KEPCO's Private 5G network and Applications: IoT-based predictive maintenance, robot-based maintenance inspection, and safety management

November 30, 2023 | By Harrison J. Son (son@netmanias.com)

Online viewer:

HTML

Comments (0)

In this blog, we will look at Korea Electric Power Corporation (KEPCO)'s case of IoT-based predictive maintenance, unmanned maintenance inspection, safety and access control by introducing wireless IoT sensors, quadruped robots, CCTV, private 5G network, and AI at substations.

<KEPCO's Shin-Jungbu Power Plant in Korea>

1. KEPCO's need for a private 5G network

Preventing accidents in advance is important due to the aging of power facilities in substations:

KEPCO has more than 900 substations, and more than 50% of the facilities in substations are over 20 years old, so it is necessary to take preventive measures against accidents.

Transformers and gas-insulated switchgear in substations are the core facilities of the power grid, and in the event of a breakdown or failure, large-scale power outages occur, causing national damage (e.g., KEPCO's fire at a substation in Daejeon in 2021 caused 67,000 households to lose power).

In December 2022, KEPCO deployed private 5G in substations, and based on this, introduced wireless IoT sensors, quadruped robots, wireless CCTV, and AI servers to build a smart facility management system that monitors facilities in real time and predicts and prevents failures through AI analysis.

Predictive maintenance for substation equipment based on private 5G networks, IoT sensors, robots, and AI

<KEPCO's private 5G network and application of preventive diagnosis and instantaneous inspection>

2. IoT-based Predictive Maintenance (Wired sensors -> Wireless sensors)

Like most power companies, KEPCO attaches various IoT sensors such as partial discharge sensors, bushing sensors, and oil and gas sensors to its facilities to monitor fault signals and operating conditions that occur inside the facilities (transformers, gas insulated switchgear, etc.) in substations.

Through this, sensing data is monitored, collected, and analyzed to detect potential defects in real time to provide real-time predictive maintenance of facility conditions.

Problem: Wired cabling between sensors and diagnostic servers

The more sensors that can monitor the status of facilities in a substation, the more advanced safety diagnostics are possible. However, currently, the data collected from the sensors attached to the equipment is transmitted to the diagnostic server through the wired infrastructure (coaxial cable + optical cable).

According to KEPCO, "Currently, more than 400 sensors are installed in a single 765 kV substation, and the construction of wired cable facilities costs more than 1 billion won (about $1million) per substation," and it is impossible to apply various sensing technologies to all 900 KEPCO substations in this way.

The sensor industry also says that sensor and diagnostic server technologies that can predict potential failures and diagnose abnormalities in the operation of facilities have existed for many years, but the wired connection between sensors and diagnostic servers has prevented the universal adoption of such solutions in practice.

Solution: Acquiring sensing data through a private 5G network

KEPCO built a private 5G network in December 2022 at the Shin-Jungbu substation and attached IoT sensors (including 5G modules) to the facilities, and the data collected from the sensors is delivered to the diagnostic server through the private 5G network.

Wireless connection between IoT sensors and diagnostic server

This eliminates the cost of wired construction by eliminating the need for wired cable construction to transmit sensor data to the diagnostic server, and also makes it possible to operate power facility management very efficiently by eliminating not only the cost but also the time required for wired facility construction and easy sensor installation.

By cost-effectively building a large-scale wireless sensing network through a private 5G network, KEPCO has smartened the operation of power facilities such as predicting the lifespan of facilities by performing diagnosis and fault analysis using AI and big data, and improved the accuracy of real-time facility diagnosis.

3. Maintenance inspection based on quadruped robot (human -> unmanned)

Problem

Currently, humans walk around twice a day to visually check the operating status of substation facilities (analog/digital gauges, temperature of facilities, leaks, cracks, etc.) and record the results manually, resulting in inaccurate result management.

In addition, safety accidents are always occurring at substation sites, which are dangerous, and unmanned maintenance inspections that incorporate new ICT technologies such as robots and AI are required.

Recently, there have been many cases of unmanned inspections using wheeled robot vehicles, cameras, and AI, but the problem is that wheeled robot vehicles have difficulty navigating steep slopes, climbing stairs, and opening and closing doors. Therefore, they are limited to facilities with horizontal or gentle slopes.

Solution: Dynamic sensing using quadruped robots

<KEPCO's unmanned maintenance inspection using private 5G network and quadruped robots>

In complex indoor and outdoor environments (steep slopes, stair climbing, door opening and closing, etc.) where wheeled vehicles are difficult/impossible to operate, autonomous quadrupedal robots walk to predefined facilities to be inspected.

The robot's camera captures facility images (analog/digital gauges, LEDs, valves, lamps, thermal images, etc.) and streams them to the AI server via a private 5G network (FHD, H.264, 30fps, 9.5Mbps). A high capacity uplink is required to deliver the camera images, which is why a private 5G network is required.

Based on the received video, the AI server reads the status values of gauges, etc. to detect facility abnormalities in real time and sends an alarm to the manager when a problem occurs.

The quadruped robot returns to the station after completing the designated facility inspection and is automatically recharged.

In this way, there is no human intervention in the entire process of moving the quadruped robot to the facilities to be inspected, taking video of the facilities, delivering the video through the private 5G network, analyzing the video on the AI server, and automatically charging the quadruped robot.

In other words, the entire process of maintenance inspection of substation facilities has been unmanned.

KEPCO has converted maintenance inspections previously performed by humans to unmanned inspections using camera-equipped quadruped robots and AI technology. As a result, KEPCO improved the working environment by minimizing the simple tasks of field workers and preventing the occurrence of safety accidents in the field.

In addition, by establishing an environment for collecting facility inspection history and status information, the company plans to utilize big data to predict the failure and life of power facilities.

Dynamic Sensing has a major advantage in that it does not require a communication module on the sensor and does not require a wired communication network between the sensor and the AI application, as a mobile device equipped with a camera captures the values/status of sensors located in various places and sends them through a mobile network, and the AI reads the image to read the sensor values/status.

This provides strong scalability by not requiring additional communication costs even when adding more sensors.

The advantage of Dynamic Sensing over the existing "fixed camera + AI method" is that it does not require a large number of cameras to eliminate blind spots.

4. Safety & Access Control

<KEPCO's safety and access monitoring (security) through AI real-time analysis of CCTV images>

High-definition CCTV is installed outside the substation, and CCTV images are transmitted to the AI server in real time through a private 5G network. The AI server automatically detects dangerous situations (falls, collapses, access to dangerous areas, etc.), fires, intrusions, etc. through deep learning video analysis and sends alarms to managers. No need for cabling even as the number of CCTVs increases.

Thank you for visiting Netmanias! Please leave your comment if you have a question or suggestion.

Comment

Related Contents