Cooperative utilization method for virtual resources of heterogeneous equipment in power generation system network target range
By generating device behavior images and using machine learning to optimize model parameters, the problem of low realism in virtual resource collaboration methods in power generation system network test ranges was solved, realizing autonomous and efficient collaboration between devices and improving the realism of the simulation environment and the effectiveness of the exercise.
Patent Information
- Application Number
- CN202511346252.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-19
- Publication Date
- 2026-01-06
AI Technical Summary
The existing virtual resource collaboration methods in power generation system network test ranges have low realism, poor flexibility, and lack of autonomy, resulting in a large gap between the simulation environment and the real production environment, distorted exercise effects, and the simulation model lacks self-evolution capabilities.
By collecting network communication messages, input/output signal timings, and state transition logic from heterogeneous devices, device behavior description files are generated, device behavior images are created, autonomous interaction and collaborative operation are achieved, and model parameters are optimized through machine learning to improve simulation realism.
Accurately simulate the behavior of heterogeneous devices to achieve autonomous and efficient collaboration between devices, improve the effectiveness and reliability of drills, and reduce subsequent maintenance costs.
Smart Images
Figure CN121284043A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of power system simulation and network security technology, and in particular to a method for the collaborative utilization of virtual resources of heterogeneous devices in a power generation system network range. Background Technology
[0002] With the increasing intelligence and networking of power systems, the cybersecurity threats they face are becoming increasingly severe. In order to effectively address these threats and improve the security protection and emergency response capabilities of power systems, cyber ranges have become an indispensable infrastructure for conducting cybersecurity technology verification, attack and defense drills, and personnel training.
[0003] A power generation system network test range needs to simulate a complex heterogeneous environment encompassing various power generation methods such as thermal, hydro, wind, photovoltaic, and nuclear power, as well as relay protection and dispatch control. Existing test range construction methods typically focus on the virtualization and pooling of basic resources such as computing, storage, and networking, and allocate them through resource scheduling strategies. However, this approach has significant limitations: First, it fails to deeply simulate the unique external behavioral logic, communication protocols, and precise timing characteristics of power industrial control equipment (such as PLCs, RTUs, and protection devices), resulting in a significant gap between the simulation environment and the real production environment, leading to distorted exercise results. Second, virtual resources lack autonomous collaborative capabilities based on power business logic; their interactions often require strong intervention from the central dispatcher, making it difficult to efficiently and realistically reproduce the propagation paths of large-scale cascading failures or complex network attacks. Third, the simulation model lacks self-evolution capabilities; its behavioral accuracy heavily relies on the completeness of the initial modeling and cannot be self-optimized based on exercise results.
[0004] Therefore, there is an urgent need in this field for a virtual resource utilization method that can accurately simulate the behavior of heterogeneous devices and enable autonomous, efficient, and realistic collaboration between devices, so as to improve the effectiveness and reliability of power generation system network test range exercises. Summary of the Invention
[0005] To address the aforementioned technical issues, namely the low realism, poor flexibility, and lack of autonomy of existing virtual resource collaboration methods in power generation system network test ranges.
[0006] This invention provides a method for the collaborative utilization of virtual resources of heterogeneous devices in a power generation system network test range, comprising the following steps:
[0007] S1: Collect network communication messages, input / output signal timing and state transition logic of various heterogeneous physical devices in the test range, parse the collected data, extract feature parameters that characterize the device behavior patterns, and generate device behavior description files;
[0008] S2: Based on the device behavior description file, create a corresponding device behavior image;
[0009] S3: Analyze the exercise script to determine the starting action, target equipment, and process; based on the type of target equipment, call its corresponding equipment behavior description file, generate an initial stimulus signal that conforms to the communication protocol of the equipment, and inject the initial stimulus signal into the corresponding equipment behavior image in the target range virtual network at a preset simulation time point;
[0010] S4: The device behavior mirror processes the initial stimulus signal according to its preset logic strategy and generates an output signal. The output signal is transmitted to the associated device behavior mirror in the virtual network to trigger a chain reaction to realize autonomous interaction and collaborative operation between multiple device behavior mirrors and generate interactive data.
[0011] S5: Collect the interaction data of all device behavior images;
[0012] S6: Compare the collected interaction data with standard data, and update the model parameters of the relevant device behavior image using machine learning algorithms based on the differences.
[0013] In some preferred embodiments, in step S1, the characteristic parameters include: response delay range, message payload structure, heartbeat interval, exception code generation conditions, and device handshake sequence.
[0014] In some preferred embodiments, in step S2, the operating mode of the device behavior mirror is set to be able to switch between high-fidelity mode and fast simulation mode.
[0015] In some preferred embodiments, in step S2, all device behavior mirroring is synchronized based on a global emulation clock.
[0016] In some preferred embodiments, in step S3, the content of the exercise script includes: attack scenarios, failure scenarios, and normal operation scenarios.
[0017] In some preferred embodiments, in step S3, the method further includes:
[0018] After generating the initial stimulus signal, the device behavior image that has been triggered with a set threshold is preloaded according to the process of the exercise script and computing resources are allocated to it.
[0019] In some preferred embodiments, in step S4, the interaction data includes: interaction messages, state change events, and timing information.
[0020] In some preferred embodiments, step S5 further includes:
[0021] The collected interactive data is then visualized and rendered to generate a system situation view.
[0022] In some preferred embodiments, the method further includes:
[0023] S7: After the control signal generated by the device behavior mirror in step S4 is securely isolated and converted into a protocol, it is sent to the real physical device for execution. The response data of the real physical device is collected and fed back to the simulation system as the input signal of the corresponding device behavior mirror.
[0024] In some preferred embodiments, the heterogeneous physical devices include a steam turbine controller, a wind turbine main controller, a photovoltaic inverter, a relay protection device, and a dispatching master station system.
[0025] As can be seen from the above, the method for collaborative utilization of virtual resources of heterogeneous devices in a power generation system network test range provided by the present invention has the following beneficial technical effects:
[0026] This invention extracts behavioral features from real equipment and constructs a device behavior mirror, achieving accurate simulation of the external behavior, communication protocols, and timing characteristics of power industrial control equipment. This greatly enhances the realism of the test range environment. Simultaneously, it abandons centralized forced scheduling and, through event-driven and standard protocols, enables the device behavior mirror to interact autonomously, efficiently and accurately reproducing the business flow and fault propagation chain of the power system. This improves the automation level and scale of the exercise. Furthermore, by collecting interaction data during the exercise and comparing it with standard data, this invention can continuously optimize the model parameters of the device behavior mirror using machine learning algorithms, making its behavior increasingly approximate that of real equipment. This achieves the self-evolution of the simulation model and reduces subsequent maintenance costs. Attached Figure Description
[0027] The features and advantages of the invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the drawings:
[0028] Figure 1 This is a flowchart of the method for collaborative utilization of virtual resources of heterogeneous devices in a power generation system network test range according to the present invention. Detailed Implementation
[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0030] Based on the problems of low realism, poor flexibility, and lack of autonomy in existing virtual resource collaboration methods in power generation system network test ranges as pointed out in the background technology, this invention provides a method for collaborative utilization of virtual resources of heterogeneous devices in power generation system network test ranges. The aim is to provide a virtual resource utilization method that can accurately simulate the behavior of heterogeneous devices and realize autonomous, efficient, and realistic collaboration between devices, so as to improve the effectiveness and reliability of power generation system network test range exercises.
[0031] like Figure 1 As shown, the method for collaborative utilization of virtual resources of heterogeneous devices in a power generation system network test range according to the present invention includes the following steps:
[0032] S1: For various heterogeneous physical devices that need to be simulated in the test range, deploy data acquisition devices to obtain their network communication messages, input and output signal timing and state transition logic during operation. Analyze and process the acquired data to extract feature parameters that can characterize the unique behavior patterns of the devices and generate standardized and structured device behavior description files.
[0033] In step S1 above, preferably, the characteristic parameters include: response delay range, message payload structure, heartbeat interval, exception code generation conditions, and device handshake sequence.
[0034] S2: Based on the device behavior description file, create corresponding device behavior images for various physical devices; the device behavior image is a software-defined simulation model, which has built-in response logic, communication protocol stack and timing control unit constructed according to the feature parameters, in order to accurately simulate the external behavior of real devices; connect all instantiated device behavior images to a unified simulation platform and synchronize them with a global simulation clock.
[0035] In step S2 above, preferably, the operating mode of the device behavior mirror is set to be able to switch between high-fidelity mode and fast simulation mode.
[0036] S3: Receive user-defined exercise scripts, which describe attack, fault, or normal operation scenarios; parse the exercise scripts using a simulation script parsing engine to determine the initial actions, target devices, and processes of the scenarios; based on the type of the target device, the engine calls its corresponding device behavior description file to generate precise control commands or data messages conforming to the device's communication protocol (such as IEC 104, Modbus, etc.) as initial stimulus signals, and injects the initial stimulus signals into the corresponding device behavior image in the target range virtual network at a preset simulation time point.
[0037] More preferably, after generating the initial stimulus signal, the device behavior image that has been triggered with a set threshold is preloaded according to the process of the exercise script and computing resources are allocated to it.
[0038] S4: Upon receiving the initial stimulus signal, the device behavior mirror processes it according to its built-in response logic and generates a new output signal. This output signal is transmitted to other associated device behavior mirrors in the virtual network in accordance with the actual power system business process and communication protocol. After receiving the signal, subsequent device behavior mirrors also respond according to their own models, thereby triggering a series of chain reactions, realizing event-driven, decentralized autonomous interaction and collaborative operation between multiple heterogeneous device behavior mirrors, jointly performing a complete exercise scenario, and generating interactive data.
[0039] In step S4 above, the interaction data includes: interaction messages, state change events, and timing information.
[0040] S5: During the simulation, a data monitor captures and records in real time the interaction messages between all device behavior images, internal state change events and timing information; at the same time, the collected data is mapped onto a predefined topology map, rendered in real time, and a dynamic system situation view is generated for users to observe and analyze.
[0041] S6: After the exercise, the interaction data recorded in step S5 will be compared and analyzed with real historical operating data or expert-recognized standard results. Based on the differences found, machine learning algorithms will be used to automatically adjust and optimize the internal model parameters of the relevant equipment behavior mirror to reduce the deviation between its behavior and the real equipment or expected behavior, thereby continuously improving the realism of the simulation model. The aforementioned real historical operating data or expert-recognized standard results can be used as standard data.
[0042] In some preferred embodiments, the method for collaborative utilization of virtual resources of heterogeneous devices in a power generation system network test range of the present invention further includes:
[0043] S7: The control signals generated by the device behavior mirror in step S4 are sent to the real physical device for execution after being securely isolated and converted through a protocol. The response data of the real physical device is collected and fed back to the simulation system as the input signal for the corresponding device behavior mirror. Through this setting, the key control signals generated by the optimized device behavior mirror in the simulation are sent to the real physical device connected to the test platform for execution after being converted through a protocol by a secure isolation device. At the same time, the response data of the real physical device is collected and fed back to the simulation system as the input for the corresponding device behavior mirror, forming a closed verification loop for calibrating the model or testing the anti-attack capability of the real device.
[0044] Preferably, the heterogeneous physical equipment includes a steam turbine controller, a wind turbine main controller, a photovoltaic inverter, a relay protection device, and a dispatching master station system.
[0045] It should be noted that heterogeneous physical devices can include intelligent electronic devices in various stages of power generation, transmission, transformation, distribution, and consumption. For example, for the digital electro-hydraulic control system (DEH) controller of a steam turbine in a thermal power plant, the data acquisition device needs to capture all control commands, status feedback, analog setpoints, and other messages exchanged between it and the distributed control system (DCS), boiler furnace safety monitoring system (FSSS), and other subsystems. The focus of the analysis is to extract its behavioral characteristics, such as: the rate of power ramp-up (MW / min) after receiving a load increase command, the overshoot after reaching the setpoint, the coordination response delay with the boiler system, and the alarm information generation rules under abnormal operating conditions such as valve jamming. These characteristics are quantified and structured and stored as a device behavior description file.
[0046] Based on the device behavior description files, corresponding device behavior images are instantiated. All device behavior images are synchronized based on the global simulation clock. At the start of the simulation, the platform reads the corresponding device behavior description files according to the requirements of the exercise scenario. For example, when simulating a microgrid containing a photovoltaic power station, the platform loads description files for devices such as photovoltaic inverters, smart combiner boxes, and energy storage converters (PCS). Based on the parameters in the files, corresponding device behavior image instances are dynamically created on the computing nodes. A photovoltaic inverter image is configured to: calculate and output active power according to its specific conversion efficiency curve after receiving simulated light intensity input; its output current harmonic content conforms to the characteristics in the description file; and its communication with the power station monitoring system follows a specific IEC 61850MMS or SunSpec protocol. All these image instances are immediately synchronized with the global simulation clock service after creation to ensure that the entire virtual power system evolves on a unified time scale. This is the foundation for achieving precise cross-device collaborative timing.
[0047] The simulation script is analyzed to determine the initial action, target equipment, and process. Based on the type of target equipment, its equipment behavior description file is invoked to generate an initial stimulus signal conforming to the equipment's communication protocol. This initial stimulus signal is then injected into the corresponding equipment behavior image in the virtual network of the test range at a specified simulation time. The simulation script defines the initial conditions and event sequence of the simulation scenario. For example, a script might describe "simulating a sudden drop in the power grid frequency to 49.2Hz to assess the primary frequency regulation response of a hydropower station's speed control system." The simulation script analysis engine will analyze this script and determine that the initial action is to inject a frequency drop signal into the controller images of all power generation units in the power grid. The engine will query the equipment behavior description files of various controllers generated in S1, such as turbine governors, steam turbine DEH, and photovoltaic inverters, to obtain their respective protocols and addresses for receiving frequency signals. Based on this, the engine will generate a series of messages conforming to DL / T 860-104 or Modbus protocols, containing a 49.2Hz frequency value, as the initial stimulus signal. This signal is then injected into all relevant controller behavior images simultaneously at a millisecond-level time during the simulation, as scheduled by the global clock.
[0048] Subsequently, the system enters the core autonomous deduction phase. The device behavior mirror processes the initial stimulus signal according to its built-in logic and generates an output signal. The output signal is transmitted to the associated device behavior mirror in the virtual network, triggering a chain reaction to realize autonomous interaction and collaborative operation among multiple device behavior mirrors and generate interactive data. After receiving the frequency drop signal, the turbine governor mirror's built-in logic, derived from the parameters in its device behavior description file regarding the dead zone, droop rate, and response speed of primary frequency regulation, is immediately triggered. It calculates that an increase in active power output is needed and then generates a "open guide vane opening" command output signal, which is sent to the turbine simulation model through the virtual network. At the same time, the steam turbine DEH mirror also receives the frequency signal, but its description file may define that it does not participate in primary frequency regulation, so it may not generate power adjustment output. After receiving the frequency signal, the photovoltaic inverter mirror can execute the "active power drop-frequency" characteristic according to the latest grid specifications, that is, automatically reduce power to support frequency recovery when the frequency drops. The entire process is entirely driven by the autonomous responses of each device's behavior mirror based on its own behavioral logic and received input signals, and they interact through a virtual network. This realistically reproduces the distributed collaborative control process based on local rules in a real power system, and generates massive amounts of interactive data, such as all command messages, status changes, and measurement changes.
[0049] To observe and optimize the simulation, the interaction data generated in step S4 is collected. The full-link data acquisition and monitoring module records all interaction data generated in stage S4, including but not limited to: the five-tuple information of each message, payload, precise send / receive timestamp, and the historical trajectory of key state variables inside the behavior image of each device, such as generator power angle, bus voltage, and switch status.
[0050] After the exercise, the interactive data collected in step S5 is compared with standard data. Based on the differences, machine learning algorithms are used to update the model parameters of the relevant equipment behavior mirror. The "standard data" can be historical operating data from the real physical system, the output results of a high-order fine model, or the ideal response determined by domain experts. For example, the action time of a certain type of relay protection device mirror under a simulated short-circuit fault, recorded in S5, is compared with the typical action time range of the real protection device of the same type. If a systematic deviation in the action time of the simulated mirror is found, the model iteration optimization management module can initiate a reinforcement learning process to automatically adjust the delay-related parameters in the mirror model, making its behavior in the next simulation closer to reality. This gives the entire test range a "continuous learning" capability, and the simulation realism can continuously evolve over time.
[0051] The beneficial effects of this invention include:
[0052] This invention extracts behavioral features from real equipment and constructs a device behavior mirror, achieving accurate simulation of the external behavior, communication protocols, and timing characteristics of power industrial control equipment. This greatly enhances the realism of the test range environment. Simultaneously, it abandons centralized forced scheduling and, through event-driven and standard protocols, enables the device behavior mirror to interact autonomously, efficiently and accurately reproducing the business flow and fault propagation chain of the power system. This improves the automation level and scale of the exercise. Furthermore, by collecting interaction data during the exercise and comparing it with standard data, this invention can continuously optimize the model parameters of the device behavior mirror using machine learning algorithms, making its behavior increasingly approximate that of real equipment. This achieves the self-evolution of the simulation model and reduces subsequent maintenance costs.
[0053] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0054] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.
[0055] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present invention as described above, which are not provided in detail for the sake of brevity.
[0056] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description.
[0057] One or more embodiments of the present invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of the present invention should be included within the scope of this disclosure.
Claims
1. A method for the collaborative utilization of virtual resources of heterogeneous devices in a power generation system network range, characterized in that, The method comprises the following steps: S1: Collecting network communication messages, input and output signal timing and state transition logic of various heterogeneous physical devices in the target field during operation, analyzing the collected data, extracting characteristic parameters representing device behavior patterns, and generating device behavior description files; S2: Creating corresponding device behavior images based on the device behavior description files; S3: Analyzing the rehearsal script to determine the starting action, target device and process; according to the type of the target device, calling its corresponding device behavior description file, generating an initial stimulus signal conforming to the communication protocol of the device, and injecting the initial stimulus signal into the corresponding device behavior image in the virtual network at the preset simulation time point; S4: The device behavior image processes the initial stimulus signal according to its preset logic strategy, and generates an output signal, which is transmitted to the associated device behavior image in the virtual network, triggering a chain reaction to achieve autonomous interaction and cooperative operation between multiple device behavior images and generating interaction data; S5: Collecting the interaction data of all device behavior images; S6: Comparing the collected interaction data with standard data, and updating the model parameters of the relevant device behavior images using machine learning algorithms according to the differences.
2. The method of claim 1, wherein, In step S1, the characteristic parameters include: response delay range, message load structure, heartbeat interval, abnormal code generation condition and device handshake sequence.
3. The method of claim 1, wherein, In step S2, the running mode of the device behavior image is set to be able to switch between high-fidelity mode and fast simulation mode.
4. The method of claim 1, wherein, In step S2, all device behavior images are synchronized based on a global simulation clock.
5. The method of claim 1, wherein, In step S3, the content of the rehearsal script includes: attack scene, fault scene and normal operation scene.
6. The method of claim 1, wherein, In step S3, the method further comprises: After generating the initial stimulus signal, preloading the device behavior image whose triggered probability reaches a set threshold according to the process of the rehearsal script and allocating computing resources to it.
7. The method of claim 1, wherein, In step S4, the interaction data includes: interaction message, state change event and timing information.
8. The method of claim 1, wherein, In step S5, the method further comprises: Visualizing the collected interaction data to generate a system situation view.
9. The method of claim 1, wherein, The method further comprises: S7: After the control signal generated by the device behavior image in step S4 is transmitted to the real physical device for execution after security isolation and protocol conversion, the response data of the real physical device is collected and fed back to the simulation system as the input signal of the corresponding device behavior image.
10. The method of claim 1, wherein, The heterogeneous physical devices include steam turbine controllers, fan main controllers, photovoltaic inverters, relay protection devices and dispatching master station systems.