A high-voltage switch cabinet robot operation ticket execution sequence checking and optimization recommendation method

Abstract

The application discloses a kind of high-voltage switchgear robot operation ticket execution sequence verification and optimization recommendation method.The method first establishes modular single-action library, and the paper operation ticket step input by user is converted into initial robot execution sequence;Subsequently, the execution sequence is simulated, and the sequence state chain containing cabinet door state, switch state, chassis position, mechanism occupation state and verification state is formed;Then, according to the safety constraint set, check the missing pre-action, cabinet door state error, verification action missing, abnormal processing missing, action sequence conflict and space interference risk;At the same time, based on the risk scoring model, the single-action risk and adjacent action transfer risk are calculated, and the candidate repair sequence is generated according to the error type, and finally the robot execution sequence with low risk, low cost and less interference is recommended through multi-objective cost function, so as to improve the operation ticket arrangement efficiency and robot execution safety.

Description

Technical Field

[0001] This invention relates to the field of power system robot control and digital execution of operation tickets, and in particular to a method for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots. Background Technology

[0002] High-voltage switchgear operation tickets typically consist of multiple paper steps, but when executed by a robot, each paper step often needs to be broken down into multiple individual actions. For example, opening and closing an air switch may involve actions such as opening the cabinet door, positioning the chassis, moving the robotic arm, visual verification, retrying after failure, and manual confirmation. For the same or similar switchgear models, multiple operation tickets may contain a large number of repeated action modules, such as opening and closing the cabinet door, moving the trolley in and out, plugging and unplugging secondary connectors, opening and closing SM10 and SM40 air switches, verifying the status of the grounding switch, positioning the moving chassis, and visual photo verification.

[0003] Current methods largely rely on manual creation of robot operation ticket programs based on paper operation tickets, which is prone to problems such as missing preliminary actions, improper timing of cabinet door closing, missing post-verification, incomplete abnormal branches, unreasonable action sequence, and the risk of institutional interference. To improve the deployment efficiency and execution safety of new operation tickets, it is necessary to automatically verify, alarm, and optimize the generated action sequence before the robot operation ticket is executed.

[0004] To address the aforementioned issues, this application proposes a method for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots. This method can complete error alarms and optimization recommendations before the robot operation tickets are executed, thereby improving the efficiency of operation ticket arrangement and the safety of robot execution.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots includes the following steps:

[0008] S1. Establish a modular single-action library. The action modules in the modular single-action library correspond to the single actions that the robot can execute on the high-voltage switchgear, including cabinet door opening and closing, circuit breaker trolley movement, air switch opening and closing, knob switching, visual verification, mobile chassis positioning, and docking platform actions. The action modules in the modular single-action library shall at least include task number, action name, action category, applicable object, executing subject, preconditions, post-execution status, working area, resource consumption, verification requirements, exception handling rules, and action cost.

[0009] S2. Obtain the high-voltage switchgear operation ticket steps input by the user, and generate an initial robot execution sequence based on the modular single-action library;

[0010] S3. Perform state simulation on the initial robot execution sequence, and recursively generate a sequence state chain according to the preconditions and post-execution states of the action module.

[0011] S4. Perform hard constraint verification on the sequence state chain based on the set of security constraints, identify arrangement errors in the execution sequence and output error alarm information;

[0012] S5. Calculate the risk value of each action module and adjacent action transfer in the initial robot execution sequence based on the risk scoring model, and generate candidate repair sequences in combination with the programming error;

[0013] S6. The candidate repair sequence is optimized and searched using a multi-objective cost function to obtain a recommended robot execution sequence. The original operation ticket steps, initial robot execution sequence, error alarm information and recommended adjustment scheme are displayed on the measurement and control software. The recommended robot execution sequence is then solidified into a program file that can be sent to the high-voltage switchgear robot controller for execution.

[0014] Preferably, in S3, the state simulation is performed in the following manner:

[0015] Let the first The system state before the step execution is as follows: , No. Each action module is Action module The set of preconditions is The set of states after execution is Then when At that time, the state is recursively calculated as follows:

[0016] ;

[0017] in, This is the set of states that are changed or deleted after the action module is executed;

[0018] when At that time, the judgment of the first A certain action module has a missing prerequisite error, and the set of missing states is denoted as:

[0019]

[0020] The system generates error alarm information and candidate repair actions based on the set of missing states.

[0021] Preferably, in S4, the system performs hard constraint verification on the sequence state chain based on the structured attributes of the action module. The structured attributes include preconditions, post-execution state, work area, resource usage, verification requirements, and exception handling rules.

[0022] For precondition constraints, the system will use action modules. Precondition set With the System state set before step execution Compare; if If the action is determined to have a missing preceding action error, then the missing state set will be determined. As a basis for error localization.

[0023] For cabinet door constraints, the system reads the corresponding cabinet door status according to the working area of ​​the action module; when the action object is located in the low-voltage room, circuit breaker room or grounding switch operating area, and the corresponding cabinet door status does not meet the action execution requirements, it is determined that there is a cabinet door status error.

[0024] For status verification constraints, the system determines whether there is a corresponding verification action after the action that changes the status of the equipment. When the knob is switched, the air switch is opened or closed, the grounding switch is opened or closed, or the position of the circuit breaker trolley changes, if no visual acquisition action or status verification action corresponding to the operation object and target status is found within the preset step size range, it is determined that there is a status verification missing error.

[0025] For abnormal branch constraints, the system determines whether an action with verification requirements has at least one processing branch after verification failure, namely re-execution, re-verification, suspension of execution, or manual confirmation; if not, it is determined that there is an error of missing abnormal processing.

[0026] For mutual exclusion constraints and spatial interference constraints of the actuators, the system determines whether there are conflicts between adjacent action modules in terms of the executing entity, mobile chassis position, working area and mechanism resource occupancy status; if there are conflicts, it determines that there is a conflict in the action sequence or a risk of actuator interference.

[0027] The system outputs error alarm information based on the above verification results and passes the error type, error location, missing conditions and candidate repair actions to the candidate repair sequence generation step.

[0028] Preferably, in S5, the risk scoring model includes a rule-based risk matrix and a data-driven risk prediction model;

[0029] The data-driven risk prediction model uses action feature vectors For input, output the first... Risk value of each action module Among them, the action feature vector It should include at least the action type, work area, cabinet door status, actuator occupancy status, target status, whether verification is required, historical failure count, retry count, and safety procedure level;

[0030] When using a logistic regression model, the risk value is calculated according to the following formula:

[0031] ;

[0032] in, For the Sigmoid function, This is the weight matrix. For bias terms;

[0033] When using a multilayer perceptron model, the risk value is calculated according to the following formula:

[0034]

[0035] The system outputs the action risk level and action transfer risk level based on the risk value.

[0036] Preferably, in step S5, the system generates a candidate repair sequence based on the mapping relationship between error type and repair rule.

[0037] When the error type is missing preceding action, the system reads the missing state set; if the missing state set includes cabinet door open state, the corresponding cabinet door open action is called according to the working area of ​​the target action and inserted before the target action.

[0038] When the error type is "inappropriate timing of cabinet door closing", the system starts from the position of the current cabinet door closing action and searches for related actions in the same cabinet area along the initial robot execution sequence. If the working area of ​​the subsequent action still belongs to the cabinet area, or its preconditions still include the cabinet door being open, the original cabinet door closing action is deleted and postponed to after the last related action and its status verification action in the same cabinet area.

[0039] When the error type is missing cabinet door closing action, the system determines the last relevant action in the current cabinet area; if subsequent actions after this action no longer require the cabinet door to remain open, the corresponding cabinet door closing action is inserted after the last relevant action and its status verification action.

[0040] When the error type is missing status verification, the system matches the corresponding visual acquisition action or status verification action according to the target action's operation object and target status, and inserts it after the target action.

[0041] When the error type is missing exception handling, the system supplements the branch by re-execution, re-verification, suspension of execution, or manual confirmation, based on the failure result of the verification action and the risk level of the action.

[0042] When the error type is action sequence conflict or actuator interference risk, the system generates candidate repair sequences for action rearrangement, duplicate positioning merging, or actuator switching reduction based on the executing entity, mobile chassis location, working area, and resource occupancy status.

[0043] Preferably, in S6, the multi-objective cost function is:

[0044] ;

[0045] in, Candidate repair sequences, The cost of action execution time. This is the risk value. The cost of movement or positioning between adjacent actions. Due to the cost of switching of implementing agencies or resource occupation, The cost of manual confirmation, For hard constraints, there are penalties for violations. These are the weighting coefficients;

[0046] The system is minimized To achieve this, a recommended robot execution sequence is selected from multiple candidate repair sequences.

[0047] This invention also provides a high-voltage switchgear robot operation ticket execution sequence verification and optimization recommendation system, comprising:

[0048] A modular action library storage module is used to store the action modules described in claim 1;

[0049] The initial sequence generation module is used to map the steps on the paper operation ticket to the initial robot execution sequence;

[0050] The state simulation and verification module is used to recursively generate the sequence state chain and output the arrangement error based on the set of safety constraints;

[0051] The risk scoring module is used to calculate the action risk value and the transfer risk value;

[0052] The candidate repair generation module is used to generate candidate repair sequences based on orchestration errors and risk scores;

[0053] A multi-objective optimization module is used to output a recommended robot execution sequence with a minimum cost function J(π).

[0054] The output module is used to display alarm information and recommended sequences on the monitoring and control software, and to generate program files that can be downloaded and executed.

[0055] By adopting the above technical solution, including: establishing a modular single-action library, defining actions such as cabinet door opening and closing, circuit breaker trolley movement, air switch opening and closing, knob switching, visual verification, mobile chassis positioning, and docking platform as action modules with task numbers, preconditions, post-execution states, work areas, resource occupancy, and abnormal branches; generating an initial robot execution sequence based on the steps of the paper operation ticket input by the user; performing state simulation on the initial robot execution sequence to form a sequence state chain including cabinet door state, opening and closing state, chassis position, mechanism occupancy state, and verification state; checking for missing preconditions, incorrect cabinet door states, missing verification actions, missing abnormal handling, conflicting action sequences, and spatial interference risks based on the safety constraint set; calculating action risks and adjacent action transfer risks based on a risk scoring model; generating candidate repair sequences based on error types, and recommending robot execution sequences with low risk, low cost, and minimal interference through a multi-objective cost function.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] This invention can detect programming errors such as missing pre-actions, incorrect cabinet door status, missing verification, and missing abnormal branches before the operation ticket is executed; it can predict action risks and transfer risks; it can automatically generate repair solutions and recommend better execution sequences based on error types; it can reduce the workload of manually compiling and modifying operation ticket programs, and improve the safety, stability, and deployment efficiency of high-voltage switchgear robot operation ticket execution. Attached Figure Description

[0058] Figure 1 This is a flowchart of the present invention;

[0059] Figure 2 This is a schematic diagram of the sequence verification, risk scoring, and optimized recommendation model in this invention;

[0060] Figure 3 This is a flowchart of the state simulation and precondition missing judgment process of the present invention;

[0061] Figure 4 This is a data flow diagram for candidate repair generation and multi-objective optimization in this invention. Detailed Implementation

[0062] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings, so that those skilled in the art can better understand the advantages and features of the present invention, thereby making a clearer definition of the scope of protection of the present invention. The embodiments described in this invention are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0063] A method for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots includes the following steps:

[0064] Step 1: Establish a modular single-action library.

[0065] Based on multiple paper operation tickets for high-voltage switchgear and verified robot execution programs, recurring robot actions are abstracted into modular single actions, and each single action is assigned a unique task number. The action modules include 1xx dual-arm collaborative actions, 2xx second robotic arm actions, 3xx first robotic arm actions, 4xx vision acquisition or status verification actions, 5xx mobile chassis actions, and 6xx docking platform actions.

[0066] Each action module records the task number, action name, preconditions, post-execution status, work area, resource usage, verification requirements, exception handling rules, and action cost.

[0067] Step 2: Generate the initial robot execution sequence.

[0068] Users create a new operation ticket and enter the paper-based steps in the system. Based on the cabinet type, cabinet number, paper-based steps, and modular single-action library, the system generates an initial robot execution sequence. This sequence serves as the basis for subsequent verification and optimization.

[0069] Step 3: Perform sequence state simulation.

[0070] The system recursively builds the state chain according to the order of actions in the initial robot execution sequence. Let the first state be... The state before step execution is The action module is The prerequisite is The status after execution is The deletion status is When satisfied hour:

[0071] ;

[0072] When the condition is not met, the system records the missing status. It also generates an alarm for missing prerequisites.

[0073] The state chain includes at least the cabinet door status, open / closed status, handcart position, ground knife status, chassis position, robotic arm occupancy status, docking platform occupancy status, verification status, and anomaly handling status.

[0074] Step 4: Perform hard constraint verification.

[0075] The system performs step-by-step verification of the state chain based on a set of safety constraints. This set of safety constraints includes precondition constraints, cabinet door status constraints, continuous operation constraints for the area, status verification constraints, abnormal branch constraints, mutual exclusion constraints for actuators, and spatial interference constraints.

[0076] For precondition constraints, the system reads the first... Action Module Precondition set and with the System state set before step execution Compare them. If satisfied... If the condition is met, the action is ready to be executed; otherwise, it is determined that the action has a missing prerequisite action error, and the missing state set is recorded.

[0077] ;

[0078] The system outputs specific alarm information based on the missing state set. For example, if the precondition for the operation of a low-voltage room air switch includes "low-voltage room cabinet door is open", but the corresponding cabinet door state in the state chain is closed or unknown, the system outputs an alarm "low-voltage room cabinet door precondition missing" and recommends inserting a low-voltage room cabinet door opening action before this action.

[0079] Regarding cabinet door status constraints, the system determines whether the corresponding cabinet door status meets the action execution requirements based on the operating area of ​​the action module. When the action object is located in the low-voltage room, circuit breaker room, or grounding switch operating area, and the corresponding cabinet door is not in the permitted state, the system determines that there is a cabinet door status error. When subsequent consecutive actions still belong to the same cabinet room area, and the corresponding cabinet door closing action has been inserted after the current action, the system determines that there is an inappropriate timing for cabinet door closing and recommends that the cabinet door closing action be postponed until after the last related action and its verification action in that area.

[0080] For status verification constraints, the system determines whether a corresponding verification action is set after an action that changes the equipment status. When the action module involves knob switching, air switch opening / closing, grounding switch opening / closing, or circuit breaker trolley position change, the system searches for visual acquisition actions or status verification actions corresponding to the operated object and target status within a preset step size range after the action. If no corresponding verification action is found, an "Status verification action missing" alarm is output, and it is recommended to insert the corresponding visual acquisition or status verification action.

[0081] For abnormal branch constraints, the system determines whether the action with verification requirements has failure handling logic. If at least one of the following processing branches—re-execution, re-verification, suspension of execution, or manual confirmation—is not configured after the verification action, the system outputs an "Abnormal handling branch missing" alarm and recommends adding the corresponding retry or manual confirmation logic.

[0082] For mutual exclusion constraints and spatial interference constraints of actuators, the system checks the executing entities, mobile chassis positions, working areas, and mechanism resource occupancy status of adjacent action modules. When there is a conflict between adjacent actions, such as a conflict in robotic arm occupancy, a failure to meet mobile chassis position requirements, a failure to meet docking platform status requirements, or a risk of motion space interference, the system outputs an alarm for "action sequence conflict" or "actuator interference risk" and includes the action segment in the candidate optimization list.

[0083] Through the above verification, the system can determine the error location, error type, missing conditions, and candidate repair actions for each programming error before the robot operation ticket is executed, providing input for subsequent risk scoring and optimization recommendations.

[0084] Step 5: Conduct a risk assessment.

[0085] System construction action risk feature vector It includes action type, work area, cabinet door status, chassis location, actuator occupancy, target status, whether verification is required, historical failure count, number of retries, and safety procedure level.

[0086] Risk scoring employs a combination of rule-based risk matrices and data-driven models.

[0087] When using a logistic regression model:

[0088] ;

[0089] When using a multilayer perceptron model:

[0090] ;

[0091] The system according to The system outputs a low-risk, medium-risk, or high-risk level. For adjacent actions, the system calculates the action transition risk by combining spatial interference, resource conflicts, and state conflicts, and includes segments that exceed the threshold in the optimization list.

[0092] Step 6: Generate candidate repair sequences.

[0093] The system generates candidate repair sequences by calling the corresponding repair rules based on the error type, error location, and missing state set output by the hard constraint verification.

[0094] When the error type is "missing preceding action", the system reads the set of missing states. .like If the information includes cabinet door opening statuses such as "Low-voltage compartment cabinet door is open," "Circuit breaker compartment cabinet door is open," or "Grounding switch operating area cabinet door is open," then the target action will be determined accordingly. In the work area, retrieve the corresponding cabinet door opening action from the modular single-action library and insert the cabinet door opening action into the module. Previously, candidate repair sequences were formed.

[0095] When the error type is "inappropriate cabinet door closing timing," the system locates the current cabinet door closing action. and from The search continues from the next action. If the subsequent action... work area and The corresponding cabinet area is consistent, or If the system still shows the cabinet door as open, it determines that the cabinet door closed too early. The system deletes the entry from the original location. Continue searching until you find the last action that still requires the cabinet door to be open. Then judge Does a corresponding status verification action exist afterward? If so, then... Insert after the status verification action; if it does not exist, then... Insert to after.

[0096] When the error type is "cabinet door closing action missing," the system starts from the last related action in the current work area and checks the work areas and preconditions of subsequent actions. If none of the subsequent actions require the cabinet door to remain open, the corresponding cabinet door closing action is inserted after the last related action and its status verification action in the work area.

[0097] When the error type is "state verification missing," the system matches the corresponding visual acquisition action or state verification action from the modular single-action library based on the target action's operation object, target state, and verification requirements, and inserts it after the target action. For example, after an air switch opening / closing action, a corresponding air switch state verification action is inserted; after a knob switching action, a corresponding knob position verification action is inserted; and after a circuit breaker truck position change, a corresponding position indication verification action is inserted.

[0098] When the error type is "missing exception handling," the system supplements at least one exception handling branch for the action based on the failure result of the verification action and the risk level of the action. For low-risk actions, candidate repair branches include re-execution and re-verification; for medium-risk actions, candidate repair branches include suspending execution and manual confirmation; for high-risk actions, candidate repair branches include immediately suspending execution and prompting manual handling.

[0099] When the error type is an action sequence conflict or actuator interference risk, the system generates candidate solutions for action rearrangement, duplicate positioning merging, or actuator switching reduction based on the executing entity of adjacent actions, mobile chassis location, work area, and resource occupancy status. If the adjusted solution still violates the hard safety constraints, it will not be included in subsequent optimization recommendations.

[0100] In this way, the system establishes a one-to-one correspondence between error types and repair rules, and inserts, deletes or moves each candidate repair action to a specified position in the initial robot execution sequence to form one or more candidate repair sequences.

[0101] Step 7: Perform multi-objective optimization recommendations.

[0102] The system constructs candidate execution graphs from candidate repair sequences. Among them, nodes This indicates the system state after a certain action has been performed. The system state includes cabinet door status, chassis position, actuator occupancy status, verification status, and abnormal branch status; (Side) This indicates the action module or candidate repair action that can continue to be executed from the current state. The system configures a comprehensive cost for each edge, which consists of action execution time, action risk, movement or positioning cost, actuator switching cost, manual confirmation cost, and hard constraint violation penalty.

[0103] For any candidate repair sequence The system calculates the overall cost using the following formula:

[0104] ;

[0105] in, For the time cost of the action, This is the risk value. The cost of movement or positioning between adjacent actions. Due to the cost of switching of implementing agencies or resource occupation, The cost of manual confirmation, For hard constraints, there are penalties for violations. Let be the weighting coefficient. If the candidate sequence violates the hard safety constraint, then let . Apply a high penalty value to prevent the candidate sequence from being recommended.

[0106] In one implementation, the system employs the A* search method to search for candidate execution graphs. The A* search method is a heuristic cost search method where search nodes represent partially generated robot action sequences and their corresponding states, and search edges represent addable action modules or candidate repair actions. The system assigns a value to each search node. Calculate the evaluation function:

[0107]

[0108] in, This represents the total cost accumulated from the initial sequence state to the current node. This represents the estimated cost of the remaining actions from the current node to completing the operation ticket. The system selects nodes from the set to be expanded each time. The smallest node continues to expand until a candidate execution sequence that satisfies all hard safety constraints and has the minimum overall cost is obtained. The remaining estimated cost... It can be estimated based on the sum of the minimum action time, the minimum risk value, and the minimum movement cost of the remaining unexecuted actions.

[0109] To illustrate how the aforementioned cost function is implemented in the robot control program, a specific numerical example is given below. Assume that for the step of "disconnecting the SM10 air switch," the system generates two candidate repair sequences:

[0110] Candidate sequences Open the low-voltage compartment cabinet door, disconnect SM10, perform SM10 visual verification, and immediately close the low-voltage compartment cabinet door;

[0111] Candidate sequences Open the low-pressure compartment cabinet door, disconnect SM10, perform SM10 visual verification, continue to perform subsequent SM40 related actions and verifications, and then close the low-pressure compartment cabinet door.

[0112] Let the weighting coefficient be , , , , , Among them, the cost of time The risk value is obtained by normalizing the estimated time of the action. The cost of moving or relocating, output by the risk scoring model. The resource switching cost is obtained by normalizing the change in chassis or robotic arm position. The cost is obtained by normalizing the number of execution entity switching times and manually confirming it. The penalty for violating hard constraints is obtained by normalizing the number of manual confirmations. The value is 0 when there is no violation, and 100 when there is a violation of a hard safety constraint.

[0113] For candidate sequences Because the low-voltage compartment door was closed immediately after SM10 verification, but subsequent SM40 low-voltage compartment operations require the door to be reopened, resulting in redundant door opening and closing and action coordination. Let the normalized cost corresponding to this sequence be: action time cost. Risk value Cost of moving or relocating Resource switching cost Manual confirmation of cost Hard constraints and penalties for violations .but:

[0114] ;

[0115] For candidate sequences Because it delays the closing action of the low-pressure chamber cabinet door until all related actions in the low-pressure chamber are completed, it reduces repeated door opening and closing and mechanism switching. Let the normalized cost corresponding to this sequence be: action time cost. Risk value Cost of moving or relocating Resource switching cost Manual confirmation of cost Hard constraints and penalties for violations .but:

[0116] ;

[0117] because The system recommends candidate sequences. The closing action of the low-pressure compartment cabinet door will be postponed until all low-pressure compartment-related actions, such as SM10 and SM40, and their verification actions are completed. If a candidate sequence has a safety hard constraint violation, such as performing the SM10 disconnect action while the low-pressure compartment cabinet door is not open, then... The overall cost of this candidate sequence increases significantly, so it is not included in the recommendation results.

[0118] Through the above processing, the system can convert time, risk, movement, resource switching, manual confirmation, and hard constraint penalties in the formula into calculable numerical terms, and recommend robot execution sequences with low risk, few repeated positioning, few cabinet door opening and closing times, few resource switching, and satisfying safety hard constraints in the measurement and control software with the goal of minimizing the overall cost.

[0119] Step 8: Output alarms and suggestions.

[0120] The system displays the original paper steps, initial robot execution sequence, alarm locations, error types, risk levels, and recommended adjustment plans on the monitoring and control software. After operator confirmation, the system solidifies the recommended sequence into an executable robot operation ticket program.

[0121] A high-voltage switchgear robot operation ticket execution sequence verification and optimization recommendation system includes:

[0122] A modular action library storage module is used to store the action modules described in claim 1;

[0123] The initial sequence generation module is used to map the steps on the paper operation ticket to the initial robot execution sequence;

[0124] The state simulation and verification module is used to recursively generate the sequence state chain and output the arrangement error based on the set of safety constraints;

[0125] The risk scoring module is used to calculate the action risk value and the transfer risk value;

[0126] The candidate repair generation module is used to generate candidate repair sequences based on orchestration errors and risk scores;

[0127] A multi-objective optimization module is used to output a recommended robot execution sequence with a minimum cost function J(π).

[0128] The output module is used to display alarm information and recommended sequences on the monitoring and control software, and to generate program files that can be downloaded and executed.

[0129] Example:

[0130] like Figure 1 As shown in the figure, this embodiment provides a method for verifying and optimizing the execution sequence of a high-voltage switchgear robot operation ticket. After the user creates a new operation ticket in the measurement and control software and enters the steps of the paper operation ticket, the system first generates an initial robot execution sequence based on a modular single-action library; then, it performs state simulation on the execution sequence to obtain the sequence state chain; next, it performs hard constraint verification based on the safety constraint set and outputs the arrangement error; then, it calculates the action risk and the adjacent action transfer risk by combining the risk scoring model; finally, it generates candidate repair sequences according to the error type and recommends the adjusted robot execution sequence through a multi-objective optimization method. After confirmation by the operator, it generates a robot operation ticket program that can be issued and executed.

[0131] like Figure 2As shown, the system takes an initial sequence of actions as input and constructs a state space and a set of hard constraints. The state space includes cabinet door state, open / closed state, chassis position, actuator occupancy state, verification state, and abnormal branch state; the set of hard constraints includes precondition constraints, interlock constraints, cabinet door constraints, spatial interference constraints, verification constraints, and abnormal handling constraints. The system determines whether each action has the conditions for execution through state simulation and calculates the action risk and action transfer risk through a risk scoring model. When a missing precondition action, incorrect cabinet door state, missing verification, missing abnormal handling, sequence conflict, or interference risk is detected, the system transmits the corresponding error information to the candidate repair generation module.

[0132] like Figure 3 As shown, during the state simulation process, the system first inputs the initial robot execution sequence. and initialize the system state. The system status includes cabinet door status, switch status, circuit breaker trolley position, grounding switch status, mobile chassis position, robotic arm occupancy status, docking platform occupancy status, verification status, and abnormal branch status. The system reads the sequence of actions... Action Module And obtain the set of preconditions for the action module from the modular single action library. Post-execution state set and the set of deletion states .

[0133] When satisfied When the system determines that the action meets the execution conditions, it recursively calculates the state according to the following formula:

[0134] ;

[0135] When not satisfied When this occurs, the system determines that the action has a missing precondition error and calculates the set of missing states:

[0136] ;

[0137] The system is based on the set of missing states. The system outputs error alarm information and passes the missing status to the candidate repair generation module. For example, if the precondition for the operation of a low-voltage room air switch includes "low-voltage room cabinet door is open", but the current status of the low-voltage room cabinet door in the current state chain is closed or unknown, the system outputs an alarm "low-voltage room cabinet door precondition missing" and uses "low-voltage room cabinet door is open" as the missing status for subsequent matching of the corresponding low-voltage room cabinet door opening action.

[0138] like Figure 4As shown, after obtaining the error alarm set, the system performs candidate repair generation and multi-objective optimization. The error alarm set includes error type, error location, missing conditions, and risk level; the modular single-action library provides action modules such as door opening, door closing, status verification, chassis positioning, retry, and manual confirmation; the error type-repair rule library is used to establish the correspondence between error types and repair actions. The candidate repair generator performs insertion, deletion, movement, or rearrangement operations on the initial robot execution sequence according to the error type, generating a candidate repair sequence set. .

[0139] Specifically, when a current action is missing and the missing state is that a cabinet door is open, the system retrieves the corresponding cabinet door opening action from the modular single-action library and inserts it before the target action. When the cabinet door closing timing is inappropriate, the system deletes the original cabinet door closing action and searches for the last related action and its status verification action in the same cabinet area, moving the cabinet door closing action after that verification action. When status verification is missing, the system inserts the corresponding visual acquisition or status verification action based on the action object and target state. When anomaly handling is missing, the system supplements the branch with re-execution, re-verification, suspension, or manual confirmation. Subsequently, the system inputs the candidate repair sequences into the risk scoring model and multi-objective cost function, calculates the action risk, transfer risk, and comprehensive cost of each candidate sequence, and outputs the recommended sequence, alarm description, adjustment reason, and comprehensive cost value.

[0140] Taking the step of "disconnecting the energy storage power supply switch SM10 of the 6kV desulfurization stage 2 04 standby bay switch" as an example, this step corresponds to the operation of the low-voltage room air switch. After the system generates the SM10 disconnection action based on the modular single-action library, it verifies its preconditions. If the low-voltage room cabinet door is not open in the current state chain, the system determines that there is a missing precondition action error and recommends inserting a low-voltage room cabinet door opening action before the SM10 disconnection action. After executing the SM10 disconnection action, the system continues to determine whether there is a corresponding state verification action afterward. If the SM10 state visual verification action is not configured, the system outputs a "state verification missing" alarm and recommends inserting the corresponding visual verification action; if the branch is not re-executed, re-verified, suspended, or manually confirmed after the verification fails, the system outputs an "abnormal handling branch missing" alarm and supplements the corresponding abnormal handling logic.

[0141] In this low-pressure room operation scenario, the system also scans for subsequent actions. If low-pressure room objects such as SM40, SM91, SM90, protection devices, or pressure plates still exist in the subsequent sequence, it is determined that the low-pressure room cabinet door still needs to remain open for subsequent actions. In this case, the system does not immediately insert a door-closing action after the SM10 action and its verification action, but continues to search for low-pressure room-related actions until the last low-pressure room-related action and its status verification action are determined, and then inserts the low-pressure room cabinet door closing action after this status verification action. If there are no more low-pressure room-related actions in the subsequent sequence, the system inserts the low-pressure room cabinet door closing action after the SM10 action and its status verification action. This process avoids the cabinet door closing too early, which could restrict subsequent actions, cause repeated door opening and closing, or interfere with the robotic arm's movement.

[0142] Furthermore, taking "moving the circuit breaker trolley from the working position to the test position and completing the grounding switch operation" as an example, this illustrates the verification and optimization process of this method when switching between multiple areas, including the trolley compartment, cable compartment, and grounding switch operation area. The paper steps for this operation typically include confirming that the circuit breaker is open, moving the circuit breaker trolley from the working position to the test position, confirming that the trolley has reached the test position, confirming that there is no voltage on the load side, performing the grounding switch opening and closing operation, and verifying the grounding switch position. When the robot executes this operation, the above paper steps need to be converted into multiple action modules, including mobile chassis positioning, circuit breaker status verification, trolley movement in and out, position indication verification, voltage testing, grounding switch operation, and grounding switch status verification.

[0143] The system first constructs a state chain based on the initial robot execution sequence. If the initial sequence involves executing the trolley outward movement before the circuit breaker is confirmed to be open, the system will proceed according to... The system determines that a "circuit breaker open status missing" error exists and recommends inserting a circuit breaker tripping status verification action before the trolley is moved out. If no position indication verification action is configured after the trolley is moved from the working position to the test position, the system determines that a "trolley position verification missing" error exists and recommends inserting a visual verification action for the trolley test position.

[0144] When the sequence continues to the grounding switch operation area, the system checks the preconditions for the grounding switch operation. If no load-side voltage verification is configured before the grounding switch operation, or if the voltage verification result is unknown, the system outputs a "Grounding Switch Operation Preconditions Missing" alarm and recommends inserting a voltage verification action and a voltage verification result confirmation branch. If no grounding switch position verification action is configured after the grounding switch operation, the system outputs a "Grounding Switch Status Verification Missing" alarm and recommends inserting a corresponding visual acquisition or status verification action.

[0145] For multi-area switching, the system determines the timing of cabinet door and chassis actions based on the continuity of the work area. When the relevant actions and verification of the truck compartment are not yet completed, the system does not recommend switching to the grounding switch operation area. When the truck compartment actions are completed and the truck compartment cabinet door no longer needs to be kept open, the system recommends closing the truck compartment cabinet door and adjusting the chassis positioning action before the grounding switch operation area. If the chassis position is switched before the truck compartment actions are completed in the initial sequence, or if the grounding switch operation area cabinet door closing action is omitted after the grounding switch operation, the system will output an "Action Sequence Conflict" or "Cabinet Door Closing Action Missing" alarm, and generate a corresponding candidate repair sequence.

[0146] For example, the system compares two candidate sequences: Candidate Sequence 1 is "handcart out, chassis switch to grounding switch area, grounding switch operation, return to handcart compartment for handcart position verification"; Candidate Sequence 2 is "circuit breaker status verification, handcart out, handcart test position verification, close handcart compartment cabinet door, chassis switch to grounding switch area, voltage verification, grounding switch operation, grounding switch position verification, close grounding switch operation area cabinet door". The system calculates the comprehensive cost based on action time, movement and positioning cost, risk value, resource switching cost, and hard constraint penalty. Candidate Sequence 1 has a higher comprehensive cost due to cross-regional travel, delayed handcart position verification, and incomplete grounding switch pre-conditions; Candidate Sequence 2 meets the pre-conditions, verification sequence, and continuous regional operation requirements, and reduces chassis travel. Therefore, the system recommends Candidate Sequence 2 as the optimized execution sequence.

[0147] During the multi-objective optimization process, the system constructs candidate execution graphs from candidate repair sequences. Among them, nodes This indicates the system state after execution reaches a certain step. This represents an executable action or candidate repair action between two adjacent system states, meaning that executing this action will transfer the system state from one node to another. Each edge is configured with action time, risk value, movement or relocation cost, actuator switching cost, and manual confirmation cost. The system calculates the cost value of each candidate sequence according to the comprehensive cost function and prioritizes recommending the sequence that satisfies all hard safety constraints and has the lowest comprehensive cost.

[0148] To verify the practical effectiveness of this method, 50 operation tickets of the same model 6kV high-voltage switchgear and similar cabinet types from a power plant were selected as test objects. These included operation tickets for hot standby to cold standby, cold standby to hot standby, hot standby to maintenance, maintenance to hot standby, and test operation. During the test, an initial robot execution sequence was first generated manually based on a modular single-action library. Then, this method was used for state simulation, hard constraint verification, risk scoring, and optimization recommendations. Statistical results showed that the 50 operation tickets contained a total of 684 paper steps and 3126 initial robot action modules. This method detected 147 arrangement problems, including 36 missing pre-actions, 29 improper cabinet door closing timings, 41 missing state verifications, 23 missing anomaly handling branches, and 18 conflicting or duplicated action sequences. After candidate repair and multi-objective optimization, corresponding adjustment suggestions were generated for all 147 problems. Among them, 139 problems could be automatically repaired by the system, while 8 problems required manual confirmation due to on-site safety conditions. Compared to manual item-by-item inspection, the average inspection time for a single operation ticket was reduced from approximately 35 minutes to approximately 8 minutes. Repeated door opening and closing actions in the initial sequence were reduced by 31.4%, repeated chassis positioning actions by 26.7%, and the rate of missing verification actions decreased from 6.0% in the manual initial programming stage to 0. These results demonstrate that this method can improve the efficiency of robot operation ticket programming and inspection, reduce the risk of hidden action omissions and sequence conflicts, and enhance the safety and stability of the operation ticket execution sequence.

[0149] In summary, the present invention achieves this through... Figure 1 The overall verification optimization process shown Figure 2 The sequence verification and risk scoring model shown Figure 3 The state simulation and precondition missing judgment process shown are as follows: Figure 4 The candidate repair generation and multi-objective optimization data flow shown realizes the localization of programming errors, identification of risk segments, and recommendation of repair sequences before the execution of robot operation tickets. Compared with manual item-by-item inspection, this method can reduce problems such as omission of pre-actions, missing verification actions, improper timing of cabinet door closing, and repeated chassis positioning. Furthermore, it reduces the difficulty of adjusting the operation ticket program through interpretable alarm information and recommended sequences, thereby improving the programming efficiency and execution reliability of high-voltage switchgear robot operation tickets.

[0150] The descriptions and practices disclosed in this invention are readily apparent and understandable to those skilled in the art, and various modifications and refinements can be made without departing from the principles of this invention. Therefore, any modifications or improvements made without departing from the spirit of this invention should also be considered within the scope of protection of this invention.

Claims

1. A method for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots, characterized in that, Includes the following steps: S1. Establish a modular single-action library. The action modules in the modular single-action library correspond to the single actions that the robot can execute on the high-voltage switchgear, including cabinet door opening and closing, circuit breaker trolley movement, air switch opening and closing, knob switching, visual verification, mobile chassis positioning, and docking platform actions. The action modules in the modular single-action library shall at least include task number, action name, action category, applicable object, executing subject, preconditions, post-execution status, working area, resource consumption, verification requirements, exception handling rules, and action cost. S2. Obtain the high-voltage switchgear operation ticket steps input by the user, and generate an initial robot execution sequence based on the modular single-action library; S3. Perform state simulation on the initial robot execution sequence, and recursively generate a sequence state chain according to the preconditions and post-execution states of the action module. S4. Perform hard constraint verification on the sequence state chain based on the set of security constraints, identify arrangement errors in the execution sequence and output error alarm information; S5. Calculate the risk value of each action module and adjacent action transfer in the initial robot execution sequence based on the risk scoring model, and generate candidate repair sequences in combination with the programming error; S6. The candidate repair sequence is optimized and searched using a multi-objective cost function to obtain a recommended robot execution sequence. The original operation ticket steps, initial robot execution sequence, error alarm information and recommended adjustment scheme are displayed on the measurement and control software. The recommended robot execution sequence is then solidified into a program file that can be sent to the high-voltage switchgear robot controller for execution.

2. The method for verifying and optimizing the execution sequence of a high-voltage switchgear robot operation ticket according to claim 1, characterized in that, In S3, the state simulation is performed in the following manner: Let the first The system state before the step execution is as follows: , No. Each action module is Action module The set of preconditions is The set of states after execution is Then when At that time, the state is recursively calculated as follows: in, This is the set of states that are changed or deleted after the action module is executed; when At that time, the judgment of the first A certain action module has a missing prerequisite error, and the set of missing states is denoted as: The system generates error alarm information and candidate repair actions based on the set of missing states.

3. The method for verifying and optimizing the execution sequence of a high-voltage switchgear robot operation ticket according to claim 1, characterized in that, In S4, the hard constraint verification performs a step-by-step judgment on the sequence state chain according to the preconditions of the action module, the post-execution state, the work area, resource usage, and verification requirements, specifically including: For the Action Module Read its set of preconditions and with the System state set before step execution Compare; when When this happens, it is determined that the action module has a missing prerequisite action error, and the missing state set is... As evidence for reporting an incident; When the action module When the operating area is a low-voltage room, circuit breaker room, or grounding switch operating area, the corresponding cabinet door status is read from the sequence status chain; if the corresponding cabinet door status does not meet the prerequisite requirements of the action module, it is determined that there is a cabinet door status error. When the action module When the action involves knob switching, air switch opening and closing, grounding switch opening and closing, or circuit breaker truck position change, the system searches for visual acquisition actions or status verification actions corresponding to the operation object and target status within a preset step size range after the action module. If no corresponding action is found, it is determined that there is a status verification missing error. When the action module When verification requirements are configured, determine whether at least one exception handling branch is configured after the corresponding verification action, such as re-execution, re-verification, pause execution, or manual confirmation; if not configured, it is determined that there is an error of missing exception handling. For adjacent action modules and A continuity assessment is conducted. If the two entities are mutually exclusive in terms of the executing entity, operating area, mobile chassis location, or the resource occupancy status of the executing agency, it is determined that there is a conflict in the sequence of actions or a risk of interference from the executing agency. Based on the above judgment results, the system outputs the error location, error type, missing conditions, and corresponding candidate repair actions.

4. The method for verifying and optimizing the execution sequence of a high-voltage switchgear robot operation ticket according to claim 1, characterized in that, The risk scoring model includes a rule-based risk matrix and a data-driven risk prediction model; The data-driven risk prediction model uses action feature vectors For input, output the first... Risk value of each action module Among them, the action feature vector It should include at least the action type, work area, cabinet door status, actuator occupancy status, target status, whether verification is required, historical failure count, retry count, and safety procedure level; When using a logistic regression model, the risk value is calculated according to the following formula: in, For the Sigmoid function, This is the weight matrix. For bias terms; When using a multilayer perceptron model, the risk value is calculated according to the following formula: The system outputs the action risk level and action transfer risk level based on the risk value.

5. The method for verifying and optimizing the execution sequence of a high-voltage switchgear robot operation ticket according to claim 1, characterized in that, In S5, the candidate repair sequence is generated according to the mapping relationship between error type and repair rule, specifically including: When the error type is missing a preceding action and the missing state set includes the cabinet door open state, the system calls the corresponding cabinet door open action from the modular single action library according to the work area of ​​the action module, and inserts the cabinet door open action before the target action. When the error type is "inappropriate cabinet door closing timing", the system starts from the current cabinet door closing action and searches for action modules associated with the same cabinet area along the initial robot execution sequence. If the working area of ​​the subsequent action is still the same cabinet area, or if the preconditions still include the cabinet door being open, the current cabinet door closing action is deleted from its original position and inserted after the last related action and its status verification action in the same cabinet area. When the error type is missing cabinet door closing action, the system starts from the last related action in the current work area and judges backward. If the subsequent actions no longer require the cabinet door to remain open, the corresponding cabinet door closing action is inserted after the last related action and its status verification action. When the error type is missing status verification, the system selects the corresponding visual acquisition action or status verification action from the modular single action library according to the target action's operation object, target status, and verification requirements, and inserts it after the target action; When the error type is missing exception handling, the system supplements at least one exception handling branch for the corresponding verification action based on the verification requirements and risk level of the target action, including re-execution, re-verification, suspension of execution, or manual confirmation. When the error type is action sequence conflict or actuator interference risk, the system generates candidate repair sequences for action sequence adjustment, duplicate positioning merging, or actuator switching reduction based on the executing entity, work area, mobile chassis position, and resource occupancy status of adjacent actions. The system combines the above candidate repair actions with the original initial robot execution sequence to form one or more candidate repair sequences, and then passes them to the multi-objective cost function for optimization selection.

6. The method for verifying and optimizing the execution sequence of a high-voltage switchgear robot operation ticket according to claim 1, characterized in that, In S6, the multi-objective cost function is: in, Candidate repair sequences, The cost of action execution time. This is the risk value. The cost of movement or positioning between adjacent actions. Due to the cost of switching of implementing agencies or resource occupation, The cost of manual confirmation, For hard constraints, there are penalties for violations. These are the weighting coefficients; The system is minimized To achieve this, a recommended robot execution sequence is selected from multiple candidate repair sequences.

7. A system for verifying and optimizing the execution sequence of operation tickets for high-voltage switchgear robots, characterized in that, include: A modular action library storage module is used to store the action modules described in claim 1; The initial sequence generation module is used to map the steps on the paper operation ticket to the initial robot execution sequence; The state simulation and verification module is used to recursively generate the sequence state chain and output the arrangement error based on the set of safety constraints; The risk scoring module is used to calculate the action risk value and the transfer risk value; The candidate repair generation module is used to generate candidate repair sequences based on orchestration errors and risk scores; A multi-objective optimization module is used to output a recommended robot execution sequence with a minimum cost function J(π). The output module is used to display alarm information and recommended sequences on the monitoring and control software, and to generate program files that can be downloaded and executed.

Images