A model management method and system for digital twin platform

Abstract

The present invention discloses a model management method and system for a digital twin platform, which relates to the fields of water conservancy engineering and information technology, including integrating a model into a microservice architecture, debugging model parameters online, and optimizing model return results; obtaining the interface call of the model through a rest interface and formulating a call strategy; analyzing the model operation status, and visualizing the use details of the model according to the call data of the model. The model management method for a digital twin platform provided by the present invention realizes the rapid listing and dynamic management of the model, improves the flexibility and management efficiency of model deployment, ensures the reliability and stability of the model service, ensures the high adaptability and accuracy of the model output through meticulous parameter adjustment, flexibly configures the use authority according to user and on-site needs, and enhances the security and stability of the model service. Through authority control and call strategy, the present invention achieves better results in terms of flexibility, security and stability.

Description

Technical Field

[0001] The present invention relates to the fields of water conservancy engineering and information technology, and in particular to a model management method for a digital twin platform. Background Art

[0002] Smart water conservancy integrates traditional theories, technologies and methods of water conservancy projects, and deeply integrates modern information technologies such as artificial intelligence, big data, the Internet of Things, cloud computing, etc., to achieve intelligent perception, precise measurement, dynamic monitoring, efficient management and scientific decision-making of water resources. Smart water conservancy aims to improve the informatization, digitization, automation and intelligence level of water conservancy facility construction and management, optimize the allocation and utilization efficiency of water resources, strengthen the ability to prevent water and drought disasters, and ensure water resource security and sustainable development.

[0003] Build the smart water conservancy system version 2.0, comprehensively improve the informationization, digitization and intelligence level of the water conservancy industry, emphasize relying on river basins as units, improve water situation monitoring and intelligent scheduling capabilities, and take smart water conservancy construction as an important path to promote high-quality development of water conservancy. The core digital twin platform of the overall framework of smart water conservancy mainly includes model platform, knowledge platform and data backplane. The model platform formulates technical standards such as model platform development, model calling, sharing and interface in accordance with the requirements of "standardization, modularization and cloud service", ensures the universal encapsulation of various models and the standardization of model interfaces, and provides unified calling services in the form of microservices for calling by units at all levels. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the existing model management methods have the following problems: low model management flexibility, low security, low operating efficiency, and optimization problems of how to effectively evaluate and adjust and improve according to actual application effects.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a model management method for a digital twin platform, comprising integrating the model into a microservice architecture, debugging the model parameters online, and optimizing the model return results; obtaining the interface call of the model through the rest interface and formulating a call strategy; analyzing the model operation status and visualizing the model usage details based on the model call data.

[0007] As a preferred solution of the model management method of the digital twin platform described in the present invention, the integration of the model into the microservice architecture includes the decision of launching and launching the model. On the basis of model startup, the startup information of the model is recorded, the model is registered to nacos for service management, and the model is dynamically launched and launched in the running state, which is expressed as:

[0008]

[0009] Among them, M pub (t) is the model release decision function, which represents the model release status at time t, with a value range of (-∞,1], λ i is the release priority weight of model i, φ i is the environmental adaptation parameter of model i, g(λ i ,φ i ) is the information filtering function, which is used to evaluate the applicability of model i in a specific environment, τ0 is the initial time, t is the current time, μ is the average life cycle of the model, σ 2 is the standard deviation of the model life cycle, and n is the total number of models.

[0010] As a preferred solution of the model management method of the digital twin platform described in the present invention, wherein: the online debugging model parameters includes optimizing the model parameters, by measuring the difference between the model prediction value and the actual value, evaluating the stability and calculation cost of the model parameters, assuming that the model parameters are P = {p1, p2, ..., p n}, the objective function is expressed as:

[0011]

[0012] in, is the model prediction function, y(x) is the actual data function, Ω is the preset domain, and the prediction error is integrated in the domain. α is the regularization parameter, which is used to control the weight of the time derivative of the model parameters. β is an exponent, 0<β<1, which is used to adjust the balance between stability and error. δ is an exponent, δ>1, which is used to amplify the influence of model complexity. γ is an exponent, γ>1, which is used to amplify the influence of prediction error. is the model parameter p i The rate of change over time, used to evaluate the stability of the parameters, λ j is the absolute value of the model eigenvalue, which is used to evaluate the complexity and computational cost of the model, and m is the number of model eigenvalues ​​to solve the optimization problem. The closer the F(p) value is to 0, the closer the model's prediction is to the actual data, the more stable the parameters are, and the lower the computational cost is. The larger the F(p) value is, the larger the model's prediction error is, the more unstable the parameters are, or the computational cost is increased.

[0013] As a preferred solution of the model management method of the digital twin platform described in the present invention, the interface call for obtaining the model includes the call efficiency. Combined with the load distribution of the model, the state of the health instance and the triggering condition of the system self-protection mechanism, the comprehensive efficiency of the model call is expressed as:

[0014]

[0015] Where N is the total number of model instances in the system, W i is the weight of the i-th model instance, T i is the average response time of the i-th model instance within the preset time, f(s,ξ i ) is the function of the load and time relationship of the i-th model instance, s is the time variable, ξ i is the call rate parameter of the i-th model, α is the adjustment coefficient, balancing the impact of the number of healthy instances on efficiency, H i is the number of healthy instances of the i-th model instance, P i The performance score of the i-th model instance. The closer the E value is to 1, the higher the running efficiency of the model call is, and the closer it is to 0, the lower the running efficiency of the model call is.

[0016] As a preferred solution of the model management method of the digital twin platform described in the present invention, the formulation of the call strategy includes authority allocation and call strategy. According to the actual needs of users and the site, the model is controlled by the framework, different users are allocated different viewing or calling permissions, and the call strategy is formulated at the same time. The complexity of the user's authority allocation to the model and the call strategy is expressed as:

[0017]

[0018] Where Γ is the gamma function, which is used to calculate the integral of the power function of the user role level u, R(u) is the role level of user u, m is the identifier of model m, and sinh is the nonlinear response to the role level.

[0019] As a preferred solution of the model management method of the digital twin platform described in the present invention, the formulation of the call strategy also includes dynamic configuration of the model, which dynamically reflects the score of the model under different configurations by smoothing the influence of the model weight through a Gaussian function and combining the adjustment effect of the protection threshold, and performs model optimization and adjustment strategies. The dynamic configuration of the model is expressed as:

[0020]

[0021] Where W(m) is the weight of model m, T(m) is the protection threshold of model m, and cosh is the change of dynamic configuration score under different protection thresholds.

[0022] As a preferred scheme of the model management method of the digital twin platform described in the present invention, wherein: the analysis of the model operation status includes statistical analysis of the model usage details based on the model call time, frequency, model call calculation results and model call method, and presenting them in a list format, a visualization page to query the model operation status, healthy instances, dynamically configure the model weights under multiple instances, select instance allocation usage quotas, and configure protection thresholds. According to different sites and different server performance differences, instance allocation weights are selected. If the model call frequency is too high, or the server fails to power on, multiple instances will crash, and the instance protection threshold is set. When the protection threshold is exceeded, the model will start self-protection. Once the protection threshold is triggered, the request will be forwarded to all service instances to avoid system crashes.

[0023] Another object of the present invention is to provide a model management system for a digital twin platform, which can integrate the model into the microservice architecture through an integrated debugging module and support online debugging of model parameters, thereby solving the current problems of low efficiency of model debugging and low flexibility of model application.

[0024] As a preferred solution for the model management system of the digital twin platform described in the present invention, it includes: an integrated debugging module, a calling strategy module, and a statistical analysis module; the integrated debugging module is used to integrate the model into the microservice architecture, debug the model parameters online, and optimize the model return results; the calling strategy module is used to obtain the interface call of the model through the rest interface and formulate a calling strategy; the statistical analysis module is used to analyze the model operation status and visualize the model usage details based on the model's calling data.

[0025] A computer device includes a memory and a processor, wherein the memory stores a computer program, and wherein the processor executes the computer program to implement a step of a model management method for a digital twin platform.

[0026] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a model management method for a digital twin platform are implemented.

[0027] Beneficial effects of the present invention: The model management method of the digital twin platform provided by the present invention realizes the rapid listing and dynamic management of models, promotes the flexible deployment and efficient management of model services, improves the flexibility and management efficiency of model deployment, ensures the reliability and stability of model services, and simplifies the management process of the model. Through careful parameter adjustment, the high adaptability and accuracy of the model output are ensured, and more reliable and accurate data support is provided to users. The usage permissions are flexibly configured according to user and site needs, and the security and stability of the model service are enhanced. Through permission control and calling strategies, the rational use of services and the efficient allocation of system resources are guaranteed. The present invention achieves better results in flexibility, security and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0029] Figure 1 An overall flow chart of a model management method for a digital twin platform provided for the first embodiment of the present invention.

[0030] Figure 2 A flowchart of an implementation method of a model management method for a digital twin platform provided in the first embodiment of the present invention.

[0031] Figure 3 An overall flow chart of a model management system for a digital twin platform provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0033] Example 1

[0034] Reference Figure 1-Figure 2 , as an embodiment of the present invention, provides a model management method for a digital twin platform, comprising:

[0035] S1: Integrate the model into the microservice architecture, debug the model parameters online, and optimize the model return results.

[0036] Furthermore, the model is integrated into the microservice architecture including the on-shelf and off-shelf decisions.

[0037] It should be noted that, based on the model startup, the model startup information is recorded, the model is registered to nacos for service management, and the model is dynamically removed from the shelves during operation, as shown in the following figure:

[0038]

[0039] Among them, M pub (t) is the model release decision function, which represents the model release status at time t, with a value range of (-∞,1], λ i is the release priority weight of model i, φ i is the environmental adaptation parameter of model i, g(λ i ,φ i ) is the information filtering function, which is used to evaluate the applicability of model i in a specific environment, τ0 is the initial time, t is the current time, μ is the average life cycle of the model, σ 2 is the standard deviation of the model life cycle, and n is the total number of models.

[0040] Furthermore, online debugging of model parameters includes optimizing model parameters.

[0041] It should be noted that by measuring the difference between the model prediction value and the actual value, the stability and computational cost of the model parameters are evaluated. Let the model parameters be P = {p1, p2, ..., p n}, the objective function is expressed as:

[0042]

[0043] in, is the model prediction function, y(x) is the actual data function, Ω is the preset domain, and the prediction error is integrated in the domain. α is the regularization parameter, which is used to control the weight of the time derivative of the model parameters. β is an exponent, 0<β<1, which is used to adjust the balance between stability and error. δ is an exponent, δ>1, which is used to amplify the influence of model complexity. γ is an exponent, γ>1, which is used to amplify the influence of prediction error. is the model parameter p i The rate of change over time, used to evaluate the stability of the parameters, λ j is the absolute value of the model eigenvalue, which is used to evaluate the complexity and computational cost of the model, and m is the number of model eigenvalues ​​to solve the optimization problem. The closer the F(p) value is to 0, the closer the model's prediction is to the actual data, the more stable the parameters are, and the lower the computational cost is. The larger the F(p) value is, the larger the model's prediction error is, the more unstable the parameters are, or the computational cost is increased.

[0044] It should also be noted that the same model supports multi-instance deployment, and a polling algorithm is used by default to implement a load balancing strategy for model services, and current limiting measures can be formulated.

[0045] S2: Get the model's interface call through the REST interface and formulate a calling strategy.

[0046] Furthermore, the interface call of the acquisition model includes the call efficiency.

[0047] It should be noted that, combined with the load distribution of the model, the status of the healthy instance and the triggering conditions of the system self-protection mechanism, the comprehensive efficiency of the model call is expressed as:

[0048]

[0049] Where N is the total number of model instances in the system, W i is the weight of the i-th model instance, T i is the average response time of the i-th model instance within the preset time, f(s,ξ i ) is the function of the load and time relationship of the i-th model instance, s is the time variable, ξ i is the call rate parameter of the i-th model, α is the adjustment coefficient, balancing the impact of the number of healthy instances on efficiency, H i is the number of healthy instances of the i-th model instance, P i The performance score of the i-th model instance. The closer the E value is to 1, the higher the running efficiency of the model call is, and the closer it is to 0, the lower the running efficiency of the model call is.

[0050] Furthermore, formulating a calling strategy includes assigning permissions and calling strategies.

[0051] It should be noted that according to the actual needs of users and sites, the model is controlled by the framework, different users are assigned different viewing or calling permissions, and calling strategies are formulated. The complexity of user permission allocation and calling strategy for the model is expressed as:

[0052]

[0053] Where Γ is the gamma function, which is used to calculate the integral of the power function of the user role level u, R(u) is the role level of user u, m is the identifier of model m, and sinh is the nonlinear response to the role level.

[0054] Furthermore, formulating the calling strategy also includes dynamic configuration of the model.

[0055] It should be noted that by smoothing the influence of the model weight through a Gaussian function and combining it with the regulation of the protection threshold, the model scores under different configurations are dynamically reflected, and the model optimization and adjustment strategy are performed. The dynamic configuration of the model is expressed as:

[0056]

[0057] Where W(m) is the weight of model m, T(m) is the protection threshold of model m, and cosh is the change of dynamic configuration score under different protection thresholds.

[0058] It should also be noted that the same model supports different permission controls, and different users and different roles are assigned different permissions, dynamically meeting the needs of project site users to assign permissions.

[0059] S3: Analyze the model operation status and visualize the model usage details based on the model call data.

[0060] Furthermore, the analysis of the model operation status includes the model calling time, frequency, model calling calculation results and model calling method.

[0061] It should be noted that the use of the statistical analysis model is detailed and presented in a list format. The visualization page queries the model's operating status, healthy instances, dynamically configures model weights under multiple instances, selects instance allocation usage quotas, and configures protection thresholds. According to different sites and server performance differences, select instance allocation weights. If the model call frequency is too high, or the server fails to power on and multiple instances go down, set the instance protection threshold. When the protection threshold is exceeded, the model will start self-protection. Once the protection threshold is triggered, the request will be forwarded to all service instances to avoid system crashes.

[0062] Example 2

[0063] One embodiment of the present invention provides a model management method for a digital twin platform. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0064] The experiment selected two groups of test objects, one group used the model platform based on SpringCloud+Nacos of the present invention (referred to as the new platform), and the other group used the traditional model management method (referred to as the traditional platform). In each group of tests, the model was used to simulate water level control and flow prediction to evaluate the performance of different management platforms in terms of processing speed, accuracy, scalability and fault tolerance. The new platform automatically recorded the model startup information and registered it to Nacos for management, while the traditional platform was manually deployed and registered and lacked dynamic management capabilities. The new platform automatically adjusted the model parameters according to real-time data to optimize performance. The traditional platform required manual intervention and had low efficiency. The new platform provided flexible permission settings and dynamic model on and off-shelf functions to improve management efficiency. The traditional platform was cumbersome and slow to respond in this regard. The new platform automatically collected running data and performed statistical analysis to optimize model performance. The traditional platform lacked systematic analysis tools and relied on empirical judgment.

[0065] As shown in Table 1, the model deployment time of the new platform is significantly shorter than that of the traditional platform, indicating that the new platform can complete the model deployment work more quickly, which is particularly important for application scenarios that require frequent updates or replacements of models. The parameter adjustment response time of the new platform is much lower than that of the traditional platform, which means that on the new platform, the adjustment and optimization of model parameters can be carried out more quickly, so that it can adapt to environmental changes or experimental needs more quickly. The response time of the new platform in permission setting is also significantly better than that of the traditional platform, which shows that the new platform is more flexible and efficient in user permission management, which helps to improve the security and manageability of the system. The average call latency of the new platform (including multi-instance deployment) is lower than that of the traditional platform, indicating that the new platform is faster in processing requests and can provide a smoother user experience. The model accuracy of the new platform is slightly higher than that of the traditional platform, especially in multi-instance deployment, the accuracy is further improved, which proves that the new platform can improve the prediction or calculation accuracy of the model while ensuring the quality of service.

[0066] Table 1 Experimental comparison table

[0067]

[0068] Example 3

[0069] Reference Figure 3 , as an embodiment of the present invention, provides a model management system for a digital twin platform, including: an integrated debugging module, a calling strategy module, and a statistical analysis module.

[0070] The integrated debugging module is used to integrate the model into the microservice architecture, debug the model parameters online, and optimize the model return results; the call strategy module is used to obtain the interface call of the model through the rest interface and formulate the call strategy; the statistical analysis module is used to analyze the model operation status and visualize the model usage details based on the model call data.

[0071] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0072] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0073] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0074] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A model management method for a digital twin platform, characterized in that: include: Integrate the model into the microservice architecture, debug model parameters online, and optimize the model return results; Obtain the model's interface call through the REST interface and formulate a call strategy; Analyze the model operation status and visualize the model usage details based on the model call data; The integration of the model into the microservice architecture includes the decision of whether to launch or remove the model. On the basis of model startup, the startup information of the model is recorded, the model is registered with nacos for service management, and the model is dynamically launched or removed from the shelves in the running state, which is expressed as: Among them, M pub (t) is the model release decision function, which represents the model release status at time t, with a value range of (-∞,1], λ i is the release priority weight of model i, φ i is the environmental adaptation parameter of model i, g(λ i ,φ i ) is the information filtering function, which is used to evaluate the applicability of model i in a specific environment, τ0 is the initial time, t is the current time, μ is the average life cycle of the model, σ 2 is the standard deviation of the model life cycle, and n is the total number of models; The online debugging of model parameters includes optimizing model parameters, evaluating the stability and computational cost of model parameters by measuring the difference between model prediction values ​​and actual values, and assuming that model parameters are P = {p1, p2, ..., p n }, the objective function is expressed as: in, is the model prediction function, y(x) is the actual data function, Ω is the preset domain, and the prediction error is integrated in the domain. α is the regularization parameter, which is used to control the weight of the time derivative of the model parameters. β is an exponent, 0<β<1, which is used to adjust the balance between stability and error. δ is an exponent, δ>1, which is used to amplify the influence of model complexity. γ is an exponent, γ>1, which is used to amplify the influence of prediction error. is the model parameter p i The rate of change over time, used to evaluate the stability of the parameters, λ j is the absolute value of the model eigenvalue, which is used to evaluate the complexity and computational cost of the model, and m is the number of model eigenvalues ​​to solve the optimization problem. The closer the F(p) value is to 0, the closer the model's prediction is to the actual data, the more stable the parameters are, and the lower the computational cost is. The larger the F(p) value is, the greater the model's prediction error is, the more unstable the parameters are, or the computational cost is increased. The interface call of the acquisition model includes the call efficiency. Combined with the load distribution of the model, the status of the healthy instance and the triggering conditions of the system self-protection mechanism, the comprehensive efficiency of the model call is expressed as: Where N is the total number of model instances in the system, W i is the weight of the i-th model instance, T i is the average response time of the i-th model instance within the preset time, f(s,ξ i ) is the function of the load and time relationship of the i-th model instance, s is the time variable, ξ i is the call rate parameter of the i-th model, α is the adjustment coefficient, balancing the impact of the number of healthy instances on efficiency, H i is the number of healthy instances of the i-th model instance, P i The performance score of the i-th model instance. The closer the E value is to 1, the higher the running efficiency of the model call is, and the closer it is to 0, the lower the running efficiency of the model call is. The formulation of the call strategy includes authority allocation and call strategy. According to the actual needs of users and the site, the model is controlled by the framework. Different users are assigned different viewing or calling permissions. At the same time, the call strategy is formulated. The complexity of the user's authority allocation to the model and the call strategy is expressed as: Where Γ is the gamma function, which is used to calculate the integral of the power function of the user role level u, R(u) is the role level of user u, m is the identifier of model m, and sinh is the nonlinear response to the role level; The formulation of the call strategy also includes dynamic configuration of the model. By smoothing the influence of the model weight through a Gaussian function and combining it with the regulation of the protection threshold, the score of the model under different configurations is dynamically reflected to optimize and adjust the model strategy. The dynamic configuration of the model is expressed as: Where W(m) is the weight of model m, T(m) is the protection threshold of model m, and cosh is the change of dynamic configuration score under different protection thresholds.

2. The model management method of the digital twin platform according to claim 1, characterized in that: The analysis of model operation status includes statistical analysis of model usage details based on the model's call time, frequency, model call calculation results and model call method, presented in a list format, and a visualization page to query the model's operation status, healthy instances, dynamically configure model weights under multiple instances, select instance allocation usage quotas, and configure protection thresholds. According to different sites and server performance differences, select instance allocation weights. If the model call frequency is too high, or the server loses power and multiple instances go down, set the instance protection threshold. When the protection threshold is exceeded, the model will start self-protection. Once the protection threshold is triggered, the request will be forwarded to all service instances to avoid system crashes.

3. A system using the model management method of the digital twin platform as claimed in any one of claims 1 to 2, characterized in that: Includes integrated debugging module, calling strategy module, and statistical analysis module; The integrated debugging module is used to integrate the model into the microservice architecture, debug the model parameters online, and optimize the model return results; The calling strategy module is used to obtain the interface call of the model through the rest interface and formulate the calling strategy; The statistical analysis module is used to analyze the operation status of the model and visualize the usage details of the model based on the call data of the model.

4. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the model management method of the digital twin platform described in any one of claims 1 to 2 are implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the model management method of the digital twin platform described in any one of claims 1 to 2 are implemented.

Images