Shaoxing Converter Station and Huawei Partner on UHV O&M Upgrade via Compute-Network Synergy

Faced with a growing number of devices, continuous data growth, and rising O&M complexity, Shaoxing Converter Station partnered with Huawei to integrate networking, computing power, and intelligent applications. This collaboration drives coordinated upgrades across inspection, analysis, monitoring, and execution, pioneering an end-to-end intelligent O&M model for UHV converter stations.

As the new power system accelerates its digital and intelligent transformation, UHV converter stations are grappling with more devices, denser data streams, and increasingly complex service coordination. The traditional O&M model, heavily reliant on manual inspections, manual analysis, and siloed systems, is struggling to deliver the required efficiency, security, and real-time responsiveness.

Located at the receiving end of the ±800KV ULTRA HIGH VOLTAGE DIRECT CURRENT SHAOXING CONVERTER STATION serves as a critical link connecting clean energy from the northwest to the economic powerhouse of eastern Zhejiang. It is also among the first digital converter station pilots of State Grid Corporation of China (SGCC). With the growing scope and intensity of intelligent inspections, equipment analysis, O&M monitoring, and high-risk operations, Shaoxing Converter Station has faced growing demands for real-time connectivity, AI computing power, and intelligent applications.

To meet these demands, Shaoxing Converter Station partnered with Huawei to build an intelligent O&M system that covers inspection, equipment analysis, digital security, and intelligent reporting. Leveraging the integrated networking, general-purpose computing, and AI computing capabilities of Huawei's compute-network convergence infrastructure, the system creates a closed-loop process of monitoring, analysis, alerting, and response. This drives the evolution of O&M from labor-intensive operations toward machine-assisted execution, data-driven decision-making, and intelligent collaboration.

Rising O&M Complexity Calls for a New Model

As drones, robot dogs, HD visual sensing terminals, and tens of thousands of sensors become part of daily operations, Shaoxing Converter Station's sensing reach is growing fast, presenting three key challenges.

1. Massive volumes of data make manual analysis impractical. The sheer volume of inspection images, equipment status data, and alarm alerts renders manual search, comparison, and interpretation highly time-consuming, while subtle anomalies are easily lost in the data flood.

2. The proliferation of terminals overwhelms the network. HD video demands high bandwidth, while sensors require low power consumption and high reliability. The traditional approach of separately building broadband and narrowband networks increases deployment and O&M complexity, making it difficult to flexibly support mobile inspection and real-time sensing.

3. High-risk operations still rely heavily on manual supervision. Tasks performed at heights, in confined spaces, or near live high-voltage equipment present significant safety risks. A single misstep in critical actions, such as engaging/disengaging protection plates on the wrong object or in the wrong sequence, can compromise equipment integrity.

To overcome these challenges, Shaoxing Converter Station no longer needs isolated additions of devices or algorithms, but rather a full-chain intelligent O&M foundation that enables networks, computing power, and applications to work in synergy.

Compute-Network Convergence Lays the Foundation for Data, Compute, and Application Synergy

Guided by the principles of machine-assisted execution, data-driven decision-making, and intelligent collaboration, Shaoxing Converter Station and Huawei jointly built system-level capabilities across three dimensions: computing power, networking, and operation management.

• Compute: Unified AI computing power enables rapid deployment of intelligent applications

Based on Huawei's Atlas intelligent edge servers, the two parties established an AI computing foundation that provides unified computational support for applications such as intelligent inspection, equipment analysis, and intelligent monitoring. Leveraging the independent Compute Architecture for Neural Networks (CANN) and a complete proprietary technology stack, algorithms, models, and applications can be deployed on demand and continuously iterated.

• Network: A converged network ensures stable connectivity for numerous terminals

The two parties jointly developed a networking solution that combines a secure private network based on WLAN Authentication and Privacy Infrastructure (WAPI), a NearLink-based domestic IoT network, and an all-optical network, enabling unified access for both broadband and narrowband devices. WAPI carries high-bandwidth services such as HD video, NearLink supports low-power, high-reliability sensing, and the all-optical network provides high-quality transmission. Together, they ensure real-time data backhaul from drones, robot dogs, visual sensing terminals, and other sensor devices.

• Application: Intelligent operation control proactively prevents errors in critical tasks

For high-risk operations such as engaging/disengaging protection plates, the two parties have built a three-in-one management and control system integrating Integrated Service Delivery Platform (ISDP), AI algorithms, and Bluetooth locks. ISDP standardizes operational procedures, AI verifies operation targets in real time, and Bluetooth locks keep non-target equipment locked out of operation, shifting safety control from post-event investigation to real-time prevention.

With powerful compute, unified connectivity, and intelligent control, the entire service chain—from sensing and transmission to analysis and response—is fully streamlined, delivering greater efficiency across the station.

O&M Transitions from Attended to Fully Intelligent Operations

With the deployment of the compute-network convergence infrastructure, Shaoxing Converter Station has not only achieved isolated efficiency gains, but also comprehensively upgraded inspection, analysis, monitoring, networking, and field operations.

1. Intelligent inspection: Efficiency nearly quadrupled

Previously, the station relied heavily on on-site manual inspections. Machine-based identification was limited in both efficiency and accuracy, requiring extensive human verification. Today, drones, robot dogs, and fixed visual sensing terminals form an integrated space-air-ground intelligent inspection system that handles routine checks with minimal human intervention. Meanwhile, AI rapidly filters out normal images and performs in-depth analysis on suspected anomalies, achieving a fault and defect identification accuracy of 98%. This has cut the entire inspection process from roughly two hours to about 30 minutes, as well as decreasing the frequency of personnel entering high-voltage and confined spaces.

2. Equipment analysis: From 3 hours to 2 minutes

In the past, O&M personnel had to manually retrieve data across multiple systems, compare curves, identify anomalies, perform analysis, and compile reports. Today, the equipment analysis agent automatically aggregates multi-source data, and completes trend comparison, anomaly identification, and report generation, delivering analysis results in about two minutes. Equipment analysis has shifted from manual, piecemeal data retrieval and periodic assessment to proactive system-driven analysis with real-time anomaly alerts.

3. Fault analysis: From experience-based reasoning to intelligent inference

Complex faults once required staff to manually filter alarms, review drawings and operating curves, and consult multiple experts. Root-cause analysis took about half an hour, and event report compilation required at least 24 hours. Today, intelligent monitoring automatically filters out associated alarms, pinpoints root causes in seconds using ontology-based knowledge graphs, and generates event reports in minutes by correlating domain expertise and historical data. The analysis process has shifted from manual clue-hunting and multi-party consultations to proactive system inference and root-cause tracing in seconds, turning individual experience into reusable organizational knowledge.

4. Networking: From multiple separate networks to one converged network

In the past, services such as control commands, status reporting, and HD video ran on separate networks and were maintained independently. Now, the converged network enables unified access for both broadband and narrowband heterogeneous terminals, with data transmission latency controlled within 10 milliseconds. It also supports synchronous backhaul of 16-channel 4K video and ensures data security through Chinese cryptographic algorithms and two-way authentication. The network has evolved from separate construction and dispersed maintenance across multiple networks to unified access, converged service transmission, and secure data transport.

5. Safety control: From manual verification to systematic prevention

Engaging and disengaging protection plates once relied mainly on manual confirmation and verification. Now, with the three-in-one management and control system (ISDP + AI algorithms + Bluetooth locks), safety control has shifted from manual verification and post-event investigation to systematic error prevention and real-time intervention. This significantly mitigates the risk of misoperation during execution.

As a result, the Shaoxing converter station has gradually streamlined the entire monitoring–analysis–alerting–response process, freeing O&M personnel from repetitive inspections, manual data retrieval, and routine verifications while enabling them to focus more on anomaly analysis and critical decision-making.

From Pilot to Benchmark: A Replicable Model for Intelligent O&M

The digital transformation of UHV converter stations is not about simply overlaying terminals, networks, computing power, and AI applications. Rather, it requires genuine synergy across sensing, connectivity, compute, and services. The practice at Shaoxing Converter Station, in partnership with Huawei, has demonstrated that a unified compute and network convergence base can extend intelligent capabilities from isolated scenarios to full-process collaboration across inspection, analysis, monitoring, and execution, building a continuously evolving intelligent O&M system.

Shifting from attended operations to machine-assisted execution, data-driven decision-making, and intelligent collaboration, Shaoxing Converter Station is charting a new course towards intelligent O&M in UHV converter stations. This effort also provides a practical foundation for replicating these capabilities to other converter stations, substations, and a broader range of power grid scenarios.

Going forward, the two parties will continue to drive innovation across intelligent inspections, equipment analysis, monitoring, and field operations, advancing AI from single-point intelligence to end-to-end intelligence and building a safer, more efficient, and more intelligent digital foundation for the new power system.