A power distribution network operation control system and method based on VR command practical training

Abstract

This invention provides a distribution network operation control system and method based on VR command and training, belonging to the field of distribution network operation control technology. To address the technical problem of the single dimension of fault behavior simulation in existing VR power operation training programs, a multi-source fault hybrid dynamic generation module is used to generate multi-source fault and cascading fault scenarios containing non-deterministic combined responses. A multi-source information conflict comprehensive judgment module presents trainees with an incomplete and non-absolutely reliable master station information interface, allowing trainees to perform multi-source information conflict comprehensive judgment. A multi-role command and collaborative interaction module allows AI characters to interact with trainees through scheduling dialogue; the AI ​​characters proactively issue violation commands, which trainees then perform command recognition, safety regulation verification, and correction operations. This invention is applied to distribution network operation interactive control.

Description

Technical Field

[0001] This invention provides a distribution network operation control system and method based on VR command and training, belonging to the field of distribution network operation control technology. Background Technology

[0002] With the maturity of Virtual Reality (VR) technology, its application in the field of power system safety operation training has become increasingly widespread, forming an operation training system with immersive equipment operation simulation as its core. This system effectively solves the prominent problems of high safety hazards, high cost, and limited scenarios in traditional training by constructing a three-dimensional simulation training environment.

[0003] However, existing technical solutions have revealed significant limitations when developing towards more advanced and intelligent training stages. Current VR power operation training solutions suffer from serious deficiencies in the simulation of fault behavior, exhibiting a singular and idealized approach. Specifically:

[0004] (1) Simulation of non-ideal behavior of lack of protection and distribution automation devices:

[0005] In actual 10kV distribution network emergency repairs, frequent occurrences include relay protection device malfunctions (such as false outputs due to no fault in overcurrent stage I), automatic transfer switch failures or logical instability, unexpected tripping of undervoltage release devices, and incorrect line selection by low-current grounding fault location devices. These malfunctions, failures, and abnormal behaviors are the main technical reasons for repair personnel's hesitation in judgment, failed test runs, and delayed power outages. However, existing VR training systems can only reproduce the ideal mapping relationship of "one fault → correct device operation," and cannot simulate the deceptive fault scenarios caused by device defects or abnormal settings. This will result in trainees lacking the ability to identify and trace the source of faults when faced with real device anomalies.

[0006] (2) Lack of simulation of non-electrical work environment risks:

[0007] Existing VR training scenarios are highly focused on simulating the operation of electrical equipment itself. However, non-electrical personal safety risks and work environment risks that frequently occur in actual emergency repair operations, such as fires in cable trenches / ring network cabinets, accumulation of toxic and harmful gases in confined spaces, and collapse of tower foundations due to external damage to underground pipelines, are completely absent from existing technical solutions. Trainees cannot experience practical aspects such as fire assessment, gas detection, ventilation evaluation, and emergency evacuation command in a simulated environment, resulting in a lack of awareness and collaborative handling capabilities regarding non-electrical risks on-site after obtaining operational qualifications.

[0008] (3) Lack of simulation of multi-role command and coordination and safety regulation authorization verification:

[0009] Existing VR training systems generally adopt a human-computer interaction paradigm where one person directly operates the equipment. However, real power distribution network emergency repair operations involve a three-level closed-loop command process of "dispatcher-repair leader-team member," which involves core professional standards such as reciting dispatch terminology, executing operation tickets, and verifying safety regulations and authority boundaries. Existing technology cannot effectively train the aforementioned non-operational but directly quality-determining command, decision-making, and communication soft skills. Furthermore, it cannot simulate real-world scenarios that require the person in charge to identify, stop, and correct, such as dispatch command errors, unauthorized requests by team members, and misjudgments of safe distances.

[0010] In summary, while current VR power operation training systems have solved the problem of "immersion" in the operating environment, they have shortcomings in core intelligent dimensions such as "dynamic intelligent generation" of training content, "deep cognitive assessment" of the operation process, and "full-element interactive simulation" of the workflow. These deficiencies make it difficult for the training effect to meet the needs of cultivating high-level and highly adaptable power operation and maintenance talents. Summary of the Invention

[0011] To address the technical problems existing in the background art, the present invention adopts the following technical solution: providing a distribution network operation control system based on VR command and training, comprising:

[0012] The multi-source fault hybrid dynamic generation module has a built-in primary equipment electrical quantity model and secondary device abnormal behavior parameter library. After generating a primary fault event, it inputs the fault characteristic quantity into the secondary device virtual behavior model, independently calculates the action behavior and timing of each protection and automation device, and outputs a nondeterministic combination response containing at least two of the following: correct action, maloperation, failure to operate, and logical disorder.

[0013] The multi-source information conflict comprehensive judgment module under the simulation of power distribution automation system has at least one of the following built-in: monitoring blind zone model, self-healing function failure model, and information false alarm / distortion model. It is used to present the main station information interface to the trainee that is not completely mapped to the real fault state. The main station information interface contains at least one of the following non-real information: communication interruption, invalid data, switch change false alarm, protection message false alarm, and telemetry data distortion.

[0014] The multi-role command and collaborative interaction module has built-in AI dispatcher and AI team member simulation objects, and a pre-set violation instruction generation engine, which is used to actively issue instructions that violate power safety work procedures during training, allowing trainees to perform instruction recognition, safety regulation verification, and correction operations as the person in charge of emergency repairs.

[0015] The intelligent decision reference module generates the theoretically optimal handling plan based on load importance classification and transfer capacity optimization. It also provides an immersive dual-track parallel playback of the trainees' actual operation, highlighting key deviation nodes and outputting quantitative comparison prompts.

[0016] The post-event intelligent review and diagnostic report automatic generation module is used to establish a three-layer attribution mapping chain of fault phenomenon-electrical quantity-device behavior, generate a multi-dimensional evaluation matrix of the trainee's operation steps, and automatically synthesize emergency repair reports.

[0017] The multi-source fault hybrid dynamic generation module is specifically configured with:

[0018] The primary-secondary joint state space model includes the electrical quantity status of primary equipment, the setting zone status of secondary devices, the activation / deactivation status of pressure plates, the communication status, and preset templates for abnormal device behavior parameters.

[0019] The fault-action decoupling generation engine is used to input fault feature quantities into the virtual behavior model of the secondary device, and combine the abnormal behavior parameter template to independently calculate the action behavior and timing of each device.

[0020] The malfunction probability weight adjustment interface is used by instructors or AI training strategy modules to dynamically adjust the probability factors of various devices malfunctioning, refusing to operate, or experiencing logical disorder.

[0021] The nondeterministic simulation unit supports mixed outputs of correct / false operation / failure to operate. It is used for multiple devices to present at least two mixed outputs among correct operation, false operation, failure to operate, and logical disorder when a single fault event triggers them. It is also used for simulating at least one of the following scenarios: reclosing due to a permanent fault triggers secondary tripping, the tie switch switching to a faulty line triggers overcurrent protection tripping of the backup line, and the automatic transfer switch operating on a faulty bus causes cascading tripping. It is also used for simulating at least one of the following scenarios: multi-device timing coupling and fault dynamic evolution chain reaction.

[0022] The pre-defined abnormal behavior parameter templates for the device in the first- and second-order joint state-space model include at least three of the following types:

[0023] The constant drift model has a correction formula as follows: ,in As a benchmark value, This is the drift coefficient;

[0024] The TA saturation model is used to correct distortion in secondary sampled values. The saturation criterion is... and Where I is the primary current, Where t is the saturation threshold and t is the duration. This is the saturation time threshold;

[0025] A communication interruption model is used to determine the online status and interruption duration of a computing device. The online probability function is: , where λ is the failure rate and t is time;

[0026] The mechanism cascade model is used to simulate the failure probability and delay time of the actuator.

[0027] The multi-source information conflict comprehensive analysis module under the power distribution automation system simulation also includes:

[0028] The monitoring blind spot model is set up so that some power distribution nodes are not covered by automation terminals or the terminals are offline. The main station interface displays the node as "communication interrupted" or "data invalid". Trainees are required to obtain the real status of the node through on-site verification simulation commands in the VR environment.

[0029] The self-healing function is missing model. In the case of at least one of the following operating conditions: feeder automation is not put into operation, self-healing function is not enabled, or master station remote control permission has not been issued, the system will not automatically perform recovery operation and the trainee is required to actively apply for remote control permission or give an order to operate on site.

[0030] The information false alarm / distortion model has a built-in message false alarm engine. Based on a probability model or instructor preset, it randomly or deterministically generates at least one type of false information, such as switch position false alarm, protection message false alarm, telemetry data distortion, and message timing disorder.

[0031] The information false alarm / distortion model is specifically configured with the following:

[0032] A switch position change false alarm sub-model is used to generate false alarms in the case of closing position reporting opening position, opening position reporting closing position, or random jitter.

[0033] The telemetry data distortion sub-model is used to generate distorted telemetry values, and the calculation formula is as follows: Where P is the actual telemetry value, The distortion coefficient has a value range of [-0.2, 0.2].

[0034] The message timing disorder sub-model is used to generate timing disorder by exchanging message timestamps;

[0035] The DTU offline and actual switch tripping composite scenario generation unit is used to generate information conflict states where the master station information deviates from the actual operating conditions in the scenario where DTU communication is interrupted but the switch has actually tripped.

[0036] The unit is equipped with a composite scenario generation unit for automatic false alarm and actual failure to operate, which is used to generate an information distortion state when the main station receives a success message but the device actually fails to operate.

[0037] The multi-role command and collaborative interaction module is specifically configured with the following:

[0038] The multi-role simulation object library contains two types of collaborative role simulation objects: AI dispatcher and AI team member. It also includes pre-made standard terminology templates and instruction logic trees that conform to regulations, including at least four typical dispatch dialogue processes such as power outage and restoration application, instruction repetition, operation permission, task change, anomaly reporting, and work termination.

[0039] The violation instruction generation engine allows AI dispatchers or AI team members to actively issue instructions that violate safety regulations or dispatch procedures during training. These instructions must include at least three types of violations, such as issuing power supply instructions without removing the grounding wire, requesting operation without a ticket, repeating errors without correction, misjudging safety distance, giving orders beyond one's authority, unlocking operations without a ticket, expired work tickets, and missing grounding wires.

[0040] In the command, decision-making and correction unit, trainees enter as the person in charge of emergency repairs. They do not have direct authority to operate equipment. They are required to issue operation commands to team members through voice or command panel, interact with dispatchers, and autonomously identify, judge and execute at least one corrective action among refusing to execute, requesting correction, and suspending operation when the AI ​​character issues illegal or erroneous commands.

[0041] The command and decision-making quantitative evaluation unit collects all voice interaction content between trainees and AI dispatchers and AI team members, as well as the timing of instruction issuance. Based on the built-in safety rule library, it performs real-time compliance verification. The evaluation dimensions include at least three of the following: instruction recitation completeness rate, violation instruction recognition rate and response accuracy rate, completeness of the two-ticket execution process, and compliance of overstepping reporting / overstepping command behavior in emergency situations.

[0042] The targeted training unit for violation instructions dynamically selects the violation type based on the student's historical violation recognition rate. When the student's violation recognition rate is below 60%, highly deceptive violation instructions are generated first, and when the student's violation recognition rate is above 60%, highly concealed or complex violation instructions are generated first.

[0043] The specific settings in the violation instruction generation engine are as follows:

[0044] The violation type library should include at least five types of violations, such as grounding wire not being disconnected and power being supplied, operation without a permit, repeated errors not being corrected, misjudgment of safety distance, commanding beyond one's level, unlocking operation without a permit, work permit expired, and grounding wire missing. Each violation instruction should be associated with the corresponding safety regulations and the correct response template.

[0045] The violation probability control unit uses a calculation formula. Calculate the probability of generating a violation instruction, where D is the system's preset difficulty level, ranging from 1 to 5;

[0046] The deceptive parameter configuration unit for illegal commands controls the difficulty of identifying illegal commands by adjusting at least one parameter among the following: the ambiguity of the illegal command's expression, its similarity to normal commands, and the concealment of the timing of its issuance.

[0047] The intelligent decision reference module is specifically configured with the following:

[0048] The load importance classification and power outage loss assessment model has a built-in user load classification label library for real-time quantification of the cumulative power outage loss weights for different handling paths. The formula for calculating the cumulative power outage loss weights is as follows: ,in User level coefficient For node active load, The total power outage duration for the node;

[0049] The recovery path optimization engine constructs an objective function with at least two of the following as joint optimization objectives: shortest power outage time, minimum impact range, and highest power restoration rate. ,in Total power restoration time Weighting for power outage losses, This refers to the number of switch operations. Using the weighted coefficients, we iterate through possible operation sequences to generate a theoretically optimal fault isolation and power restoration reference scheme;

[0050] The immersive dual-track comparison and review unit is used to play back the trainee's actual handling process and the AI-generated optimal solution in parallel in a VR environment. When the difference between key indicators exceeds the limit, the operation node is highlighted and a quantitative comparison prompt is output.

[0051] The scheme deviation analysis unit calculates the order deviation score and locates the specific deviation node by detecting the longest common subsequence between the student's operation sequence and the optimal scheme operation sequence.

[0052] The specific settings in the post-event intelligent review and diagnostic report automatic generation module are as follows:

[0053] The fault cause reasoning and operation evaluation report automatic generation module has a built-in three-layer attribution mapping chain of fault phenomenon-electrical quantity-device behavior. It reverse maps at least one of the surface phenomena observed by trainees in the VR environment, such as power outage, abnormal noise, smoke, protection action signal, and switch change signal, to the fault type of primary equipment, the action behavior of secondary device, and the logic state of automation system.

[0054] The multi-dimensional evaluation matrix generation unit evaluates trainees' entire operation in a granular and quantitative manner based on at least four dimensions, including operation sequence, safety compliance, command decision correctness, power restoration efficiency, and risk identification ability. Each deduction item is associated with a specific operation node and violation / error type, and specific clauses of safety regulations or dispatch procedures are cited as the basis for judgment.

[0055] The standardized emergency repair report generation unit automatically generates a standardized report after training, which includes basic fault information, fault cause analysis, trainee operation records, scores and evaluations, and comparative improvement suggestions. The standardized report includes at least the fault occurrence time, equipment name, fault type, protection action information, electrical principle or device behavior logic explanation corresponding to the abnormal phenomenon, all command issuance and voice interaction records displayed on a timeline, score and deduction details for each dimension, violation clause index, deviation analysis from the optimal solution, and targeted training suggestions.

[0056] A method for interactive control of a power distribution network operation control system based on VR command and control training includes the following interactive control steps:

[0057] Step 1: Generate multi-source fault and cascading fault scenarios containing nondeterministic combined responses through the multi-source fault hybrid dynamic generation module;

[0058] Step 2: Through the multi-source information conflict comprehensive analysis module, the main station information interface is presented to the trainees as incomplete and not absolutely reliable, and the trainees perform multi-source information conflict comprehensive analysis.

[0059] Step 3: Through the multi-role command and collaborative interaction module, the AI ​​role and the trainee conduct scheduling dialogue interaction. The AI ​​role actively issues violation instructions, and the trainee executes instruction recognition, safety regulation verification and correction operations.

[0060] Step 4: Through the intelligent decision reference module, the theoretically optimal solution is generated, and after the trainee completes the solution, the trainee's actual operation and the optimal solution are played back in an immersive dual-track parallel process, and quantitative comparison prompts are output.

[0061] Step 5: Through the post-event intelligent review module, the trainees' operations are quantitatively scored from multiple dimensions, and a standardized emergency repair report is automatically generated.

[0062] The advantages of this invention compared to existing technologies are as follows: This invention provides a distribution network operation control system and method based on VR command and training. By setting up a dynamic fault generation engine, it can simulate unpredictable and complex fault combinations in real-world work, solving the fundamental problem of fixed training scenarios in traditional training. By setting up an AI examiner module, it can perform multi-dimensional real-time analysis and quantitative scoring of the operation process, solving the problems of superficiality and mechanization in traditional evaluation methods. By introducing an AI dispatcher interaction module, it simulates the standardized work process from instruction issuance to task reporting, solving the defects of fragmented training processes and single interaction in traditional training. By adopting fault-device behavior decoupling modeling, it achieves high-fidelity simulation of dynamic, temporal, and multi-device coupled complex working conditions such as reclosing to a permanent fault and the impact tripping of backup lines caused by the transfer of faulty lines to tie switches. By setting up a fully automatic fault cause reasoning and operation evaluation report generation module, it automatically converts the millisecond-level operation behavior data, voice interaction commands, and equipment action responses collected during VR training into technical analysis documents that conform to production operation specifications. Attached Figure Description

[0063] The present invention will be further described below with reference to the accompanying drawings:

[0064] Figure 1 This is a schematic diagram of the structure of the power distribution network operation control system based on VR command training according to the present invention;

[0065] Figure 2 This is a schematic diagram of the workflow of the multi-source fault hybrid dynamic generation module of the present invention;

[0066] Figure 3 This is a schematic diagram of the workflow of the multi-source information conflict comprehensive judgment module under the simulation of the power distribution automation system of the present invention;

[0067] Figure 4 This is a schematic diagram of the workflow of the multi-role command and collaborative interaction module of the present invention;

[0068] Figure 5 This is a schematic diagram of the workflow of the intelligent decision reference module of the present invention;

[0069] Figure 6 This is a schematic diagram of the workflow of the intelligent post-event review and diagnostic report automatic generation module of the present invention. Detailed Implementation

[0070] like Figures 1 to 6As shown, this invention provides a distribution network operation control system and method based on VR command training. Specifically, it relates to a VR emergency repair command training interactive control scheme for simulating multi-source faults and device malfunctions in distribution networks. It adopts an immersive virtual reality (VR) training system and method that integrates relay protection action behavior simulation, distribution automation terminal logic simulation, and multi-role command collaboration in on-site emergency repair, so as to realize decision-making training for emergency repair personnel under complex distribution network fault backgrounds.

[0071] The distribution network operation control method based on VR command training provided by this invention mainly includes:

[0072] Step 1: Generate multi-source fault and cascading fault scenarios containing nondeterministic combined responses through the multi-source fault hybrid dynamic generation module;

[0073] Step 2: Through the multi-source information conflict comprehensive analysis module, the main station information interface is presented to the trainees as incomplete and not absolutely reliable, and the trainees perform multi-source information conflict comprehensive analysis.

[0074] Step 3: Through the multi-role command and collaborative interaction module, the AI ​​role and the trainee conduct scheduling dialogue interaction. The AI ​​role actively issues violation instructions, and the trainee executes instruction recognition, safety regulation verification and correction operations.

[0075] Step 4: Through the intelligent decision reference module, the theoretically optimal solution is generated, and after the trainee completes the solution, the trainee's actual operation and the optimal solution are played back in an immersive dual-track parallel process, and quantitative comparison prompts are output.

[0076] Step 5: Through the post-event intelligent review module, the trainees' operations are quantitatively scored from multiple dimensions, and a standardized emergency repair report is automatically generated.

[0077] All of the above functional modules are integrated into a unified virtual scene, realizing a closed loop from "visual presentation" to "practical interaction" through VR devices. In the VR environment, trainees enter a high-precision three-dimensional power distribution scene through a head-mounted display, complete various operations of emergency repair command with the help of dual controllers or data gloves, and achieve remote collaboration through voice communication and gesture commands.

[0078] Specifically, in step 1, to address the technical shortcomings of existing VR training systems that can only simulate an ideal, singular mapping of "fault → correct device action" and are completely unable to simulate protection malfunctions, automation terminal failures, and abnormal device behavior, this invention proposes to construct a multi-source fault hybrid dynamic generation module, which specifically includes:

[0079] Establish a joint primary and secondary state space model for the distribution network. This joint state space should include at least: ① Primary equipment layer: electrical quantity status (voltage, current, frequency) and physical damage status (external damage, overheating, toxic gas concentration) of lines, cables, transformers, and switchgear; ② Secondary equipment layer: setting areas of microprocessor-based protection devices, distribution automation terminals (DTU / FTU / TTU), automatic transfer switches, undervoltage release devices, pressure plate activation / deactivation status, sampling channel status, communication status, and actuator status; ③ Pre-set templates for abnormal device behavior parameters.

[0080] Establish a fault-action decoupling generation engine. After generating electrical faults such as short circuits, grounding, and open circuits, or non-electrical events such as fires and toxic gases on the primary side, the dynamic fault generation engine does not directly preset the corresponding device's action result. Instead, it inputs the fault characteristic quantities (fault current amplitude, zero-sequence voltage, harmonic content, and duration) into the secondary equipment behavior model. Combined with the virtual health status parameters set at the current device level, the behavior model independently calculates whether each protection / automation device is activated, whether it outputs an alert, and the action sequence.

[0081] The system provides an interface for adjusting the probability weight of malfunctions. Instructors or AI training strategy modules can dynamically adjust the probability factors for various devices to malfunction, refuse to operate, or exhibit logical instability, allowing for a gradual increase in training difficulty. Trainees cannot complete operations by simply memorizing scripts; they must deduce the device's behavioral logic based on malfunction phenomena, identify abnormal characteristics, and then correct their decisions. This achieves a fundamental leap from training in operational proficiency to training in fault tracing and command decision-making capabilities.

[0082] Supports nondeterministic simulation with mixed outputs of correct / false tripping / failure to trip. Based on the above decoupling mechanism, this invention supports multiple devices exhibiting differentiated and non-ideal response behaviors triggered by a single fault event. Typical output scenarios include, but are not limited to: ① A single-phase ground fault occurs on the line, the A-set protection correctly selects the line and trips, the B-set protection fails to trip due to setting drift, and the automatic transfer switch does not operate because charging is not completed; ② A cable terminal head breaks down and short-circuits, the overcurrent stage I operates correctly, but the undervoltage release device causes the non-faulty feeder on the low-voltage side to trip falsely due to improper delay setting; ③ A fire occurs inside the ring main unit, the temperature sensor has alarmed, but the DTU does not send the signal to the master station due to communication interruption.

[0083] Specifically, in step 1, the device abnormal behavior parameter template of the primary-secondary joint state-space model of the distribution network includes at least three of the following types:

[0084] The correction formula for the constant drift model is as follows: ,in As a benchmark value, This is the drift coefficient;

[0085] The TA saturation model is used to correct distortion in quadratic sampled values. Its saturation criterion is: and Where I is the primary current, Where t is the saturation threshold and t is the duration. This is the saturation time threshold;

[0086] A communication interruption model is used to determine the online status and interruption duration of a computing device; its online probability function is: , where λ is the failure rate and t is time;

[0087] The mechanism cascade model is used to simulate the failure probability and delay time of the actuator.

[0088] Specifically, in step 2, to address the technical shortcomings of existing VR training systems that pre-define distribution automation (DA) systems as "omniscient, omnipotent, and absolutely reliable" ideal observation tools, completely detached from the real-world challenges of "blind spots, lack of self-healing, and false alarms," ​​the multi-source information conflict comprehensive judgment module for distribution automation system simulation proposed in this invention specifically includes:

[0089] The monitoring blind spot model is a model that assumes that some power distribution nodes are not covered by automation terminals or that the terminals are offline. The main station interface displays the node as "communication interrupted" or "data invalid". Trainees must obtain the actual status of the node through on-site verification simulation commands in the VR environment.

[0090] The self-healing function missing model is a preset operating condition where feeder automation is not put into operation, self-healing function is not enabled, or master station remote control permission is not issued. The system will not automatically perform recovery operation, and trainees must actively apply for remote control permission or give on-site operation orders.

[0091] The information false alarm / distortion model has a built-in message false alarm engine. Based on a probability model or instructor preset, it randomly or deterministically generates at least one type of false information, such as switch position false alarm, protection message false alarm, telemetry data distortion, and message timing disorder.

[0092] Specifically, in step 2, the false alarm / distortion model further includes:

[0093] A switch position change false alarm sub-model is used to generate false alarms in the case of closing position reporting opening position, opening position reporting closing position, or random jitter.

[0094] The telemetry data distortion sub-model is used to generate distorted telemetry values, and its calculation formula is as follows: Where P is the actual telemetry value, The distortion coefficient has a value range of [-0.2, 0.2].

[0095] The message timing disorder sub-model is used to generate timing disorder by exchanging message timestamps;

[0096] The DTU offline and actual switch tripping composite scenario generation unit is used to generate information conflict states where the master station information deviates from the actual operating conditions in the scenario where DTU communication is interrupted but the switch has actually tripped.

[0097] The unit is equipped with a composite scenario generation unit for automatic false alarm and actual failure to operate, which is used to generate an information distortion state when the main station receives a success message but the device actually fails to operate.

[0098] Specifically, in step 3, to address the technical shortcomings of existing VR training systems such as single interactive roles, fragmented workflows, and superficial safety training, the multi-role command and collaborative interaction module proposed in this invention further includes:

[0099] The library contains a multi-role simulation object library, which includes two types of collaborative role simulation objects: AI dispatcher and AI team member. It also includes pre-made standard terminology templates and instruction logic trees that conform to the "State Grid Corporation of China Power Safety Work Regulations" and "Dispatch Operation Management Regulations". It covers at least four typical dispatch dialogue processes, including power outage and restoration application, instruction repetition, operation permission, task change, anomaly reporting, and operation termination.

[0100] The violation instruction generation engine allows AI dispatchers or AI team members to actively issue instructions that violate safety regulations or dispatch procedures during training. These instructions must include at least three types of violations, such as issuing power supply instructions without removing the grounding wire, requesting operation without a ticket, repeating errors without correction, misjudging safety distance, giving orders beyond one's authority, unlocking operations without a ticket, expired work tickets, and missing grounding wires.

[0101] The Command Decision and Correction Unit is where trainees enter as the person in charge of emergency repairs. They do not have direct access to equipment and must issue operation commands to team members via voice or command panel, interact with the dispatcher, and autonomously identify, judge, and execute at least one corrective action among refusing to execute, requesting correction, and suspending the operation when the AI ​​character issues illegal or erroneous commands.

[0102] The command and decision-making quantitative evaluation unit collects all voice interactions between trainees and AI dispatchers and AI team members, as well as the timing of command issuance. It performs real-time compliance verification based on the built-in safety rule library. The evaluation dimensions include at least three of the following: completeness of command recitation, rate of identification and correctness of response to illegal commands, completeness of the two-ticket execution process, and compliance of reporting / commanding behavior beyond the level in emergency situations.

[0103] The targeted training unit for violation instructions dynamically selects the violation type based on the student's historical violation recognition rate. When the student's violation recognition rate is below 60%, highly deceptive violation instructions are generated first; when the student's violation recognition rate is above 60%, highly concealed or complex violation instructions are generated first.

[0104] Specifically, in step 3, the violation instruction generation engine further includes:

[0105] The violation type library contains at least five types of violations, including grounding wire not disconnected and power supply, operation without a permit, repeated errors not corrected, misjudgment of safety distance, commanding beyond one's level, unlocking operation without a permit, work permit expired, and grounding wire missing. Each type of violation instruction is associated with the corresponding safety regulations and the correct response template.

[0106] The violation probability control unit, this unit is Calculate the probability of generating a violation command, where D is the system's preset difficulty level, ranging from 1 to 5;

[0107] The unit configures parameters for deceptive illegal commands. This unit controls the difficulty of identifying illegal commands by adjusting at least one parameter among the following: the ambiguity of the expression of the illegal command, the similarity to the normal command, and the concealment of the timing of issuance.

[0108] Specifically, in step 4, to address the technical blind spot of existing VR training systems that can only judge trainees' operations as "right" or "wrong" and cannot provide better handling strategies for comparative teaching, the intelligent decision reference module of this invention further includes:

[0109] The load importance classification and power outage loss assessment model has a built-in user load classification label library for real-time quantification of the cumulative power outage loss weights for different handling paths. The formula for calculating the cumulative power outage loss weights is as follows: ,in User level coefficient For node active load, The total power outage duration for the node;

[0110] The recovery path optimization engine constructs an objective function that uses at least two of the following as joint optimization objectives: shortest power outage time, minimum impact range, and highest power restoration rate. ,in Total power restoration time Weighting for power outage losses, This refers to the number of switch operations. Using the weighted coefficients, we iterate through possible operation sequences to generate a theoretically optimal fault isolation and power restoration reference scheme;

[0111] The immersive dual-track comparison and review unit is used to play back the trainee's actual handling process and the AI-generated optimal solution in parallel in a VR environment. When the difference between key indicators exceeds the limit, the operation node is highlighted and a quantitative comparison prompt is output.

[0112] The scheme deviation analysis unit calculates the order deviation score and locates the specific deviation node by detecting the longest common subsequence between the student's operation sequence and the optimal scheme operation sequence.

[0113] Specifically, in step 4, to address the technical shortcomings of existing VR training systems that only output total scores or simple deductions and cannot perform in-depth attribution analysis of the causes of fault phenomena and the decision-making logic of trainees, this invention proposes a post-event intelligent review and diagnostic report automatic generation module, which specifically includes:

[0114] The fault cause reasoning and operation evaluation report automatic generation module has a built-in three-layer attribution mapping chain of fault phenomenon-electrical quantity-device behavior. It can reverse map at least one of the surface phenomena observed by trainees in the VR environment, such as power outage, abnormal noise, smoke, protection action signal, and switch change signal, to the fault type of primary equipment, the action behavior of secondary device, and the logical state of automation system.

[0115] The multi-dimensional evaluation matrix generation unit evaluates trainees' entire operation based on at least four dimensions: operation sequence, safety compliance, command decision correctness, power restoration efficiency, and risk identification ability. Each deduction item is associated with a specific operation node and violation / error type, and specific clauses of safety regulations or dispatch procedures are cited as the basis for judgment.

[0116] The standardized emergency repair report generation unit automatically generates a standardized report after training, which includes basic fault information, fault cause analysis, trainee operation records, scores and evaluations, and comparative improvement suggestions. The report includes at least the fault occurrence time, equipment name, fault type, protection action information, electrical principle or device behavior logic explanation corresponding to the abnormal phenomenon, all command issuance and voice interaction records displayed on a timeline, scores and deduction details for each dimension, violation clause index, deviation analysis from the optimal solution, and targeted training suggestions.

[0117] Specifically, based on the aforementioned fault-device behavior decoupling mechanism, this invention further supports high-fidelity simulation of secondary cascading faults during the dynamic reconfiguration of distribution networks. Typical complex operating condition simulation examples include reclosing tripping and tie line impact cascading faults. Typical embodiments include:

[0118] Example 1: A line fault overlaps with a permanent fault, causing the protection to trip again.

[0119] The system simulates the entire dynamic process of an overhead line transient fault → correct protection tripping → reclosing initiation → reclosing onto a permanent fault → accelerated protection tripping. Trainees need to determine, in the VR environment, whether the fault nature has changed from transient to permanent based on the reclosing action signal, fault current duration, and number of trips, and adjust the emergency repair strategy in a timely manner (from forcefully restoring power to requesting line patrol) to avoid blindly and repeatedly forcefully restoring power, which could lead to equipment damage or escalation of the situation.

[0120] Example 2: The tie switch switches to the faulty line, and the backup line is tripped by the impact:

[0121] The system simulates a scenario where a faulty line has been isolated. Trainees or dispatchers close the tie switch, attempting to transfer the load from an adjacent feeder to the non-faulty section. However, due to incomplete fault isolation or unidentified hidden grounding points in the faulty section, the overcurrent protection of the backup line trips the moment the transfer switch is closed, causing a complex interlocking situation of secondary power outage in the non-faulty area and failure of the transfer. In this example, the system can be configured to detect multiple causes such as excessively high sensitivity of the backup line's protection settings or fluctuations in the grounding resistance of the hidden grounding points in the faulty section. Trainees are required to quickly trace the source after the transfer failure: determine whether the fault isolation is incomplete, the backup line settings are improperly coordinated, or the tie switch is closed in an abnormal state, and accordingly develop a new segmented troubleshooting and recovery plan.

[0122] Specifically, based on the simulation model of the "three non-ideal characteristics" of power distribution automation systems, the system can generate complex and confusing scenarios where the information from the automation master station deviates significantly from or even contradicts the actual on-site operating conditions. Typical examples are as follows:

[0123] Example 3: DTU offline + actual circuit breaker trip, no information from the master station.

[0124] A permanent fault occurred on the line, and the protection tripped correctly. However, due to a communication interruption with the DTU in that bay, the distribution automation master station still displayed the switch as closed and the current as zero (the sampler is on the switching power supply side). Trainees faced with this conflicting information—a user reporting a power outage while the master station showed the equipment as operational—need to make a comprehensive judgment: was it a user-side switch trip, a blown TV fuse, or a DTU offline with the actual switch already tripped? They then needed to decide whether to order an on-site verification of the equipment's actual status.

[0125] Example 4: False alarm "backup self-deployment successful", but actual backup deployment fails to activate.

[0126] The master station receives a successful automatic transfer switch (ATS) notification, indicating that the backup power switch is closed and the lost-power bus has been re-energized. However, the ATS may actually fail to operate due to incomplete charging, and the notification may be a false alarm caused by the device or communication interference. If trainees blindly trust the master station's information, they may misjudge the power restoration and prematurely evacuate the site. The system requires trainees to cross-reference information from multiple sources, such as load curve changes, on-site personnel reports, and local indicator lights on protection devices, to identify distorted automation information and take corrective measures.

[0127] The above examples elevate the training scope from single-fault handling to multi-fault coupling, dynamic evolution, and secondary accident prevention, closely approximating the emergency repair scenarios under extremely complex operating conditions of the power distribution network.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power distribution network operation control system based on VR command and training, characterized in that: include: The multi-source fault hybrid dynamic generation module has a built-in primary equipment electrical quantity model and secondary device abnormal behavior parameter library. After generating a primary fault event, it inputs the fault characteristic quantity into the secondary device virtual behavior model, independently calculates the action behavior and timing of each protection and automation device, and outputs a nondeterministic combination response containing at least two of the following: correct action, maloperation, failure to operate, and logical disorder. The multi-source information conflict comprehensive judgment module under the simulation of power distribution automation system has at least one of the following built-in: monitoring blind zone model, self-healing function failure model, and information false alarm / distortion model. It is used to present the main station information interface to the trainee that is not completely mapped to the real fault state. The main station information interface contains at least one of the following non-real information: communication interruption, invalid data, switch change false alarm, protection message false alarm, and telemetry data distortion. The multi-role command and collaborative interaction module has built-in AI dispatcher and AI team member simulation objects, and a pre-set violation instruction generation engine, which is used to actively issue instructions that violate power safety work procedures during training, allowing trainees to perform instruction recognition, safety regulation verification, and correction operations as the person in charge of emergency repairs. The intelligent decision reference module generates the theoretically optimal handling plan based on load importance classification and transfer capacity optimization. It also provides an immersive dual-track parallel playback of the trainees' actual operation, highlighting key deviation nodes and outputting quantitative comparison prompts. The post-event intelligent review and diagnostic report automatic generation module is used to establish a three-layer attribution mapping chain of fault phenomenon-electrical quantity-device behavior, generate a multi-dimensional evaluation matrix of the trainee's operation steps, and automatically synthesize emergency repair reports.

2. The distribution network operation control system based on VR command training according to claim 1, characterized in that: The multi-source fault hybrid dynamic generation module is specifically configured with: The primary-secondary joint state space model includes the electrical quantity status of primary equipment, the setting zone status of secondary devices, the activation / deactivation status of pressure plates, the communication status, and preset templates for abnormal device behavior parameters. The fault-action decoupling generation engine is used to input fault feature quantities into the virtual behavior model of the secondary device, and combine the abnormal behavior parameter template to independently calculate the action behavior and timing of each device. The malfunction probability weight adjustment interface is used by instructors or AI training strategy modules to dynamically adjust the probability factors of various devices malfunctioning, refusing to operate, or experiencing logical disorder. The nondeterministic simulation unit supports mixed outputs of correct / false operation / failure to operate. It is used for multiple devices to present at least two mixed outputs among correct operation, false operation, failure to operate, and logical disorder when a single fault event triggers them. It is also used for simulating at least one of the following scenarios: reclosing due to a permanent fault triggers secondary tripping, the tie switch switching to a faulty line triggers overcurrent protection tripping of the backup line, and the automatic transfer switch operating on a faulty bus causes cascading tripping. It is also used for simulating at least one of the following scenarios: multi-device timing coupling and fault dynamic evolution chain reaction.

3. A distribution network operation control system based on VR command and training according to claim 2, characterized in that: The pre-defined abnormal behavior parameter templates for the device in the first- and second-order joint state-space model include at least three of the following types: The constant drift model has a correction formula as follows: ,in As a benchmark value, This is the drift coefficient; The TA saturation model is used to correct distortion in secondary sampled values. The saturation criterion is... and Where I is the primary current, Where t is the saturation threshold and t is the duration. This is the saturation time threshold; A communication interruption model is used to determine the online status and interruption duration of a computing device. The online probability function is: , where λ is the failure rate and t is time; The mechanism cascade model is used to simulate the failure probability and delay time of the actuator.

4. A distribution network operation control system based on VR command and training according to claim 3, characterized in that: The multi-source information conflict comprehensive analysis module under the power distribution automation system simulation also includes: The monitoring blind spot model assumes that some power distribution nodes are not covered by automation terminals or are offline. The main station interface displays the node as "communication interrupted" or "data invalid". Trainees are required to obtain the actual status of the node through on-site verification simulation commands in the VR environment. The self-healing function is missing model. In the case of at least one of the following operating conditions: feeder automation is not put into operation, self-healing function is not enabled, or master station remote control permission has not been issued, the system will not automatically perform recovery operation and the trainee is required to actively apply for remote control permission or give an order to operate on site. The information false alarm / distortion model has a built-in message false alarm engine. Based on a probability model or instructor preset, it randomly or deterministically generates at least one type of false information, such as switch position false alarm, protection message false alarm, telemetry data distortion, and message timing disorder.

5. A distribution network operation control system based on VR command and training according to claim 4, characterized in that: The information false alarm / distortion model is specifically configured with the following: A switch position change false alarm sub-model is used to generate false alarms in the case of closing position reporting opening position, opening position reporting closing position, or random jitter. The telemetry data distortion sub-model is used to generate distorted telemetry values, and the calculation formula is as follows: Where P is the actual telemetry value, The distortion coefficient has a value range of [-0.2, 0.2]. The message timing disorder sub-model is used to generate timing disorder by exchanging message timestamps; The DTU offline and actual switch tripping composite scenario generation unit is used to generate information conflict states where the master station information deviates from the actual operating conditions in the scenario where DTU communication is interrupted but the switch has actually tripped. The unit is equipped with a composite scenario generation unit for automatic false alarm and actual failure to operate, which is used to generate an information distortion state when the main station receives a success message but the device actually fails to operate.

6. A distribution network operation control system based on VR command training according to claim 5, characterized in that: The multi-role command and collaborative interaction module is specifically configured with the following: The multi-role simulation object library contains two types of collaborative role simulation objects: AI dispatcher and AI team member. It also includes pre-made standard terminology templates and instruction logic trees that conform to regulations, including at least four typical dispatch dialogue processes such as power outage and restoration application, instruction repetition, operation permission, task change, anomaly reporting, and work termination. The violation instruction generation engine allows AI dispatchers or AI team members to actively issue instructions that violate safety regulations or dispatch procedures during training. These instructions must include at least three types of violations, such as issuing power supply instructions without removing the grounding wire, requesting operation without a ticket, repeating errors without correction, misjudging safety distance, giving orders beyond one's authority, unlocking operations without a ticket, expired work tickets, and missing grounding wires. In the command, decision-making and correction unit, trainees enter as the person in charge of emergency repairs. They do not have direct authority to operate equipment. They are required to issue operation commands to team members through voice or command panel, interact with dispatchers, and autonomously identify, judge and execute at least one corrective action among refusing to execute, requesting correction, and suspending operation when the AI ​​character issues illegal or erroneous commands. The command and decision-making quantitative evaluation unit collects all voice interaction content between trainees and AI dispatchers and AI team members, as well as the timing of instruction issuance. Based on the built-in safety rule library, it performs real-time compliance verification. The evaluation dimensions include at least three of the following: instruction recitation completeness rate, violation instruction recognition rate and response accuracy rate, completeness of the two-ticket execution process, and compliance of overstepping reporting / overstepping command behavior in emergency situations. The targeted training unit for violation instructions dynamically selects the violation type based on the student's historical violation recognition rate. When the student's violation recognition rate is below 60%, highly deceptive violation instructions are generated first, and when the student's violation recognition rate is above 60%, highly concealed or complex violation instructions are generated first.

7. A distribution network operation control system based on VR command training according to claim 6, characterized in that: The specific settings in the violation instruction generation engine are as follows: The violation type library should include at least five types of violations, such as grounding wire not being disconnected and power being supplied, operation without a permit, repeated errors not being corrected, misjudgment of safety distance, commanding beyond one's level, unlocking operation without a permit, work permit expired, and grounding wire missing. Each violation instruction should be associated with the corresponding safety regulations and the correct response template. The violation probability control unit uses a calculation formula. Calculate the probability of generating a violation instruction, where D is the system's preset difficulty level, ranging from 1 to 5; The deceptive parameter configuration unit for illegal commands controls the difficulty of identifying illegal commands by adjusting at least one parameter among the following: the ambiguity of the illegal command's expression, its similarity to normal commands, and the concealment of the timing of its issuance.

8. A distribution network operation control system based on VR command training according to claim 7, characterized in that: The intelligent decision reference module is specifically configured with the following: The load importance classification and power outage loss assessment model has a built-in user load classification label library for real-time quantification of the cumulative power outage loss weights for different handling paths. The formula for calculating the cumulative power outage loss weights is as follows: ,in User level coefficient For node active load, The total power outage duration for the node; The recovery path optimization engine constructs an objective function with at least two of the following as joint optimization objectives: shortest power outage time, minimum impact range, and highest power restoration rate. ,in Total power restoration time Weighting for power outage losses, This refers to the number of switch operations. Using the weighted coefficients, we iterate through possible operation sequences to generate a theoretically optimal fault isolation and power restoration reference scheme; The immersive dual-track comparison and review unit is used to play back the trainee's actual handling process and the AI-generated optimal solution in parallel in a VR environment. When the difference between key indicators exceeds the limit, the operation node is highlighted and a quantitative comparison prompt is output. The scheme deviation analysis unit calculates the order deviation score and locates the specific deviation node by detecting the longest common subsequence between the student's operation sequence and the optimal scheme operation sequence.

9. A distribution network operation control system based on VR command and training according to claim 8, characterized in that: The specific settings in the post-event intelligent review and diagnostic report automatic generation module are as follows: The fault cause reasoning and operation evaluation report automatic generation module has a built-in three-layer attribution mapping chain of fault phenomenon-electrical quantity-device behavior. It reverse maps at least one of the surface phenomena observed by trainees in the VR environment, such as power outage, abnormal noise, smoke, protection action signal, and switch change signal, to the fault type of primary equipment, the action behavior of secondary device, and the logic state of automation system. The multi-dimensional evaluation matrix generation unit evaluates trainees' entire operation in a granular and quantitative manner based on at least four dimensions, including operation sequence, safety compliance, command decision correctness, power restoration efficiency, and risk identification ability. Each deduction item is associated with a specific operation node and violation / error type, and specific clauses of safety regulations or dispatch procedures are cited as the basis for judgment. The standardized emergency repair report generation unit automatically generates a standardized report after training, which includes basic fault information, fault cause analysis, trainee operation records, scores and evaluations, and comparative improvement suggestions. The standardized report includes at least the fault occurrence time, equipment name, fault type, protection action information, electrical principle or device behavior logic explanation corresponding to the abnormal phenomenon, all command issuance and voice interaction records displayed on a timeline, score and deduction details for each dimension, violation clause index, deviation analysis from the optimal solution, and targeted training suggestions.

10. The method for interactive control of a distribution network operation control system based on VR command training as described in claim 9, characterized in that: The interactive control steps include the following: Step 1: Generate multi-source fault and cascading fault scenarios containing nondeterministic combined responses through the multi-source fault hybrid dynamic generation module; Step 2: Through the multi-source information conflict comprehensive analysis module, the main station information interface is presented to the trainees as incomplete and not absolutely reliable, and the trainees perform multi-source information conflict comprehensive analysis. Step 3: Through the multi-role command and collaborative interaction module, the AI ​​role and the trainee conduct scheduling dialogue interaction. The AI ​​role actively issues violation instructions, and the trainee executes instruction recognition, safety regulation verification and correction operations. Step 4: Through the intelligent decision reference module, the theoretically optimal solution is generated, and after the trainee completes the solution, the trainee's actual operation and the optimal solution are played back in an immersive dual-track parallel process, and quantitative comparison prompts are output. Step 5: Through the post-event intelligent review module, the trainees' operations are quantitatively scored from multiple dimensions, and a standardized emergency repair report is automatically generated.

Images