DL/T 2908-2025 in English

Standard Development Background and Technology Evolution Analysis

DL/T 2908-2025, "Technical Requirements for Unmanned Aerial Vehicle Image Recognition Systems for Power Grid Equipment," is the first dedicated technical standard for drone image recognition systems in the power industry, marking a new stage in the standardization of smart grid inspection technology. This standard, issued by the National Energy Administration on June 30, 2025, will officially take effect on December 30, 2025. It is managed by the Energy Industry Technical Committee for Standardization of Smart Grid Inspection Equipment (NEA/TC 41).

With the widespread adoption of drone technology in power grid inspection, traditional manual inspection models are gradually transitioning towards intelligent and automated systems. This standard builds on the technological advancements of advanced data acquisition devices such as visible light cameras and infrared thermal imagers, combined with edge computing and artificial intelligence technologies, to establish a comprehensive drone image recognition technology system for power grid equipment. The standard drafting units include Guangdong Power Grid, State Grid Jiangsu Electric Power Research Institute, DJI and other leading companies in the industry, ensuring the advancement and practicality of the technical requirements.

---

System Architecture and Core Technology Requirements

The standard clearly states that the drone image recognition system for power grid equipment consists of three core parts: data acquisition, processing and application implementation. The system architecture adopts a hybrid model that combines centralized processing with edge computing, which not only ensures the high efficiency of large-scale image processing, but also meets the low latency requirements of on-site real-time analysis.

Centralized Processing Layer: Inference Servers and Storage Systems; Storage ≥ 1PB, Transfer Rate ≥ 8Gb/s; Batch Image Recognition and Model Training; Application Service Layer: Application Servers and Management Platform; Concurrent Users ≥ 500, Availability ≥ 99.9%; Task Management and Statistical Analysis; In terms of hardware configuration, the standard establishes differentiated edge device configuration requirements based on annual image processing volume: Edge devices are mandatory for annual processing volume exceeding 80 million images, while edge devices are recommended for processing between 20 and 80 million images. This tiered configuration strategy takes into account both system performance requirements and construction cost optimization.

---

In-Depth Analysis of Functional Requirements

Image Processing Functional System

The functional requirements specified in the standard cover the entire process, from image acquisition to analysis report generation. The centralized processing system must have core functions such as model management, publishing, image organization, compression, and upload. Model management supports both image import and model import, ensuring unified management and version control of algorithm models.

Image compression must be performed while maintaining a reduction in recognition accuracy and recall below 1%. This metric reflects the standard's balanced consideration of image quality and storage efficiency. Furthermore, the system must preserve key metadata such as image acquisition coordinates and temperature matrix to provide complete information support for subsequent data analysis and fault diagnosis.

Edge Device Functional Characteristics

As a crucial component of the system, edge devices must implement functions such as automatic image compression and upload, model synchronization, and inference result transmission. The standard specifically emphasizes cloud-edge collaboration, supporting automatic and elastic adjustment of the number of inference service nodes based on service load, demonstrating the elastic scalability of modern distributed systems.

---

Comparison of performance indicators and technical requirements

System level	Core components	Technical indicators	Application scenarios
Data acquisition layer	Visible light camera, infrared thermal imager	Pixel ≥ 20Mp, resolution ≥ 640×480px	Equipment appearance inspection, fever detection
Edge processing layer	Smart gateway, airborne computing chip	Computing power ≥ 12TOPS, memory ≥ 32GB	On-site real-time analysis, data preprocessing

Performance dimensions	Standard requirements	Technical significance	Test methods
Response performance	Page loading ≤3-5s, image recognition ≤5s	Ensure user experience and inspection efficiency	GB/T 39788 Performance Test
System stability	Resource utilization ≤70%, availability ≥99.9%	Ensure continuous and reliable operation of the system	Continuous monitoring and stress testing
Recognition accuracy	Error rate ≤1%, manual review error rate ≤0.1%	Ensure defect recognition accuracy	Accuracy and recall rate calculation
Concurrency	Maximum user access ≥500	Support large-scale collaborative operations	Load testing and concurrency testing

Of particular note among the performance requirements is that the average turnaround time for the system to recognize a single image is less than 5 seconds. This indicator places high demands on algorithm optimization and hardware configuration. At the same time, the requirement for annual system availability of no less than 99.9% (i.e., annual downtime of no more than 8.76 hours) reflects the power industry's stringent standards for system reliability. Chapter 8 of the standard defines the system's security requirements, which must comply with information security technical standards such as GB/T 20270 and GB/T 36572, and meet the requirements of Level 2 cybersecurity protection. This means that the system must pass the corresponding level of cybersecurity protection assessment to ensure the security of the entire data transmission, storage, and processing process. Regarding data synchronization, the system must regularly exchange data with third-party systems, including grid equipment records, defect and hidden danger categories, and patrol mission status. This data integration requirement reflects the standard's emphasis on system interoperability and lays the foundation for building a unified smart grid inspection platform.

---

Implementation Recommendations and Best Practices

System Construction Planning

It is recommended that all units conduct a detailed business needs analysis in the early stages of system construction and rationally plan edge device configuration based on the annual image processing volume. Large power grid companies with annual processing volumes exceeding 80 million images should adopt a distributed edge computing architecture and deploy intelligent gateway devices at all levels of substations and line inspection centers.

Technology Selection Considerations

When selecting hardware, focus on the effective pixel count and optical zoom capability of the visible light camera, as well as the temperature measurement sensitivity and resolution of the infrared thermal imager. Furthermore, the computing power configuration of the inference server should match the expected image processing volume to avoid under- or over-provisioning of resources.

Operation and Maintenance Management Strategy

Establish a comprehensive model update mechanism and regularly optimize and upgrade the recognition algorithm. Furthermore, formulate a strict data backup and disaster recovery plan to ensure rapid resumption of business operations in the event of a system failure. It is recommended to establish a cross-departmental collaborative working mechanism to integrate professional forces such as operation and inspection, information, and security to jointly promote the efficient operation of the system.

The technical requirements of this standard provide comprehensive specifications for the design, construction, and acceptance of drone image recognition systems for power grid equipment. It will effectively promote the standardization and large-scale application of intelligent inspection technologies in the power industry and provide technical support for the safe and stable operation of the power grid.