Equipment cluster simulation deduction method and system based on digital twin drive
By using a digital twin-driven device cluster simulation method, dynamic utility functions and meta-reinforcement learning models are employed to automatically adjust simulation strategies, thus solving the problems of insufficient adaptability and intelligence in existing device cluster simulation methods and achieving efficient and accurate simulation decision-making and resource allocation.
Patent Information
- Application Number
- CN202510816269.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-18
- Publication Date
- 2025-11-07
AI Technical Summary
Existing equipment cluster simulation methods are not adaptable enough to complex operating environments and dynamic changes. They require manual adjustment of model parameters, have slow response times, and are difficult to achieve intelligent decision-making and efficient resource allocation.
The device cluster simulation and extrapolation method based on digital twins automatically selects simulation strategy configuration by defining dynamic utility functions and meta-reinforcement learning models, and optimizes the simulation strategy by combining continuous learning mechanisms to adapt to changes in operational objectives.
It enables real-time adjustment of simulation strategies, improves the timeliness and accuracy of decision-making, optimizes the use of computing resources, balances simulation accuracy and cost, and supports agile production and intelligent decision-making.
Smart Images
Figure CN120911243A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of device cluster management, and particularly relates to a device cluster simulation deduction method and system based on digital twin driving. BACKGROUND
[0002] With the deep integration of new generation information technology and manufacturing industry, digital twin technology as a key enabling technology to realize the interaction and symbiosis of physical world and information world has become the focus of attention in academia and industry. Digital twin constructs an accurate digital mirror image of physical entities in virtual space, integrates multi-physical, multi-scale and multi-probability simulation processes, and uses Internet of Things (IoT), big data analysis, artificial intelligence (AI) and other technologies to realize the state perception, diagnosis and prediction, optimization control and collaborative interaction of the whole life cycle of physical entities. At present, in the field of device cluster management and operation, the application of digital twin technology shows great potential. It can provide unprecedented insight for the design verification, performance evaluation, fault diagnosis, operation decision and production scheduling of complex device systems. Traditional device cluster simulation deduction methods often rely on pre-embedded models and fixed simulation parameters, which can simulate cluster behavior to some extent, but in the face of increasingly complex operating environments, dynamically changing production targets and evolution of device states, their adaptability and intelligence level have been difficult to meet the needs of modern industry. Especially in dealing with large-scale, heterogeneous device clusters, how to efficiently configure simulation resources, balance simulation accuracy and computing efficiency, and extract effective information from simulation results to guide actual operation has become a key bottleneck restricting its effectiveness.
[0003] Although the existing technology has made certain progress in device cluster simulation, there are still some significant shortcomings. The existing simulation deduction methods have insufficient adaptability to the dynamic evolution of operation targets. When external market demand, internal production strategy or device health status changes, resulting in the migration of operation targets, traditional simulation systems often need to manually reconfigure model parameters, adjust simulation logic, or even modify the code. This process not only consumes time and effort, but also responds slowly, making it difficult to support agile production and intelligent decision-making. In addition, the existing methods still lack the ability to intelligently configure and optimize simulation strategies. When faced with a complex simulation strategy space, how to automatically find the optimal or near-optimal simulation configuration according to the current specific operation target and the system state reflected by the digital twin, in order to obtain the most valuable simulation results for decision-making under limited computing resources, is a highly challenging problem, and most existing methods use trial-and-error methods, which are inefficient and difficult to guarantee optimality. SUMMARY
[0004] This section is intended to introduce some aspects of one or more embodiments of the present application, which are described below in the detail section. This section is not intended to limit the application in any way, but to provide insight into possible embodiments of the application.
[0005] In view of the above existing problems, the present application is proposed. Therefore, the present application provides a device cluster simulation deduction method based on digital twin driving, which is used to solve the problems mentioned in the background art.
[0006] To solve the above technical problems, the present application provides the following technical solutions:
[0007] In a first aspect, the present application provides a device cluster simulation deduction method based on digital twin driving, comprising:
[0008] Based on the digital twin of the device cluster, the dynamic utility function is defined for the current evolution operation target to evaluate the contribution of the simulation strategy and the simulation result to the operation target.
[0009] According to the dynamic utility function, a machine learning model is used to select a simulation strategy configuration for the current evolution operation target from a preset simulation strategy space, and the simulation strategy configuration is applied to execute simulation deduction on the digital twin.
[0010] Through the simulation deduction result and the evaluation result of the dynamic utility function, a continuous learning mechanism is created to iteratively optimize the machine learning model to enhance the simulation strategy configuration capability of the machine learning model for future evolution targets.
[0011] As a preferred scheme of the device cluster simulation deduction method based on digital twin driving, the dynamic utility function is defined, comprising:
[0012] The evolution operation target is mapped to target parameters including key performance indicators, target values, target types and constraint conditions through semantic analysis.
[0013] The dynamic utility function is constructed based on the target parameters, and the dynamic utility function is the expected achievement degree of the simulation deduction result and the cost benefit of the strategy configuration required to execute the simulation deduction under the current evolution operation target.
[0014] As a preferred scheme of the device cluster simulation deduction method based on digital twin driving, the machine learning model is an adaptive simulation strategy configuration based on meta-reinforcement learning training, and the training process includes learning how to select effective simulation strategy configuration according to different evolution operation targets.
[0015] As a preferred scheme of the device cluster simulation and deduction method based on digital twin driving provided in the application, wherein: the preset simulation strategy space at least contains one or more optional configurations of the model fidelity level of different parts in the digital twin, the time step of simulation calculation, and the priority of resource allocation.
[0016] As a preferred scheme of the device cluster simulation and deduction method based on digital twin driving provided in the application, wherein: in the process of performing simulation and deduction on the digital twin, the prediction uncertainty of the digital twin and the current evolution operation target output and the potential influence of the prediction on the dynamic utility function are evaluated, and the simulation parameters or the simulation fidelity level are selected or adjusted.
[0017] As a preferred scheme of the device cluster simulation and deduction method based on digital twin driving provided in the application, wherein: the continuous learning mechanism comprises:
[0018] The simulation strategy configuration knowledge learned under the current evolution operation target is applied to the simulation strategy configuration process of a new or changed evolution operation target through knowledge distillation.
[0019] As a preferred scheme of the device cluster simulation and deduction method based on digital twin driving provided in the application, wherein: the operation feedback of the device cluster includes the operation data of the device and the achieved value of the key performance indicator;
[0020] The digital twin of the device cluster is calibrated through the achieved value, and is used as the input of the continuous learning mechanism to optimize the machine learning model.
[0021] In a second aspect, the application provides a device cluster simulation and deduction system based on digital twin driving, comprising:
[0022] A target utility definition module is configured to define a dynamic utility function based on the digital twin of the device cluster, to evaluate the contribution of the simulation strategy and the simulation result to the current evolution operation target;
[0023] A simulation strategy configuration and digital twin simulation deduction module is configured to select a simulation strategy configuration for the current evolution operation target from a preset simulation strategy space according to the dynamic utility function by using a machine learning model, and perform simulation deduction on the digital twin by applying the simulation strategy configuration.
[0024] A simulation strategy optimization module is configured to create a continuous learning mechanism to iteratively optimize the machine learning model through the simulation deduction result and the evaluation result of the dynamic utility function, so as to enhance the simulation strategy configuration capability of the machine learning model for future evolution targets.
[0025] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements any step of the above method when executing the computer program.
[0026] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and wherein the computer program implements any step of the above method when executed by a processor.
[0027] Compared with the prior art, the present application has the following beneficial effects:
[0028] 1. The present application can represent and quantify the current evolution operation target in real time by constructing a dynamic utility function, and can adjust the simulation strategy according to the change of the target by using an adaptive simulation strategy configurator based on meta-reinforcement learning, thereby avoiding the hysteresis and inefficiency of manual adjustment in the traditional method, making the simulation deduction always serve the most important operation target, and improving the timeliness and accuracy of decision-making.
[0029] 2. By selecting a simulation strategy mechanism, the present application can direct the computing resources to the simulation area that is most sensitive to the current operation target and has the highest information value, thereby avoiding blind exploration in irrelevant or low-value areas, and at the same time, the adaptive simulation strategy configurator can select the optimal or suboptimal simulation configuration under the premise of meeting the target demand, thereby balancing the simulation accuracy and the computing cost. BRIEF DESCRIPTION OF DRAWINGS
[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on these drawings. Among them:
[0031] Figure 1 The overall flowchart of the device cluster simulation deduction method based on digital twin driving according to an embodiment of the present application. DETAILED DESCRIPTION
[0032] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.
[0033] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.
[0034] It should also be noted that, as used in the specification and in the claims, the article "a", "an", or "the" is intended to mean that there is one or more of the features or elements. Unless otherwise indicated, the use of the terms "coupled" and / or "connected", means that two or more elements, either directly or indirectly, are linked in
[0035] The present application is described in detail below with reference to the attached drawing figures, wherein the implementations of the present application are shown by way of examples as well as specific embodiments. As those skilled in the art will understand, the application described herein can be embodied in many different forms. In the description of the present application that follows, parallel aspects of the present application are described with reference made to the drawings, in which the drawings themselves should not be considered limiting of the present application's scope as these drawings are provided merely for explanation and understanding.
[0036] In the description of the present application, it should be noted that the terms "upper and lower", "inner and outer", and the like, indicate the positional or spatial relationship based on the orientation or position relationship shown in the drawings, and are merely for the purpose of facilitating the description of the present application and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first, second or third" are for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0037] Unless otherwise defined, the terms "mounting", "connected", "connecting" are to be construed broadly, for example: can be fixed connection, detachable connection or integral connection; can also be mechanical connection, electrical connection or direct connection, can also be indirectly connected through intermediate medium, or can be internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.
[0038] Embodiment 1
[0039] Reference Figure 1 For the first embodiment of the present application, the embodiment provides a device cluster simulation deduction method based on digital twin driving, comprising:
[0040] S1, based on the digital twin of the device cluster, for the current evolution operation target, define a dynamic utility function to evaluate the contribution of the simulation strategy and simulation results to the operation target;
[0041] It needs to be explained that the digital twin of the device cluster is a virtual model for real-time mapping of the state, behavior and operating environment of the physical device cluster. In practical applications, the evolution operation goals (such as efficiency, cost, reliability, etc.) of the physical device cluster are usually complex and diverse, and may also involve multiple dimensions (such as efficiency key performance indicators, cost constraint conditions, etc.). Therefore, in order to ensure that the definitions of these goals and dimensions are consistent and unambiguous, an ontology library needs to be created through a standardized semantic model to systematically represent these goals and dimensions.
[0042] Specifically, an ontology library based on OWL (Web Ontology Language) is created using an ontology modeling tool (Protégé) to define goal types, key performance indicators and constraint conditions.
[0043] Specifically, for the goal types defined in the ontology library, the evolution operation goals of the device cluster are defined as efficiency goals, cost goals and reliability goals, and each goal type includes its subcategories, such as production efficiency and energy efficiency for efficiency goals.
[0044] Specifically, for the key performance indicators defined in the ontology library, each defined goal type is assigned a quantitative indicator, such as "unit product energy consumption (kWh / piece)" for energy efficiency and "mean time between failures (MTBF)" for reliability.
[0045] Specifically, for the constraint conditions defined in the ontology library, they include physical constraints (maximum load of devices), time constraints (target achievement deadline) and resource constraints (computing resource limitations).
[0046] In addition, the time attribute of the ontology library can be defined, i.e., adding a time dimension to the goal type, such as short-term (1 week), medium-term (1 month) and long-term (1 year), or supporting dynamic adjustment. This time dimension can use a structured input interface to receive evolution operation goals input by upper-level business systems or operators (for example, in the next month, reduce unit product energy consumption by 5% while ensuring that the production line A uptime is not less than 95%).
[0047] Further, the evolution operation goals are mapped to goal parameters including key performance indicators, target values, goal types and constraint conditions through semantic parsing.
[0048] Specifically, semantic analysis is performed by a Chinese word segmentation tool (such as Jieba) to segment the input, extract keywords (in the above example, the keywords are “activation rate”, “energy consumption”, and “reduce 5%”), and then use a pre-trained language model (such as BERT) for intent recognition and entity extraction to map the input to the target type, key performance indicators, and constraints in the ontology library, for example, to generate the target parameter: {target type: cost target, key performance indicator: unit product energy consumption, target value: reduce 5%, constraint condition: activation rate ≥ 95%, time attribute: medium term} through input semantic analysis;
[0049] In addition, a reasoning engine (such as HermiT) can be used to perform semantic matching on the target parameters generated by the above semantic analysis to verify the logical consistency of the input semantic analysis and the generated target parameters. If the input semantic analysis is incomplete (such as missing keywords), the user is prompted to supplement the information;
[0050] Furthermore, based on the target parameters, a dynamic utility function is constructed, which is the expected degree of achievement of the simulation and deduction result under the current evolutionary operation target and the cost-benefit of the strategy configuration required to perform the simulation and deduction;
[0051] It needs to be explained that the dynamic utility function is mainly used to measure whether the simulation and deduction result meets the operation target (such as reducing energy consumption), while considering the resource consumption (such as computation time, hardware occupation) of the simulation process. The function principle is to compare the simulation result with the evolutionary operation target, calculate the achievable degree of the evolutionary operation target, and by removing the simulation cost and possible constraint violation penalty, a comprehensive score is obtained, based on which the optimal simulation strategy can be selected;
[0052] Specifically, the utility score of the optimal simulation strategy is defined as 1;
[0053] Further, the simulation result is compared with the target value to evaluate the completion of each key performance indicator. For each indicator, the closeness of the simulation result to the target value is calculated, and then the achievable degree is converted to a value between 0 and 1 to ensure comparability between different key performance indicators (such as unit product energy consumption and mean time between failures);
[0054] Specifically, the conversion of the achievable degree is performed by a normalization method, which is not described here;
[0055] Specifically, the system monitoring tool (such as the top command of Linux) is used to record the CPU and time, and the simulation cost is calculated, for example, the CPU usage is 80% (normalized to 0.8), and the running time is 10 seconds (normalized to 0.1), and the cost formula can be expressed as: cost = 0.5 * CPU usage + 0.01 * running time = 0.5 * 0.8 + 0.01 * 10 = 0.4 + 0.1 = 0.5, wherein 0.5 is the CPU usage weight, and 0.01 represents the running time weight;
[0056] Specifically, the simulation cost formula can be expressed as:
[0057] cost = ω1 * CPU_u + ω2 * R_t
[0058] Wherein, cost represents the cost, CPU_u represents the CPU usage, R_t represents the running time, and ω1 and ω2 are the CPU usage weight and the running time weight, respectively;
[0059] Specifically, the simulation result can be obtained through the constraint condition, and if the simulation result is in breach, that is, the constraint condition is not established, the penalty points are deducted;
[0060] Specifically, the setting of the penalty points is set through the breach amount, that is, the threshold value of the constraint condition (such as the starting rate < 95% under the condition of starting rate ≥ 95%) and the penalty weight, wherein the penalty weight is used to control the punishment degree, and for the constraint condition with an upward trend in the number of breaches, the penalty weight is increased, and for the constraint condition with a downward trend in the number of breaches, the penalty weight is reduced;
[0061] Specifically, the penalty point formula can be expressed as:
[0062] P = D * ω3
[0063] Wherein, P represents the penalty points, D represents the breach amount, and ω3 represents the penalty weight;
[0064] Further, when the value of the comprehensive score is close to the utility score of the simulation strategy, it indicates that the contribution of the simulation strategy is high, and the evolution operation target is easy to achieve and the cost is low;
[0065] S2, according to the dynamic utility function, a machine learning model is used to select a simulation strategy configuration for the current evolution operation target from a preset simulation strategy space, and the simulation strategy configuration is applied to perform simulation deduction on the digital twin;
[0066] Specifically, the machine learning model is a meta-reinforcement learning model;
[0067] Further, define that each task corresponds to a specific evolution operation target and a dynamic utility function associated therewith, and store the evolution operation target and the dynamic utility function in the ontology library to support quick retrieval;
[0068] Specifically, the simulation strategy configuration parameters include model simulation fidelity levels: high fidelity (physical model), medium fidelity (data driven), and low fidelity (simplified model), time step, solver, and resource allocation priority, and the defined simulation strategy configuration parameters are generated in different combinations to generate a simulation strategy space;
[0069] Further, the meta-strategy is trained by the MAML algorithm to quickly adapt to new tasks, and the meta-strategy is trained using historical evolution operation targets and digital twin interaction data, and the training process includes inner loop (task adaptation) and outer loop (meta-optimization);
[0070] Specifically, the MAML algorithm trains the meta-strategy as follows:
[0071] Randomly select a batch of tasks from the historical evolution operation target distribution or procedurally generated tasks;
[0072] In the inner loop, the following operations are sequentially performed:
[0073] For each task, initialize the strategy parameters of the task, i.e., copy the meta-strategy parameters;
[0074] In the second layer of the inner loop, set the task strategy and the digital twin interaction;
[0075] Select a combination of simulation strategy space;
[0076] Run the simulation through the digital twin, input the digital twin state and the simulation strategy configuration parameters, and output the key performance indicators;
[0077] Calculate the dynamic utility function as the reward;
[0078] Collect the simulation strategy configuration parameters and the dynamic utility function, calculate the loss value of the data simulation strategy configuration parameters and the dynamic utility function, and update the learning rate based on the loss value;
[0079] Specifically, in the outer loop, evaluate the strategy performance of all tasks, calculate the meta-loss value, and update the meta-strategy parameters based on the meta-loss value;
[0080] Specifically, repeat the inner loop and the outer loop until the meta-strategy parameters converge;
[0081] Further, in order to evaluate the prediction uncertainty of the digital twin and the current evolution operation target output and the potential impact of the prediction on the dynamic utility function, several configuration parameters are randomly or uniformly selected from the simulation strategy space, the simulation is run on the digital twin, the simulation strategy configuration parameters and the dynamic utility function are collected to form a dataset in the aforementioned manner, a collection function is selected for evaluating the value of each simulation strategy configuration parameter to determine the next simulation point;
[0082] Specifically, the collection function can usually be selected as UCB, EI, PI or ES;
[0083] Specifically, the selected collection function is trained by the dataset to output the mean and variance of the collection function;
[0084] Specifically, for the simulation strategy space, all simulation strategy configuration parameters under the collection function are calculated, and the highest parameter of the simulation strategy configuration parameter under the collection function is selected as the input item of the next simulation strategy configuration parameter to replace the original simulation strategy configuration parameter, and the rest remains unchanged, and the dataset is updated. If there are more than two simulation strategy configuration parameters with the same parameters, adjust the model simulation fidelity level, such as from high level to medium level;
[0085] S3, through the simulation deduction result and the evaluation result of the dynamic utility function, a continuous learning mechanism is created to iteratively optimize the machine learning model to enhance the simulation strategy configuration capability of the machine learning model for future evolution targets;
[0086] Further, the simulation strategy configuration knowledge learned under the current evolution operation target is applied to the simulation strategy configuration process of new or changed evolution operation targets through knowledge distillation;
[0087] Specifically, when new target parameters appear, similar historical evolution operation targets need to be retrieved from the knowledge base, and the similarity is calculated based on the following methods: cosine similarity or Euclidean distance of target parameters, similarity of key performance indicators, and Jaccard similarity of key performance indicator set;
[0088] Further, for knowledge distillation, the teacher network is defined as the probability distribution of outputting different simulation strategy configuration parameters, and the student network is defined as the strategy network after the meta-strategy quickly adapts or initializes;
[0089] Specifically, for the teacher network, the probability distribution of outputting different simulation strategy configuration parameters is softened by using the Softmax function;
[0090] Specifically, the difference between the student network output and the teacher network output is calculated by distillation loss, which encourages the student to imitate the probability distribution of the teacher rather than a single selection;
[0091] Specifically, the calculated distillation loss value is simulated on the digital twin, the simulation result of the distillation loss value and the historical simulation result are added to the data set, and the acquisition function is reselected to achieve the purpose of the meta-reinforcement learning model.
[0092] Further, the embodiment also provides a device cluster simulation deduction system based on digital twin driving, comprising:
[0093] A target utility definition module is configured to define a dynamic utility function based on the digital twin of the device cluster, to evaluate the contribution of the simulation strategy and the simulation result to the current evolution operation target;
[0094] A simulation strategy configuration and digital twin simulation deduction module is configured to select a simulation strategy configuration for the current evolution operation target from a preset simulation strategy space according to the dynamic utility function, and apply the simulation strategy configuration to perform simulation deduction on the digital twin;
[0095] A simulation strategy optimization module is configured to create a continuous learning mechanism to iteratively optimize the machine learning model based on the simulation deduction result and the evaluation result of the dynamic utility function, to enhance the simulation strategy configuration capability of the machine learning model for future evolution targets.
[0096] The embodiment also provides a computer device suitable for the device cluster simulation deduction method based on digital twin driving, comprising:
[0097] A memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the device cluster simulation deduction method based on digital twin driving proposed in the above embodiment.
[0098] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.
[0099] The embodiment further provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the device cluster simulation deduction method based on digital twin driving proposed in the above embodiment.
[0100] The storage medium proposed in the embodiment belongs to the same inventive concept as the data storage method proposed in the above embodiment, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.
[0101] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages, such as object-oriented programming languages Java and interpreted scripting language JavaScript.
[0102] The present application is described with reference to flowcharts and / or block diagrams according to the methods, devices (systems), and computer program products of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that carries out the functions specified in one or more flows and / or blocks.
[0103] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that carries out the functions specified in one or more flows and / or blocks.
[0104] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1 The flowchart blocks or blocks in the flowchart can also represent a
[0105] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the preferred embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to cover all such variations and modifications as fall within the scope of the application.
[0106] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.
Claims
1. A method for simulation and deduction of a device cluster based on digital twin driving, characterized in that, The method comprises the following steps: a digital twin based on a device cluster is used to define a dynamic utility function to evaluate the contribution of a simulation strategy and a simulation result to a current evolution operation target; a machine learning model is used to select a simulation strategy configuration for the current evolution operation target from a preset simulation strategy space according to the dynamic utility function, and the simulation strategy configuration is applied to perform simulation deduction on the digital twin; a continuous learning mechanism is created to iteratively optimize the machine learning model based on the simulation deduction result and the evaluation result of the dynamic utility function, so as to enhance the simulation strategy configuration capability of the machine learning model for future evolution targets.
2. The digital-twin-driven device cluster simulation and inference method of claim 1, wherein, The definition of the dynamic utility function comprises the following steps: the evolution operation target is mapped to target parameters including key performance indicators, target values, target types and constraint conditions through semantic analysis; the dynamic utility function is constructed based on the target parameters, and the dynamic utility function is the expected achievement degree of the simulation deduction result and the cost benefit of the strategy configuration required to perform the simulation deduction under the current evolution operation target.
3. The digital-twin-driven device cluster simulation and inference method of claim 1 or 2, wherein, The machine learning model is an adaptive simulation strategy configuration model trained based on meta-reinforcement learning, and the training process comprises the following steps: learning how to select an effective simulation strategy configuration according to different evolution operation targets.
4. The digital-twin-driven device cluster simulation and inference method of claim 1, wherein, The preset simulation strategy space at least contains the optional configurations of one or more of the following: model fidelity levels of different parts in the digital twin, time steps of simulation calculation and priority of resource allocation.
5. The digital-twin-driven device cluster simulation and inference method of claim 1, wherein, During the process of performing simulation deduction on the digital twin, the prediction uncertainty of the digital twin and the current evolution operation target output is evaluated, and the potential impact of the prediction on the dynamic utility function is evaluated, and the simulation parameters are selected or the simulation fidelity level is adjusted.
6. The digital-twin-driven device cluster simulation and inference method of claim 1, wherein, The continuous learning mechanism comprises the following steps: the simulation strategy configuration knowledge learned under the current evolution operation target is applied to the simulation strategy configuration process of new or changed evolution operation targets through knowledge distillation.
7. A digital-twin-driven device cluster simulation deduction system based on any one of claims 1-6, wherein, The method comprises the following steps: a target utility definition module is configured to define a dynamic utility function based on a digital twin of a device cluster to evaluate the contribution of a simulation strategy and a simulation result to a current evolution operation target; a simulation strategy configuration and digital twin simulation deduction module is configured to use a machine learning model to select a simulation strategy configuration for the current evolution operation target from a preset simulation strategy space according to the dynamic utility function, and apply the simulation strategy configuration to perform simulation deduction on the digital twin; a simulation strategy optimization module is configured to create a continuous learning mechanism to iteratively optimize the machine learning model based on the simulation deduction result and the evaluation result of the dynamic utility function, so as to enhance the simulation strategy configuration capability of the machine learning model for future evolution targets.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the method of any one of claims 1-6.
9. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the method of any one of claims 1-6.