Design scheme and key technologies of intelligent robot inspection system for substations

introduction

With the increase in power grid scale and voltage level, the requirements for power supply safety and reliability have become more stringent, and the normal operation of substations has become an important link in ensuring the power supply safety of the power system [1-2]. At present, China's power grid mainly adopts manual inspection, that is, manual inspection and recording of substation equipment in operation. Manual inspection has disadvantages such as high labor intensity, low work efficiency, and unstable inspection quality. In addition, severe weather conditions also pose a health hazard to the inspectors [3-5]. In recent years, the use of robotic inspection to replace manual inspection has become a hot topic in the development of substation inspection [6-8].

Foreign countries started research on substation robots earlier. Japan began developing substation inspection robots in the 1980s, which used visible light and infrared sensors to inspect substation equipment and automatically process the inspection data [9-10]. The substation inspection robot developed in Quebec, Canada, achieved remote monitoring and could be controlled and operated remotely in real time from the background [11]. Brazil developed a hotspot monitoring robot that walks on a track at high altitude in substations and achieved infrared detection of abnormal heating of substation equipment [12].

In 1999, Shandong Electric Power Company was the first in China to start research on substation inspection robots, and in 2004, it successfully developed the first functional prototype. Later, with the support of the National 863 Program, it developed a substation inspection robot[13]. In 2012, Chongqing Electric Power Company successfully applied the robot to carry out autonomous inspection operations at the 500kV Banan Substation. In 2014, the robot developed by Zhejiang Guozhi Robotics Co., Ltd. was put into operation at the Ruian Substation. At present, China has made great progress in the field of substation robot inspection, but there are still bottlenecks in technologies such as multi-sensor integrated detection, four-wheel drive mobile platform, integrated navigation and precise alignment, and accurate fault diagnosis[14-15].

Based on the aforementioned key technical challenges, a substation robotic intelligent inspection system was developed. This paper introduces the overall design scheme and components of the inspection system, discusses the key technologies of each part, explains the implementation details of the substation robotic inspection project, and describes its application in the 220kV Qingyuan substation and the 500kV Wuyi substation. It also analyzes the problems encountered in the inspection application and proposes solutions. The developed substation robotic intelligent inspection system has been able to complete inspection tasks well in subsequent inspection applications, achieving good inspection results and showing promising prospects for widespread application.

1 Overall Design of Robot Inspection System

1.1 Components of the Inspection System

The robot inspection system consists of a robot body, a charging system, a wireless transmission system, a robot body monitoring backend, and auxiliary facilities. The system composition diagram is shown in Figure 1. The system has the following features: (1) It uses a trackless navigation method to achieve rapid deployment and facilitates inter-station allocation; (2) It adopts four-wheel independent drive, adapts to various complex environments, provides high-definition infrared and visible light video images, and achieves a temperature measurement accuracy of 0.5℃; (3) It adopts a navigation scheme based on a combination of lidar and inertial navigation with precise terrain matching, achieving a positioning accuracy of 1cm; (4) It features ultrasonic anti-collision, providing high reliability and safety, and can move in all directions on the spot, providing greater ease of use for inspection.

1.2 Robot Body

The robot body consists of external structural components, motion control system, power supply system, sensor system and navigation system.

1.2.1 External structural components

The robot's external structure is designed based on the principles of simplicity, practicality, robustness, and reliability, combined with excellent planar cutting technology, taking into account the requirements of weight, stability, and protection level; the surface is treated with powder coating and anodizing processes, which have strong anti-corrosion performance; the robot structure uses a large amount of aluminum alloy material, and the weight is less than 100kg.

1.2.2 Motion Control System

The robot motion control system mainly consists of a motion controller, motor driver, motor, reducer, wheels, ultrasonic obstacle avoidance module, manual remote control module, status indicator lights, etc., as shown in Figure 2.

The motion control system mainly realizes the communication with the monitoring backend and the control functions of the vehicle body and gimbal. It receives and uploads the status information of the vehicle body and gimbal in real time. Its workflow is shown in Figure 3.

To meet the outdoor operation requirements of substations, the robot's chassis utilizes a four-wheel drive system. PID control and PMSM vector control algorithms are applied to the vehicle's motion control, achieving precise speed control and rapid torque response, ensuring the maturity and stability of the control algorithms. Independent four-wheel drive and flexible matching control enable zero turning radius, 360° rotation on the spot, flexible path planning, and strong environmental adaptability. The drive motors are low-resistance, high-torque motors, providing a wide speed range, high efficiency, and good reliability. The robot's maximum operating speed reaches 1.1 m/s, it can overcome obstacles up to 10 cm thick, and its climbing ability reaches 25°.

1.2.3 Power Supply System

(1) Battery selection. Lithium iron phosphate batteries are used for power supply. The rated voltage of the battery is 36V and the battery capacity is 50Ah. In order to meet the safety requirements of battery charging, discharging and storage and transportation, the battery is installed in a special battery box made of explosion-proof and flame-retardant materials.

(2) Power Management System Design. The lithium battery pack power management system (BMS) adopts centralized management. The BMS consists of a main control unit (CMU) and several monitoring units (BMUs). The BMU detects and manages the voltage and temperature of the battery modules and transmits the data to the CMU. The CMU detects the total voltage, total current and insulation of the lithium battery pack, and is responsible for communicating with the robot control system and charger to protect the battery pack during charging and discharging.

(3) Improved BMS performance. Slight differences inevitably exist in the parameters of various lithium batteries. These differences can be amplified by internal resistance, self-discharge, and the increasing number of charge-discharge cycles, reducing battery life and potentially posing power safety hazards. By implementing equalization management through a BMS, the battery pack will maintain better consistency, extending battery life, reducing costs, ensuring a minimum range of 5 hours on a single charge, and improving system reliability and stability.

1.2.4 Sensor System

(1) Visible Light Detection. The visible light camera is used to observe the appearance of equipment and read instrument readings. It features automatic or manual focusing, a video resolution of 1080p, and an optical zoom of up to 30x. Employing an automatic aperture design, it automatically controls the lens aperture by detecting the average video signal value, thus obtaining standard video signal levels under different illumination conditions. A high-intensity LED light and a wiper are installed on the robot's pan-tilt-zoom (PTZ) platform to enable visible light detection at night and in rainy weather. The light and wiper are controlled by a local monitoring system or a remote centralized control system.

(2) Infrared thermal imaging detection. The infrared detector of the infrared thermal imager receives the radiant heat from the object, converts it into an electrical signal, and after subsequent amplification, filtering, analog-to-digital conversion, and CPU processing, displays it on the image display. In actual temperature measurement, a high-precision blackbody is first used for calibration to find the correspondence between the blackbody temperature and the image grayscale value. The infrared thermal imager has an autofocus function and can overlay and display the location and temperature value of the highest temperature point in the real-time image. The thermal sensitivity of the infrared thermal imager is better than 50mK, and the temperature measurement accuracy is better than 2K.

(3) Sound Detection. The robot is equipped with a loudspeaker and microphone, enabling two-way voice communication with the monitoring backend and on-site sound collection. Simultaneously, by collecting normal and abnormal sounds from the operating equipment, the robot extracts sound feature parameters and establishes a normal and abnormal sound model library. The noise data collected by the robot is transmitted to the control backend, where operating status is identified based on audio diagnostic software and the model library, abnormal sounds are detected, and alarms are issued. The diagnostic process is shown in Figure 4.

The sound analysis software mainly consists of signal processing, signal feature extraction, and signal display. The signal processing section includes endpoint detection, framing, and windowing. Endpoint detection detects the effective speech components in the input sound signal using cepstral signatures; framing divides the original speech signal into smaller segments and performs frame shifting; windowing smooths the spectrum and prevents high-frequency leakage after framing. The signal feature extraction process includes FFT power spectrum calculation, bandpass filter bank, logarithmic energy calculation, discrete cosine transform, and extraction of first-order difference MFCC coefficients. The display section shows the digital signal graphs of each frequency in the measured sound.

(4) Omnidirectional Intelligent Gimbal. The gimbal is mounted above the robot platform to carry visible light, infrared, and sound sensors. Driven by a DC servo motor, the gimbal has two independent rotational degrees of freedom: horizontal and vertical. The gimbal's pitch frame is equipped with infrared and visible light windows. The core of the gimbal's motion control uses a digital signal processor (DSP) chip, which is responsible for the motor motion control of the horizontal and pitch degrees of freedom and communication with the interface conversion module. The main performance indicators of the gimbal meet the following requirements: ≥10,000 preset gimbal positions; vertical motion range –30° to +90°; continuous horizontal motion range 0° to +360°; positioning accuracy ±0.1°; horizontal rotation speed 0.01~60°/s; vertical rotation speed 0.01~30°/s.

1.2.5 Navigation System

The robot relies on a combination of LiDAR, inertial navigation, and odometer for navigation. Utilizing multi-sensor fusion technology, it obtains vehicle positioning information, enabling autonomous walking and docking according to preset routes and locations. The LiDAR uses the SICK LMS511 high-performance outdoor LiDAR scanner, with a measurement range of up to 80m, a scanning range of 190°, a resolution of 0.1667°, and a scanning frequency of up to 25Hz. It can operate in harsh environments ranging from -30℃ to +55℃. The inertial navigation system provides three-axis attitude angles (or angular rates) and acceleration information, with a resolution of 0.05° and an error of 1.5°. The odometer information includes the vehicle's current coordinates, calculated from the vehicle's kinematic model and information such as the rotational speed and displacement of the four wheels, with an error within 3%.

1.3 Charging System

The charging room consists of a charging cabinet, charging base, wireless communication equipment, and an automatic roller shutter door. The charging cabinet and charging base are used for automatic docking and charging of robots. The wireless communication equipment uses the same wireless bridge and antenna as the local monitoring backend, and the antenna is installed on the top of the charging room.

The robot's working states are divided into three types: inspection, charging, and idle. Upon receiving an inspection command, the robot checks whether the battery power is sufficient. If it is sufficient, it enters inspection mode and begins to execute the inspection task; otherwise, it refuses to execute and issues an alarm. After completing the inspection, the robot returns to the charging room. The robot monitors the battery power in real time during inspections, and if the power is insufficient, it returns to the charging room to recharge. The charging process is fully automated.

1.4 Wireless Transmission System

The robot achieves two-way, real-time information exchange with the local monitoring backend via a wireless bridge. The exchanged information includes the robot's status and images, voice, and indicative data from the monitored equipment. The robot utilizes a high-quality outdoor dedicated digital wireless bridge in the 5.8GHz band, enabling real-time transmission of multiple video, audio, and data over long distances, up to 10km. The data transmission error rate is ≤10⁻⁶, data transmission latency is ≤20ms, and image transmission latency is ≤300ms. Since this frequency band wireless bridge does not require a wireless license, it is easier to deploy than other wired network equipment. The robot receives control commands from the monitoring backend via the wireless bridge, performing actions such as pan-tilt rotation, equipment detection, vehicle movement, and automatic charging. It also monitors the robot's status and reports various warnings and alarms. In the event of communication interruption or abnormal received messages, images, voice, and data are not lost; the system will issue an alarm and automatically resume transmission after communication is restored.

1.5 Local Monitoring Backend

The local monitoring backend consists of a computer, wireless communication equipment, monitoring and analysis software, and a database. The robot connects to the monitoring backend via a wireless local area network and uses the TCP/IP protocol for data exchange, as shown in Figure 5.

The monitoring backend software is developed using the C++ language and is based on the .NET architecture. It can run across platforms on various versions of the Windows operating system. The software composition is shown in Figure 6.

(1) The real-time monitoring module is responsible for viewing image information, vehicle status information, vehicle movement information, battery status information, on-site weather information, and inspection task information during the robot's operation.

(2) The task planning module is divided into three modes: routine, special inspection, and remote inspection. The task mode can be switched at any time. According to the substation inspection requirements, the routine task planning can generate several inspection tasks in advance for daily regular inspections; the special inspection task planning can generate temporary inspection tasks in real time for special inspections.

(3) The remote control module can remotely control the robot to a designated location and perform a designated action in real time. This module can control the gimbal's orientation and pitch, as well as the robot's speed and direction, via a handle.

(4) The configuration center module includes three sub-interfaces: device configuration, map configuration, and basic configuration. The device configuration interface includes infrared configuration, visible light configuration, vehicle configuration, and gimbal configuration.

(5) The historical query and data analysis module can store, diagnose and query information such as visible light images, infrared images, sound and meter readings, equipment location status, and oil level of oiling equipment.

1.6 Environmental adaptability

The robot is designed according to the extreme environmental climate of substations in various regions of the country. It is designed to withstand harsh weather conditions such as storms, heavy rain, heat and humidity, high altitude, and cold, as well as strong electric and magnetic fields in substations. Through "three-proof" design, windproof design, electromagnetic compatibility, shock resistance design, and temperature adaptability design, the robot can ensure long-term reliable, safe and stable operation under different climatic conditions.

1.6.1 Three-proof design

The robot's outer shell is coated with an electrostatic powder coating, providing corrosion, water, and oxidation protection. The internal sensors and control systems utilize a modular design and standardized production. The robot features an integrated structural design, ensuring waterproof and dustproof capabilities. The entire robot meets the IP54 design requirements of GB4208, with a maximum wading depth greater than 10cm.

1.6.2 Windproof design

The robot adopts a four-wheel drive chassis structure with a low center of gravity, which facilitates stable operation on the ground. Its compact design and high sealing allow it to withstand winds up to force 10.

1.6.3 Electromagnetic Compatibility Design

The robot's electronic components, power supply, communication modules, etc., are shielded and isolated. Key signals are designed with impedance matching, and each device module adopts an equipotential common ground design. The input and output interfaces are filtered and protected to ensure the signal integrity, safety and reliability of each module.

1.6.4 Vibration-resistant design

During the inspection of the substation, the robot will inevitably experience a certain degree of vibration due to the influence of the road environment. To address potential issues such as loosening of fixing screws and breakage of components, the following anti-vibration measures are taken: (1) Add spring washers, toothed washers, apply thread adhesive, and use anti-loosening nuts to all fasteners to improve the fastening effect and strength of bolts and screws; (2) Optimize the design of broken parts to improve the strength of components; (3) Add protective sleeves, vibration damping springs, and other measures to reduce the effect of external forces on pipeline connection parts.

1.6.5 Temperature Adaptability Design

To ensure normal operation and long-term storage in hot or cold environments, all parts and components of the robot are selected from industrial-grade products with a wide temperature range. A heat dissipation fan and heating plate are installed inside the gimbal housing to automatically dissipate or heat the environment inside the housing, which is beneficial for the normal operation of the visible light camera and infrared detector under different temperature conditions. The robot's operating environment temperature is -25℃ to +55℃, the storage environment temperature is -30℃ to +65℃, and the relative humidity of both the operating and storage environments is 5% to 95% (without condensation).

2. Substation Robot Inspection Project Implementation

2.1 Substation Site Survey

The on-site survey phase of the substation robot inspection project is divided into the following steps: (1) Select the optimal installation location of the charging room based on the distribution of equipment, features and inspection convenience in the station; (2) Select the installation location of wireless communication facilities based on the actual situation of the substation building; (3) Determine whether the road needs to be modified based on the road connectivity, step height and other factors; (4) Preliminarily determine the inspection route based on the equipment layout in the station.

2.2 Installation of Inspection Equipment

2.2.1 Installation of charging room

The robot charging room is equipped with an automatic charging device and an access control system that can automatically open and close. The charging room has external dimensions of 2.0m (width) × 2.5m (length) × 2.8m (height), adopts an integrated box structure, and is installed on the open ground in the high-voltage equipment area of ​​the substation. The location is higher than the main road in the station, and the foundation is naturally sloped to connect with the road in the station. The site selection principles for the charging room are: (1) close to the main control room for convenient infrastructure construction and commissioning; (2) choose a flat ground and avoid areas with obvious potholes; (3) not too far away from the inspection area.

2.2.2 Installation of Wireless Bridge and Weather Sensor

The wireless bridge and micro-weather sensor are auxiliary equipment for robot inspection, respectively undertaking wireless communication and weather monitoring functions. Both are designed for outdoor use and have an IP55 protection rating. To facilitate wired connection with the back-end monitoring equipment, the wireless bridge and weather sensor are installed on the top floor of the main control building in the station, with power and data cables running along the wall conduits and connected to the monitoring back-end.

2.2.3 Installation of the monitoring backend

The monitoring backend consists of a computer, router, mouse, keyboard, loudspeaker, and microphone. The computer and wireless bridge are connected to the router, which can connect to the organization's local area network.

2.3 Inspection Planning

2.3.1 Inspection Route Planning

Based on the location of the equipment to be inspected and the condition of the inspection roads, technicians plan the inspection route within the station, starting from the charging room, to achieve the optimal inspection path.

2.3.2 Inspection Map Construction

The inspection map is automatically generated by the robot itself using backend control software. It requires no modification to the substation's internal environment or road surface, and does not affect the normal operation of substation equipment and facilities. The steps for constructing the substation inspection map are as follows.

(1) Select the map origin: The origin is usually selected near the charging room.

(2) Initialize the lidar device: Enable the robot lidar sensor transmission function to complete the lidar initialization.

(3) Collect map construction data: The remote-controlled robot walks around the entire substation according to the planned inspection route. The robot automatically records the geographical information of all equipment and buildings, thereby completing the collection of the entire map construction data.

(4) Automatic map generation: After completing the data collection for map construction, start the map generation program to automatically generate a substation inspection map.

(5) Set inspection points and inspection routes: After the map is built, set inspection points and optimize inspection routes according to the type and quantity of substation inspection equipment.

(6) Inspection and testing: Test the inspection points of the substation equipment according to the calibrated optimal route, find omissions, adjust the inspection route, and ensure full coverage of equipment inspection points.

The robot inspection process is shown in Figure 7.

2.4 Inspection Application

2.4.1220kV Qingyuan Substation Application

In December 2015, the robot charging station, wireless bridge, and monitoring backend were deployed at the 220kV Qingyuan substation. In January 2016, the robot began laser navigation map scanning, inspection map creation, and inspection task point collection. Inspection application was completed in April. After four months of on-site debugging and application, robot deployment and task point collection were completed, followed by full-station inspection debugging. The inspection tasks covered 2490 inspection points in the main transformer area, 220kV area, and 110kV area, achieving a coverage rate of 94.1%. The distribution of inspection task points is shown in Figure 8.

2.4.2500kV Wuyi Station Application

In September 2016, the robot charging room, wireless bridge, and monitoring backend were deployed. In October, the inspection deployment for the main transformer and 500kV area within the station was completed and inspections were carried out. After two weeks of on-site debugging, the communication interference problem of the robot in the strong electromagnetic field environment of the 500kV area was resolved, and the robot could reliably communicate with the backend system in all 500kV areas. Simultaneously, the deployment and task point collection for the first series of areas of the main transformer and 500kV were completed, followed by the execution of inspections in the deployed areas. The inspection tasks included 64 visible light inspection task points, 273 infrared inspection task points, and 10 high-definition auxiliary observation points. The high-definition auxiliary observation points included external images of equipment such as switch control boxes and main transformers.

3. Problem Analysis and Solutions in Inspection Application

During the inspection and commissioning of robots in substations, some technical problems were found that still need to be solved, such as the robot derailing due to interference from external objects, unstable wireless signal connection, insufficient accuracy of meter readings, and the robot getting stuck when walking and turning.

3.1 Robot derailment during inspection

During robot inspections, external interference, such as people watching, caused the robot to deviate from its inspection route. Analysis of sensor and communication data recorded by the navigation computer revealed that the robot's odometer output occasionally jumped at the moment of the anomaly. Because the jump value was greater than 40m, it happened to reach the set coordinates of the next turning point, causing the program to misjudge and make the robot mistakenly believe that it had reached the turning point and execute a turning maneuver. Subsequently, as the robot moved, the odometer data continued to accumulate according to the jump value, causing the robot to continue moving forward after turning.

The solutions are as follows: (1) Add mileage data jump judgment and abnormal data processing programs to the vehicle control board and navigation computer. When the change in mileage data is greater than the set application change, the vehicle control board filters out abnormal data, and the navigation computer replaces the abnormal data with the predicted value accumulated from the previous frame of normal data; (2) Use the iterative nearest point algorithm for the early map stitching and robot localization during the navigation process of the inspection robot. This algorithm has the characteristics of simple logic, high accuracy, easy implementation, and inherent stability and robustness. After adopting the above measures, the anti-interference ability of the robot during operation is greatly enhanced, and the walking of the zigzag path at any angle is realized.

3.2 Unstable wireless communication

During the initial debugging phase of the robot's operation, some issues were observed, such as choppy display on the backend screen and reduced data transmission speed in certain patrol sections. After on-site observation and analysis, it was found that the cause was that these sections were located in blind spots of the directional receiving antenna coverage, resulting in weak communication signals, which caused the backend screen to freeze and the data transmission speed to decrease.

By replacing the antenna with a high-power, wide-angle antenna and adjusting parameters such as signal gain, signal frequency, and communication mode, the communication quality was greatly improved, ensuring the normal operation of the robot.

3.3 Inaccurate meter readings

During the inspection, there are many SF6 pressure gauges that need to be identified and read. However, the robot has a low accuracy rate in identifying such instruments. After analysis, it was found that this is because the pointers of the SF6 pressure gauges installed in the substation are small, with the lower half of the pointer being white and the upper half being black. The dial scale is also black, making it difficult for the instrument identification software to distinguish between the pointer and the scale, thus resulting in inaccurate readings.

After collecting a large number of SF6 pressure gauge samples from all stations, the pointer reading algorithm of the instrument identification software was re-optimized, and the accuracy of the robot's SF6 pressure gauge identification was greatly improved in subsequent inspection tasks.

3.4 Stuck when walking and turning

During a stationary turn, the robot experienced some wheels becoming airborne, causing the turning motion to become stuck. After on-site observation and analysis, the cause was determined to be the uneven road surface and the robot's solid tires, which have low elasticity, causing some wheels to become airborne during the turn. Since the robot relies on all four wheels for turning, the airborne wheels resulted in insufficient turning power, causing the turn to become stuck.

By replacing solid tires with more elastic pneumatic tires, the chances of individual wheels being suspended in the air due to uneven road surfaces were greatly reduced, and no such problems were found during subsequent inspections.

4 Conclusion

Intelligent robot inspection of substations is an important direction for the development of power inspection models, but many technical bottlenecks still exist in practical applications. This paper develops an intelligent robot inspection system for substations, introducing the system design scheme and key technologies of each component from aspects such as the robot body, charging system, wireless transmission system, local monitoring backend, and environmental adaptability. This inspection system has significant advantages such as rapid deployment, strong adaptability, accurate data collection, high positioning accuracy, and ultrasonic anti-collision, and all key performance indicators meet the requirements of intelligent substation inspection tasks.

This paper introduces the implementation steps of the substation robot inspection project from aspects such as on-site survey, equipment installation, inspection planning, and inspection application. Through actual inspections of 220kV and 500kV substations in the power grid, it points out the problems found in the robot inspection and debugging process, analyzes and solves them. In subsequent inspection applications, it has successfully completed various inspection tasks and achieved remarkable results. The developed inspection system has good prospects for promotion and application.