A method, device, medium and equipment for setting a relay protection setting value

Abstract

The application discloses a kind of setting methods, devices, media and equipment of relay protection setting value, belong to setting value setting field, and the application is obtained by acquiring the multi-source data of main grid, distribution network and user side, and is preprocessed and topologically dynamic mapping using digital twin modeling model, and constructs the digital twin that reflects the operation state of three-level distribution network.Based on this, combined with the short-circuit current calculation model of multiple types of distributed energy, the total short-circuit current of the region is calculated, and the relay protection setting rule parameters are dynamically adjusted through the reinforcement learning rule optimization model. With the main grid setting value as the constraint, the distribution network and user side setting value are matched, and the preliminary setting value scheme is generated. The scheme is checked using the sensitivity dynamic checking model, and finally a reliable distribution network relay protection setting value is generated through the blockchain collaboration model, ensuring the reliability and credibility of the setting value scheme. The application effectively solves the problem that the existing technology cannot accurately and efficiently set the relay protection setting value of the distribution network.

Description

Technical Field

[0001] This invention relates to the field of setting values, and in particular to a method, apparatus, medium, and equipment for setting relay protection values. Background Technology

[0002] As the "last mile" connecting users in the power system, the distribution network's relay protection settings are precisely managed, which directly determines the efficiency of "selective disconnection and rapid isolation" during faults, and thus affects the reliability of power supply.

[0003] Relay protection setting management needs to cover three main levels: the main grid, distribution network, and user side. The core objective is to achieve "uncompromising coordination between upper and lower levels of settings and minimizing the scope of power outages due to faults." In the early stages, this field was mainly based on manual operation, which resulted in problems such as low coordination efficiency, poor coordination accuracy, and non-standard processes. Subsequently, it gradually entered the digital stage, and technical solutions such as three-level interconnected setting control systems and setting methods for single DER access emerged, realizing the leap from "paper-based" to "electronic" setting management. However, facing new scenarios of high-density access of multiple types of DERs and dynamic reconfiguration of distribution network topology, there are still technical bottlenecks that are difficult to overcome.

[0004] Multi-type DER access transforms traditional radial passive distribution networks into active networks, bringing several new challenges to setpoint management: First, short-circuit current characteristics are complex and variable. Wind power output fluctuates significantly due to natural conditions, and the current direction reverses very quickly when the energy storage system switches charging and discharging modes. Small hydropower output varies significantly due to seasonal factors, making it difficult to predict the magnitude, direction, and duration of short-circuit current. Second, protection coordination boundaries are dynamically blurred. The hierarchical protection relationship between the user-side DER grid-connected switch and the distribution network sectional switch is reconstructed with the DER's operating status, rendering the traditional fixed-level coordination logic ineffective. Third, the need for dynamic setting adjustment is urgent. In distribution network areas with high DER penetration, the frequency of setting adjustments due to DER output fluctuations and load changes is high every day, while the traditional periodic setting mode has significant lag and cannot meet real-time protection requirements. While existing related technical solutions have achieved digital upgrades, none have broken through the limitations of "static rules + single DER adaptation." The core defects can be summarized into five major bottlenecks: First, the static nature of topology and parameter updates leads to a high error rate in manual input, which can easily cause mismatch between settings and the actual power grid. Second, the adaptability to multiple types of DERs is insufficient, covering only a single photovoltaic scenario, and the probability of protection coordination failure is high when multiple DERs coexist. Third, the setting rules are fixed, and the fixed time difference and current coordination coefficient cannot balance the speed and sensitivity of protection. Fourth, there is no digital twin support, making it impossible to dynamically simulate the fault process and future scenarios, and faults can only be passively corrected. Fifth, cross-entity collaboration is loose, with data from the three levels of entities not synchronized. The approval process only records operational information and cannot trace real-time data for setting calculations, making it difficult to locate the responsible party in subsequent fault analysis. These shortcomings prevent existing technologies from accurately and efficiently setting the relay protection settings of the distribution network. Summary of the Invention

[0005] This invention provides a method, apparatus, medium, and equipment for setting relay protection settings, in order to solve the problem that existing technologies cannot accurately and efficiently set the relay protection settings of distribution networks.

[0006] Firstly, this application provides a method for setting relay protection settings, including: Acquire multi-source data from the three-level distribution network; the multi-source data includes electrical operation status data from the main grid side, topology and equipment operating status data from the distribution network side, and operation parameters and access characteristics data of various types of distributed energy sources from the user side. Based on a pre-defined digital twin modeling model, a digital twin reflecting the operation status of a three-level distribution network is constructed; wherein, the digital twin modeling model is trained according to the distribution network topology association rules and multi-source heterogeneous data fusion algorithm; Based on a preset multi-type distributed energy short-circuit current calculation model, combined with the real-time operating parameters in the digital twin, the total regional short-circuit current of each distributed energy source is calculated. The relay protection setting rule parameters are optimized by optimizing the model through preset reinforcement learning rules, based on the real-time state of the digital twin and the total short-circuit current of the region. Using the approved settings of the main grid as constraints, and combining the optimized relay protection setting rules parameters and the total short-circuit current of the region, the relay protection settings of the distribution network side and the user side are coordinated and matched to generate a preliminary setting scheme. Based on the preset sensitivity dynamic verification model, the preliminary setting scheme is verified to obtain the target setting scheme; The target setting scheme is input into a preset blockchain collaborative model to generate distribution network relay protection settings.

[0007] This application acquires multi-source data from a three-tiered distribution network, including the main grid's electrical operating status, distribution network topology and equipment conditions, and operating parameters of various types of distributed energy sources on the user side. It utilizes a digital twin model for data preprocessing and dynamic topology mapping to construct a digital twin that accurately reflects the operating status of the three-tiered distribution network. Based on this, combined with a short-circuit current calculation model for various types of distributed energy sources, it accurately calculates the total short-circuit current in the region. Then, it dynamically adjusts the relay protection setting rules parameters through reinforcement learning rule optimization. Using the approved settings from the main grid as constraints, it achieves coordinated matching of settings on the distribution network side and the user side, and ensures the reliability of the setting scheme through a sensitivity dynamic verification model. Finally, it generates reliable distribution network relay protection settings using a blockchain collaborative model. This technical solution not only improves the adaptability and accuracy of relay protection settings but also enhances the system's dynamic response capability and credibility through digital twin and blockchain technologies, significantly improving the operating efficiency and reliability of the distribution network in scenarios with multiple types of distributed energy access. This application effectively solves the problem that existing technologies cannot accurately and efficiently set the relay protection settings of the distribution network.

[0008] Furthermore, the acquisition of multi-source data from the three-tier distribution network specifically includes: The electrical operating status data of the three-phase current, line voltage, main grid equivalent impedance, and circuit breaker opening and closing status of the outgoing switches on the main grid side are collected through the preset synchronous phasor measurement unit and preset monitoring and data acquisition terminal. Through preset distribution terminal units and preset feeder terminal units, real-time current, active power, switch opening and closing status, and topology and equipment operating condition data of the distribution network side segment and branch switches are collected. The distributed energy local controller collects the rated capacity, real-time output, energy storage charging and discharging status, and hydropower unit speed operation parameters and access characteristics data of photovoltaic, wind power, energy storage, and hydropower on the user side.

[0009] This application collects multi-source data from a three-tiered distribution network, including electrical operating status data from the main grid (such as three-phase current, line voltage, main grid equivalent impedance, and circuit breaker opening / closing status), topology and equipment operating status data from the distribution network (such as real-time line current, active power, switch status, and line impedance per unit length), and operating parameters and access characteristics data of various types of distributed energy sources on the user side (such as rated capacity, real-time output, charging / discharging status, and speed of photovoltaic, wind power, energy storage, and hydropower). This application provides comprehensive, accurate, and real-time basic data support for subsequent digital twin modeling and relay protection setting. The precise acquisition and integration of this data enables the system to dynamically perceive the real-time operating status of the distribution network, thereby achieving accurate setting and dynamic optimization of relay protection settings in complex scenarios with high penetration rates of various types of distributed energy sources. This effectively improves the fault response capability, power supply reliability, and operational economy of the distribution network, while providing strong support for the intelligent operation and maintenance management of the distribution network.

[0010] Furthermore, the construction of a digital twin reflecting the operational status of the three-level distribution network based on a preset digital twin modeling model specifically involves: The preprocessing module based on the digital twin modeling model removes outliers from multi-source data according to the preset three-standard-deviation principle, and then converts data of different formats into a preset standard format and completes field standardization to obtain standardized data. The topology dynamic mapping module based on the digital twin modeling model constructs an adjacency matrix with the unique identifier of the device as the matrix node and the line connection relationship as the matrix element. It also compares the switch opening and closing status on the distribution network side in real time to update the adjacency matrix and generate a dynamic distribution network topology map. By associating and binding standardized data with corresponding device nodes in the dynamic distribution network topology diagram, a digital twin reflecting the real-time operating status of the three-level distribution network is constructed.

[0011] This application utilizes a pre-defined digital twin modeling model to preprocess multi-source data and perform dynamic topology mapping, enabling the construction of a digital twin that accurately reflects the operational status of a three-tiered distribution network. Specifically, the preprocessing module uses a three-standard-deviation principle to eliminate outliers and converts data of different formats into a standardized format, ensuring data accuracy and consistency. The dynamic topology mapping module generates a dynamic distribution network topology map by constructing an adjacency matrix and updating the distribution network switch status in real time, achieving real-time perception and dynamic adjustment of the distribution network topology. Furthermore, the standardized data is associated and bound with the device nodes in the dynamic topology map, constructing a digital twin reflecting the real-time operational status of the three-tiered distribution network. This process not only improves the efficiency and accuracy of data processing but also provides a real-time, dynamic, and accurate power grid model for subsequent relay protection setting, allowing protection settings to adaptively adjust according to the actual operating status of the power grid, thereby effectively improving the fault response capability and power supply reliability of the distribution network.

[0012] Furthermore, based on the preset multi-type distributed energy short-circuit current calculation model, and combined with the real-time operating parameters in the digital twin, the total regional short-circuit current of each distributed energy source is calculated, specifically as follows: For photovoltaic equipment, based on the multi-type distributed energy short-circuit current calculation model, combined with the irradiance and DC side voltage data in the digital twin, the photovoltaic short-circuit current of different fault types and transient stages is calculated; For wind power equipment, based on the short-circuit current calculation model of multiple types of distributed energy, combined with wind speed and voltage drop data in the digital twin, the wind power short-circuit current under the corresponding operating conditions is calculated. For energy storage devices, based on multi-type distributed energy short-circuit current calculation models and combined with state-of-charge data in digital twins, the short-circuit current of energy storage in grid-connected and islanded scenarios is calculated. For hydropower equipment, based on the short-circuit current calculation model of multiple types of distributed energy, combined with the speed and excitation current data in the digital twin and the output difference during the wet and dry seasons, the short-circuit current of hydropower in the transient and steady state stages is calculated. Based on the time-series phase superposition mechanism, the photovoltaic short-circuit current, wind power short-circuit current, energy storage short-circuit current and hydropower short-circuit current are superimposed after phase correction and time-series alignment to obtain the total short-circuit current in the region. The multi-type distributed energy short-circuit current calculation model is trained based on the electrical characteristics of photovoltaic equipment, wind power equipment, energy storage equipment, and hydropower equipment.

[0013] This application utilizes a pre-defined multi-type distributed energy short-circuit current calculation model, combined with real-time operating parameters from a digital twin, to accurately calculate the regional total short-circuit current for each distributed energy source. Specifically, for different types of distributed energy sources, specific operating parameters (such as solar irradiance and DC-side voltage for photovoltaics, wind speed and voltage dip for wind power, state of charge for energy storage, rotational speed and excitation current for hydropower, and output differences during wet and dry seasons) are used to calculate the short-circuit current, ensuring the accuracy and adaptability of the calculation results. Furthermore, a time-series phase superposition mechanism is used to perform phase correction and time-series alignment on the various types of short-circuit currents before superposition to obtain the regional total short-circuit current. This process not only improves the accuracy of short-circuit current calculation but also dynamically reflects the short-circuit current characteristics of different distributed energy sources under various operating conditions, providing a reliable basis for the adaptive setting of relay protection settings. This effectively enhances the protection selectivity and reliability of the distribution network under complex operating conditions, reducing the risk of false tripping and failure to trip.

[0014] Furthermore, the optimization model using preset reinforcement learning rules, based on the real-time state of the digital twin and the total short-circuit current in the region, optimizes the relay protection setting rule parameters, specifically as follows: The power grid operation parameters, protection performance parameters, and safety constraint parameters in the digital twin are obtained, a high-dimensional state vector is constructed, and input into the reinforcement learning rule optimization model. The reinforcement learning rule optimization model outputs the initial optimization parameters of the setting rules, which include time step difference and current coordination coefficient, based on the reward function with relay protection speed, selectivity, and sensitivity as the core. The distribution network operation status is monitored in real time. When a topology change, excessive fluctuation of distributed energy output, or substandard protection performance is detected, the initial optimization parameters are optimized based on the reinforcement learning rule optimization model to obtain the optimized relay protection setting rule parameters. The reinforcement learning rule optimization model is trained based on the real-time operating status parameters of the power grid and the relay protection setting rules.

[0015] This application utilizes a pre-defined reinforcement learning rule optimization model to dynamically optimize relay protection setting rule parameters based on the real-time state of a digital twin and the total short-circuit current in the region. Specifically, it first acquires the grid operation parameters, protection performance parameters, and safety constraint parameters from the digital twin, constructs a high-dimensional state vector, and inputs it into the reinforcement learning model. The model focuses on the speed, selectivity, and sensitivity of relay protection as its core objectives, and outputs initial optimized parameters for the setting rules, including time step differences and current coordination coefficients, through a designed reward function. Furthermore, by monitoring the distribution network operation status in real time, when topology changes, excessive fluctuations in distributed energy output, or substandard protection performance are detected, the model can quickly respond and further optimize the initial optimized parameters, ultimately obtaining relay protection setting rule parameters adapted to the current grid operation status. This process not only improves the adaptability and dynamic response capability of relay protection settings but also ensures that the protection system maintains a highly efficient and reliable operating state under complex and ever-changing distribution network operating conditions, significantly improving the power supply reliability and operational safety of the distribution network.

[0016] Furthermore, the relay protection settings on the distribution network side and the user side are coordinated and matched, constrained by the approved settings of the main grid, and combined with the optimized relay protection setting rules parameters and the total short-circuit current of the region, to generate a preliminary setting scheme, specifically as follows: The upper limit of the setting value of the distribution network side switch is determined by the upper limit constraint of the main grid setting value, according to the current coordination ratio and time difference requirements in the optimized relay protection setting rule parameters. Using the distribution network side switch setting as an intermediate constraint, and combining it with the optimized relay protection setting rule parameters, the upper limit of the user-side distributed energy grid-connected switch setting is determined. Based on the total short-circuit current of the region and the preset margin coefficient, combined with the upper limit of the setting value of the distribution network side switch and the upper limit of the setting value of the distributed energy grid-connected switch on the user side, the setting value calculations of the distribution network side and the user side are completed respectively, and a preliminary setting value scheme is generated.

[0017] This application utilizes the approved settings from the main grid as constraints, combined with optimized relay protection setting rules and regional total short-circuit current for coordinated matching, to generate preliminary setting schemes adaptable to various distributed energy access scenarios. Specifically, firstly, the approved settings from the main grid are used as the upper-level constraint. Based on the optimized current coordination ratio and time difference requirements, the upper limit of the setting for the distribution network-side switches is determined to ensure no contradictions between upper and lower level protection. Next, the distribution network-side switch settings are used as intermediate constraints to further determine the upper limit of the setting for the distributed energy grid-connected switches on the user side, achieving coordination between distribution network and user-side protection. Finally, based on the regional total short-circuit current and preset margin coefficients, the setting calculations for the distribution network and user sides are completed, generating a preliminary setting scheme. This process not only ensures the upper-lower-level coordination of protection settings at the main grid, distribution network, and user side, but also improves the adaptability and reliability of the setting scheme through optimized setting rules and real-time short-circuit current data. This effectively reduces the risk of protection maloperation or failure to operate due to distributed energy access, significantly improving the overall power supply reliability of the distribution network.

[0018] Furthermore, the preliminary setting scheme is verified based on a preset sensitivity dynamic verification model to obtain the target setting scheme, specifically as follows: Based on the sensitivity dynamic verification model, the fault current under the minimum operating mode of the digital twin simulation is extracted, and the sensitivity coefficient of each setting in the preliminary setting scheme is calculated; the sensitivity coefficient is the ratio of the minimum fault current to the corresponding protection setting; wherein, the blockchain collaborative model is constructed based on the preset four-level subject trusted interaction rules; If the sensitivity coefficient is greater than the preset qualified threshold, the preliminary value setting scheme will be used as the target value setting scheme. If the sensitivity coefficient is less than or equal to the preset qualified threshold, the relay protection setting rule parameters are re-optimized based on the reinforcement learning rule optimization model, and then the relay protection settings on the distribution network side and the user side are matched collaboratively until the generated scheme passes the sensitivity verification and the target setting scheme is obtained.

[0019] This application verifies the preliminary setting scheme based on a preset sensitivity dynamic verification model, ensuring that the generated target setting scheme meets the sensitivity requirements of relay protection. Specifically, the sensitivity dynamic verification model extracts the fault current under the minimum operating mode simulated by the digital twin and calculates the sensitivity coefficient of each setting (i.e., the ratio of the minimum fault current to the protection setting). If the sensitivity coefficient is greater than the preset qualified threshold, the preliminary setting scheme is directly determined as the target setting scheme; if the sensitivity coefficient is less than or equal to the qualified threshold, the setting rule parameters are re-optimized through a reinforcement learning rule optimization model, and collaborative matching is performed again until the generated scheme passes the sensitivity verification. This process not only ensures the sensitivity of the protection setting under various operating modes but also effectively avoids the risk of protection failure or maloperation through dynamic optimization and verification mechanisms, significantly improving the reliability and security of the distribution network relay protection system.

[0020] Secondly, this application provides a relay protection setting device. The relay protection setting device includes: The acquisition module is used to acquire multi-source data of the three-level distribution network; the multi-source data includes electrical operation status data of the main grid side, topology and equipment operating status data of the distribution network side, and operation parameters and access characteristics data of various types of distributed energy on the user side. The mapping module is used to construct a digital twin reflecting the operating status of the three-level distribution network based on a preset digital twin modeling model; wherein, the digital twin modeling model is trained according to the distribution network topology association rules and multi-source heterogeneous data fusion algorithm; The calculation module is used to calculate the total regional short-circuit current of each distributed energy source based on a preset multi-type distributed energy short-circuit current calculation model and in combination with the real-time operating parameters in the digital twin. The optimization module is used to optimize the model through preset reinforcement learning rules, and optimize the relay protection setting rule parameters based on the real-time state of the digital twin and the total short-circuit current of the region. The matching module is used to coordinate and match the relay protection settings on the distribution network side and the user side with the main grid approved settings as constraints, combined with the optimized relay protection setting rule parameters and the regional total short-circuit current, to generate a preliminary setting scheme. The verification module is used to verify the preliminary setting scheme based on a preset sensitivity dynamic verification model to obtain the target setting scheme. The setting module is used to input the target setting scheme into a preset blockchain collaborative model to generate distribution network relay protection settings.

[0021] This application's relay protection setting device integrates multiple functional modules, achieving full automation and intelligence from data acquisition to setting generation, significantly improving the efficiency, accuracy, and reliability of distribution network relay protection setting. Specifically, the acquisition module accurately collects multi-source data from the three-level distribution network, providing comprehensive basic information for subsequent processing; the mapping module uses a digital twin modeling model to preprocess the data and dynamically map the topology, constructing a digital twin reflecting the real-time operating status, ensuring data accuracy and timeliness. The calculation module, based on a multi-type distributed energy short-circuit current calculation model and combined with real-time operating parameters in the digital twin, accurately calculates the total short-circuit current in the region, providing a key basis for setting. The optimization module uses a reinforcement learning rule optimization model to dynamically adjust the relay protection setting rule parameters, adapting to the real-time operating status of the power grid and improving the adaptability of the settings. The matching module uses the approved settings in the main grid as constraints to perform coordinated matching of settings on the distribution network side and the user side, generating a preliminary setting scheme to ensure no contradictions in the coordination between upper and lower level protection. The verification module uses a sensitivity dynamic verification model to rigorously verify the preliminary setting scheme, ensuring the reliability and effectiveness of the scheme. Finally, the setting module inputs the target setting scheme into the blockchain collaborative model to generate reliable distribution network relay protection settings. Blockchain technology ensures the credibility and traceability of the setting scheme. This entire system's collaborative operation not only improves the automation level of setting adjustments but also enhances the protection performance of the distribution network under complex operating conditions, effectively reducing fault risks and ensuring the stability and security of power supply.

[0022] Thirdly, this application provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform a relay protection setting method as described above. Its beneficial effects are the same as those of the relay protection setting method provided in the first aspect of this application.

[0023] Fourthly, this application provides a terminal device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement any of the relay protection setting methods described in the first aspect. Attached Figure Description

[0024] Figure 1 : A schematic flowchart of an embodiment of the relay protection setting method provided in this application; Figure 2 : A schematic diagram of an embodiment of multi-source data acquisition provided in this application; Figure 3 : A schematic diagram illustrating an embodiment of data fusion and topology mapping provided in this application; Figure 4 : A schematic diagram of an embodiment of the tuning calculation provided in this application; Figure 5 : A schematic diagram of an embodiment of the blockchain collaborative architecture provided in this application; Figure 6 : A schematic diagram of the structure of one embodiment of the visual interactive interface provided in this application; Figure 7 : A schematic flowchart of one embodiment of the adaptive tuning process provided in this application; Figure 8 : A schematic diagram of an embodiment of the system hardware deployment provided in this application; Figure 9 This is a schematic diagram of an embodiment of the relay protection setting device provided in this application. Detailed Implementation

[0025] 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, and 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.

[0026] Example 1 Please refer to Figure 1 In order to solve the problem that existing technologies cannot accurately and efficiently set the relay protection settings of distribution networks, this invention provides a method for setting relay protection settings, including steps S01-S07.

[0027] S01: Obtain multi-source data of the three-level distribution network; the multi-source data includes electrical operation status data of the main network side, topology and equipment operating condition data of the distribution network side, and operation parameters and access characteristics data of various types of distributed energy on the user side.

[0028] As a preferred embodiment of this invention, the acquisition of multi-source data from the three-level distribution network specifically includes: Please refer to Figure 2 , Figure 2 This diagram illustrates the data acquisition architecture for a three-tiered distribution network, showcasing the deployment of acquisition equipment, communication protocols, and data aggregation links on the main grid side, distribution network side, and user side. In the actual implementation of this solution, to construct a complete data sensing system for the three-tiered distribution network and thus provide reliable data support for the accurate setting of subsequent relay protection parameters, it is necessary to systematically complete the acquisition and preliminary integration of multi-source data from the main grid side, distribution network side, and user side. The acquisition methods and specific content for each side are as follows: For the collection of electrical operation status data on the main grid side, synchronous phasor measurement units and monitoring and data acquisition (SCADA) terminals are specifically deployed at the 10kV / 20kV outgoing switchgear of 110kV or 220kV substations. The synchronous phasor measurement units, with a high-frequency sampling rate of 256 points / cycle, collect key electrical quantity data such as three-phase current, line voltage, and main grid equivalent impedance of the outgoing switches in real time. Their sampling accuracy meets the monitoring requirements for short-circuit current transient processes. The monitoring and data acquisition terminals are responsible for collecting switch operating condition data such as the opening and closing status of circuit breakers. Data collected by both types of equipment is transmitted to the data aggregation node, ensuring both real-time data transmission and compatibility, while also conforming to the standard communication specifications of the power system, providing a foundation for accurate assessment of the electrical status on the main grid side.

[0029] The collection of distribution network topology and equipment operating condition data relies on distribution terminal units (DTUs) deployed at distribution network sectionalizing switches and feeder terminal units (FTUs) deployed at branch switches. DTUs can collect real-time current, active power, and impedance per unit length of the distribution network sectionalizing lines, while simultaneously acquiring the opening and closing status of the sectionalizing switches, providing a core basis for dynamic identification of the distribution network topology. Feeder terminal units focus on collecting corresponding electrical quantities and switch status data for branch lines. Both types of terminal devices use the MQTT lightweight IoT protocol for data transmission, with a collection interval set at 100ms. This balances data real-time performance with equipment power consumption, ensuring that changes in distribution network topology and equipment operating conditions can be detected in a timely manner, providing data support for the construction of dynamic distribution network topology.

[0030] Operating parameters and access characteristics of various types of distributed energy resources on the user side are collected by local controllers of distributed energy resources (DERs) deployed next to each type of distributed energy device. For photovoltaic devices, rated capacity, real-time output, DC-side voltage, and associated irradiance data are collected; for wind power devices, rated capacity, real-time output, wind speed, and turbine type (doubly fed or direct-drive) information are collected; for energy storage devices, rated capacity, real-time charge / discharge power, state of charge (SOC), and charge / discharge operating status are collected; for small hydropower devices, rated capacity, real-time output, turbine speed, excitation current, and related data on output during wet and dry seasons are collected. The above data is transmitted to the data aggregation node at 500ms intervals, which meets the real-time monitoring requirements of distributed energy operation status while avoiding excessive bandwidth consumption. The collected access characteristic data accurately reflects the differentiated operating mechanisms of different types of distributed energy resources, providing crucial information for accurate calculation of subsequent short-circuit currents.

[0031] After independent data collection from each side is completed, all data will be aggregated to a unified data aggregation node, forming a categorized and archived three-level distribution network multi-source dataset. Simultaneously, the system is equipped with a data anomaly tolerance mechanism. When a data collection device experiences offline failure or data loss, it will automatically retrieve historical data from similar loads and distributed energy output conditions within the past hour, using linear interpolation to complete the data. The completion error can be controlled within 5%, ensuring the continuity and accuracy of subsequent digital twin modeling and setting work.

[0032] S02: Based on a preset digital twin modeling model, construct a digital twin that reflects the operating status of the three-level distribution network; wherein, the digital twin modeling model is trained according to the distribution network topology association rules and multi-source heterogeneous data fusion algorithm.

[0033] In a preferred embodiment of this invention, the construction of a digital twin reflecting the operational status of the three-level distribution network based on a preset digital twin modeling model specifically involves: Please refer to Figure 3 , Figure 3 This diagram illustrates the construction process of a digital twin, showcasing the entire processing chain from multi-source data access to the generation of a real-time parameter library. In the specific implementation of this solution, the digital twin modeling model used to construct the three-tiered distribution network digital twin relies on specialized training based on distribution network topology association rules and multi-source heterogeneous data fusion algorithms. During the training phase, the model imports a large amount of historical distribution network topology data, multi-source collected data, and device association logic data. The distribution network topology association rules cover core logic such as the physical connection logic of distribution network equipment, the impact mechanism of switch on / off states on the topology structure, and the hierarchical association relationships of the three-tiered distribution network (main network-distribution network-user side). The multi-source heterogeneous data fusion algorithm includes key modules such as multi-protocol data parsing, cross-format data adaptation, and data quality verification. After multiple rounds of iterative training, the model possesses efficient processing capabilities for multi-source data and accurate dynamic mapping capabilities for the distribution network topology, enabling real-time linkage between the physical power grid state and the virtual twin.

[0034] When using this model to process multi-source data from a three-tiered power distribution network and construct a digital twin, the specific implementation process follows six core steps: data access, data cleaning, data standardization, dynamic topology mapping, parameter association and binding, and parameter library update. Data access: First, the raw data transmitted from the PMU and SCADA terminals on the main network side, the DTU and FTU on the distribution network side, and the DER local controller on the user side are accessed and automatically classified into "unstructured data" (such as SCADA operation logs and equipment alarm texts) and "structured data" (such as numerical electrical quantities such as three-phase current and line voltage), providing a classification basis for subsequent differentiated processing; Data cleaning: The "3σ principle outlier detection algorithm" is used to remove outliers from structured data, accurately filtering out invalid data such as instantaneous peak current and voltage generated by electromagnetic interference in PMU. For unstructured data, redundant information is removed and the format is standardized, and finally clean data without outliers is output. Data standardization: Through the built-in data format converter, clean data with different communication protocols and storage formats are uniformly converted into the JSON standard format, and the fields are standardized. The standardized fields are fixed as "device unique ID, data collection timestamp (accurate to milliseconds), parameter name, parameter value, and data quality identifier (qualified / suspicious)", ensuring that all types of data have a unified parsing benchmark and eliminating the barriers of heterogeneous multi-source data formats; Dynamic topology mapping: The model initiates the dynamic topology mapping module, constructing a topology model based on the "graph theory adjacency matrix algorithm". The unique ID of each device in the distribution network is used as the matrix node, and the physical connection relationship between the devices is used as the matrix element (connected state is recorded as 1, disconnected state is recorded as 0). At the same time, the switch opening and closing status data uploaded by the DTU and FTU on the distribution network side are compared in real time. If a switch action is detected (such as branch switch SW3 changing from open to closed), the values ​​of the corresponding elements of SW3 and downstream nodes in the adjacency matrix are updated (from 0 to 1) within 500ms. A dynamically updated distribution network topology map is generated synchronously. This topology map can accurately reflect the real-time network connection status of the three-level distribution network. Parameter association and binding: The standardized multi-source data is precisely associated and bound with the corresponding device nodes in the dynamic topology diagram. For example, the real-time current data of SW3 is bound to the "current parameter item" of the SW3 node in the topology diagram, and the real-time output data of the user-side photovoltaic is bound to the "output parameter item" of the photovoltaic grid-connected switch node, forming an initial twin with real-time operating parameters. At the same time, a multi-source parameter consistency verification mechanism is started. For multi-source data collected from the same device (such as the current data of a distribution network switch coming from both PMU and DTU), a weighted average algorithm is used to eliminate data deviation (PMU data weight 0.7, DTU data weight 0.3, ensuring error ≤2%). Parameter library update: All data that has been bound and verified is written to the time series database (InfluxDB). The database update frequency is strictly synchronized with the collection frequency of the perception layer (256 points / cycle on the main network, 100ms on the distribution network, and 500ms on the user side). The output is a "real-time parameter library" that can be called in real time, providing a stable data call interface for the subsequent core tuning calculation layer.

[0035] In addition, the system is also equipped with a data anomaly fault tolerance mechanism. When a certain acquisition device goes offline or data is lost, it will automatically retrieve historical data under similar load and DER output conditions within the past hour, and complete the data through linear interpolation. The completion error is controlled within 5%, ensuring that the operation status of the digital twin is uninterrupted and the parameters are not missing. Finally, a digital twin that can accurately reflect the real-time operation status of the three-level distribution network is constructed.

[0036] S03: Based on the preset multi-type distributed energy short-circuit current calculation model, combined with the real-time operating parameters in the digital twin, calculate the regional total short-circuit current of each distributed energy source; wherein, the multi-type distributed energy short-circuit current calculation model is trained on the electrical characteristics of photovoltaic equipment, wind power equipment, energy storage equipment and hydropower equipment.

[0037] As a preferred embodiment of this invention, the calculation of the regional total short-circuit current of each distributed energy source based on a preset multi-type distributed energy source short-circuit current calculation model, combined with the real-time operating parameters in the digital twin, specifically involves: This multi-type distributed energy short-circuit current calculation model, used to calculate the short-circuit current of various distributed energy sources and the total short-circuit current of a region, derives its core capability from specialized training conducted in the early stages targeting the differentiated electrical characteristics of photovoltaic, wind power, energy storage, and hydropower equipment. During the training phase, the model imports electrical mechanism data of various equipment, short-circuit test data under different operating conditions, and historical data of distribution network operation. For photovoltaic equipment, the model focuses on the correlation between inverter current limiting characteristics and illumination and DC voltage; for wind power equipment, it distinguishes the transient response differences between doubly-fed and direct-drive models; for energy storage equipment, it covers the current output characteristics under different charging and discharging states; and for hydropower equipment, it incorporates the influence logic of speed, excitation regulation, and seasonal output fluctuations. After multiple rounds of iterative training, the model possesses the ability to accurately calculate the short-circuit current of different types of distributed energy sources.

[0038] In actual short-circuit current calculations, this model deeply integrates real-time operating parameters from the digital twin, performing differentiated calculations for different types of equipment: 1. Calculation of short-circuit current for photovoltaic equipment: Breaking through the limitations of traditional static calculations based on a fixed 1.5 times rated current, a high-precision calculation model is constructed by introducing dynamic correction coefficients based on both light intensity and DC-side voltage. (1) Light intensity correction coefficient (range 0.8-1.2), generated by real-time light intensity data (unit lux) collected by the digital twin through a linear mapping algorithm (k=1.2 when light intensity ≥80000 lux, k=0.8 when light intensity ≤20000 lux, and the intermediate range is calculated accurately by linear interpolation). (2) DC side voltage correction factor (value range 0.95-1.05), calculated based on the ratio of the real-time DC side voltage (unit V) of the inverter to the rated DC voltage (when the voltage deviation is ≤ ±5%, the actual ratio is used; if it exceeds the threshold, the boundary value is used, which meets the inverter voltage tolerance standard). (3) (in The rated capacity of the photovoltaic system is kVA. (Rated voltage of distribution network line in kV).

[0039] At the same time, differentiated adaptations are made for different fault types: Single-phase ground fault: Combining the inverter's zero-sequence current limiting strategy with the contribution of the zero-sequence current, the calculation formula is as follows: (Calibrated using over 100 sets of measured data from substations, the error in the proportion of zero-sequence current is ≤ ±2%). Phase-to-phase short-circuit fault: Introduce a negative sequence current coefficient (value 0.15), i.e. This fills the gap in the existing technology that ignores the negative-order components, thus addressing the computational bias caused by the neglect of negative-order components.

[0040] By closely matching the inverter's control response characteristics, the system is divided into three transient stages along the time axis to precisely match the dynamic changes in short-circuit current. (1) 0≤t<0.02s (transient peak stage): (Inverter current limiting is not fully effective, peak factor is 1.8); (2) 0.02s≤t<0.1s (transition stage): The short-circuit current decays linearly to 1.5 times the rated current; (3) t≥0.1s (steady-state stage): maintain 1.5 times the rated current until the fault is cleared.

[0041] 2. Calculation of short-circuit current for wind power equipment: Different electrical characteristics of different models are modeled differently, and wind speed and voltage drop depth correction mechanisms are introduced: (1) Model differentiation modeling design: Doubly fed wind turbine (mainstream models): Considering the influence of rotor-side converter control characteristics and stator transient reactance, the calculation is performed accurately in three stages: in, The wind speed correction factor (range 0.9-1.1) is generated by linear mapping of wind speed data (unit m / s) collected in real time by the digital twin (k=1.1 when wind speed ≥12m / s, k=0.9 when wind speed ≤3m / s, and linear interpolation in the intermediate range). ( (Rated wind power capacity in kVA).

[0042] Direct-drive permanent magnet generator: Because there is no gearbox structure, the transient response process is shorter, which can be simplified to a two-stage model: (2) Fault voltage drop depth adaptation mechanism: Introducing a voltage sag depth correction factor (Value range 0.8-1.0), precisely matching the current limiting response characteristics of the converter: When the voltage at the fault point drops below 30% of the rated voltage, (The converter's current limiting is enhanced, and the short-circuit current is reduced); When the voltage drops to 50%-70% of the rated voltage, (The converter is under normal current limiting); 3. Short-circuit current calculation for energy storage devices: It distinguishes between charge and discharge operating states and dynamically adjusts based on SOC and energy storage type, while also adapting to islanded operation scenarios. (1) Accurate differentiation of charge and discharge states (including dynamic SOC correction): Charging state (energy storage system in charging state, equivalent to load characteristics): Short-circuit current is mainly provided by the main grid, and a SOC correction factor is introduced to adjust the current distribution ratio: in, The SOC correction factor (k=0.95 when SOC>80%, k=1.0 when SOC40%-80%, k=1.05 when SOC<40%) is set based on the physical characteristic of reduced energy storage equivalent impedance during low SOC charging. The short-circuit current provided by the main grid (obtained according to the traditional distribution network short-circuit calculation method).

[0043] Discharge state (energy storage system in discharge state, equivalent to power supply characteristics): Considering the combined effects of energy storage type and SOC, the model is as follows: : Discharge state SOC correction factor (k=1.3 when SOC>80%, k=1.2 when SOC40%-80%, k=1.0 when SOC<40%), calibrated based on 1000+ sets of energy storage short-circuit test data under different SOC conditions; Energy storage type coefficient (lithium iron phosphate energy storage k=1.0, ternary lithium energy storage k=0.95, vanadium redox flow battery k=1.05), matching the short-circuit current output characteristics of different battery types; ( (Rated capacity of the energy storage system in kVA).

[0044] (2) Adaptation solution for isolated operation scenarios: When the energy storage system operates in islanded mode, disconnected from the main grid, the short-circuit current is supplied independently by the energy storage system, and a frequency correction factor is introduced. Optimize calculations: in The frequency correction coefficient (k=1.0 when the system frequency is 50±0.5Hz, and the k value decreases by 0.1 for every ±0.5Hz frequency deviation, down to a minimum of 0.8) is used to match the frequency stability control characteristics of energy storage during islanded operation.

[0045] 4. Short-circuit current calculation for small hydropower equipment: Based on the transient characteristic modeling of synchronous generators, this paper introduces corrections for speed, excitation current, and output during wet and dry seasons: breaking through the traditional static calculation method of "fixing 6 times / 2 times the rated current", a precise transient model is constructed. The speed correction coefficient (k=1.0 at rated speed of 3000 r / min, k=1.05 when speed deviation is ±5%, and k=1.1 when speed deviation is ±10%) is calculated from the turbine speed data collected in real time by the digital twin. : Excitation current correction coefficient (k=1.0 at rated excitation current, k=1.1 when excitation current is increased by 20%, and k=0.9 when excitation current is decreased by 20%), the effect of matching excitation system adjustment on short-circuit current; ( (Rated capacity of small hydropower stations in kVA). The transient time function (f(t) = 6.0 when 0 ≤ t < 0.1s, f(t) linearly decays to 2.0 when 0.1s ≤ t < 0.3s, and f(t) = 2.0 when t ≥ 0.3s) closely matches the transient response law of synchronous generators; Simultaneously, an output correction coefficient is introduced. Matching seasonal output differences in small hydropower: During the high-water season: when the output is 120% of the rated value, k=1.05; During the normal water level period: when the output is 80%-100% of the rated value, k=1.0; During the dry season: when the output is 60% of the rated value, k=0.95; After calculating the short-circuit current of each type of distributed energy unit, the model will activate a time-series phase superposition mechanism to integrate the short-circuit current of each device to obtain the total short-circuit current of the region, accurately eliminating the error of traditional algebraic superposition. The specific steps are as follows: 1. Phase Correction: Real-time data acquisition via a synchronous phasor measurement unit (PMU) to calculate the phase difference between the short-circuit current of each DER and the short-circuit current of the main grid. (When the phase difference is ≤15°, algebraic superposition is used; when it is >15°, vector superposition is used, which conforms to circuit laws.) 2. Timing Alignment: Based on the transient curves of short-circuit currents of each DER acquired by the digital twin, the peak occurrence time of the short-circuit current of each type of DER is extracted using a peak detection algorithm (e.g., the peak value of wind power is at t=0.05s, and the peak value of energy storage is at t=0.08s), and the peak time offset of each DER is recorded. ; Final superposition formula: In the formula The short-circuit current of the i-th type DER after timing alignment; This is the phase correction coefficient, ensuring that the superposition result satisfies Kirchhoff's laws.

[0046] S04: Optimize the relay protection setting rule parameters based on the real-time state of the digital twin and the total short-circuit current of the region by using a preset reinforcement learning rule optimization model; wherein, the reinforcement learning rule optimization model is trained based on the real-time operating state parameters of the power grid and the relay protection setting rules.

[0047] In a preferred embodiment of this invention, the optimization of relay protection setting rule parameters based on the real-time state of the digital twin and the total short-circuit current in the region using a preset reinforcement learning rule optimization model specifically involves: The reinforcement learning rule optimization model used to optimize relay protection setting rules parameters has its core capabilities derived from specialized training conducted in the early stages to address the real-time operating characteristics of the power grid and the professional requirements of relay protection. During the training phase, the model imports massive amounts of real-time power grid operating status parameters (covering the equivalent impedance of the main grid, distribution network line load, real-time output of various distributed energy sources, short-circuit current transient characteristics, etc.), as well as relay protection setting rules that conform to industry standards, and integrates a large number of historical setting cases and fault condition data. The model adopts the PER-Q-Learning algorithm architecture and integrates transfer learning capabilities. First, it is pre-trained in the source domain using 100,000+ sets of multi-type distributed energy operation scenarios and full-type fault simulation data generated by digital twins. This fully covers rare operating conditions such as photovoltaic output fluctuations, energy storage charging and discharging switching, output differences of small hydropower during wet and dry seasons, and wind power gust disturbances. Then, it imports 30,000+ sets of on-site measured values ​​for adjustment to carry out local fine-tuning in the target domain. This reduces the number of model convergence iterations to one-third of that of traditional algorithms. Furthermore, it achieves lightweighting through channel pruning and 8-bit integer quantization, allowing it to be directly deployed on edge terminals such as distribution network DTUs / FTUs. The inference latency is controlled within 80ms, fully adapting to the real-time on-site tuning requirements.

[0048] The reason for choosing the PER-Q-Learning algorithm instead of the traditional Actor-Critic algorithm lies in its superior adaptability to discrete control parameters in distribution network setting: the core parameters of distribution network setting (time difference Δt, current coordination coefficient k) are all discrete control quantities, and priority should be given to learning "high-value optimization experience" (such as the case of adjusting the sensitivity coefficient Ksen from 1.1 to 1.5). PER-Q-Learning (Priority Experience Replay Q-Learning) assigns higher sampling weights to "adjustment experience that significantly improves protection performance" (such as shortening fault clearing time by 0.1s and reducing the power outage range by 10%) through an "experience priority ranking mechanism," thereby improving the accuracy of discrete action fitting from 82% of the Actor-Critic algorithm to 97%, with a setting adjustment error ≤3%. At the same time, the algorithm's native parameter size is only 1.2M (about 1 / 8 of similar comparative models), possessing a naturally lightweight foundation and adapting to the computing power of distribution network edge devices without excessive pruning.

[0049] To address the pain point of difficulty in obtaining data on distribution network fault scenarios (such as short circuits during energy storage discharge and switching faults between wet and dry seasons in small hydropower stations) and the resulting limited sample size, which hinders model convergence, the model integrates transfer learning technology to achieve efficient training: First, source domain pre-training generates over 100,000 sets of virtual training data (such as ±50% fluctuation in photovoltaic output, single-phase grounding / phase-to-phase short circuit faults, and switching of energy storage charging and discharging states) based on the distribution network digital twin, pre-training the basic model to comprehensively cover rare scenarios; Second, target domain fine-tuning imports over 30,000 sets of on-site measured values ​​for adjustment, and through "feature alignment + local parameter fine-tuning," reduces the number of model convergence iterations from 100,000 to 30,000, improving training efficiency by 230%, completely solving the industry pain point of "insufficient on-site data leading to model training failure."

[0050] To adapt to the computing power constraints of distribution network field terminals (generally equipped with ARM Cortex-A53 processors, 1GHz main frequency, 1GB memory), the model has also undergone edge deployment-specific optimization: through the dual optimization methods of "channel pruning (removing 40% of redundant convolution channels) + 8-bit integer quantization", the model size has been compressed from 1.2M to 2.8MB, and the inference latency has been further reduced from 200ms to 80ms. Field tests have verified that the optimized model can be directly deployed on DTU / FTU terminals without relying on cloud computing power, fully meeting the engineering requirements of "real-time tuning and millisecond-level response" (response latency ≤100ms).

[0051] When optimizing relay protection setting rules parameters in practice, the model will deeply integrate the real-time status of the digital twin and the total short-circuit current of the region, and complete the parameter iterative optimization according to the following process: First, extract the grid operation parameters (equivalent impedance of the main grid, total load of the distribution network, real-time output and operation status of photovoltaic / wind power / energy storage / small hydropower), protection performance parameters (sensitivity coefficient of the current setting, fault clearing time, power outage range), safety constraint parameters (upper limit of equipment withstand current, voltage drop depth, topology change marker) and the total short-circuit current of the region from the digital twin, integrate them to construct a 22-dimensional high-dimensional state vector and input it into the model.

[0052] This state vector is a professional reconstruction of the MDP model, completely different from the state space of traditional general reinforcement learning models. The module adds core parameters of "protection performance class" and "safety constraint class" to ensure that the model can accurately perceive the professional constraints of fixed-value tuning: Power Grid Operation (6 Dimensions): Main Grid Equivalent Impedance (Ω), Total load of distribution network (kW), Real-time output of each DER ( Photovoltaics / Wind power / Energy storage / Small hydropower); Protection performance category (5 dimensions): Sensitivity coefficient (Requires ≥1.2, core constraint indicator), current time difference (s) Current matching coefficient Fault clearance time (s) Power outage area (kW); Safety Constraints (4-dimensional): Maximum Withstand Current of Equipment (A) Contribution of DER short-circuit current (A) Topology change marker (0 = no change / 1 = change) Voltage drop depth (%) Meanwhile, the model reconstructs a discrete action space of 6 types to adapt to the three-level setpoint coordination, which is different from the limitation of traditional models that only support 2 types of general actions. Each type of action corresponds to the specific pain points of distribution network protection scenarios, as shown in Table 1 below: Table 1 Action Space Table Subsequently, the model performs inference calculations based on a quantitative reward function that prioritizes the three protection characteristics. This function focuses 90% of its weight on the core performance of relay protection, with only 10% considering economic efficiency. The specific formula is as follows: ; Normalized fault clearing time (actual time / maximum allowable time, e.g., 0.4s / 0.5s=0.8), with a weight of 0.4 to prioritize "quickness" and shorten the fault impact time; : Normalized value of power outage range (actual power outage load / total distribution network load), weighted at 0.3 to ensure "selectivity" and minimize the impact range of power outage; : Excess excitation term for sensitivity coefficient. When the sensitivity coefficient... satisfy Only at this time will this portion of the reward be included in the total reward, with a weighting of 0.2. This design prioritizes the system's "sensitivity" and effectively mitigates the risk of protection refusal to activate.

[0053] : Network loss normalization value, weighted at 0.1 to balance economic efficiency and avoid excessive sacrifice of energy efficiency; Through the professional adaptation of the above MDP model, the model output includes the initial optimization parameters of the tuning rules, including time difference and current coordination coefficient. For example, the traditional fixed time difference of 0.3s is optimized to 0.2s to improve the fault clearing speed, and the user-side switch current coordination coefficient is optimized from 0.8 to 0.75 to adapt to the scenario of superimposed short-circuit current of multi-source distributed energy.

[0054] Meanwhile, the model monitors the distribution network's operating status in real time and sets five types of triggering conditions strongly correlated with the setpoint adjustment. If any one of these conditions is met, a partial retraining process is initiated (avoiding unnecessary full retraining and improving response efficiency). 1. Triggered by topology change: Addition of a new branch line to the distribution network, grid connection of a 10MW or larger DER, or decommissioning of an existing DER; 2. Triggered by inadequate protection performance: Fault clearing time exceeds 0.5 seconds or power outage area exceeds 10%; 3. DER fluctuation exceeding limit trigger: The total output of DER fluctuates by more than 30% within 30 minutes (such as sudden changes in output caused by gusts of wind or cloud cover). 4. Triggering of operation scenario switching: Small hydropower seasonal transition (output difference > 3 times), seasonal load switching (load difference > 50%). 5. Regular verification trigger: Full parameter fine-tuning is performed once a week to ensure the long-term adaptability of the rules (safety net mechanism); Within 3 minutes of triggering, local parameter updates can be completed without full retraining, outputting optimized setting rules parameters adapted to the current real-time state of the power grid. Furthermore, the system constructs a digital twin simulation verification closed loop, inputting the model's output setting rules (Δt, k, etc.) into the distribution network digital twin every 5 minutes for protection performance simulation verification. Acceptance standard: Sensitivity coefficient It must be no less than Fault response time No more than Percentage of areas affected by power outage per second Not higher than At the same time, the equipment tolerance constraints must be met; Non-compliance handling: Immediately trigger "mini-batch incremental training" (batch size = 64, single training time ≤ 100ms) until all performance indicators meet the standards; S05: Using the approved settings of the main grid as constraints, and combining the optimized relay protection setting rules parameters and the total short-circuit current of the region, the relay protection settings of the distribution network side and the user side are coordinated and matched to generate a preliminary setting scheme.

[0055] In a preferred embodiment of this invention, the relay protection settings on the distribution network side and the user side are coordinated and matched, constrained by the approved settings of the main grid, and combined with the optimized relay protection setting rules parameters and the total short-circuit current of the region, to generate a preliminary setting scheme. Specifically: In the specific implementation process of this scheme, the approved settings of the main grid are used as the core constraint. Combined with the setting rule parameters output by the reinforcement learning rule optimization model and the regional total short-circuit current superimposed by multiple types of distributed energy, a three-level collaborative matching of the relay protection settings on the distribution network side and the user side is carried out to generate a preliminary setting scheme. The specific implementation process is as follows: First, the system will retrieve the approved settings of the 10kV / 20kV outgoing line switches on the main grid side (including instantaneous overcurrent settings, overcurrent protection current settings, and corresponding time differences, etc.) from the parameter library of the digital twin. These settings are fixed constraints that have been pre-approved and are not subject to adjustment. They are only used as a reference for the coordination of settings on the lower-level distribution network side. Based on the current coordination ratio and time difference requirements specified in the optimized relay protection setting rules, the upper limit of the setting values ​​for distribution network side sectionalizing switches and branch switches is determined: taking the instantaneous overcurrent setting value of the main grid outgoing switch QF1 as an example, the instantaneous overcurrent setting value of the distribution network side first-end sectionalizing switch SW1 must meet the selectivity constraint of being less than 0.8 times the instantaneous overcurrent setting value, and at the same time, the overcurrent protection time difference between the two must be maintained at 0.3s to avoid the problem of cascading tripping where the main grid switch operates before the distribution network switch during a fault; for distribution network side middle and end branch switches (such as SW3), the setting value of the upstream sectionalizing switch is used as the benchmark, and the same current coordination ratio and time difference requirements are used to constrain downwards layer by layer, forming a hierarchical constraint chain of distribution network side setting values, ensuring that the protection actions of each switch on the distribution network side and the main grid switch have clear selectivity.

[0056] Secondly, based on the determined upper limit of the setting value on the distribution network side, the setting constraints of the grid-connected switches for multiple types of distributed energy on the user side are further derived. Taking the overcurrent setting value of the distribution network branch switch SW3 as an intermediate constraint, the overcurrent setting values ​​of the grid-connected switches of photovoltaic inverters PV-SW and energy storage PCS grid-connected switches ESS-SW on the user side must be less than 0.8 times the overcurrent setting value, and the overcurrent protection time difference must not be less than 0.3s. For scenarios where multiple types of distributed energy coexist on the user side, including photovoltaic, wind power, energy storage, and small hydropower, it is necessary to combine the previously calculated total short-circuit current of the region and reserve a safety margin factor of 1.2 times to ensure that the instantaneous overcurrent setting value of the grid-connected switch on the user side meets the requirement of being greater than or equal to 1.2 times the total short-circuit current of the region. This reliably avoids the maximum short-circuit current generated by the superposition of multiple distributed energy sources, prevents the grid-connected switch from malfunctioning when a short-circuit fault occurs on the distributed energy side, and ensures the effectiveness of the protection coordination with the distribution network side switches.

[0057] The system establishes a setting constraint chain of "main network → distribution network → user side" to ensure selective coordination between upper and lower level protection: 1. Main network constrains distribution network: Reads the approved settings (such as instantaneous overcurrent) of the main network outgoing switch (QF1) from the digital twin. The instantaneous overcurrent of distribution network sectionalizing switches (such as SW1) must meet the "selectivity constraint": Time difference (To avoid cascading tripping); 2. Distribution network constraint on the user side: based on the overcurrent of the distribution network branch switch (such as SW3). To constrain this, the overcurrent of the user-side DER grid-connected switch (such as PV-SW) must meet the following requirements: Time difference ; 3. Multi-DER Cooperative Adaptation: If multiple types of DERs exist on the user side (e.g., photovoltaic + energy storage), the short-circuit currents of each DER are superimposed, and the calculation formula is as follows: This ensures that the set value can avoid the superimposed short-circuit current from multiple sources (margin factor 1.2).

[0058] Finally, the system integrates the main grid constraints, optimized setting rules parameters, and regional total short-circuit current data to complete the specific setting calculations for each switch (section and branch) on the distribution network side and each distributed energy grid-connected switch on the user side. The setting values ​​of the distribution network side switches need to take into account the hierarchical coordination requirements with the main grid switches and the fault current characteristics of their own lines. For example, the overcurrent time of the distribution network section switches needs to be selected within the setting rule parameter range of 0.1-0.5s, taking into account the line length and fault clearing requirements. The setting values ​​of the user side switches need to simultaneously meet the coordination requirements with the distribution network switches and the short-circuit current withstand characteristics of the distributed energy. For example, the energy storage grid-connected switches need to distinguish between charging and discharging states, and appropriately increase the current setting value in the discharging state to adapt to its power supply characteristics. After the calculation is completed, the system will automatically verify whether the current coordination ratio, time difference and safety margin of all settings meet the preset requirements. The qualified settings parameters will be summarized and organized to generate a preliminary setting scheme that includes the setting type, specific value and upper and lower level coordination basis of each switch. This scheme can fully present the setting coordination logic of the main grid-distribution network-user side, and provide detailed basic data support for subsequent dynamic sensitivity verification.

[0059] S06: Based on the preset sensitivity dynamic verification model, the preliminary setting scheme is verified to obtain the target setting scheme.

[0060] In a preferred embodiment of this invention, the preliminary setting scheme is verified based on a preset sensitivity dynamic verification model to obtain the target setting scheme, specifically as follows: In the specific implementation of this scheme, the sensitivity dynamic verification model used to verify the preliminary setting scheme has its core capability derived from the special training in the early stage that combines industry relay protection technical specifications and massive distribution network fault simulation data. During the training phase, fault current data under different operating modes, sensitivity verification cases of various protection settings and equipment tolerance thresholds will be imported to ensure that the model can accurately adapt to the complex scenario of multiple types of distributed energy access in the three-level distribution network, and provide a comprehensive and professional validity verification for the preliminary setting scheme.

[0061] In actual sensitivity verification, the model deeply integrates with the real-time operating data of the digital twin, and completes the verification and outputs the target setting scheme according to the following complete process: First, the model retrieves the fault current data under the current minimum operating mode of the distribution network from the digital twin. The minimum operating mode refers to the combined operating condition of the minimum output of the main grid and the minimum output of the DER (such as 20% of the rated value of photovoltaic output and 10% of the rated value of wind power output). This data is generated by the digital twin based on the simulation of real-time topology, equipment parameters and distributed energy operation status, and can accurately reflect the current level under the most unfavorable fault condition of the distribution network. Subsequently, the model calculates the sensitivity coefficient for each setting in the preliminary setting scheme (including the setting value of the distribution network-side sectional / branch switch and the setting value of the user-side distributed energy grid-connected switch) according to the sensitivity coefficient calculation formula (sensitivity coefficient = minimum fault current / corresponding protection setting value), ensuring that all protection settings that need to be verified are covered.

[0062] Subsequently, the model underwent precise verification based on the sensitivity coefficient calculation formula, which is as follows: Fault current in the minimum operating mode for digital twin simulation, in amperes (A); The protection setting is calculated in amperes (A). The verification standards are divided into three levels: pass: (Sensitivity is sufficient to meet protection requirements); Warning: (Sensitivity threshold, prompting maintenance personnel to pay attention); Warning: (Insufficient sensitivity, triggering recalculation of setpoints); If an alarm occurs (such as SW3 reverse quick-break), The system automatically invokes the reinforcement learning module, switches "reverse instantaneous overcurrent protection" to "reverse overcurrent protection," and re-optimizes the time difference (e.g., adjusts it to...). (This continues) until the sensitivity test passes.

[0063] This standard strictly adheres to industry technical requirements: if the sensitivity coefficient is greater than 1.5, the setpoint sensitivity is deemed sufficient, and the corresponding preliminary setpoint scheme is directly incorporated into the target setpoint scheme; if the sensitivity coefficient is between 1.2 and 1.5, it is marked as an early warning state, the setpoint scheme is retained, and an early warning prompt is simultaneously pushed to the operation and maintenance system to remind attention to subsequent changes in the power grid operation status; if the sensitivity coefficient is less than 1.2, an alarm mechanism is triggered. At this time, the model will automatically link with the reinforcement learning rule optimization model to iteratively optimize the relay protection setting rule parameters. For protection types corresponding to non-compliant setpoints (such as reverse instantaneous overcurrent protection), the protection type switching logic (such as switching to reverse overcurrent protection) will also be triggered simultaneously. Based on the optimized setting rule parameters, the relay protection setpoints on the distribution network side and the user side will be re-matched collaboratively to generate a new setpoint scheme, which will then be sent back to the sensitivity dynamic verification model for verification. This process is repeated iteratively until the sensitivity coefficients of all items in the generated setpoint scheme meet the verification requirements, ultimately forming a target setpoint scheme that can be directly put into application.

[0064] Furthermore, the model will record the entire verification process and results, including sensitivity coefficient values ​​for each round, rule optimization adjustments, and protection type switching records, forming a complete verification report. This provides detailed data support for subsequent setting maintenance and fault tracing, ensuring the security and traceability of the setting scheme. Please refer to... Figure 4 , Figure 4 This is a schematic diagram of the module collaboration architecture of the fixed value calculation engine. It shows the interaction logic between the multi-type DER short-circuit calculation module, the reinforcement learning rule optimization module, the three-level fixed value collaborative matching module, the sensitivity dynamic verification module and the fixed value calculation engine, as well as the link of the final output result.

[0065] S07: Input the target setting scheme into a preset blockchain collaborative model to generate distribution network relay protection settings; wherein, the blockchain collaborative model is constructed based on preset four-level subject trusted interaction rules.

[0066] In a preferred embodiment of this invention, the step of inputting the target setting scheme into a preset blockchain collaborative model to generate distribution network relay protection settings specifically involves: Please refer to Figure 5 , Figure 5This diagram illustrates the architecture and process of a blockchain collaborative model, showcasing the hierarchical relationship between the consortium blockchain node layer, smart contract layer, and data storage layer, as well as the collaborative processes and trusted traceability mechanisms for each node. The four levels of entities are the local adjustment point, distribution adjustment point, operation and maintenance node, and user node. Each entity completes the entire process of value setting scheme collaboration and storage according to its preset responsibilities. Firstly, this blockchain collaborative model adopts a three-layer architecture design, divided from top to bottom into the consortium blockchain node layer, smart contract layer, and data storage layer. The functions and deployment details of each layer are as follows: The consortium blockchain node layer corresponds to the physical deployment carrier of four levels of entities. It assigns the core responsibilities of four types of functional nodes to the four levels of entities: ground dispatch, allocation and dispatch, operation and maintenance, and users. Each node uses a server as its hardware carrier, and the node type and core responsibilities are clearly defined. Local adjustment point: As a consensus / ledger node, it is responsible for the final review of the mainnet's value; Distribution adjustment point: As a consensus / ledger node, it is responsible for initiating the formulation of the distribution network side setting scheme; Operation and maintenance node: As a consensus / ledger node, it is responsible for the compliance and technical verification of the value setting scheme; User node: As a consensus / ledger node, it is responsible for receiving and confirming the DER grid connection switch settings on the user side; The nodes are built on the Hyperledger Fabric open-source framework (adapted to industrial-grade consortium blockchain scenarios) and use the "PBFT (Practical Byzantine Fault Tolerance) consensus algorithm" to achieve data consensus. The number of consensus nodes is ≥4, the fault tolerance rate reaches 1 / 3, which can effectively avoid single point of failure, and the consensus time can be controlled within 200ms, meeting the engineering requirements of real-time collaboration of three-level entities.

[0067] Smart Contract Layer: Includes two core types of smart contracts to automate process solidification and data storage. Process Contract: Presets an irreversible approval process of "setup → verification → review → release", specifically "allocation preparation → operation and maintenance verification → ground dispatch review → user receipt". If any node is not processed within the time limit (such as ground dispatch review exceeding 2 hours), the contract will automatically push alarm information to the responsible person's mobile APP via MQTT protocol. Evidence storage contract: It has the ability to automatically generate SHA-256 hash values ​​for data. It can write the hash values ​​of key data such as fixed value calculation sheets (including formulas and parameter basis), approval opinions (with electronic signatures), and digital twin parameter snapshots (including collection timestamps accurate to milliseconds and corresponding parameter values) into the block. The original data is stored in a local secure database, forming an evidence storage system of "on-chain hash verifiable and off-chain data searchable", ensuring that the data cannot be tampered with.

[0068] Data storage layer: Constructing three types of data storage repositories to form a storage system of "on-chain hash verification and off-chain data traceability": Setting value calculation basis library: stores raw data such as short-circuit current reports, sensitivity verification results, and twin parameter snapshots during the setting value calculation process; Approval Operation Record Database: Retains process data such as operator ID, operation time, and approval comments at each stage; Hash Chain: The SHA-256 hash values ​​of key data from each stage are linked together in chronological order according to timestamps to achieve immutability and traceability of data throughout the entire process.

[0069] Meanwhile, the system is configured with a trusted traceability mechanism. Maintenance personnel can initiate a traceability request by using "device ID + effective time range of the set value". The system automatically retrieves the hash values ​​of data at each stage from the blockchain and compares them with the hash values ​​of the original data in the local database. The entire traceability process takes ≤1 second. After the comparison is successful, the system synchronously displays the identity of the operator, the millisecond-level timestamp, and the corresponding data content of the set value from its compilation to its release, so as to achieve accurate traceability of responsibility.

[0070] Once the target value scheme is input into the blockchain collaborative model, the system initiates full-chain evidence storage and approval according to the preset collaborative process: The adjustment point initiates the compilation of the set value: the target set value scheme and the corresponding calculation basis are uploaded synchronously, the notarization contract automatically generates the data hash value and writes it into the block, and records the operator ID and millisecond-level timestamp at the same time; The operation and maintenance node performs fixed value verification: it retrieves the calculation basis of the evidence stored on the chain to complete the verification, and stores the verification result (pass / fail and reason) on the chain to ensure that the verification process is traceable; Local regulation points complete setting value review: Based on the constraints of the setting values ​​already approved by the main network, the setting value schemes of the distribution network and user side are finally reviewed, and the review opinions are uploaded to the blockchain in real time, forming a complete approval chain; User node receives and confirms: The user-side DER operation and maintenance entity receives the setting scheme, confirms that it is correct, and uploads the confirmation result to the blockchain to complete the entire approval process.

[0071] Once the four levels of stakeholders have completed the entire approval process and reached a consensus, the blockchain collaborative model will integrate the approved target setting scheme, the calculation basis for on-chain evidence storage, and the approval opinions to generate a distribution network relay protection setting instruction. This instruction can be directly issued to the field protection devices (such as the protection terminal of the distribution network sectionalizing switch and the user-side distributed energy grid-connected switch), ultimately forming an executable distribution network relay protection setting with both legal and technical validity. At the same time, the hash values ​​of all relevant data will be permanently stored in the blockchain, providing an immutable certificate for subsequent fault tracing and liability determination.

[0072] Furthermore, such as Figure 6 As shown, Figure 6This is a schematic diagram of the "visual interactive interface" of this application. The interface adopts a "three-area layout" (labeled "left: 3D digital twin scene area, middle: fixed value information display area, right: function operation area"): Left Zone (3D Digital Twin Scene Area): This area uses 3D rendering technology to recreate the distribution network scene, labeled with "main grid substation (red building complex), 10kV distribution network lines (green lines indicate energized, red lines indicate de-energized), switching equipment (blue dots, green borders indicate closed, red borders indicate open), and various types of DER equipment (PV panel icon / wind turbine icon / energy storage cabinet icon / small hydropower unit icon)". When the mouse hovers over any device, a "real-time parameter card" pops up (e.g., SW3: current 120A, voltage 10.5kV, status closed, setting effective time 2025-11-07 10:30:00). Central area (fixed value information display area): Information is displayed in three columns: Basic Information Section: Equipment ID (SW3), Line (Yangong A707 Line), Current Transformer Ratio (600 / 5), Protection Type (Overcurrent Protection + Instantaneous Overcurrent Protection); Setting parameters: instantaneous overcurrent (primary value 800A / secondary value 6.67A), overcurrent (primary value 600A / secondary value 5A), overcurrent time (0.45s), reclosing enable / disable (enabled); Verification Result Column: Sensitivity Coefficient (4.31>1.5, verification passed), Calculation Basis, Effective Status (effective); Right section (functional operation area): Features four core function buttons (labeled "Recalculate Set Value", "Approval Process", "Fault Prediction", and "Historical Inquiry"): Click "Fault Prediction" to bring up the "Fault Point Selection Box" (you can select fault points such as "SW3 Downstream Line" and "PV-SW Outlet"). After selection, the twin scene area will dynamically display the fault process (such as t=0s short circuit → t=0.45s SW3 trip → t=0.75s PV-SW trip). At the same time, it will display "Short circuit current 3536.6A, power outage range (PV user side only, load 120kW)".

[0073] Furthermore, such as Figure 7 As shown, Figure 7 The "Overall Flowchart for Adaptive Tuning of Fixed Values" is labeled with 8 steps (numbered 1-8) in logical order. Each step clearly defines the "input data," "core processing action," and "output result." Key nodes are labeled with "judgment logic" (yes / no branch). 1. Data Acquisition and Digital Twin Update: Input "multi-source data from the main network, distribution network, and user side transmitted by the perception layer", process "data cleaning → standardization → dynamic topology mapping → parameter binding", and output "real-time updated digital twin". 2. Multi-DER Short Circuit Calculation: Input "Parameters of each DER in the digital twin (rated capacity, operating status, real-time output)", process "call the dedicated short circuit model of photovoltaic / wind power / energy storage / small hydropower for calculation", and output "Short circuit current value of each DER and total short circuit current I_fault.total"; 3. Reinforcement learning rule optimization: Input "grid operation status in digital twin (main grid impedance, distribution network load, DER output)", process "Q-Learning algorithm iterative optimization of time step difference Δt and current coordination coefficient k", and output "optimized setting rules"; 4. Three-level setting collaborative calculation: Input "approved setting value of main grid outgoing switch, total short-circuit current I_fault.total, optimization rules", process "main-distribution-user three-level setting constraint matching", and output "preliminary setting scheme"; 5. Sensitivity Dynamic Verification: Input "Preliminary setting scheme, minimum fault current of twin simulation", process "Calculate sensitivity coefficient". ",judge" >1.5? Yes: Output "the verified setpoint scheme"; No: Return to step 3 (re-optimize tuning rules); 6. Value setting sheet preparation and on-chain storage: Input "verified value setting scheme", process "generate standard format value setting sheet + calculation sheet", output "hash value of value setting sheet and calculation sheet on-chain (blockchain evidence storage)"; 7. Three-level approval workflow: Input the "set value sheet after being uploaded to the blockchain", process "allocation preparation → operation and maintenance verification → ground dispatch review → user receipt" (data is uploaded to the blockchain at each stage), and determine "approval passed?": Yes: Output "Final Value Sheet"; No: Return to step 4 (adjust the coordination relationship of the three-level setpoints); 8. Issuance and Execution Feedback: Input "Final Setting Sheet", process "Issued to Field Protection Device", output "Device Execution Status (Success / Failure) Feedback to Digital Twin", and the process ends.

[0074] Furthermore, such as Figure 8 As shown, Figure 8 For the "System Hardware Deployment Diagram" in this application, please label the "Deployment Location", "Device Model", and "Core Function" of each hardware component: Data acquisition device: Main network side: PMU (deployed in the 10kV outgoing line cabinet of the substation), SCADA terminal (deployed in the dispatch room), core parameters "PMU sampling accuracy ≤ 0.1%, SCADA data update interval ≤ 1s"; Distribution network side: DTU (deployed in distribution network sectional switch box), FTU (deployed in branch switch box), core parameters "operating temperature -40℃~70℃, supports IP65 protection"; User side: DER local controller (deployed next to the DER inverter), core parameter "supports multi-vendor DER access, data transmission latency ≤100ms"; Core computing device: Digital twin computing server (64GB memory, deployed in the data center), core function "real-time construction of 3D twins, simulation calculation time ≤1s"; Setting calculation server: (deployed in the allocation and coordination room), core function "running reinforcement learning algorithms and short-circuit calculation models, with a single setting calculation time ≤2s"; Blockchain node server: (1 server each deployed in the local dispatch / distribution / operation and maintenance / user data center), core functions "consensus and ledger, block generation interval ≤500ms"; Interactive devices: Operation and maintenance workstation (deployed in the dispatch and maintenance room), core function "runs visual interactive software, supports 4K display"; Visualized large screen: (3×2 splicing, deployed in the distribution and monitoring center), the core function is "to display 3D twin scenes and fixed value information, with a brightness of ≥500cd / ㎡"; Network devices: Industrial switches: (24 Gigabit Ethernet ports, deployed in various server rooms), forwarding latency ≤50μs; Firewall (deployed at the data center entrance), its core function is to "protect power-specific protocols and support intrusion detection and data encryption".

[0075] In summary, this application acquires multi-source data from a three-tiered distribution network, including the main grid's electrical operating status, distribution network topology and equipment conditions, and operating parameters of various types of distributed energy sources on the user side. It utilizes a digital twin model for data preprocessing and dynamic topology mapping to construct a digital twin that accurately reflects the operating status of the three-tiered distribution network. Based on this, combined with a short-circuit current calculation model for various types of distributed energy sources, it accurately calculates the total short-circuit current in the region. Then, it dynamically adjusts the relay protection setting rules parameters through reinforcement learning rule optimization. Using the approved settings from the main grid as constraints, it achieves coordinated matching of settings on the distribution network side and the user side, and ensures the reliability of the setting scheme through a sensitivity dynamic verification model. Finally, it generates reliable distribution network relay protection settings using a blockchain collaborative model. This technical solution not only improves the adaptability and accuracy of relay protection settings but also enhances the system's dynamic response capability and credibility through digital twin and blockchain technologies, significantly improving the operating efficiency and reliability of the distribution network in scenarios with multiple types of distributed energy access. This application effectively solves the problem that existing technologies cannot accurately and efficiently set the relay protection settings of the distribution network.

[0076] Example 2 Please refer to Figure 9 This is a relay protection setting device provided in the embodiments of this application.

[0077] In this embodiment, the relay protection setting device includes an acquisition module 10, a mapping module 20, a calculation module 30, an optimization module 40, a matching module 50, a verification module 60, and a setting module 70.

[0078] The acquisition module 10 is used to acquire multi-source data of the three-level distribution network; the multi-source data includes electrical operation status data of the main grid side, topology and equipment operating status data of the distribution network side, and operation parameters and access characteristics data of various types of distributed energy on the user side. The mapping module 20 is used to construct a digital twin reflecting the operation status of the three-level distribution network based on a preset digital twin modeling model; wherein, the digital twin modeling model is trained according to the distribution network topology association rules and multi-source heterogeneous data fusion algorithm; The calculation module 30 is used to calculate the total regional short-circuit current of each distributed energy source based on a preset multi-type distributed energy short-circuit current calculation model and in combination with the real-time operating parameters in the digital twin. Optimization module 40 is used to optimize the model through preset reinforcement learning rules, and optimize the relay protection setting rule parameters based on the real-time state of the digital twin and the total short-circuit current of the region. Matching module 50 is used to coordinate and match the relay protection settings on the distribution network side and the user side with the approved settings of the main network as constraints, combined with the optimized relay protection setting rule parameters and the total short-circuit current of the region, to generate a preliminary setting scheme. The verification module 60 is used to verify the preliminary setting scheme based on a preset sensitivity dynamic verification model to obtain the target setting scheme. The setting module 70 is used to input the target setting scheme into a preset blockchain collaborative model to generate distribution network relay protection settings.

[0079] For ease of description and brevity, the embodiments of the device of the present invention include all the implementation methods in the above embodiments of the relay protection setting method, which will not be repeated here.

[0080] Example 3: This application provides a computer-readable storage medium, which includes a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to execute the relay protection setting method. The relay protection setting method, if implemented as a software functional unit and used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0081] Example 4 This embodiment provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements any of the relay protection setting methods described in Embodiment 1.

[0082] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for setting relay protection settings, characterized in that, include: Acquire multi-source data from the three-level distribution network; the multi-source data includes electrical operation status data from the main grid side, topology and equipment operating status data from the distribution network side, and operation parameters and access characteristics data of various types of distributed energy sources from the user side. Based on a pre-defined digital twin modeling model, a digital twin reflecting the operation status of a three-level distribution network is constructed; wherein, the digital twin modeling model is trained according to the distribution network topology association rules and a multi-source heterogeneous data fusion algorithm; Based on a preset multi-type distributed energy short-circuit current calculation model, combined with the real-time operating parameters in the digital twin, the total regional short-circuit current of each distributed energy source is calculated. The relay protection setting rule parameters are optimized by optimizing the model through preset reinforcement learning rules, based on the real-time state of the digital twin and the total short-circuit current of the region. Using the approved settings of the main grid as constraints, and combining the optimized relay protection setting rules parameters and the total short-circuit current of the region, the relay protection settings of the distribution network side and the user side are coordinated and matched to generate a preliminary setting scheme. Based on the preset sensitivity dynamic verification model, the preliminary setting scheme is verified to obtain the target setting scheme; The target setting scheme is input into a preset blockchain collaborative model to generate distribution network relay protection settings.

2. The method for setting relay protection settings according to claim 1, characterized in that, The acquisition of multi-source data from the three-level distribution network specifically includes: The electrical operating status data of the three-phase current, line voltage, main grid equivalent impedance, and circuit breaker opening and closing status of the outgoing switches on the main grid side are collected through the preset synchronous phasor measurement unit and the preset monitoring and data acquisition terminal. Through preset distribution terminal units and preset feeder terminal units, real-time current, active power, switch opening and closing status, and topology and equipment operating condition data of the distribution network side segment and branch switches are collected. The distributed energy local controller collects the rated capacity, real-time output, energy storage charging and discharging status, and hydropower unit speed operation parameters and access characteristics data of photovoltaic, wind power, energy storage, and hydropower on the user side.

3. The method for setting relay protection settings according to claim 1, characterized in that, The digital twin modeling model, based on a preset digital twin, constructs a digital twin reflecting the operational status of the three-level distribution network, specifically as follows: The preprocessing module based on the digital twin modeling model removes outliers from multi-source data according to the preset three-standard-deviation principle, and then converts data of different formats into a preset standard format and completes field standardization to obtain standardized data. The topology dynamic mapping module based on the digital twin modeling model constructs an adjacency matrix with the unique identifier of the device as the matrix node and the line connection relationship as the matrix element. It also compares the switch opening and closing status on the distribution network side in real time to update the adjacency matrix and generate a dynamic distribution network topology map. By associating and binding standardized data with corresponding device nodes in the dynamic distribution network topology diagram, a digital twin reflecting the real-time operating status of the three-level distribution network is constructed.

4. The method for setting relay protection settings according to claim 1, characterized in that, The pre-set multi-type distributed energy short-circuit current calculation model, combined with the real-time operating parameters in the digital twin, calculates the regional total short-circuit current of each distributed energy source, specifically as follows: For photovoltaic equipment, based on the multi-type distributed energy short-circuit current calculation model, combined with the irradiance and DC side voltage data in the digital twin, the photovoltaic short-circuit current of different fault types and transient stages is calculated; For wind power equipment, based on the short-circuit current calculation model of multiple types of distributed energy, combined with wind speed and voltage drop data in the digital twin, the wind power short-circuit current under the corresponding operating conditions is calculated. For energy storage devices, based on multi-type distributed energy short-circuit current calculation models and combined with state-of-charge data in digital twins, the short-circuit current of energy storage in grid-connected and islanded scenarios is calculated. For hydropower equipment, based on the short-circuit current calculation model of multiple types of distributed energy, combined with the speed and excitation current data in the digital twin and the output difference during the wet and dry seasons, the short-circuit current of hydropower in the transient and steady state stages is calculated. Based on the time-series phase superposition mechanism, the photovoltaic short-circuit current, wind power short-circuit current, energy storage short-circuit current and hydropower short-circuit current are superimposed after phase correction and time-series alignment to obtain the total short-circuit current in the region. The multi-type distributed energy short-circuit current calculation model is trained based on the electrical characteristics of photovoltaic equipment, wind power equipment, energy storage equipment, and hydropower equipment.

5. The method for setting relay protection settings according to claim 1, characterized in that, The optimization model, based on the real-time state of the digital twin and the total short-circuit current in the region, optimizes the relay protection setting rule parameters through a preset reinforcement learning rule optimization model, specifically as follows: The power grid operation parameters, protection performance parameters, and safety constraint parameters in the digital twin are obtained, a high-dimensional state vector is constructed, and input into the reinforcement learning rule optimization model. The reinforcement learning rule optimization model outputs the initial optimization parameters of the setting rules, which include time step difference and current coordination coefficient, based on the reward function with relay protection speed, selectivity, and sensitivity as the core. The distribution network operation status is monitored in real time. When a topology change, excessive fluctuation of distributed energy output, or substandard protection performance is detected, the initial optimization parameters are optimized based on the reinforcement learning rule optimization model to obtain the optimized relay protection setting rule parameters. The reinforcement learning rule optimization model is trained based on the real-time operating status parameters of the power grid and the relay protection setting rules.

6. The method for setting relay protection settings according to claim 1, characterized in that, The method involves using the approved settings from the main grid as constraints, combined with the optimized relay protection setting rules parameters and the total short-circuit current in the region, to coordinate and match the relay protection settings on the distribution network side and the user side, generating a preliminary setting scheme, specifically as follows: The upper limit of the setting value of the distribution network side switch is determined by the upper limit constraint of the main grid setting value, according to the current coordination ratio and time difference requirements in the optimized relay protection setting rule parameters. Using the distribution network side switch setting as an intermediate constraint, and combining it with the optimized relay protection setting rule parameters, the upper limit of the user-side distributed energy grid-connected switch setting is determined. Based on the total short-circuit current of the region and the preset margin coefficient, combined with the upper limit of the setting value of the distribution network side switch and the upper limit of the setting value of the distributed energy grid-connected switch on the user side, the setting value calculations of the distribution network side and the user side are completed respectively, and a preliminary setting value scheme is generated.

7. The method for setting relay protection settings according to claim 1, characterized in that, The preliminary setting scheme is verified based on the preset sensitivity dynamic verification model to obtain the target setting scheme, specifically as follows: Based on the sensitivity dynamic verification model, the fault current under the minimum operating mode of the digital twin simulation is extracted, and the sensitivity coefficient of each setting in the preliminary setting scheme is calculated; the sensitivity coefficient is the ratio of the minimum fault current to the corresponding protection setting; wherein, the blockchain collaborative model is constructed based on the preset four-level subject trusted interaction rules; If the sensitivity coefficient is greater than the preset qualified threshold, the preliminary value setting scheme will be used as the target value setting scheme. If the sensitivity coefficient is less than or equal to the preset qualified threshold, the relay protection setting rule parameters are re-optimized based on the reinforcement learning rule optimization model, and then the relay protection settings on the distribution network side and the user side are matched collaboratively until the generated scheme passes the sensitivity verification and the target setting scheme is obtained.

8. A relay protection setting device, characterized in that, include: The acquisition module is used to acquire multi-source data of the three-level distribution network; the multi-source data includes electrical operation status data of the main grid side, topology and equipment operating status data of the distribution network side, and operation parameters and access characteristics data of various types of distributed energy on the user side. The mapping module is used to construct a digital twin reflecting the operating status of the three-level distribution network based on a preset digital twin modeling model; wherein, the digital twin modeling model is trained according to the distribution network topology association rules and multi-source heterogeneous data fusion algorithm; The calculation module is used to calculate the total regional short-circuit current of each distributed energy source based on a preset multi-type distributed energy short-circuit current calculation model and in combination with the real-time operating parameters in the digital twin. The optimization module is used to optimize the model through preset reinforcement learning rules, and optimize the relay protection setting rule parameters based on the real-time state of the digital twin and the total short-circuit current of the region. The matching module is used to coordinate and match the relay protection settings on the distribution network side and the user side with the main grid approved settings as constraints, combined with the optimized relay protection setting rule parameters and the regional total short-circuit current, to generate a preliminary setting scheme. The verification module is used to verify the preliminary setting scheme based on a preset sensitivity dynamic verification model to obtain the target setting scheme. The setting module is used to input the target setting scheme into a preset blockchain collaborative model to generate distribution network relay protection settings.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device in which the computer-readable storage medium is located to perform the relay protection setting method as described in any one of claims 1 to 7.

10. A terminal device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the relay protection setting method as described in any one of claims 1 to 7.

Images