français
English
Deutsch
español
العربية
中文
1. Project Background
In today's rapidly evolving digital economy, data centers serve as the "digital foundation" that handles core functions including computing power output, data storage, and business operations. Their stable, efficient, and green operation directly impacts corporate business continuity and market competitiveness. As the core facility of the national hub node in the "East Data West Computing" project, Guizhou Telecom Data Center has undergone a significant expansion since its construction began in 2013, growing from six to fifteen machine rooms. With a total planned investment exceeding 1 billion yuan and a floor area of 300,000 square meters, the center currently operates 15 data center buildings housing 14,000 racks and 120,000 servers. It has become the largest and most dynamic national computing power node in southern China. However, with the surge in AI computing demands, continuous expansion and upgrades have led to increased rack capacity, resulting in a sharp rise in high-density servers, liquid cooling systems, UPS power supplies, and variable-frequency air conditioners. These non-linear loads have gradually caused a series of power quality issues after operation, severely affecting operational efficiency and business security, necessitating comprehensive governance measures.
II. Hazards of Power Quality Issues
l Severe harmonic pollution has caused frequent equipment failures. Devices such as UPS systems, switching power supplies, and liquid cooling circulation systems generate excessive harmonics, resulting in grid voltage distortion with a harmonic distortion index (THDi) reaching up to 24%. This has led to frequent instantaneous server cluster outages and false alarms in precision monitoring equipment. A notable incident involved a temporary disruption in computing power scheduling due to harmonic interference, directly impacting the eastern computing power's ability to handle business. These issues have caused brand damage and economic losses for the center, severely eroding customer trust.
l Excessive reactive power loss and soaring operational costs: Data centers experience significant load fluctuations, particularly during AI server cluster startups and liquid cooling system activation, which generate substantial reactive power surges. This results in a power factor as low as 0.81, not only increasing line losses (11.5% loss rate) but also incurring hefty reactive power penalties from grid companies, adding over 1.7 million yuan in annual electricity costs. Moreover, reactive power loss causes equipment overheating, further straining cooling systems and hindering PUE optimization. These issues contradict the center's green and low-carbon development goals, as well as the "dual carbon" strategy and requirements for green data center construction.
l Reduced equipment lifespan and increased O&M costs: Harmonic currents continuously impact critical components including transformers, distribution cabinets, server power modules, and liquid cooling systems, accelerating insulation degradation. This results in a 30% reduction in Mean Time Between Failures (MTBF) and a 22% increase in equipment failure rates, with annual O&M maintenance costs rising by nearly one million yuan. Moreover, traditional capacitor compensation devices exhibit slow response times, fail to adapt to load characteristics, and deliver suboptimal mitigation effects, making them inadequate for meeting the center's high-reliability operational requirements.
III. Solutions
To address the core challenges of Guizhou Telecom's data center—characterized by high density, high load, high reliability, and green operation—along with the operational features of liquid cooling systems and intelligent computing equipment, a comprehensive on-site survey, load data analysis, and multi-solution comparison were conducted. The final decision was made to adopt a collaborative governance solution combining Active Power Filter (APF) and Static Var Generator (SVG). This approach achieves dual optimization of harmonic mitigation and reactive power compensation, ensuring stability, energy efficiency, and scalability while meeting the center's green development and computing capacity expansion needs. The core deployment strategy of this solution is as follows:
This renovation project replaces the existing capacitor-based reactive power compensation system with APF and SVG equipment. Compared to traditional capacitor compensation, the new system offers higher efficiency, faster response, compact size, and longer lifespan. It perfectly meets the requirements of Guizhou Telecom's data center for compact space, high reliability, and green energy efficiency, while also being compatible with the operational characteristics of liquid cooling systems.
Active Power Filter (APF): Deployed with 10 modular APF units delivering a total compensation capacity of 4000A. Featuring intelligent coupling anti-resonance technology and patented heat dissipation design, it achieves 99% filtering efficiency with under 5ms response time. The system precisely captures and eliminates 2nd to 51st harmonics, particularly addressing neutral line 3rd harmonic accumulation for effective mitigation and fire hazard prevention. Supporting 4G/WiFi remote monitoring, it integrates with the central intelligent operation and maintenance platform, enabling real-time harmonic data visualization for "touch-screen management" that meets the center's unmanned operation requirements.
Static Var Generator (SVG): The system features 10 modular SVG units with a total compensation capacity of 3000 kvar. Utilizing IGBT power modules and instantaneous reactive power control technology, it achieves sub-5ms response time for continuous dynamic reactive power compensation, maintaining a power factor above 0.99. The system also provides three-phase imbalance compensation, flexibly adapting to transient load characteristics in data centers. It is particularly effective in mitigating reactive power surges during AI server cluster startup and liquid cooling system activation. Even under generator-fed power supply modes, it ensures stable compensation performance to guarantee reliable power supply stability.
By integrating Guizhou Telecom's data center power distribution architecture and building layout, we implement a' centralized management + zoned compensation' deployment model. This approach, tailored to the load characteristics of liquid-cooled and intelligent computing rooms, ensures comprehensive coverage with no blind spots in management effectiveness.
At the main low-voltage distribution inlet, APF and SVG units are centrally deployed to address harmonic pollution and reactive power loss across the entire data center. This ensures compliance with power quality standards at the Public Connection Point (PCC), thereby supporting stable overall power supply.
In high-load areas like liquid-cooled server rooms and intelligent computing facilities, deploy small APF modules in zones to address localized high-harmonic loads (e.g., UPS clusters, liquid cooling systems, and high-frequency switching power supplies). This prevents harmonic accumulation and safeguards critical equipment.
All devices are integrated into the data center's intelligent operation and maintenance platform, enabling coordinated control of APF and SVG, real-time data monitoring, fault alerts, and remote maintenance. This seamless integration with the existing O&M system reduces operational costs and aligns with the trend of unmanned maintenance.
Compared to the traditional LC filter + SVC compensation approach, the APF-SVG collaborative solution delivers three core advantages, perfectly meeting the high-density, high-reliability, and green requirements of Guizhou Telecom's data center.
Highly intelligent: Enables dynamic load tracking and automatic compensation parameter adjustment without manual intervention. It adapts to the high load fluctuations in data centers and the operational characteristics of liquid cooling systems and intelligent computing devices, preventing both overcompensation and undercompensation to ensure stable power quality.
Enhanced robustness: The APF employs a multi-module parallel architecture, where individual module failures do not compromise overall system performance. The SVG features comprehensive overvoltage, overcurrent, and overheating protection, with an average time between failures (MTBF) of up to 80,000 hours and a design lifespan of 10 years, perfectly meeting the requirements for continuous and stable operation.
Excellent scalability: The modular design enables online capacity expansion with 20% reserved compensation capacity, meeting future demands for 5,000 rack expansions and computing power upgrades in data centers. This eliminates the need to rebuild power distribution systems, reduces upgrade costs, and continuously enhances the center's capacity to support the' East Data West Computing' initiative.
IV. Application Outcomes
After the APF-SVG collaborative governance solution was implemented, 12 months of continuous operation and data monitoring demonstrated that all power quality metrics met national standards. This resolved the project's core challenges during its initial phase, achieving multidimensional improvements in stability, energy efficiency, compliance, and operational maintenance. The solution aligns perfectly with Guizhou Telecom Data Center's goals of green, intelligent, and scalable development, delivering significant economic and social value. Key achievements include:
The APF equipment efficiently filters out harmonics, maintaining the power grid's harmonic distortion rate (THDi) below 3.2%—well below the national standard of 5%. Meanwhile, the SVG precisely compensates reactive power, boosting the power factor to over 0.98 and eliminating voltage distortion and flicker entirely. Since deployment, the server cluster, liquid cooling system, and intelligent computing equipment have experienced zero outages due to power quality issues, with service interruption rates dropping to zero and equipment failure rates plummeting by 70%. This has ensured stable operations for core services including eastern computing capacity, local government services, and enterprise cloud services, significantly enhancing customer satisfaction. By avoiding economic losses and brand risks caused by service interruptions, the center has further solidified its position as a national-level computing node.
The system has achieved a 14.5% reduction in line losses, slashing reactive power losses by 90% and saving 1.7 million yuan in annual electricity costs with an 8.5-month payback period. Reduced equipment heat generation and cooling system load have optimized the center's PUE from 1.2 to 1.15, cutting annual power consumption by over 900,000 kWh. These improvements further support the "dual carbon" strategy and strengthen the green computing hub's competitive edge. Additionally, 30% longer equipment lifespan and 750,000 yuan annual savings in O&M costs, combined with enhanced operational efficiency, have significantly reduced operational burdens and boosted the center's profitability.
All power quality metrics fully comply with the updated national standard "Power Quality-Harmonics in Public Power Grids", successfully avoiding rectification and penalty risks while passing the power grid company's specialized inspection. The modular design with reserved redundancy capacity seamlessly accommodates the data center's 5,000-rack expansion needs without requiring substantial additional investment in power distribution system reconstruction. This provides reliable power support for enhancing intelligent computing capabilities and strengthening the "East Data West Computing" project's capacity, enabling the center to achieve continuous breakthroughs in computing power scale.
By integrating with the existing intelligent operation and maintenance platform at the access center, the system enables remote monitoring, fault alerts, and parameter adjustments for APF and SVG devices. This eliminates the need for on-site personnel, allowing a single operator to handle multiple tasks and reducing labor costs by 40%. The modular design of the equipment supports hot-swapping, simplifying maintenance and replacement processes. This further enhances operational efficiency, reduces maintenance workload, and synergizes with the center's intelligent operation and maintenance system to elevate overall performance.
V. Case Summary and Promotion Value
This case study, centered on Guizhou Telecom's data center, demonstrates that the integrated application of Active Power Filter (APF) and Static Var Generator (SVG) represents the optimal solution for addressing power quality issues in high-density, high-load, and high-reliability green data centers, including harmonic pollution and reactive power loss. It is particularly suitable for scenarios such as data centers, intelligent computing centers, and liquid-cooled data centers supporting the "East Data West Computing" hub node. Compared to traditional solutions, this approach not only achieves precise power quality management but also creates sustained economic value for data centers through energy conservation, extended equipment lifespan, and improved operational efficiency. Additionally, it supports the development of green data centers and aligns with the "Dual Carbon" strategy, perfectly matching the strategic positioning of Guizhou Telecom's data center.
As a benchmark for data center construction and a core component of the "East Data West Computing" initiative, Guizhou Telecom's data center has established replicable and scalable best practices for power quality management in large-scale data centers, cloud computing hubs, intelligent computing centers, and disaster recovery facilities nationwide. With the initiative's deepening implementation and surging AI computing demands, data centers now require increasingly stringent power quality standards. APF (Active Power Filter) and SVG (Static Var Generator) systems have evolved from optional equipment to essential safeguards. Whether for new data center planning or existing facility upgrades, the integrated APF-SVG governance solution precisely addresses operational needs, ensuring stable power infrastructure for high-quality digital economy development while safeguarding the reliable output of every computing resource.
réseau ipv6 pris en charge
