CN-121978932-A - Control method and system for power transmission line maintenance robot

CN121978932ACN 121978932 ACN121978932 ACN 121978932ACN-121978932-A

Abstract

The invention relates to the technical field of intelligent operation and maintenance of power equipment, and particularly discloses a control method and a control system of a power transmission line maintenance robot, wherein the control method comprises the following steps of: the method comprises the steps of constructing a high-precision digital twin body, synchronizing the high-precision digital twin body with real-time data of an overhaul robot, acquiring historical data of the overhaul robot to drive digital twin body to simulate and optimize, generating a light student model through knowledge distillation to be deployed to an edge computing unit, and carrying out extreme scene self-adaptive control by utilizing a bidirectional distillation mechanism. According to the invention, through the closed-loop interaction of the digital twin body of the transmission line and the physical entity of the maintenance robot and the parameter standardization mechanism, unified control of the multi-model maintenance robot is realized, and the self-adaptive control capability, the operation safety and the multi-equipment cooperative efficiency of the transmission line maintenance robot in a complex environment are remarkably improved.

Inventors

LIU JIE
ZHANG GUOCHUN
Yao Sumao
WANG XIAOPING
ZHANG LIANGCAI
ZHANG CHAOBIN

Assignees

深圳市输变电工程有限公司

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. The control method of the power transmission line maintenance robot is characterized by comprising the following steps of: s1, collecting geometrical structure and dynamic environment parameters of a power transmission line, integrating a wire mechanical simulation module, a robot kinematics module and an environment disturbance simulation module, and constructing a power transmission line digital twin body which interacts with a power transmission line maintenance robot in real time; S2, training a teacher model based on a reinforcement learning algorithm, optimizing climbing, walking and obstacle avoidance strategies of the power transmission line maintenance robot, compressing the teacher model into a light student model through knowledge distillation, and deploying the light student model to an edge calculation unit of the power transmission line maintenance robot; S3, when the power transmission line maintenance robot detects environmental disturbance, activating a multi-physical field coupling simulation function of a digital twin body of the power transmission line to generate extreme scene data, optimizing a control instruction in real time through a bi-directional distillation mechanism, and dynamically adjusting the grabbing force and the walking path of the robot by utilizing a vision-force sense fusion algorithm; S4, generating a motion control instruction of the power transmission line maintenance robot based on the power transmission line digital twin body and the student model, and realizing data interaction between the power transmission line digital twin body and the power transmission line maintenance robot through a hybrid communication architecture.
2. The method for controlling the power transmission line maintenance robot according to claim 1, wherein the step S1 is to construct a power transmission line digital twin body which interacts with the power transmission line maintenance robot in real time, and the specific implementation manner includes the steps of: s11, conducting joint modeling on the mechanical characteristics of the lead of the power transmission line and the dynamic changes of the motion trail and the environmental interference of the power transmission line maintenance robot, so that the lead mechanical simulation module, the robot kinematics module and the environmental disturbance simulation module are coupled and run in real time; S12, bidirectional synchronization is carried out on a control instruction of the power transmission line maintenance robot and mechanical parameters of a digital twin body of the power transmission line through an industrial real-time communication protocol, so that the simulation environment is ensured to be consistent with a real operation state; S13, dynamically adjusting simulation environment parameters according to meteorological monitoring data, so that the digital twin body of the power transmission line can simulate wire deformation and environment change under extreme working conditions in advance.
3. The method for controlling the power transmission line maintenance robot according to claim 1, wherein the step S1 extracts key features through a reverse distillation algorithm to dynamically optimize the simulation precision of the digital twin body of the power transmission line, and the specific implementation manner comprises the steps of: s14, synchronously acquiring stress changes of the wires, environmental parameters and action data of the transmission line maintenance robot through an onboard sensor of the transmission line maintenance robot, and ensuring the alignment of all data time stamps by using a standard time synchronization technology; S15, extracting wire deformation characteristics, barrier space distribution characteristics and meteorological interference characteristics from historical operation data of the power transmission line maintenance robot, and screening out characteristics with high contribution to simulation environment optimization through a reverse distillation technology; s16, adopting a teacher-student dual-model architecture, injecting the optimized feature set into a digital twin body of the power transmission line, and dynamically updating the mechanical simulation parameters of the lead and the simulation rules of environmental disturbance.
4. The method for controlling the power transmission line maintenance robot according to claim 1, wherein the specific implementation manner of step S2 includes: s21, simulating a real line environment in a digital twin body of the power transmission line, wherein the real line environment comprises different wire shapes, inclination angles, common obstacles and different weather conditions, and provides a comprehensive training scene for the robot; s22, setting comprehensive evaluation standards, and simultaneously considering the factors such as the moving speed, the power consumption, the body stability, the safety distance and the like of the power transmission line maintenance robot, so that the power transmission line maintenance robot can rapidly complete tasks and simultaneously keep low energy consumption and high safety; S23, adopting an intelligent learning method to enable the power transmission line maintenance robot to continuously try different action strategies in a virtual environment, recording successful experience and optimizing a decision process, and gradually improving autonomous operation capability in a complex line environment; s24, regularly comparing the operation performance of the power transmission line maintenance robot in a real line with the simulation result in the digital twin body, identifying the difference and automatically adjusting the learning model to ensure that the trained control strategy can effectively adapt to the actual working environment; S25, in the digital twin environment, real-time operation feedback data of the transmission line maintenance robot are utilized to continuously adjust training parameters of a teacher model, and the output strategy of the training parameters is ensured to be dynamically matched with the real environment; S26, extracting core decision logic of a teacher model into a student model through feature compression and calculation graph simplification technology, reserving key obstacle avoidance rules and path planning capacity, and reducing model volume and calculation complexity; and S27, adapting the student model to an edge calculation unit of the power transmission line maintenance robot, and ensuring that the student model can still respond to control instructions in real time under a low-calculation-force environment through instruction set optimization and memory allocation strategies under the limitation of hardware resources of the edge calculation unit.
5. The method for controlling the power transmission line maintenance robot according to claim 1, wherein the step S3 of activating the multi-physical field coupling simulation function of the digital twin body of the power transmission line to generate extreme scene data, and optimizing the control command in real time by a bi-directional distillation mechanism, comprises the following steps: S31, monitoring surrounding environment parameters of a power transmission line maintenance robot in real time, and automatically triggering multi-physical field coupling simulation of a digital twin body of the power transmission line when wind speed mutation, temperature shock or abnormal swing of a lead are detected; S32, synchronously simulating interaction of an electromagnetic field, a fluid field and a structural mechanical field in a digital twin body of the power transmission line, and predicting deformation trend and dynamic response of the lead under extreme conditions; s33, comparing and analyzing the extreme scene data generated by simulation with the current state of the power transmission line maintenance robot, extracting optimal decision features through a bidirectional knowledge distillation mechanism, and dynamically adjusting the priority and execution sequence of the control instructions; S34, based on the environmental change rate and the amplitude, the update frequency of the control instruction is adaptively adjusted, and the operation continuity of the power transmission line maintenance robot is maintained on the premise of ensuring safety.
6. The method for controlling the power transmission line maintenance robot according to claim 1, wherein in the step S3, the grasping force and the traveling path of the power transmission line maintenance robot are dynamically adjusted by using a vision-force sense fusion algorithm, and the specific implementation manner includes: S35, recognizing the surface state, diameter change and obstacle position of the wire in real time through a visual sensor carried by the transmission line maintenance robot, and monitoring the contact force distribution of the grabbing mechanism and the wire through a force sensor carried by the transmission line maintenance robot; S36, establishing a mapping relation between visual features and force sense feedback, and dynamically calculating the optimal grabbing force and distribution according to the surface roughness, the icing degree or the rust state of the lead; S37, when abnormal shaking or external interference of the lead is detected, planning a fine adjustment direction and a step length of a walking path of the power transmission line maintenance robot in real time based on a visual positioning result and force feedback data; s38, setting a multi-level safety threshold, and automatically starting an emergency protection strategy when the grabbing force or the attitude angle exceeds a preset range, wherein the emergency protection strategy comprises the steps of reducing the moving speed, increasing the grabbing point or starting a temporary anchoring mechanism.
7. The method for controlling the power transmission line maintenance robot according to claim 1, wherein the specific implementation manner of step S4 includes: s41, fusing a real-time environment state provided by a digital twin body of the power transmission line with decision output of a student model on an edge calculation unit to generate a motion trail and joint control parameters of a power transmission line maintenance robot suitable for a current operation scene; S42, adopting a layered hybrid communication mechanism, wherein an edge layer deployed on an operation site of the power transmission line maintenance robot uses a deterministic real-time network to transmit key instructions and safety signals, so as to ensure the timeliness of control response; s43, establishing a communication quality monitoring and self-adaptive mechanism, dynamically adjusting a data transmission strategy according to a network state, and preferentially guaranteeing the transmission reliability of a safety control instruction when the communication quality is reduced; s44, designing a data consistency verification mechanism, and ensuring that data interaction between the digital twin body of the power transmission line and the power transmission line maintenance robot is accurate through time stamp alignment and state synchronization verification, thereby providing a reliable basis for closed-loop control.
8. The control method of a transmission line inspection robot according to claim 1, characterized in that the control method further comprises the steps of: s5, unifying heterogeneous parameters and data of the power transmission line maintenance robots of different models, wherein the specific implementation mode comprises the following steps: s51, constructing a mapping relation table between physical parameters, control instructions and sensor data of the power transmission line maintenance robots of different models and a standard parameter system; S52, carrying out format conversion and dimension unification on the collected heterogeneous data of the power transmission line maintenance robots of different types according to a standard data model, and generating a standardized data set for construction and model training of a digital twin body of the power transmission line; And S53, according to the type of the accessed power transmission line maintenance robot, the corresponding mapping relation is automatically loaded, and parameter thresholds and processing logics of power transmission line digital twin body construction, model training, extreme scene processing and real-time control are dynamically adjusted, so that the power transmission line maintenance robots of different types can work cooperatively under a unified control method.
9. A control system for a transmission line inspection robot, comprising: the digital twin body construction module is used for integrating the wire mechanical simulation module, the robot kinematics module and the environment disturbance simulation module, acquiring the geometric structure and the dynamic environment parameters of the power transmission line and constructing a power transmission line digital twin body capable of interacting with a power transmission line maintenance robot in real time; The model training module is used for training a teacher model based on a reinforcement learning algorithm, optimizing climbing, walking and obstacle avoidance strategies of the power transmission line maintenance robot, compressing the teacher model into a light student model through knowledge distillation, and deploying the light student model to an edge computing unit of the power transmission line maintenance robot; the extreme scene processing module activates the multi-physical field coupling simulation function to generate extreme scene data when the environmental disturbance is detected, optimizes the control instruction in real time through a bi-directional distillation mechanism, and dynamically adjusts the grabbing force and the walking path of the robot by utilizing a vision-force sense fusion algorithm; And the real-time control module is used for generating a motion control instruction of the power transmission line maintenance robot based on the power transmission line digital twin body and the student model, and realizing data interaction between the power transmission line digital twin body and the power transmission line maintenance robot through the hybrid communication architecture.
10. The control system of a transmission line inspection robot according to claim 9, further comprising: The parameter standardization module is used for carrying out unified processing on heterogeneous parameters and data of the power transmission line maintenance robots of different models, and specifically comprises the following steps: The parameter mapping unit is used for establishing a mapping relation table among physical parameters, control instructions, sensor data and a standard parameter system of the power transmission line maintenance robots of different models; The data conversion unit is used for respectively carrying out format conversion and dimension unification on the collected heterogeneous data of the power transmission line maintenance robots of different types according to a standard data model, and generating a standardized data set for construction and model training of a digital twin body of the power transmission line; The self-adaptive configuration unit is used for automatically loading the corresponding mapping relation according to the type of the accessed power transmission line maintenance robot, dynamically adjusting the parameter threshold values and processing logic of the power transmission line digital twin body building module, the model training module, the extreme scene processing module and the real-time control module, and ensuring that the power transmission line maintenance robots of different types can cooperatively operate under a unified control system.

Description

Control method and system for power transmission line maintenance robot Technical Field The invention relates to the technical field of intelligent operation and maintenance of power equipment, in particular to a control method and a control system of a power transmission line maintenance robot. More specifically, the invention belongs to the crossing technical field of combining the electric power robot technology with digital twin and artificial intelligence, and is particularly suitable for intelligent control, environment adaptation and multi-type unified management of a high-voltage/ultra-high-voltage overhead transmission line maintenance robot. Background With the continuous expansion of the power grid scale in China, the mileage of the high-voltage and ultra-high-voltage transmission line is continuously increased, and higher requirements are put on line overhaul and maintenance. The traditional power transmission line overhaul mainly depends on manual climbing or helicopter auxiliary operation, so that safety risks such as high falling and electric shock exist, efficiency is low, the power transmission line overhaul is obviously restricted by weather conditions, and the requirements of safe and stable operation of a modern power grid are difficult to meet. As an intelligent alternative, transmission line maintenance robots have received a lot of attention in recent years, but still face a lot of technical challenges in practical applications. At present, a power transmission line overhauling robot control system mainly has the following problems that firstly, the environment of a power transmission line is complex and changeable, a wire can dynamically deform under the influence of factors such as wind load, icing, temperature change and the like, but the existing robot control system is mostly designed based on a static environment model, and lacks accurate simulation of the mechanical characteristics and the environmental disturbance of the wire, so that the adaptability in actual operation is insufficient, for example, under extreme working conditions such as strong wind or icing, the robot cannot adapt to the dynamic change of the wire in time, and is out of control or falls. Secondly, the existing inspection robot learning control strategy mostly adopts an off-line training mode, and the training environment is different from the real environment, so that various extreme working conditions are difficult to fully cover. Meanwhile, the complex deep learning model has large calculation amount, is difficult to be directly deployed on the robot edge calculation unit, and the simple control model is difficult to deal with complex environments, so that the performance and the efficiency are difficult to balance, and the adaptability to newly encountered obstacle types is poor. Thirdly, the application of the existing digital twin technology in the maintenance of the electric power equipment is still in a primary stage, the equipment state monitoring is mostly focused on instead of the active control, the closed-loop linkage with the maintenance robot control strategy is not considered, the simulation of the multi-physical field coupling effect is lacking, and the influence of the extreme working condition on the operation safety cannot be accurately predicted. In addition, because the power transmission line maintenance robots produced by different manufacturers have significant differences in the aspects of structure, sensor configuration, control interfaces and the like, the control systems are difficult to unify. The power grid enterprises often need to be provided with independent control systems for robots of different models, and the operation and maintenance complexity and cost are increased. At present, no effective method is available for realizing collaborative operation and unified management of multiple types of maintenance robots. In recent years, reinforcement learning has demonstrated potential in the field of inspection robot control, but because of the high risk of the transmission line work environment, it is neither safe nor realistic to directly perform exploratory learning in a real environment. While knowledge distillation and other model compression techniques have been successfully applied in the fields of image recognition and the like, in the field of electric robot control, how to significantly reduce model scale while maintaining complex decision-making capability still lacks an effective solution. In view of the foregoing, there is a need in the industry for a method and a system for controlling a power transmission line maintenance robot that integrates digital twinning, reinforcement learning and edge computing, which can accurately simulate the mechanical characteristics and environmental disturbance of a wire, adapt to extreme working conditions, realize unified management of multiple types of robots, and provide real-time, safe and efficient control decisions under l