A transformer intelligent protection control method, device, equipment and storage medium

Abstract

The application discloses a transformer intelligent protection control method and device, equipment and storage medium, relates to the field of transformer protection control, and comprises the following steps: acquiring the communication performance index of a heterogeneous physical communication interface; evaluating the communication quality of the communication interface according to the communication performance index, obtaining an evaluation result, combining the evaluation result and the communication performance index, and generating a link quality portrait; identifying and recording the type and attribute of to-be-sent service data to obtain an identification result, and generating a quality of service label according to the identification result; inputting the label and the link quality portrait into a reinforcement learning decision model to select a target action from a preset action space; and executing the target action to protect the data transmission of the service data on a target communication link; the target communication link is a communication link between the transformer and a preset opposite end node. The application can significantly improve the reliability, real-time performance and overall communication of the transformer protection control service data transmission in a complex network environment.

Description

Technical Field

[0001] This application relates to the field of transformer protection and control, and in particular to a method, device, equipment and storage medium for intelligent protection and control of transformers. Background Technology

[0002] Currently, with the development trend of large-scale grid connection of distributed power sources and massive access of flexible loads, the operation mode of distribution networks has undergone fundamental changes. Distribution transformers are no longer limited to the traditional basic functions of stepping up and down the voltage of electricity, and are gradually evolving from traditional power conversion nodes into intelligent nodes that need to support bidirectional power flow, real-time status interaction and collaborative control.

[0003] The protection and control unit of a transformer needs to exchange data frequently and in multiple directions with the upper-level dispatching master station, nearby distributed energy management systems, fault indicators, and other adjacent protection devices. This includes real-time electrical parameters, non-electrical status, protection action signals, and fault recording files. This interaction process relies on the underlying communication network. However, the communication environment of a distribution network typically exhibits significant heterogeneity and dynamism, often employing a hybrid network of technologies such as fiber optics, power line carrier, public wireless networks, and private networks. These networks inherently vary greatly in transmission delay, bandwidth, and reliability. Furthermore, because the equipment comes from different eras and manufacturers, the network may contain multiple communication protocols, including Modbus RTU, IEC 61850 MMSI, and various proprietary TCP (Transmission Control Protocol) protocols. In addition, the network topology may dynamically change due to adjustments in operating methods or fault reconstruction, and external environmental interference can lead to wireless signal attenuation or fluctuations in carrier communication quality. Therefore, ensuring the continuous reliability, efficiency, and real-time performance of the communication links between the protection and control unit and various heterogeneous systems is a technical problem that needs to be solved.

[0004] However, current transformer protection control units suffer from significant shortcomings in communication compatibility and adaptive capabilities, making it difficult to meet the demands of the aforementioned complex and dynamic environments. The core technical problem lies in the rigid communication architecture and lack of intelligent scheduling capabilities, specifically including the following aspects: First, current solutions mostly employ fixed, single communication methods or simple dual-channel static hot standby configurations, failing to dynamically optimize and switch based on real-time network conditions and service requirements. Second, in terms of protocol processing, they typically rely on pre-configured static protocol conversion mapping tables or external protocol conversion gateways. This approach not only fails to adapt to dynamic changes in protocol types of peer devices in the network but also introduces additional conversion delays and single-point-of-failure risks. Third, they lack real-time perception and quantitative evaluation mechanisms for communication link quality and do not possess intelligent traffic scheduling functions based on service quality level requirements. Fourth, protection tripping and fault signals... Critical services with extremely high real-time and reliability requirements, such as signaling, and non-critical services with bandwidth requirements but relatively low real-time requirements, such as periodic status monitoring and file transfer, cannot be differentiated in communication resource competition. This can lead to delays or loss of critical protection command transmission when communication conditions deteriorate, resulting in protection failure and an expansion of the fault range. Fifth, poor status information uploading can distort remote monitoring and make operation and maintenance difficult. In addition, the rigid communication mode also severely restricts the realization of advanced applications such as regional collaboration and intelligent distributed protection among multiple protection devices in the distribution network, which require low latency and high reliability information synchronization. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a transformer intelligent protection and control method, device, equipment, and storage medium, which can accurately and in real-time perceive the status of the communication interface, achieve precise matching of communication needs and network resources, and significantly improve the reliability, real-time performance, and overall communication efficiency of transformer protection and control business data transmission in complex, multi-protocol, and dynamically changing network environments. This solves the problem of unreliable communication in heterogeneous and dynamic distribution network environments, thereby improving the stability and power supply reliability of the distribution network, reducing power outage losses, and ensuring power supply security. The specific solution is as follows: In a first aspect, this application discloses a transformer intelligent protection control method, applied to a protection control unit in a transformer, comprising: According to a preset period, data packets are exchanged with a preset peer node in the target distribution network through a heterogeneous physical communication interface to obtain the communication performance indicators of the heterogeneous physical communication interface. The communication quality of the heterogeneous physical communication interface is evaluated based on the communication performance indicators to obtain the communication quality evaluation result. The communication quality evaluation result and the communication performance indicators are then combined to generate a dynamically updated link quality profile. The system uses a pre-defined business feature knowledge base to parse the business data to be sent, so as to identify and record the type and attributes of the business data to be sent, obtain the business data identification result, and generate a service quality label for the business data to be sent based on the business data identification result. The service quality label and the link quality profile are input into the trained reinforcement learning decision model to select a target action from a preset action space according to a preset decision criterion. The preset decision criterion is to maximize the expected cumulative reward, which is the expected cumulative reward of the reward function dynamically configured according to the service quality label. The target action includes a sending interface, an encapsulation protocol, and a redundant sending strategy. The target action is executed to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

[0006] Optionally, the step of evaluating the communication quality of the heterogeneous physical communication interface based on the communication performance indicators to obtain a communication quality evaluation result includes: Normalize each of the communication performance indicators corresponding to the heterogeneous physical communication interface to obtain multiple normalized performance indicators. The communication quality evaluation result of the heterogeneous physical communication interface is obtained by weighted fusion of multiple normalized performance indicators.

[0007] Optionally, combining the communication quality assessment results and the communication performance indicators to generate a dynamically updated link quality profile includes: Data entries are created based on the communication quality assessment results and the normalized performance indicators to obtain the current data entry; The current data entry replaces the old data entry stored in the previous cycle of the heterogeneous physical communication interface, generating a dynamically updated link quality profile.

[0008] Optionally, generating a quality of service label for the service data to be sent based on the service data identification result includes: Based on the business data identification result, the corresponding service quality constraint value is read from the preset business feature knowledge base, and the service quality constraint value is used as the service quality label of the business data to be sent; the service quality label includes an urgency value, a reliability requirement value, and a tolerance latency value.

[0009] Optionally, before inputting the service quality label and the link quality profile into the trained reinforcement learning decision model, the method further includes: Based on the business data identification results and the service quality tags, generate business data identification statistics; Accordingly, the step of selecting a target action from a preset action space according to preset decision criteria includes: Using a value assessment function and based on the service quality label and the link quality profile, the expected value of the cumulative return obtained after performing each action in the preset action space is evaluated to obtain the expected cumulative return; The maximum value is determined from the multiple expected cumulative returns, and the action corresponding to the maximum value is taken as the target action; Accordingly, after executing the target action, the process further includes: Obtain the execution result of the target action, obtain the execution result information, and generate a decision log based on the execution result information, the service quality label, the link quality profile, and the target action; The link quality profile, the service data identification statistics, and the decision log are uploaded to the remote management platform through the heterogeneous physical communication interface. The system receives predictive policy instructions from the remote management platform and integrates these instructions as prior knowledge into the reinforcement learning decision model to adjust the decision tendency of the model.

[0010] Optionally, the intelligent protection and control method for the transformer further includes: The reward weight coefficients in the reward function of the reinforcement learning decision model are dynamically adjusted based on the service quality labels. The immediate reward value for this decision is calculated based on the execution result information and using the reward function; Accordingly, the step of generating decision logs based on the execution result information, the service quality label, the link quality profile, and the target action includes: Based on the execution result information, the service quality label, the link quality profile, the target action, and the instant reward value, structured data entries are created to obtain the decision log.

[0011] Optionally, the step of exchanging data packets with a preset peer node in the target distribution network according to a preset period and through a heterogeneous physical communication interface to obtain the communication performance indicators of the heterogeneous physical communication interface includes: According to a preset period, and through a heterogeneous physical communication interface, the probe data packets are sent multiple times to a preset peer node in the target distribution network, and the response data packets returned by the preset peer node are received. Physical layer performance metrics are read from the heterogeneous physical communication interface; the physical layer performance metrics include signal strength and bit error rate. Based on the time difference between sending the probe data packet and receiving the response data packet, the multiple one-way delays between the protection control unit and the preset peer node are calculated to obtain the delay sequence. The latency jitter of the heterogeneous physical communication interface is obtained based on the absolute value of the difference between two adjacent one-way delays in the latency sequence. The total number of probe data packets sent and the total number of response data packets received within the preset period are counted to obtain the total number of packets sent and the total number of packets received. The ratio of the total number of packets sent to the total number of packets received is then calculated to obtain the packet loss rate. Generate communication performance metrics that include the physical layer performance metrics and the network layer performance metrics; the network layer performance metrics include the delay sequence, the delay jitter, and the packet loss rate.

[0012] Secondly, this application discloses a transformer intelligent protection and control device, applied to a protection and control unit in a transformer, comprising: An interaction module is used to interact with a preset peer node in the target distribution network through a heterogeneous physical communication interface according to a preset period, so as to obtain the communication performance indicators of the heterogeneous physical communication interface. An evaluation module is used to evaluate the communication quality of the heterogeneous physical communication interface based on the communication performance indicators, and obtain a communication quality evaluation result. The combination module is used to combine the communication quality assessment results and the communication performance indicators to generate a dynamically updated link quality profile. The parsing module is used to parse the business data to be sent using a preset business feature knowledge base, so as to identify and record the type and attributes of the business data to be sent, and obtain the business data identification result. The generation module is used to generate a service quality label for the service data to be sent based on the service data identification result; The input module is used to input the service quality label and the link quality profile into the trained reinforcement learning decision model, so as to select a target action from a preset action space according to a preset decision criterion; the preset decision criterion is to maximize the expected cumulative return, and the expected cumulative return is the expected cumulative return of the reward function dynamically configured according to the service quality label; The sending module is used to execute the target action to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

[0013] Thirdly, this application discloses an electronic device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the aforementioned intelligent protection and control method for transformers.

[0014] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned intelligent protection and control method for transformers.

[0015] As can be seen, the protection control unit applied in this application to a transformer includes: exchanging data packets with a preset peer node in the target distribution network according to a preset period and through a heterogeneous physical communication interface to obtain the communication performance indicators of the heterogeneous physical communication interface; evaluating the communication quality of the heterogeneous physical communication interface based on the communication performance indicators to obtain a communication quality evaluation result, and combining the communication quality evaluation result and the communication performance indicators to generate a dynamically updated link quality profile; parsing the service data to be sent using a preset service feature knowledge base to identify and record the type and attributes of the service data to be sent, obtaining a service data identification result, and based on the... The service data identification result generates a service quality label for the service data to be sent; the service quality label and the link quality profile are input into the trained reinforcement learning decision model to select a target action from a preset action space according to a preset decision criterion; the preset decision criterion is to maximize the expected cumulative return, and the expected cumulative return is the expected cumulative return of the reward function dynamically configured according to the service quality label; the target action includes a sending interface, an encapsulation protocol, and a redundant sending strategy; the target action is executed to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

[0016] This application acquires communication performance indicators of heterogeneous physical communication interfaces according to a preset period, evaluates the communication quality of the interfaces based on these indicators, and then generates a dynamically updated link quality profile based on the evaluation results and communication performance indicators, thereby achieving accurate real-time perception of the communication interface status. Furthermore, this application identifies the service data to be transmitted and generates a service quality label based on the identification results, thus enabling precise matching of communication needs and network resources. In addition, this application uses a reinforcement learning decision model to adaptively determine the optimal action for transmitting service data based on real-time network status and service needs, and executes the optimal action to protect the data transmission of the service data to be transmitted between the transformer and the peer node. This significantly improves the reliability, real-time performance, and overall communication efficiency of transformer protection and control service data transmission in complex, multi-protocol, and dynamically changing network environments. Simultaneously, it solves the problem of unreliable communication in heterogeneous dynamic environments of distribution networks, improves the stability and power supply reliability of the distribution network, reduces power outage losses, and ensures power supply security. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a flowchart of a transformer intelligent protection and control method disclosed in this application; Figure 2 This is a schematic diagram of the structure of a transformer intelligent protection and control device disclosed in this application; Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] This application discloses an intelligent protection and control method for transformers, applied to the protection and control unit in a transformer. (See also...) Figure 1 As shown, the method includes: Step S11: According to a preset period, data packets are exchanged with a preset peer node in the target distribution network through a heterogeneous physical communication interface to obtain the communication performance indicators of the heterogeneous physical communication interface.

[0021] It should be noted that the intelligent protection and control scheme for transformers proposed in this application is specifically applied to the protection and control unit in the transformer, which is used to fundamentally improve the communication reliability, real-time performance and intelligence of the transformer in a dynamic heterogeneous network environment.

[0022] In this embodiment, the protection control unit can, according to a preset cycle, interact with preset peer nodes in the target distribution network (referring to key points in the distribution network that connect multiple circuits or devices, such as transformers, switchgear, and other electrical equipment) through a pre-integrated heterogeneous physical communication interface, and collect communication performance indicators of each interface in the heterogeneous physical communication interface, such as physical layer performance indicators and network layer performance indicators. The number of heterogeneous physical communication interfaces can be one or more (e.g., two), and can be selected according to the actual application scenario.

[0023] In one specific implementation, the step of exchanging data packets with a preset peer node in the target distribution network through a heterogeneous physical communication interface at a preset period to obtain the communication performance indicators of the heterogeneous physical communication interface may specifically include: sending probe data packets multiple times to the preset peer node in the target distribution network through the heterogeneous physical communication interface at a preset period, and receiving response data packets returned by the preset peer node; reading physical layer performance indicators from the heterogeneous physical communication interface; the physical layer performance indicators include signal strength and bit error rate; and calculating the protection based on the time difference between sending the probe data packets and receiving the response data packets. Multiple one-way delays between the control unit and the preset peer node are used to obtain a delay sequence; the delay jitter of the heterogeneous physical communication interface is obtained based on the absolute value of the difference between two adjacent one-way delays in the delay sequence; the number of all probe data packets sent and the number of all response data packets received within the preset period are counted to obtain the total number of sent packets and the total number of received packets, and the ratio of the total number of sent packets to the total number of received packets is calculated to obtain the packet loss rate; a communication performance index including the physical layer performance index and the network layer performance index is generated; the network layer performance index includes the delay sequence, the delay jitter, and the packet loss rate. Specifically, the protection control unit sends probe data packets containing timestamps to preset external peer nodes in the distribution network according to a preset cycle and through two pre-integrated heterogeneous physical communication interfaces, and receives response data packets returned from the peer nodes. Then, it reads physical layer performance indicators, such as signal strength and bit error rate, from the two heterogeneous physical communication interfaces. Simultaneously, it calculates the one-way delay or round-trip delay multiple times based on the time difference between the sent probe data packets and the received response data packets to obtain a delay sequence. For example, the total transmission time is obtained by subtracting the sending timestamp of the probe data packets from the receiving timestamp of the response data packets. If the processing delay value (fixed value) of the peer node is known, the preset processing delay value is subtracted from the total transmission time to obtain the network round-trip delay. Alternatively, in a symmetrical path, half of the total transmission time can be directly used as the estimated one-way delay. It should be noted that the processing delay value can be preset based on the equipment model or empirical value of the peer node, such as 5 milliseconds.

[0024] Furthermore, based on the delay sequence obtained from multiple consecutive calculations, delay jitter and packet loss rate are statistically analyzed. For example, the absolute value of the difference between two consecutive one-way delay values ​​in the delay sequence is calculated, and then the average of these absolute values ​​is calculated to obtain the average delay jitter; alternatively, the maximum value among all absolute values ​​is obtained to obtain the maximum delay jitter. Next, the number of probe packets sent within a preset probe period is counted to obtain the total number of sent packets. Then, the number of successfully received response packets (i.e., the total number of received packets) is subtracted from the total number of sent packets to obtain the number of lost packets. The ratio of the number of lost packets to the total number of sent packets is the packet loss rate. The packet loss rate can be used to evaluate the quality of the current communication link. For example, when the packet loss rate exceeds 20% in 10 consecutive probe packets, the current quality of the link can be determined to be unreliable. Finally, physical layer performance indicators and network layer performance indicators (such as delay sequence, delay jitter, and packet loss rate) are used as communication performance indicators.

[0025] Step S12: Evaluate the communication quality of the heterogeneous physical communication interface according to the communication performance indicators, obtain the communication quality evaluation result, and combine the communication quality evaluation result and the communication performance indicators to generate a dynamically updated link quality profile.

[0026] In this embodiment, after obtaining the communication performance indicators that include physical layer performance indicators and network layer performance indicators, the quality of the communication link between the current protection control unit of the corresponding heterogeneous physical communication interface and the peer node is further scored in real time according to the communication performance indicators to obtain the communication quality assessment result. Then, the above communication quality assessment result and communication performance indicators are combined to generate a dynamically updated link quality profile.

[0027] Specifically, the step of evaluating the communication quality of the heterogeneous physical communication interface based on the communication performance indicators to obtain the communication quality evaluation result may include: normalizing each of the communication performance indicators corresponding to the heterogeneous physical communication interface to obtain multiple normalized performance indicators; and weightedly fusing the multiple normalized performance indicators to obtain the communication quality evaluation result of the heterogeneous physical communication interface. In this embodiment, the physical layer performance indicators and network layer performance indicators of each heterogeneous physical communication interface can be normalized to obtain corresponding normalized values. Specifically, a theoretical optimal value and a theoretical worst value are pre-set for each indicator, which can be determined based on the maximum and minimum values ​​or empirical thresholds in communication standards and historical data. For example, for signal strength, the theoretical optimal value can be set to 10 dB above the receiver sensitivity threshold, and the theoretical worst value can be set to the receiver sensitivity threshold. For latency, the theoretical optimal value can be set to 1 millisecond, and the theoretical worst value can be set to 100 milliseconds. Then, the actual measured values ​​of each indicator are mapped to a numerical range of 0 to 1. The specific mapping rule can be: when the actual measured value is equal to or greater than the theoretical optimal value, the normalization result is one; when the actual measured value is equal to or less than the theoretical worst value, the normalization result is zero; when the actual measured value is between the theoretical optimal value and the theoretical worst value, its normalized value is calculated by linear or nonlinear interpolation.

[0028] Furthermore, based on a preset set of weighting coefficients, the normalized values ​​of each indicator are weighted and summed to calculate the real-time score (i.e., communication quality assessment result) for each heterogeneous physical communication interface. The set of weighting coefficients is pre-configured or dynamically adjusted according to the sensitivity of different service types to communication quality dimensions; for example, for protection tripping services requiring low latency, the weighting coefficients for one-way latency and latency jitter are set higher; while for file transfer services requiring high reliability, the weighting coefficients for packet loss rate and bit error rate are set higher. In the specific calculation process, the normalized value of each indicator is multiplied by its preset weighting coefficient, and then the weighted results of all indicators are summed to obtain the real-time score of the heterogeneous physical communication interface; the sum of the weighting coefficients of all indicators is 1.

[0029] In this embodiment, combining the communication quality assessment results and the communication performance indicators to generate a dynamically updated link quality profile may specifically include: creating data entries based on the communication quality assessment results and the normalized performance indicators to obtain the current data entry; replacing the old data entries stored in the previous cycle of the heterogeneous physical communication interface with the current data entry to generate a dynamically updated link quality profile. In this embodiment, the real-time scores for all heterogeneous physical communication interfaces obtained in the current calculation cycle can be combined with the normalized values ​​of various indicators, and the multi-dimensional data set covering all heterogeneous physical communication interfaces can be updated to form a dynamically updated link quality profile. Specifically, a data structure (such as a list or dictionary) can be created first, and then a data entry can be created for each heterogeneous physical communication interface. This entry should contain at least two parts: the first part is the real-time score calculated for the interface in the current period, and the second part is a subset (such as an array or key-value pairs) used to store the normalized values ​​of the interface's various physical layer performance indicators and network layer performance indicators. Next, the old data entries stored in the previous period are replaced with the new data entries for all heterogeneous physical communication interfaces obtained in the current calculation period, thus completing the update of the link quality profile. It should be noted that this data structure allows subsequent steps to efficiently retrieve and compare the overall quality and details of various dimensions of different interfaces.

[0030] Step S13: Use a preset business feature knowledge base to parse the business data to be sent, so as to identify and record the type and attributes of the business data to be sent, obtain the business data identification result, and generate a service quality label for the business data to be sent based on the business data identification result.

[0031] In this embodiment, the business data to be sent is first parsed using a preset business feature knowledge base, thereby identifying and recording the type and attributes of the business data to be sent, obtaining the business data identification result, and then generating a corresponding service quality label for the business data to be sent based on the business data identification result.

[0032] It should be noted that this application pre-stores a business characteristic knowledge base in the protection control unit. This knowledge base defines a set of identification characteristic rules for each known business data type. These rules include at least one or more combinations of the source module identifier, data load feature code, and data generation event type of the business data. Among them, the business data types include, but are not limited to, protection trip commands, fault recording files, real-time sampling data, periodic equipment status reports, and remote configuration instructions. For example, the identification characteristic rule for protection trip commands is: the source module identifier is "overcurrent protection logic unit", the data load feature code is a specific command word (such as 0xAA55), and the generation event type is "protection action output".

[0033] By using the aforementioned business feature knowledge base to parse the business data to be sent, its source module identifier, data payload content, and generation event type can be extracted. Then, the extracted information is matched one by one with the identification feature rules stored in the business feature knowledge base. For example, the extracted source module identifier, data payload feature code, and generation event type are compared with the corresponding fields specified by a certain identification feature rule recorded in the knowledge base. A successful match is determined when all compared fields conform to the rule's definition. Specifically, for the source module identifier, the fields must be completely equal; for the data payload feature code, it must be consistent with the predefined feature code in the rule or conform to a specific mask; for the generation event type, it must belong to one of the allowed event types enumerated in the rule. When a match is successful, the business data to be sent is determined to belong to the business data type corresponding to the identification feature rule, and the relevant definition information (i.e., attribute information) of that business data type is obtained accordingly, thus obtaining the business data identification result.

[0034] Specifically, generating a service quality label for the service data to be sent based on the service data identification result may include: reading the corresponding service quality constraint value from the preset service feature knowledge base based on the service data identification result, and using the service quality constraint value as the service quality label for the service data to be sent; the service quality label includes an urgency level value, a reliability requirement value, and a tolerance latency value. It should be noted that in the service feature knowledge base, a set of service quality constraint values ​​is pre-set for each type of service data. These constraint values ​​may specifically include the urgency level, the minimum reliability requirement threshold, and the maximum tolerance latency threshold, etc.; for example, for a "protection trip command," its service quality constraint values ​​are: the highest urgency level (e.g., value 9 or 10), the minimum reliability requirement threshold is 99.99% (i.e., the allowed packet loss rate is no higher than 0.01%), and the maximum tolerance latency threshold is 20 milliseconds; for "periodic device status reporting," its service quality constraint values ​​are: low urgency level (e.g., value 1), the minimum reliability requirement threshold is 95%, and the maximum tolerance latency threshold is 5 seconds.

[0035] In this embodiment, after identifying the type of business data to be sent, a set of service quality constraint values ​​corresponding to that type can be read from the business feature knowledge base. This data is then used to generate a data structure containing specific values ​​for three dimensions: urgency, reliability requirements, and tolerable latency. This data structure serves as the service quality label for the business data to be sent. For example, a data structure with three fields can be created. The read "urgency level" value is filled into the "urgency" field, the "minimum reliability requirement threshold" value is filled into the "reliability requirement" field, and the "maximum tolerable latency threshold" value is filled into the "tolerable latency" field, thus completing the label instantiation. The data structure can be in the form of tuples, structs, or JSON (JavaScript Object Notation) objects.

[0036] Step S14: Input the service quality label and the link quality profile into the trained reinforcement learning decision model to select a target action from a preset action space according to a preset decision criterion; the preset decision criterion is to maximize the expected cumulative return, and the expected cumulative return is the expected cumulative return of the reward function dynamically configured according to the service quality label; the target action includes a sending interface, an encapsulation protocol, and a redundant sending strategy.

[0037] In this embodiment, the service quality label and link quality profile are first concatenated to generate an input state vector. This vector contains two parts of data. The first part of the data comes from the link quality profile. Specifically, it extracts the real-time scores of all heterogeneous physical communication interfaces from the link quality profile and arranges them according to a preset interface order to form the first sub-vector. For example, the real-time scores recorded for each heterogeneous physical communication interface are read sequentially from the data structure storing the link quality profile, and these scores are filled into a one-dimensional array in strict accordance with a preset interface number order. This array is the first sub-vector. The interface number order can be a fixed logical number of the physical interface, such as interface 1 (Ethernet) and interface 2 (RS-485). The second part of the data comes from the service quality tags. Specifically, the urgency level, reliability requirement, and tolerance latency values ​​are extracted from the service quality tags and arranged in a preset order to form the second sub-vector. For example, the data structure of the service quality tags is parsed to read the values ​​of the "urgency level", "reliability requirement", and "tolerance latency" fields. These values ​​are then filled into another one-dimensional array in a fixed order, such as urgency level, reliability requirement, and tolerance latency. This array is the second sub-vector.

[0038] After obtaining the input state vector containing the first and second sub-vectors, this input vector is fed into the trained reinforcement learning decision model for decision-making. During each decision, the model can first read and combine the first and second sub-vectors from memory to form a complete input state vector. For example, all elements of the second sub-vector can be appended sequentially to all elements of the first sub-vector to generate a new, longer composite one-dimensional array. This new array is the complete input state vector. Then, based on this input state vector and according to preset decision criteria, an optimal target action is selected from a preset action space. The preset decision criteria are to maximize the expected cumulative reward, which is the expected cumulative reward of the reward function dynamically configured according to the aforementioned service quality label. Target actions include, but are not limited to, sending interfaces, encapsulation protocols, and redundant sending strategies.

[0039] The reward function evaluates the immediate effect of the executed action. Its calculation logic includes three basic items: latency reduction reward, reliability improvement reward, and resource consumption penalty. Furthermore, the reward function is dynamically configured, adjusting the weighting coefficients of these three basic items in the reward calculation based on the urgency, reliability requirement, and tolerable latency values ​​in the service quality label of the current business data to be sent. Specifically, a base weight is set for each of the three basic items in the reward function. When processing specific business data, the "urgency" value in the service quality label is read first, and the base weight of the "latency reduction reward" item is increased proportionally. Then, the "reliability requirement" value is read, and the base weight of the "reliability improvement reward" item is increased proportionally. Finally, the "tolerable latency" value is read; the larger this value, the smaller the weight of the "latency reduction reward" item is. After adjustment, it is still necessary to ensure that the sum of the three weighting coefficients remains one. For example, the basic weights can be set as follows: the weight of the latency reduction reward item = 0.4, the weight of the reliability improvement reward item = 0.4, and the weight of the resource consumption penalty item = 0.2. For services with an urgency level of 9 (maximum 10), the weight of the "latency reduction reward" item can be increased to 0.7, and the other two items can be decreased accordingly. For services with a tolerance latency of 5 seconds, the weight of the "latency reduction reward" item can be decreased to 0.1, and the other two items can be increased accordingly.

[0040] Specifically, before inputting the service quality label and the link quality profile into the trained reinforcement learning decision model, the method further includes: generating business data identification statistics based on the business data identification results and the service quality label. It should be noted that before inputting the service quality label and link quality profile into the reinforcement learning decision model, it is necessary to record the type of business data identified, the identification time, and the final generated service quality label. These records are then appended to a historical record list. Periodic statistical analysis of the historical record list yields the business data identification statistics. For example, by iterating through the historical record list at a fixed time period, the number of occurrences of each business data type, the average or maximum value of the "tolerable latency" field in each type of service quality label, and the distribution of the "urgency level" field are counted within that period. The statistical results are then summarized into business data identification statistics. The periodic statistical analysis period can be set according to the actual application scenario, such as 1 minute, 5 minutes, or 1 hour. The specific content of the business data identification statistics may include: the traffic share of various services within the period, the distribution of average latency requirements, and the frequency of occurrence of high-urgency-level services. This information is used for network behavior analysis.

[0041] In this embodiment, selecting a target action from a preset action space according to preset decision criteria may specifically include: using a value evaluation function and based on the service quality label and the link quality profile, evaluating the expected value of the cumulative return obtained after executing each action in the preset action space to obtain the expected cumulative return; determining the maximum value from multiple expected cumulative returns, and taking the action corresponding to the maximum value as the target action. It should be noted that this application pre-sets an action space, which is a set containing multiple selectable actions, and sets a corresponding sending interface, encapsulation protocol, and redundant sending strategy for each action; for example, the action space contains the following actions: Action A = {Interface: Ethernet, Protocol: IEC61850, Redundancy: No}; Action B = {Interface: RS-485, Protocol: Modbus, Redundancy: No}; Action C = {Interface: Ethernet & RS-485, Protocol: IEC61850 & Modbus, Redundancy: Yes}.

[0042] Furthermore, the reinforcement learning decision model in this application maintains a value evaluation function. This function evaluates the expected value of the future cumulative reward obtained by performing any action given an input state. The function receives two inputs: an input state vector representing the current environment and an encoding of a specific action to be evaluated. Internally, the function simulates a series of possible state transitions and rewards after performing the action through a series of numerical calculations (which can be implemented based on neural networks or lookup tables). The rewards obtained from these simulations are accumulated with time decay, and finally output a scalar value (i.e., the expected value) representing the expected long-term return. Here, time decay means that the reward at the k-th step in the future needs to be multiplied by a decay coefficient less than 1 raised to the power of k when accumulating, for example, the decay coefficient γ is 0.9.

[0043] Specifically, the process of selecting the target action is as follows: Based on the current input state vector, the expected value of the future cumulative reward corresponding to each action in the action space is calculated using the value evaluation function, and the action with the highest expected value is selected as the optimal action; for example, the encoding of each action in the preset action space is input into the value evaluation function along with the current input state vector in turn to obtain a series of expected values ​​of future cumulative rewards. Then, the magnitude of all expected values ​​is compared, the largest expected value is found, and the action corresponding to the largest value is determined as the optimal action to be executed (i.e., the target action).

[0044] Step S15: Execute the target action to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

[0045] In this embodiment, after selecting the optimal target action, the service data to be sent is sent from the transformer to the preset peer node according to the sending interface, encapsulation protocol and redundant sending strategy in the target action, thereby protecting the secure transmission of the service data to be sent between the transformer and the preset peer node.

[0046] Furthermore, after executing the target action, the process may further include: obtaining the execution result of the target action, obtaining execution result information, and generating a decision log based on the execution result information, the service quality label, the link quality profile, and the target action; uploading the link quality profile, the business data identification statistics, and the decision log to a remote management platform through the heterogeneous physical communication interface; receiving predictive policy instructions issued by the remote management platform, and integrating the predictive policy instructions as prior knowledge into the reinforcement learning decision model to adjust the decision tendency of the reinforcement learning decision model. In this embodiment, after selecting and executing the optimal action, the result information after the target action is executed can be collected, and then the input state vector of this decision, the identifier of the selected target action, the execution result information, and the calculated immediate reward value, along with the timestamp, are recorded as a structured data entry to generate a decision log. Specifically, the execution result information may include: the actual end-to-end latency of this communication, whether an acknowledgment message was received from the other party, and whether redundant transmission was enabled; for example, the actual latency reduction = (tolerable latency threshold - actual latency) / tolerable latency threshold; the reliability improvement is 1 when the delivery is successful, and 0 otherwise; the resource consumption is 1 when redundant transmission is used, and 0 otherwise; the weight of the resource consumption penalty item can be fixed at -0.3.

[0047] Specifically, when generating decision logs, a new data record can be created first. The fields of this record should include at least: timestamp, copy of the input state vector, optimal action identifier, result information, and calculated immediate reward value.

[0048] The process of obtaining the immediate reward value is as follows: The reward weight coefficients in the reward function of the reinforcement learning decision model are dynamically adjusted based on the service quality label. Then, the immediate reward value for this decision is calculated based on the execution result information and using the reward function. For example, after executing the target action, the actual latency reduction, reliability improvement (e.g., whether delivery was successful), and resource consumption (e.g., whether redundant transmission was used) of this communication are calculated based on the execution result information. Then, the actual latency reduction is multiplied by the adjusted weight of the "latency reduction reward" item to obtain the latency reward score; the reliability improvement is multiplied by the adjusted weight of the "reliability improvement reward" item to obtain the reliability reward score; and the resource consumption is multiplied by the weight of the "resource consumption penalty" item (usually a negative value) to obtain the resource penalty score. Finally, the sum of the latency reward score, reliability reward score, and resource penalty score is calculated to obtain the immediate reward value for this decision.

[0049] Furthermore, the aforementioned link quality profile, service data identification statistics, and decision logs can be uploaded to the remote management platform via a heterogeneous physical communication interface according to a preset cycle or event triggering conditions. For example, a timer can be maintained within the protection control unit, automatically triggering data upload when the timer reaches the preset reporting cycle (e.g., 5 minutes). Alternatively, event-driven upload can be triggered immediately when the frequency of high-urgency-level services in the service data identification statistics exceeds a preset threshold, or when the number of newly recorded abnormal actions in the decision log exceeds a preset threshold. The reporting cycle can be set according to the actual application scenario, such as 5 minutes; "high-urgency-level services" can be defined as services with an "urgency level" field value greater than or equal to 7 (out of 10 levels) in the service quality label; its frequency threshold can be set to "more than 3 times per minute"; "abnormal actions" can be defined as actions recorded in the decision log that result in a negative immediate reward value; and its quantity threshold can be specifically set to "3 abnormal actions occurring in 10 consecutive decisions".

[0050] It should be noted that before uploading the above information to the remote management platform, the data to be uploaded can be packaged and encapsulated into an upload message format that meets the requirements of the remote management platform. For example, the complete data structure of the latest link quality profile can be read from memory, and the records of business data identification statistics and decision logs stored locally can be retrieved. Based on the timestamp of each record, all new records with times later than the last successful upload timestamp can be filtered out. Then, the link quality profile, the filtered new business data identification statistics records, and the new decision log records can be organized according to the data field order and encoding method specified by the remote management platform, and message headers, length checksums, and other information can be added to form a complete upload message. The encoding method can be JSON, XML (Extensible Markup Language), or a custom binary format. The last successful upload timestamp refers to the timestamp that is updated after each successful confirmation response is received from the remote management platform.

[0051] In addition, when multiple heterogeneous physical communication interfaces exist, the heterogeneous physical communication interface that meets the current comprehensive communication quality requirements can be selected from the current link quality profile, and then the upload message can be sent to the remote management platform through that interface. For example, the latest generated link quality profile can be read, and the real-time score recorded for each heterogeneous physical communication interface can be traversed. Based on the preset comprehensive communication quality threshold, all interfaces with real-time scores not lower than the threshold can be added to the candidate interface set. If the candidate interface set is not empty, the interface with the highest real-time score can be selected as the sending interface for this upload. If the candidate interface set is empty, the interface with the highest real-time score among all interfaces can be selected, that is, it is not necessary to reach the threshold. After selecting the interface, the encapsulated upload message can be sent out through the communication protocol stack of that interface. The overall communication quality threshold can be set based on experience, such as 0.6. The communication protocol stack refers to the software module that matches the selected interface and the remote management platform's peer protocol, such as the TCP / IP (Transmission Control Protocol / Internet Protocol) protocol stack and HTTP / HTTPS (Hypertext Transfer Protocol Secure) client based on the selected Ethernet interface, or the ModbusRTU (a serial communication protocol) master station protocol stack based on the selected RS-485 interface.

[0052] Next, the system receives predictive policy instructions from the remote management platform after receiving the uploaded message. These instructions are then integrated into the reinforcement learning decision model as prior knowledge to adjust its decision-making tendency. Specifically, after sending the uploaded message, a response receiving process with a timeout mechanism can be initiated. This process continuously monitors the heterogeneous physical communication interface currently used by the reinforcement learning decision model, or dedicated to the management channel (which could be a pre-specified physical interface specifically for communicating with the remote management platform, such as an Ethernet interface), for a preset waiting time (e.g., 30 seconds) to receive possible response messages. Upon receiving a predictive policy instruction, the instruction undergoes security verification and parsing to extract the policy parameters or policy update rules contained within. For example, the received message undergoes format verification, digital signature verification, or checksum verification to ensure its integrity and source credibility. Once verification is successful, the message payload is parsed according to the protocol agreed upon with the remote management platform to extract the core policy instruction content. The policy parameters can be recommended actions for specific service types (such as protection trip commands) (e.g., prioritizing Ethernet interfaces and the IEC61850 protocol), or suggestions for adjusting weighting coefficients in the value assessment function (e.g., reducing latency rewards for all services within the next hour). (Item weight increased by 10%). The policy update rule can be a set of "state-action" mapping relationships used to directly expand or modify the prior knowledge of the reinforcement learning decision model. For example, the policy parameters for "protection trip command" may be: {"business type": "Trip_Command", "recommended action": {"interface": "Ethernet", "protocol": "IEC61850"}}; the adjustment suggestion for the weight coefficient can be: {"effective period": "13:00-14:00", "adjustment item": "latency reduction reward weight", "adjustment range": "+0.2"}; the policy update rule can be a set of pre-compiled if-then rules.

[0053] Finally, the parsed policy parameters or policy update rules are converted into a format that the reinforcement learning decision model can load and understand. This format is then used to update the model's prior knowledge base or directly adjust the internal parameters of the value evaluation function, ensuring that the model's decision-making tendency is guided by predictive policy instructions. For example, when the policy parameter is a recommended action, it is encoded as a "state-action" pair recognizable by the reinforcement learning decision model and assigned a relatively high initial expected value (which can be set slightly higher than the average reward value that the model can explore under the current policy to accelerate policy adoption). This is then inserted into or updated in the model's prior knowledge base (e.g., a specific entry in the Q-table). If the policy parameter is a weight adjustment suggestion, the weight parameters used in calculating the expected value of future cumulative rewards in the value evaluation function are directly modified according to the suggested magnitude. If the policy parameter is a policy update rule, this new mapping relationship partially covers (meaning that when the new rule conflicts with the original model's decision, within the specific state range defined by the rule, the new rule takes precedence) the model's original policy logic. After the above update is completed, when the reinforcement learning decision model encounters the same or similar state in the future, the expected value of the action calculated by its value evaluation function will change due to the change in prior knowledge, or its decision function will be directly guided by the new rules, thereby adjusting its decision tendency in the direction of the expected predictive policy instruction.

[0054] This application achieves accurate real-time perception of the status of heterogeneous communication interfaces by periodically constructing dynamically updated link quality profiles. Through intelligent identification and service quality tagging of business data, it achieves precise matching of communication needs and network resources. Furthermore, by integrating a reinforcement learning decision model, it can adaptively select the optimal communication interface, protocol, and redundancy strategy based on real-time network status and business requirements. This significantly improves the reliability, real-time performance, and overall communication efficiency of transformer protection and control business data transmission in complex, multi-protocol, and dynamically changing network environments. In addition, by uploading local operating data to a remote management platform and integrating the issued predictive strategies, this application enables the model to continuously learn and optimize, enhancing its long-term environmental adaptability and decision-making intelligence. Moreover, by avoiding the expansion of the fault scope caused by delayed or lost protection commands, it can prevent significant economic losses. By ensuring the real-time transmission of critical services, it improves grid stability and power supply reliability, reducing power outage losses. By intelligently utilizing communication resources, it reduces dependence on single high-cost links, optimizes operation and maintenance costs, and thus achieves higher cost-effectiveness in ensuring power supply security in complex environments, demonstrating significant life-cycle benefits.

[0055] As can be seen, this application embodiment acquires communication performance indicators of heterogeneous physical communication interfaces according to a preset period, evaluates the communication quality of the interfaces based on the communication performance indicators, and then generates a dynamically updated link quality profile based on the evaluation results and communication performance indicators, thereby achieving accurate real-time perception of the communication interface status. Furthermore, this application embodiment identifies the service data to be sent and generates a service quality label based on the identification results, thus achieving accurate matching between communication needs and network resources. In addition, this application embodiment uses a reinforcement learning decision model and adaptively decides the optimal action for sending service data based on real-time network status and service needs, and executes the optimal action to protect the data transmission of the service data to be sent between the transformer and the peer node. This significantly improves the reliability, real-time performance, and overall communication efficiency of transformer protection control service data transmission in a complex, multi-protocol, and dynamically changing network environment. At the same time, by real-time perception of network status, intelligent scheduling of services, and remote continuous optimization, it solves the problem of unreliable communication in the heterogeneous dynamic environment of the distribution network, improves the stability and power supply reliability of the distribution network, reduces power outage losses, and ensures power supply security.

[0056] Accordingly, this application also discloses a transformer intelligent protection control device, applied to the protection control unit in a transformer, see [link to relevant documentation]. Figure 2 As shown, the device includes: The interaction module 11 is used to interact with a preset peer node in the target distribution network through a heterogeneous physical communication interface according to a preset period, so as to obtain the communication performance index of the heterogeneous physical communication interface. Evaluation module 12 is used to evaluate the communication quality of the heterogeneous physical communication interface according to the communication performance indicators, and obtain the communication quality evaluation result; The combination module 13 is used to combine the communication quality assessment results and the communication performance indicators to generate a dynamically updated link quality profile. The parsing module 14 is used to parse the business data to be sent using a preset business feature knowledge base, so as to identify and record the type and attributes of the business data to be sent, and obtain the business data identification result. The generation module 15 is used to generate a service quality label for the service data to be sent based on the service data identification result; Input module 16 is used to input the service quality label and the link quality profile into the trained reinforcement learning decision model, so as to select a target action from a preset action space according to a preset decision criterion; the preset decision criterion is to maximize the expected cumulative return, and the expected cumulative return is the expected cumulative return of the reward function dynamically configured according to the service quality label; The sending module 17 is used to execute the target action to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

[0057] The specific workflow of each of the above modules can be found in the relevant content disclosed in the foregoing embodiments, and will not be repeated here.

[0058] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0059] Figure 3 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the transformer intelligent protection control method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0060] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0061] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0062] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the transformer intelligent protection control method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include a computer program capable of performing other specific tasks.

[0063] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned intelligent protection and control method for transformers. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0064] Furthermore, embodiments of this application also disclose a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the transformer intelligent protection control method disclosed above.

[0065] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0066] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0067] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0068] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0069] The above provides a detailed description of a transformer intelligent protection control method, device, equipment, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for intelligent protection control of a transformer, characterized in that, Protection and control units used in transformers include: According to a preset period, data packets are exchanged with a preset peer node in the target distribution network through a heterogeneous physical communication interface to obtain the communication performance indicators of the heterogeneous physical communication interface. The communication quality of the heterogeneous physical communication interface is evaluated based on the communication performance indicators to obtain the communication quality evaluation result. The communication quality evaluation result and the communication performance indicators are then combined to generate a dynamically updated link quality profile. The system uses a pre-defined business feature knowledge base to parse the business data to be sent, so as to identify and record the type and attributes of the business data to be sent, obtain the business data identification result, and generate a service quality label for the business data to be sent based on the business data identification result. The service quality label and the link quality profile are input into the trained reinforcement learning decision model to select a target action from a preset action space according to a preset decision criterion. The preset decision criterion is to maximize the expected cumulative reward, which is the expected cumulative reward of the reward function dynamically configured according to the service quality label. The target action includes a sending interface, an encapsulation protocol, and a redundant sending strategy. The target action is executed to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

2. The intelligent protection control method for transformer as claimed in claim 1 wherein, The step of evaluating the communication quality of the heterogeneous physical communication interface based on the communication performance indicators to obtain the communication quality evaluation result includes: Normalize each of the communication performance indicators corresponding to the heterogeneous physical communication interface to obtain multiple normalized performance indicators. The communication quality evaluation result of the heterogeneous physical communication interface is obtained by weighted fusion of multiple normalized performance indicators.

3. The intelligent protection control method of transformer according to claim 2, characterized in that, The step of combining the communication quality assessment results and the communication performance indicators to generate a dynamically updated link quality profile includes: Data entries are created based on the communication quality assessment results and the normalized performance indicators to obtain the current data entry; The current data entry replaces the old data entry stored in the previous cycle of the heterogeneous physical communication interface, generating a dynamically updated link quality profile.

4. The intelligent protection control method for transformer as claimed in claim 3 wherein, The step of generating a quality of service label for the service data to be sent based on the service data identification result includes: Based on the business data identification result, the corresponding service quality constraint value is read from the preset business feature knowledge base, and the service quality constraint value is used as the service quality label of the business data to be sent; the service quality label includes an urgency value, a reliability requirement value, and a tolerance latency value.

5. The intelligent protection control method for transformer as claimed in claim 1 wherein, Before inputting the service quality label and the link quality profile into the trained reinforcement learning decision model, the method further includes: Based on the business data identification results and the service quality tags, generate business data identification statistics; Accordingly, the step of selecting a target action from a preset action space according to preset decision criteria includes: Using a value assessment function and based on the service quality label and the link quality profile, the expected value of the cumulative return obtained after performing each action in the preset action space is evaluated to obtain the expected cumulative return; The maximum value is determined from the multiple expected cumulative returns, and the action corresponding to the maximum value is taken as the target action; Accordingly, after executing the target action, the process further includes: Obtain the execution result of the target action, obtain the execution result information, and generate a decision log based on the execution result information, the service quality label, the link quality profile, and the target action; The link quality profile, the service data identification statistics, and the decision log are uploaded to the remote management platform through the heterogeneous physical communication interface. The system receives predictive policy instructions from the remote management platform and integrates these instructions as prior knowledge into the reinforcement learning decision model to adjust the decision tendency of the model.

6. The intelligent protection control method for transformer as claimed in claim 5 wherein, Also includes: The reward weight coefficients in the reward function of the reinforcement learning decision model are dynamically adjusted based on the service quality labels. The immediate reward value for this decision is calculated based on the execution result information and using the reward function; Accordingly, the step of generating decision logs based on the execution result information, the service quality label, the link quality profile, and the target action includes: Based on the execution result information, the service quality label, the link quality profile, the target action, and the instant reward value, structured data entries are created to obtain the decision log.

7. The intelligent protection control method of transformer according to any one of claims 1 to 6, characterized in that, The step of exchanging data packets with a preset peer node in the target distribution network according to a preset period and through a heterogeneous physical communication interface to obtain the communication performance indicators of the heterogeneous physical communication interface includes: According to a preset period, and through a heterogeneous physical communication interface, the probe data packets are sent multiple times to a preset peer node in the target distribution network, and the response data packets returned by the preset peer node are received. Physical layer performance metrics are read from the heterogeneous physical communication interface; the physical layer performance metrics include signal strength and bit error rate. Based on the time difference between sending the probe data packet and receiving the response data packet, the multiple one-way delays between the protection control unit and the preset peer node are calculated to obtain the delay sequence. The latency jitter of the heterogeneous physical communication interface is obtained based on the absolute value of the difference between two adjacent one-way delays in the latency sequence. The total number of probe data packets sent and the total number of response data packets received within the preset period are counted to obtain the total number of packets sent and the total number of packets received. The ratio of the total number of packets sent to the total number of packets received is then calculated to obtain the packet loss rate. Generate communication performance metrics that include the physical layer performance metrics and the network layer performance metrics; the network layer performance metrics include the delay sequence, the delay jitter, and the packet loss rate.

8. A transformer intelligent protection control device, characterized in that, Protection and control units used in transformers include: An interaction module is used to interact with a preset peer node in the target distribution network through a heterogeneous physical communication interface according to a preset period, so as to obtain the communication performance indicators of the heterogeneous physical communication interface. An evaluation module is used to evaluate the communication quality of the heterogeneous physical communication interface based on the communication performance indicators, and obtain a communication quality evaluation result. The combination module is used to combine the communication quality assessment results and the communication performance indicators to generate a dynamically updated link quality profile. The parsing module is used to parse the business data to be sent using a preset business feature knowledge base, so as to identify and record the type and attributes of the business data to be sent, and obtain the business data identification result. The generation module is used to generate a service quality label for the service data to be sent based on the service data identification result; The input module is used to input the service quality label and the link quality profile into the trained reinforcement learning decision model, so as to select a target action from a preset action space according to a preset decision criterion; the preset decision criterion is to maximize the expected cumulative return, and the expected cumulative return is the expected cumulative return of the reward function dynamically configured according to the service quality label; The sending module is used to execute the target action to protect the data transmission of the service data to be sent on the target communication link; the target communication link is the communication link between the transformer and the preset peer node.

9. An electronic device, characterized in that, It includes a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the transformer intelligent protection control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer programs; wherein, when the computer programs are executed by a processor, they implement the transformer intelligent protection control method as described in any one of claims 1 to 7.

Images