at each scale provides the cloud platform with dynamic features at different levels and granularities, which helps capture complex patterns of resource usage, load changes, etc.

[0053] Step S32: performing hierarchical clustering based on the preliminary topological structure of the cloud platform, wherein the hierarchical clustering is performed on the scale nodes, and the scale nodes after the hierarchical clustering are hierarchically divided by an adaptive algorithm to obtain a preliminary topological structure of the multi-level cloud platform;

[0054] It can be understood that this step uses a hierarchical clustering algorithm to take scale nodes as input and divide them into multiple clusters based on the strength or similarity of interactions between nodes (such as resource sharing, load fluctuation, and data flow). When calculating node similarity, metrics such as cosine similarity and Pearson correlation coefficient can be used to measure the strength of relationships between different nodes in terms of resource consumption, task allocation, and other aspects. The clustering results reflect the similarity in performance and resource usage among different functional modules in the cloud platform (such as computing modules, storage modules, and network modules).

[0055] Step S33: Mapping the preliminary topological structure of the multi-level cloud platform based on a geometric mapping method in a hyperbolic space to obtain its corresponding hyperbolic space mapping result;

[0056] It can be understood that this step first uses geometric mapping methods in hyperbolic space to geometrically transform the nodes and edges in the multi-level topology graph, mapping the cloud platform's topology from traditional Euclidean space to hyperbolic space. The geometric model of hyperbolic space is implemented using the Lorentz model. During the mapping process, the position of each scale node in the cloud platform in hyperbolic space is adjusted based on its importance, function, and interaction relationship in the topology graph. The distance between nodes (i.e., similarity) reflects their functional relationship and interaction strength, while the connections between nodes (i.e., edges) represent their resource sharing, data flow, or dependency relationships. After mapping the nodes in the multi-level topology graph into hyperbolic space, geometric methods are used to adjust the relative positions of the nodes, so that nodes with strong interactions are mapped closer in space, while nodes with weaker interactions are mapped farther away. This process not only reflects the strength of relationships between nodes but also avoids information loss due to spatial constraints, ensuring that the topological structure retains as complete information as possible.

[0057] Step S34: Based on the hyperbolic space mapping result, the similarity and relative position between each node of the cloud platform are calculated, and graph optimization processing is performed based on the similarity and relative position between each node of the cloud platform to obtain a spatiotemporal multi-scale topological graph of the cloud platform.

[0058] It's understandable that this step first calculates the Euclidean distance between each pair of scale nodes in hyperbolic space. Scale nodes with smaller distances have higher similarity. Furthermore, the similarity calculation formula can be adjusted based on the intensity of functional interactions between scale nodes (such as load demand and resource consumption), so that functional and resource dependencies can be taken into account when calculating the distance between scale nodes. For example, the distance between a storage scale node and a compute scale node is not just geometric distance; it also includes the resource usage relationship and data flow pattern between them.

[0059] The similarity calculation formula is as follows:

[0060] ;

[0061] in, Represents a scale node and scale nodes The similarity between Represents the weight factor used to adjust the influence of geometric distance on similarity. Represents a scale node and scale nodes Euclidean distance in hyperbolic space, Represents the weight factor used to adjust the impact of functional interaction strength on similarity, Represents a scale node and scale nodes The strength of functional interaction between them.

[0062] The relative positions of scale nodes are determined through geometric mapping in hyperbolic space. The coordinates of each scale node in space reflect its functional relationship and interaction characteristics with other scale nodes. By calculating the relative positions of scale nodes, we can further understand the priority of each scale node in resource allocation on the cloud platform or their key role in task scheduling. For example, the close distance between the core computing scale node and the storage scale node and bandwidth scale node indicates a strong functional coupling relationship between them.

[0063] This step uses the shortest path algorithm for graph optimization. Edge weights between scale nodes are adjusted based on a similarity matrix, with edges between scale nodes with higher weights representing stronger interactions. The relative positions of scale nodes are also considered during the optimization process. The goal is to make the entire topology more efficient by adjusting the layout and connectivity of scale nodes. For example, scale nodes with similar resource requirements can be clustered together to reduce resource conflicts between them, or data transmission paths can be adjusted to reduce latency. By combining calculations based on scale node similarity and relative positions with graph optimization, the cloud platform's spatiotemporal multiscale topology map accurately reflects the relationships and interactions between scale nodes. Hyperbolic space mapping optimizes the representation of relationships between scale nodes, enhancing the topology map's expressiveness in complex dynamic environments. The optimized spatiotemporal topology map not only improves the efficiency of resource management and task scheduling but also provides a solid foundation for subsequent fault detection, anomaly prediction, and adaptive optimization.

[0064] Step S4: constructing a nonlinear delay differential equation model based on the spatiotemporal multiscale topological graph of the cloud platform, and combining it with a singular spectrum analysis method to obtain a dynamic prediction model of the cloud platform;

[0065] It can be understood that by combining nonlinear delay differential equations with singular spectrum analysis, the cloud platform's dynamic prediction model can simultaneously capture the complex nonlinear dynamic behavior and time delay effects in the system. This model not only provides accurate state predictions but also helps the cloud platform identify potential resource bottlenecks, load fluctuations, and failure risks in advance, thereby enabling intelligent resource scheduling and fault prevention. In this step, step S4 includes steps S41, S42, and S43.

[0066] Step S41: constructing a nonlinear delay differential equation model based on the spatiotemporal multiscale topological graph of the cloud platform, wherein the nonlinear delay differential equation model simulates the dynamic changes of the states of nodes at each scale in the cloud platform;

[0067] It's understandable that in this step, when constructing the nonlinear delay differential equation model, each scale node is represented as a dynamic variable, and the changes in node state are described by the equation. These dynamic variables are affected not only by their current state but also by the delay effects of other scale nodes. This step defines the interactions between scale nodes based on the multi-scale topology of the cloud platform. For example, the load of a compute-scale node (computing device) affects the bandwidth requirements of a storage-scale node (storage device), and this impact takes time to manifest itself on the storage-scale node. The model introduces a delay term to capture this latency effect, reflecting the dynamic interactions between nodes.

[0068] For each scale node, a nonlinear delay differential equation is used to describe how its state changes over time. For example, for a compute scale node, the equation involves the calculation of the current load, the impact of upstream tasks on its load, and the response of downstream storage scale nodes to changes in its load. By introducing nonlinear terms (such as power terms and exponential terms), the equation can capture the nonlinear characteristics of resource demand in the cloud platform. Furthermore, through delay terms, the model can simulate the indirect impact of resource demand or task scheduling on other parts of the system. By constructing a nonlinear delay differential equation model, it is possible to effectively simulate the dynamic changes in the state of each scale node in the cloud platform, especially in terms of inter-node interactions and time delay effects. This model can capture the nonlinear characteristics of cloud platform resource consumption, load fluctuations, and task scheduling, and reflect the interactive influences between nodes.

[0069] Among them, the nonlinear delay differential equation model is as follows:

[0070] ;

[0071] in, Indicates the status of the scale node The rate of change of resource demand, Indicates the status of the current scale node and delayed status Here, It is the delay time, which means that the interaction between nodes takes some time to manifest. This part can take into account nonlinear characteristics such as load fluctuation and task scheduling. Represents the delay state The impact on the current node, this part represents the self-feedback mechanism in the system.

[0072] Step S42: performing singular spectrum decomposition on the time series data of the spatiotemporal multi-scale topological graph of the cloud platform to obtain the main dynamic patterns of the time series data;

[0073] It can be understood that this step extracts time series data from the spatiotemporal multiscale topology graph as input for singular spectrum decomposition. Each time series represents the behavioral pattern of a node or nodes within the cloud platform over a specific time period. These sequences reflect fluctuations in load, bandwidth consumption, or computing tasks. By inputting this time series data into the singular spectrum decomposition method, the dynamic patterns of the cloud platform system at various time scales can be extracted. Through singular spectrum decomposition, the cloud platform's time series data is decomposed into several principal components, each representing an independent dynamic pattern in the system. Principal dynamic patterns are the most representative patterns in the system that significantly influence key factors such as resource consumption, load fluctuations, and task scheduling. These patterns can be periodic (such as daily or weekly load peaks) or non-periodic (such as sudden failures or temporary load surges). These dynamic patterns, including both periodic changes and sudden fluctuations, provide a comprehensive understanding of the cloud platform's operational status and behavior.

[0074] Step S43: Taking the main dynamic pattern of the time series data as input and combining it with a nonlinear delay differential equation model to construct a dynamic prediction model of the cloud platform.

[0075] It can be understood that this step combines the main dynamic patterns of time series data with a nonlinear delay differential equation model to construct a cloud platform dynamic prediction model that can comprehensively and accurately capture the complex dynamic interactions and delay effects between nodes at all scales in the cloud platform. This model can not only predict the cloud platform's resource requirements, load fluctuations, and task scheduling changes, but also help the cloud platform optimize resource allocation, improve system stability, and prevent failures. In this step, step S43 includes steps S431, S432, and S433.

[0076] Step S431: using the main dynamic mode of the time series data as input data, introducing the input data into a nonlinear delay differential equation model to perform spatiotemporal state prediction, and obtaining an initial dynamic prediction model;

[0077] It's understandable that in this step, the dynamic patterns extracted from singular spectrum analysis are input into the nonlinear delay differential equation model. The state of each scale node in the equation (such as load, storage requirements, and computing resource consumption) depends on both the current state and historical state, and there is a certain time delay in the interactions between nodes. For example, load fluctuations on compute nodes can affect the performance of storage nodes, and this impact takes time to manifest. The nonlinear term captures the complex nonlinear feedback effects in the cloud platform (such as resource contention and task scheduling).

[0078] During model training, historical cloud platform data and key dynamic patterns are used as input. Numerical methods (such as the Euler method) are used to solve nonlinear delay differential equations, resulting in predictions of the cloud platform's future state. Through repeated training and optimization, the model gradually adjusts its parameters, improving prediction accuracy. Preliminary prediction results include changes in the load of compute-scale nodes, the capacity requirements of storage-scale nodes, and the bandwidth of network-scale nodes. These results provide a basis for resource and task scheduling in the cloud platform. Compute-scale nodes represent computing devices, storage-scale nodes represent storage devices, and network-scale nodes represent network transmission devices.

[0079] Step S432: training an initial dynamic prediction model using a preset time series in the cloud platform historical data and a preset historical dynamic pattern to obtain a trained initial dynamic prediction model;

[0080] It can be understood that this step involves inputting time series from historical data (such as load data, energy consumption data, and task scheduling data for nodes at various scales on the cloud platform) into the initial dynamic prediction model. This time series data provides the actual state of the cloud platform at different points in time, helping the model learn the dynamic changes of the cloud platform under varying loads, resource demands, task scheduling, and other conditions. Through training, the model adjusts its internal parameters to more accurately align its predictions with actual data. Combined with dynamic pattern input, the model can more accurately understand the operating patterns of the cloud platform. Dynamic patterns typically include cyclical changes (such as daily load fluctuations), trend-based changes (such as gradually increasing resource demand), and sudden fluctuations (such as high load or failure conditions). By combining these patterns with historical data, the model can learn the long-term development trends and short-term fluctuation patterns of the cloud platform, thereby improving its prediction accuracy for future states.

[0081] Step S433: Adjust the delay parameter in the nonlinear delay differential equation in the initial dynamic prediction model based on the particle swarm optimization algorithm to obtain the dynamic prediction model of the cloud platform.

[0082] It can be understood that in this step, the dynamic prediction model of the cloud platform is obtained by adjusting the delay parameter in the nonlinear delay differential equation in the initial dynamic prediction model based on the particle swarm optimization step S433 and the particle swarm optimization algorithm.

[0083] In the initially constructed dynamic prediction model, a nonlinear delay differential equation is used to simulate the dynamic changes in the states of nodes at various scales within the cloud platform. The delay parameter is a key factor in determining the responsiveness and interaction strength of system behavior. The delay parameter represents the time delay required for interactions between nodes within the cloud platform. For example, the demand for storage resources from a computing task takes a certain amount of time to impact a storage node. Therefore, accurately setting the delay parameter is crucial to the model's prediction accuracy. Adjusting the delay parameter in the nonlinear delay differential equation model using a particle swarm optimization algorithm significantly improves the prediction accuracy of the cloud platform dynamic prediction model. The optimized model accurately simulates the interactions and feedback effects between nodes at various scales within the cloud platform, particularly the impact of time delay on resource demand and load fluctuations.

[0084] Step S5: pre-processing the multi-dimensional data based on the cloud platform, and inputting the pre-processing results as input data into the dynamic prediction model of the cloud platform for prediction, thereby obtaining the state prediction results of the cloud platform in a preset time period;

[0085] It can be understood that by combining fractional Brownian motion and multifractal detrended fluctuation analysis, the cloud platform's dynamic prediction model can more accurately capture dynamic changes in multiple scales, such as resource demand and load fluctuations, and excels in handling long-term dependencies and sudden fluctuations. The optimized input data provides the model with more detailed dynamic features, enabling the cloud platform to more accurately predict future states, optimize resource scheduling and load balancing, and thus improve the stability, efficiency, and reliability of the entire system. In this step, step S5 includes steps S51, S52, S53, and S54.

[0086] Step S51: constructing a fractional Brownian motion model based on the multidimensional data of the cloud platform, wherein the multidimensional data of the cloud platform is modeled as a random process with long-term dependence through the fractional Brownian motion modeling technology, thereby obtaining feature data with long-range dependence characteristics;

[0087] It's understandable that in this step, the fractional Brownian motion model is used to simulate the multidimensional data of the cloud platform (such as computing load, storage usage, and bandwidth requirements). This data exhibits strong temporal correlation. For example, the load of compute-scale nodes can remain high for a period of time, impacting the performance requirements of storage-scale nodes. By modeling this time series data as a random process with long-range dependencies, the fractional Brownian motion model can capture the long-term correlations between these scale nodes. When constructing the fractional Brownian motion model, the multidimensional data of the cloud platform is decomposed into multiple time series, each representing the state changes of a node. For example, the load data of compute-scale nodes, the storage usage of storage-scale nodes, and the bandwidth usage of network-scale nodes are input into the fractional Brownian motion model.

[0088] Through model training, the long-term dependency characteristics of each time series are obtained. Due to the self-similarity of the fractional Brownian motion model, it is able to capture the long-range dependencies and inherent patterns in these time series. The fractional Brownian motion modeling technique introduces the Hurst exponent (a measure of long-range dependencies) to quantify long-term dependencies in the data. The Hurst exponent ranges from 0 to 1 and reflects the strength of the long-term dependencies in the time series. For example, a Hurst exponent close to 1 indicates strong long-term dependencies, while a value close to 0.5 indicates strong short-term dependencies. By calculating the Hurst exponent of resource fluctuations at each scale node, we can understand the fluctuation pattern and resource consumption trends of that node in the future time period.

[0089] Step S52: performing multifractal detrended fluctuation analysis on the multidimensional data of the cloud platform, wherein adaptive fluctuation scales are extracted from the fluctuations of the multidimensional data of the cloud platform and the fluctuations after detrending are calculated at each adaptive scale to obtain fluctuation characteristics of the multidimensional data of the cloud platform at different scales;

[0090] It's understood that the multifractal detrended fluctuation analysis (MF-DFA) used in this step is an analytical method for processing non-stationary time series. It is particularly well-suited for capturing data with complex fluctuations and multi-scale characteristics. MF-DFA extracts fluctuation characteristics at different scales from a time series and removes the trend component from the data, enabling the analysis results to more accurately reflect the true fluctuation patterns of the system. In cloud platforms, resource consumption and load fluctuations often exhibit multi-scale fluctuations. MF-DFA effectively captures these fluctuations and reveals dynamic changes at different time scales. The specific operational process is as follows: MF-DFA segments the time series data into multiple windows, removes the trend component within each window, and calculates the fluctuation characteristics at each scale. By analyzing these fluctuation characteristics, the fluctuation patterns at different time scales in the cloud platform are revealed. MF-DFA is particularly well-suited for data with long-range dependencies and non-stationarity. Detrending is a key step in MF-DFA. First, at each adaptive fluctuation scale, a trend (such as a linear trend or a polynomial trend) is fitted to remove long-term variations in the data. Detrending the data leaves only the fluctuating components—changes in resource demand or load. This process eliminates stationary trends in the data, making the fluctuations more prominent. This allows subsequent analysis and modeling to focus on the system's dynamic fluctuations rather than its long-term trends. Calculating the fluctuation characteristics (Hurst exponent) at each scale involves calculating the detrended fluctuation amplitudes using the MF-DFA method. These amplitudes represent the intensity of fluctuations on the cloud platform over different time periods. For example, in the short term (hourly), the load on a compute-scale node exhibits sudden fluctuations, while in the long term (weekly), the load exhibits steady growth. Analyzing these fluctuation characteristics provides a clearer understanding of the dynamic changes in cloud platform resource usage.

[0091] Step S53: Fusing the feature data with long-range dependency characteristics and the fluctuation features of the multidimensional data of the cloud platform at different scales to generate multiple composite feature vectors, each of which contains the long-term fluctuation pattern and local sudden fluctuation pattern of the cloud platform at different time scales;

[0092] Step S54: input the composite feature vector into a cloud platform dynamic prediction model for prediction, and obtain a state prediction result of the cloud platform in a preset time period.

[0093] It can be understood that by inputting the composite feature vector into the cloud platform's dynamic prediction model, this step enables the cloud platform to accurately predict future state changes, particularly in terms of load fluctuations, resource demand, and task scheduling. The composite feature vector provides multi-scale information about the cloud platform's resource demand and load fluctuations, enabling the prediction model to more accurately reflect both long-term trends and sudden fluctuations in resource demand. Based on these prediction results, the dynamic prediction model can proactively optimize resource allocation and task scheduling, improving the cloud platform's adaptability, efficiency, and stability in complex and dynamic environments.

[0094] Step S6: Obtain a preset operation and maintenance plan based on the state prediction result of the cloud platform in a preset time period to perform operation and maintenance.

[0095] It should be understood that in this step, the system first selects the most suitable O&M solution from a library of pre-set O&M solutions based on the predicted cloud platform state (such as future load fluctuations, storage requirements, and bandwidth consumption). For example, if a compute node is predicted to experience an impending load peak, the system will select an O&M solution that expands computing resources; if a storage node is predicted to face a bottleneck, a data migration or capacity expansion solution will be selected. Each O&M solution includes a series of strategies, such as resource expansion, load balancing, fault warning, and task scheduling optimization. Once the most suitable O&M solution is selected, the system will automatically perform O&M operations based on the solution's specific strategies. These operations include resource scheduling, task optimization, load balancing, and fault prevention, ensuring the cloud platform operates efficiently and smoothly in the face of future state changes. By obtaining pre-set O&M solutions based on the predicted cloud platform state over a preset time period, the cloud platform can implement intelligent, adaptive O&M, proactively addressing load fluctuations, resource bottlenecks, and failure risks. Automated execution and adjustment of O&M solutions significantly improves the efficiency of resource scheduling and load balancing, while reducing human intervention and lowering O&M costs. In this way, the cloud platform can maintain efficient and stable operation in a complex dynamic environment, ensuring optimal utilization and reliability of system resources. Example 2:

[0096] according to Figure 2 As shown, this embodiment provides a cloud platform operation and maintenance system, which includes:

[0097] An acquisition unit 701 is configured to acquire multi-dimensional data from a cloud platform, wherein the multi-dimensional data includes energy consumption data, environmental change data, computing task data, storage data, and abnormal event data.

[0098] An integration unit 702 is configured to integrate the multi-dimensional data of the cloud platform into a high-order tensor based on a Tucker decomposition method, and perform non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform;

[0099] A mapping unit 703 is configured to construct a multi-scale topological map of the high-order tensor structure of the cloud platform and perform geometric mapping processing in a hyperbolic space to generate a spatiotemporal multi-scale topological map of the cloud platform;

[0100] A construction unit 704 is configured to construct a nonlinear delay differential equation model based on the spatiotemporal multiscale topological graph of the cloud platform, and combine the nonlinear delay differential equation model with a singular spectrum analysis method to obtain a dynamic prediction model of the cloud platform;

[0101] The prediction unit 705 is used to perform preprocessing based on the multi-dimensional data of the cloud platform, and input the preprocessing results as input data into the dynamic prediction model of the cloud platform for prediction, thereby obtaining the state prediction results of the cloud platform in a preset time period;

[0102] The operation and maintenance unit 706 is configured to obtain a preset operation and maintenance plan for operation and maintenance based on the status prediction result of the cloud platform in a preset time period.

[0103] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

[0104] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A cloud platform operation and maintenance method, characterized in that: include: Acquire multi-dimensional data from the cloud platform, including energy consumption data, environmental change data, computing task data, storage data, and abnormal event data; Integrating the multi-dimensional data of the cloud platform into a high-order tensor based on the Tucker decomposition method, and performing non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform; Performing multi-scale topological graph construction and geometric mapping processing in hyperbolic space on the high-order tensor structure of the cloud platform to generate a spatiotemporal multi-scale topological graph of the cloud platform; A nonlinear delay differential equation model is constructed based on the spatiotemporal multiscale topological graph of the cloud platform, and combined with a singular spectrum analysis method to obtain a dynamic prediction model of the cloud platform; Preprocess the multidimensional data based on the cloud platform, and input the preprocessing results as input data into the dynamic prediction model of the cloud platform for prediction, and obtain the status prediction results of the cloud platform in the preset time period; Obtaining a preset operation and maintenance plan based on the status prediction results of the cloud platform in a preset time period for operation and maintenance; The multi-dimensional data of the cloud platform is integrated into a high-order tensor based on the Tucker decomposition method, and the high-order tensor is non-uniformly sampled to construct a high-order tensor structure of the cloud platform, including: Based on the Tucker decomposition method, the multi-dimensional data of the cloud platform is represented as the product of a set of core tensors and multiple factor matrices, where each factor matrix represents the feature space of data of different dimensions, thereby obtaining the multi-dimensional interaction features of the cloud platform; Performing non-uniform sampling on the multi-dimensional interaction features of the cloud platform, and screening the multi-dimensional interaction features to obtain a data set that meets preset conditions; A high-resolution spatiotemporal graph is constructed based on a dataset that meets preset conditions, where each functional node represents a specific function of the cloud platform, and each functional edge represents the interaction between different functional nodes; Based on tensor construction technology, data sets of different dimensions that meet preset conditions in high-resolution spatiotemporal graphs are merged together to form a high-order tensor structure.

2. The cloud platform operation and maintenance method according to claim 1, characterized in that: The high-order tensor structure of the cloud platform is subjected to multi-scale topological graph construction and geometric mapping processing in hyperbolic space, including: Performing a multi-scale topological graph construction process on the high-order tensor structure of the cloud platform, wherein the data of the high-order tensor is decomposed into a graph structure of multiple scales, wherein at each scale, each scale node represents a different physical module of the cloud platform, and each scale edge represents an interaction relationship between nodes, thereby obtaining a preliminary topological structure of the cloud platform; Based on the preliminary topological structure of the cloud platform, hierarchical clustering is performed. The scale nodes are hierarchically clustered and then hierarchically divided into different levels using an adaptive algorithm to obtain a preliminary topological structure of the multi-level cloud platform. Based on the geometric mapping method in hyperbolic space, the preliminary topological structure of the multi-level cloud platform is mapped and processed to obtain its corresponding hyperbolic space mapping result; Based on the hyperbolic space mapping result, the similarity and relative position between each node of the cloud platform are calculated, and based on the similarity and relative position between each node of the cloud platform, graph optimization processing is performed to obtain a spatiotemporal multi-scale topological graph of the cloud platform.

3. The cloud platform operation and maintenance method according to claim 1, characterized in that: A nonlinear delay differential equation model is constructed based on the spatiotemporal multiscale topological graph of the cloud platform and combined with a singular spectrum analysis method, including: A nonlinear delay differential equation model is constructed based on the spatiotemporal multiscale topological graph of the cloud platform, wherein the nonlinear delay differential equation model simulates the dynamic changes in the states of nodes at each scale in the cloud platform; Performing singular spectrum decomposition on the time series data of the spatiotemporal multi-scale topological graph of the cloud platform to obtain the main dynamic patterns of the time series data; The main dynamic patterns of the time series data are used as input and combined with a nonlinear delay differential equation model to construct a dynamic prediction model of the cloud platform.

4. The cloud platform operation and maintenance method according to claim 3, characterized in that: The main dynamic patterns of the time series data are used as input and combined with the nonlinear delay differential equation model to construct a dynamic prediction model of the cloud platform, including: Taking the main dynamic mode of the time series data as input data, introducing the input data into a nonlinear delay differential equation model to perform spatiotemporal state prediction, thereby obtaining an initial dynamic prediction model; The initial dynamic prediction model is trained by using the time series in the preset cloud platform historical data and the preset historical dynamic pattern to obtain the trained initial dynamic prediction model; The delay parameters in the nonlinear delay differential equation in the initial dynamic prediction model are adjusted based on the particle swarm optimization algorithm to obtain the dynamic prediction model of the cloud platform.

5. A cloud platform operation and maintenance system, characterized in that: include: An acquisition unit, configured to acquire multi-dimensional data from the cloud platform, the multi-dimensional data including energy consumption data, environmental change data, computing task data, storage data, and abnormal event data; An integration unit, configured to integrate the multi-dimensional data of the cloud platform into a high-order tensor based on a Tucker decomposition method, and perform non-uniform sampling on the high-order tensor to construct a high-order tensor structure of the cloud platform; A mapping unit, configured to construct a multi-scale topological map of the high-order tensor structure of the cloud platform and perform geometric mapping processing in a hyperbolic space to generate a spatiotemporal multi-scale topological map of the cloud platform; A construction unit is used to construct a nonlinear delay differential equation model based on the spatiotemporal multiscale topological graph of the cloud platform, and combine it with a singular spectrum analysis method to obtain a dynamic prediction model of the cloud platform; The prediction unit is used to pre-process the multi-dimensional data of the cloud platform and input the pre-processing results as input data into the dynamic prediction model of the cloud platform for prediction, thereby obtaining the state prediction results of the cloud platform in a preset time period; An operation and maintenance unit, configured to obtain a preset operation and maintenance plan for operation and maintenance based on the status prediction result of the cloud platform in a preset time period; Wherein, the integration unit includes: a first integration subunit, configured to represent the multidimensional data of the cloud platform as a product of a set of core tensors and a plurality of factor matrices based on a Tucker decomposition method, wherein each factor matrix represents a feature space of data of different dimensions, thereby obtaining a multidimensional interaction feature of the cloud platform; The second integration subunit is configured to perform non-uniform sampling on the multi-dimensional interaction features of the cloud platform, and filter the multi-dimensional interaction features to obtain a data set that meets preset conditions; The third integration subunit is used to construct a high-resolution spatiotemporal graph based on a dataset that meets preset conditions. Each functional node represents a specific function of the cloud platform, and each functional edge represents the interaction between different functional nodes. The fourth integration subunit is used to merge data sets of different dimensions that meet preset conditions in the high-resolution spatiotemporal graph based on tensor construction technology to form a high-order tensor structure.

6. The cloud platform operation and maintenance system according to claim 5, characterized in that: The mapping unit includes: a first mapping subunit, configured to construct a multi-scale topological graph of the high-order tensor structure of the cloud platform, wherein the data of the high-order tensor is decomposed into a graph structure of multiple scales, wherein at each scale, each scale node represents a different physical module of the cloud platform, and each scale edge represents an interaction relationship between nodes, thereby obtaining a preliminary topological structure of the cloud platform; The second mapping subunit is used to perform hierarchical clustering based on the preliminary topological structure of the cloud platform, wherein the hierarchical clustering is performed on the scale nodes, and the scale nodes after the hierarchical clustering are hierarchically divided by an adaptive algorithm to obtain a preliminary topological structure of the multi-level cloud platform; The third mapping subunit is used to map the preliminary topological structure of the multi-level cloud platform based on the geometric mapping method in the hyperbolic space to obtain its corresponding hyperbolic space mapping result; The fourth mapping subunit is used to calculate the similarity and relative position between each node of the cloud platform based on the hyperbolic space mapping result, and perform graph optimization processing based on the similarity and relative position between each node of the cloud platform to obtain a spatiotemporal multi-scale topological map of the cloud platform.

7. The cloud platform operation and maintenance system according to claim 5, characterized in that: The building block comprises: A first construction subunit is configured to construct a nonlinear delay differential equation model based on the spatiotemporal multiscale topological graph of the cloud platform, wherein the nonlinear delay differential equation model simulates the dynamic changes in the states of nodes at each scale in the cloud platform; The second construction subunit is used to perform singular spectrum decomposition on the time series data of the spatiotemporal multi-scale topological graph of the cloud platform to obtain the main dynamic mode of the time series data; The third construction subunit is used to take the main dynamic mode of the time series data as input, and combine it with the nonlinear delay differential equation model to construct a dynamic prediction model of the cloud platform.

8. The cloud platform operation and maintenance system according to claim 7, characterized in that: The third building block comprises: a fourth construction subunit, configured to take the main dynamic mode of the time series data as input data, introduce the input data into a nonlinear delay differential equation model to perform spatiotemporal state prediction, and obtain an initial dynamic prediction model; A fifth construction subunit is configured to train an initial dynamic prediction model using a preset time series in the cloud platform historical data and a preset historical dynamic pattern to obtain a trained initial dynamic prediction model; The sixth construction subunit is used to adjust the delay parameter in the nonlinear delay differential equation in the initial dynamic prediction model based on the particle swarm optimization algorithm to obtain the dynamic prediction model of the cloud platform.

Images