Medium-Voltage Switchgear in Data Centers

Architecture, Protection, Redundancy, and O&M for Mission-Critical Power Distribution

1. Introduction: The First Line of Mission-Critical Power

Every megawatt that reaches a server rack passes first through one asset: the medium-voltage switchgear (MVSG). Positioned immediately downstream of the utility service entrance or on-site generation, the MVSG is the first major distribution asset in the data center power chain and the point at which raw utility power becomes managed, protected, and distributable capacity. Its role goes far beyond opening and closing circuits. A properly engineered MVSG receives, isolates, protects, controls, meters, and distributes medium-voltage power so that critical IT loads can run continuously, 24 hours a day, 365 days a year.



As data center capacities climb from tens to hundreds of megawatts — driven by AI training clusters, high-density colocation, and hyperscale campuses — the medium-voltage layer has become the defining reliability boundary of the facility. Decisions made at the MVSG level (bus architecture, protection philosophy, arc-flash mitigation, automation logic, and maintainability provisions) directly determine whether a facility can meet its concurrent-maintainability and fault-tolerance objectives. A single misapplied relay setting or a poorly conceived interlocking scheme at medium voltage can defeat millions of dollars of downstream redundancy in UPS and generator plants.

This article presents a comprehensive engineering treatment of MV switchgear in data centers: where it sits in the power chain, how a lineup is organized section by section, the redundancy architectures used to achieve Tier III and Tier IV objectives, equipment classes and ratings, protection and control design, automation and transfer schemes, and the operations and maintenance (O&M) program required to keep the system healthy across a 25–30 year service life.

Keentel Insight

The MVSG is the highest-consequence single point in the distribution chain. Downstream, redundancy multiplies — multiple transformers, UPS modules, and PDUs share the load. Upstream of the MV bus, a design or settings error affects everything at once. Engineering rigor at this level pays the highest reliability dividend per dollar in the entire facility.

2. Where the MVSG Sits in the Data Center Power Chain

A typical utility-fed data center power chain flows from the utility interconnection through medium-voltage distribution, transformation, low-voltage distribution, power conditioning, and finally to the IT load. The MVSG occupies the pivotal position between the sources (utility and generators) and everything the facility owns downstream.

Two trends are pushing more engineering weight onto the MV layer. First, campus scale: at 100+ MW, distributing at 480 V is physically impossible, so nearly all inter-building and even in-building backbone distribution is now at medium voltage, sometimes with MV UPS or MV-connected generation. Second, on-site generation integration: standby diesel or gas generation, and increasingly fuel cells and BESS, are paralleled at the MV bus, making the MVSG the point of common coupling for multiple sources and the natural home for source-transfer automation.

3. Anatomy of an MV Switchgear Lineup

An MV switchgear lineup is an integrated assembly of standardized vertical sections (cubicles), each performing a specific function, connected by a continuous busbar system and supervised by a protection, control, and monitoring layer. Understanding the lineup section by section is the foundation for safe operation, effective maintenance, and fast fault response.

3.1 Incoming Sections

Incoming sections receive power from the utility network or from on-site generation and connect it to the busbar. Each incoming section typically contains a withdrawable (drawout) vacuum circuit breaker, current transformers (CTs) for protection and metering, voltage transformers (VTs) for protection, synchronizing, and metering, and surge protection. In dual-source architectures there are two incoming sections — one per source — each sized to carry the full facility load when the other source is out of service.

The incoming breaker is the interface between the external source and everything the facility owns. Its protection package must coordinate upward with the utility's protection (or the upstream main transformer protection) and downward with feeder protection on the same bus. Directional overcurrent and reverse-power elements are commonly applied where on-site generation can back-feed toward the source.

O&M focus:

source availability and power quality monitoring, breaker status and readiness, alarm and trip-event monitoring, protection relay health, and preventive maintenance on the breaker and primary connections.

3.2 Busbar System

The busbar is the copper (or aluminum) backbone of the lineup. It carries the total load current of every connected feeder and distributes power along the length of the switchgear. Bus is insulated and braced to withstand the mechanical forces of maximum through-fault current — forces that scale with the square of the current. Bus joints are the most common thermal weak point in aging switchgear: a joint that loses clamping force develops resistance, heats, oxidizes, and accelerates toward failure.

O&M focus:

infrared or continuous thermal monitoring of joints and connections, verification of connection tightness at planned outages, insulation condition assessment (including partial discharge survey on higher-voltage lineups), and arc-flash risk management for any work near the bus zone.

3.3 Bus Coupler (Bus Tie)

The bus coupler — also called the bus tie — connects two busbar sections so load can be transferred between them. It is the keystone of the classic main-tie-main architecture. In a dual-bus data center configuration, the tie normally operates open, so each incoming source serves its own bus section independently; the tie closes only under specific, controlled conditions — typically loss of one source, or a planned transfer to release one source for maintenance.

Tie operation raises two engineering questions that must be answered in design, not discovered in operation. First, paralleling: if the tie closes while both incomings are closed (closed-transition transfer), the two sources are momentarily paralleled, which requires synchronism-check supervision (device 25) and a short-circuit duty check, because paralleled sources can raise available fault current beyond equipment ratings. Second, interlocking: a two-out-of-three logic (of the two mains and the tie, only two may be closed at once) is commonly enforced by a combination of electrical interlocks and mechanical key interlocks, unless the design has been explicitly rated and studied for continuous parallel operation.

O&M focus:

automatic and manual operation testing, interlock and transfer-logic validation, synchronizing scheme checks where applicable, and periodic verification of the transfer conditions and timers against the current facility configuration.

3.4 Outgoing Feeder Sections

Outgoing feeders supply medium-voltage power to the unit substation transformers that in turn feed critical LV systems — UPS inputs, mechanical cooling plant, and house power. Each feeder section contains its own breaker, CTs, protection relay, and metering, so every downstream transformer is individually protected and its loading individually visible. Feeder protection must sit below the incoming protection and above the transformer's own protection in the coordination hierarchy, so a transformer or cable fault trips only the affected feeder and nothing else.

O&M focus:

load monitoring and trending against feeder and transformer ratings, protection settings management under configuration control, breaker operation counts and condition, and disciplined fault investigation whenever a feeder relay operates.

3.5 Instrument Transformers and Metering

CTs and VTs are the sensory system of the switchgear. CT ratio, accuracy class, and saturation performance must be matched to the protection functions they serve; a CT that saturates during a close-in fault can delay or misdirect a trip. VTs provide the voltage references for protection (27/59/81 elements), synchronism check, power metering, and automation logic. Data center owners increasingly specify revenue-class metering on incomings and feeder-level power-quality monitoring to support capacity management and SLA reporting.

3.6 Protection, Control, and Monitoring

The protection, control, and monitoring layer is the brain of the MVSG. Modern lineups use microprocessor-based multifunction relays on every breaker, communicating over a station network to a SCADA or electrical power monitoring system (EPMS). The system continuously monitors currents, voltages, breaker status, and equipment health; detects faults; and ensures selective isolation — clearing the faulted zone without disturbing healthy parts of the system. Section 6 treats protection design in depth.

O&M focus:

relay health and self-diagnostic supervision, event and alarm log review, settings validation against the coordination study of record, firmware and configuration management, and communication-network status.

Keentel Insight

Treat the lineup as a set of protection zones, not a row of cabinets. Every maintenance plan, switching order, and arc-flash boundary should be written zone by zone — incoming zone, bus zone, feeder zone — because that is how the protection sees the system and how faults will actually be cleared.

4. Redundancy Architectures: From Single Bus to 2N

The single-line architecture of the MV system is the single most consequential reliability decision in the electrical design. It determines whether the facility can survive a source loss without interruption, whether any MV component can be maintained without shedding critical load, and how faults propagate — or are contained.

4.1 Common MV Architectures

4.2 Matching Architecture to Availability Objectives

Concurrent maintainability — the ability to take any component out of service on a planned basis without impacting critical load — requires that every MV element (incoming breaker, bus section, tie, feeder) have a maintenance bypass path. Main-tie-main achieves this for sources and, with dual-ended transformer feeds, for feeders; the bus sections themselves become the residual constraint, which is why higher-availability designs sectionalize further or move to full 2N.

Fault tolerance — riding through an unplanned failure of any single component without load impact — is a stricter test. A bus fault on a shared MTM lineup interrupts every load on that bus section until transfer completes; only a true 2N separation (independent lineups, separate rooms or at least fire-rated separation, no shared control power or automation) contains an MV bus failure without any critical-load exposure. This is the fundamental reason fault-tolerant facilities carry the A/B separation all the way up to the medium-voltage and source level, not just from the UPS down.

Physical and systems separation matters as much as the single-line diagram: independent DC control-power batteries per lineup, physically separated raceway for A and B feeders, separate relay and automation networks, and no common-mode dependencies in the EPMS. Many facilities that are 2N on paper are N on control power.

Keentel Insight

Audit the auxiliary systems with the same rigor as the primary system. Shared station batteries, a common automation PLC, or A and B feeders in the same duct bank are the classic hidden single points of failure that quietly convert a 2N design into an N design.

5. Equipment Classes, Ratings, and Construction

5.1 Metal-Clad vs. Metal-Enclosed

In the ANSI/IEEE world, metal-clad switchgear (IEEE C37.20.2) provides compartmentalization of the breaker, bus, and cable sections with grounded metal barriers, drawout vacuum breakers, insulated bus, and automatic shutters — features that directly support maintainability and internal fault containment. Metal-enclosed interrupter switchgear (IEEE C37.20.3) uses load-interrupter switches and fuses in a simpler, lower-cost construction, generally applied at smaller sites or for simple transformer primary protection rather than as the main distribution asset of a large data center. IEC-market equivalents are classified under IEC 62271-200, with loss-of-service-continuity (LSC) categories and partition classes defining comparable compartmentalization concepts.

5.2 Insulation Medium: AIS, GIS, and SF6-Free Designs

Air-insulated switchgear (AIS) remains the default for 5–15 kV class data center lineups. Gas-insulated switchgear (GIS) offers a dramatically smaller footprint, sealed primary parts that are immune to dust and humidity, and reduced maintenance — attractive at 34.5 kV and above, in space-constrained urban facilities, or in harsh environments. The industry is actively moving away from SF6 toward vacuum interruption with clean-air or fluoronitrile-based insulation; owners specifying new campuses should weigh the sustainability reporting impact of SF6 inventory against the maturity and serviceability of the newer gas-free platforms.

5.3 Core Ratings

Two rating pitfalls recur in data center practice. First, momentary paralleling during closed-transition transfer: with both utility incomings and the tie closed, fault current from both sources sums on the bus, and the equipment must be rated for that condition even if it lasts only cycles. Second, growth: AI-era campuses routinely double planned density mid-design; specifying bus continuous current and fault duty for the ultimate phase — not the day-one phase — is far cheaper than replacing switchgear in a live facility.

5.4 Arc-Resistant Construction and Personnel Safety

Arc-resistant switchgear per IEEE C37.20.7 is tested to contain and redirect the pressure wave and incandescent gases of an internal arcing fault through plenums and vents away from operators. Type 2B accessibility maintains protection with low-voltage instrument compartment doors open — important because relay work is the most common live-lineup activity. Arc resistance complements, but does not replace, arc-energy reduction in the protection design (Section 6.4): the enclosure protects people from the blast; the protection limits the energy and the damage.

6. Protection and Control Design

6.1 Protection Philosophy: Selectivity First

Data center MV protection is designed around one governing principle: a fault must be cleared by the nearest upstream device, and only that device, as fast as the coordination allows. Selectivity preserves the maximum amount of healthy system; speed limits equipment damage and arc energy. These two goals trade against each other in classic time-coordinated schemes, which is why modern designs supplement time-overcurrent coordination with bus differential, zone-selective interlocking, and arc-flash detection to get both.



6.2 Core Protection Functions

6.3 Coordination in a Compressed System

Data center MV systems are electrically short — from utility incoming to transformer primary may be only two or three coordination steps — yet the consequences of miscoordination are extreme. The coordination study must reconcile utility upstream requirements (which cap how slow the incoming can be) with transformer damage curves and inrush (which floor how fast a feeder can be). Bus differential (87B) is the cleanest way out of the squeeze: it clears bus faults in a few cycles regardless of the time-overcurrent ladder, letting feeder and incoming overcurrent settings coordinate comfortably. Zone-selective interlocking (ZSI) achieves a similar effect at lower cost by letting a feeder relay block the incoming's fast element when the fault is in the feeder's zone.

6.4 Arc-Flash Mitigation

Incident energy at the MV lineup is governed by fault current and clearing time (per IEEE 1584 methods). Because fault current is fixed by the system, clearing time is the lever. Effective mitigation layers include arc-flash detection relays combining light and current criteria (clearing in 2–4 cycles), maintenance-mode switches that temporarily enable an instantaneous element while personnel work near the equipment, bus differential protection, remote racking and remote switching so operators stand outside the arc-flash boundary, and arc-resistant construction as the last line of defense. Every layer should be reflected in the facility's arc-flash study and labeling, and the maintenance-mode procedure must be written into switching orders — a mitigation feature nobody engages protects nobody.

6.5 Automation, Transfer Schemes, and Interlocking

The main-tie-main automatic transfer scheme is the signature automation of the data center MVSG. On loss of a source (supervised by 27 elements with security timers to ride through remote faults), the scheme opens the dead incoming and closes the tie, re-energizing the affected bus from the healthy source in a defined time. Design decisions that must be made deliberately include open vs. closed transition for planned transfers, transfer inhibit conditions (bus fault lockout — never transfer onto a faulted bus), load-shed or staged-restart supervision for transformer inrush on re-energization, and manual return-to-normal philosophy. Generator integration adds paralleling logic, load ramping, and protection mode changes (grounding and fault-level differences between utility and generator sources).

Interlocking enforces the safe operating states: electrical interlocks in the transfer logic, mechanical key interlocks as the human-proof backstop, and drawout position interlocks native to metal-clad construction. On IEC 61850-based lineups, GOOSE messaging increasingly carries interlocking and ZSI signals over the station network; the design must then treat network health as protection-critical, with message supervision and defined fail-safe states.

Keentel Insight



Automatic transfer schemes fail in the corners, not the center. The scheme that works perfectly in the factory test can misoperate on a real system event — a slow-decaying bus residual voltage, a simultaneous utility dip on both sources, a VT fuse failure. Demand a documented cause-and-effect matrix and test every row of it during commissioning, including the ugly rows.

7. Operations and Maintenance: Keeping the Backbone Healthy

7.1 Why MVSG O&M Is Different in a Data Center

Data center MV switchgear runs continuously loaded, cannot tolerate casual outages, and is operated by teams whose daily work is dominated by mechanical and IT systems. The O&M program must therefore be deliberate: condition-based where possible, outage-efficient when de-energized work is required, and procedural everywhere. The payoff is direct — MV switchgear maintained on a disciplined program is among the most reliable equipment in the facility; neglected, its failures are among the most catastrophic.



7.2 Section-by-Section O&M Priorities

7.3 Condition Monitoring Technologies

• Thermal — Continuous thermal sensing at bus joints and cable terminations (or IR windows for safe periodic scans) catches the dominant slow-failure mode of the bus system.
• Partial discharge — Partial discharge (PD) monitoring — online sensors or periodic surveys — detects insulation degradation years before failure, particularly valuable at 15 kV class and above and in humid environments.
• Breaker analytics — Breaker condition indicators — operation counts, charging-motor current signatures, travel/timing analysis — move breaker overhaul from calendar-based to condition-based.
• Control power — DC control power is the silent kingmaker: a failed station battery makes every protection and transfer function inoperative simultaneously. Battery monitoring and load testing belong at the top of the program, not the bottom.

7.4 Safe Work Practices

MV switchgear work is governed by NFPA 70E arc-flash and shock protection requirements. Racking breakers is historically the highest-risk routine task; remote racking devices and arc-resistant construction have transformed that risk profile and should be standard in new facilities. Every switching evolution — planned transfer, isolation for maintenance, return to normal — should follow a written, peer-reviewed switching order, with the maintenance-mode protection setting engaged whenever personnel work within the arc-flash boundary of energized equipment.

7.5 What Disciplined MVSG O&M Delivers

• Continuous power to critical IT loads through source disturbances and planned work.
• Reduced risk of unplanned downtime from the highest-consequence failure modes in the facility.
• Safe switching and maintenance with predictable, bounded risk to personnel.
• Extended equipment lifecycle — 25–30 years is achievable on a condition-based program.
• Demonstrable compliance with reliability and safety frameworks (Uptime Institute objectives, ISO management systems, IEEE/NETA maintenance standards, NFPA 70E).

8. Commissioning: Proving the System Before It Matters

Commissioning is where the MVSG design is proven — or where its latent defects are found at the only acceptable time. A rigorous program spans five levels: factory acceptance testing (FAT) of the assembled lineup including functional automation tests; site acceptance and pre-energization testing per ANSI/NETA ATS (insulation resistance, contact resistance, breaker timing, CT/VT ratio and polarity, primary or secondary injection of every protection element); protection settings validation against the coordination study of record; functional performance testing of the complete transfer scheme against the cause-and-effect matrix, including failure-mode rows; and integrated systems testing (IST), where the MVSG, generators, UPS, and mechanical plant are exercised together through pull-the-plug scenarios at full facility load banks.



Settings management deserves special emphasis: the relay settings energized on day one must be the studied settings, under configuration control, with as-left files archived. A surprising fraction of real-world misoperations trace to settings that drifted from the study — placeholder settings never updated, or field changes never fed back into the study.

Keentel Insight

Insist that the MV transfer scheme be tested at the switchgear with primary sources, not only simulated in the factory. Utility source behavior, VT fusing, DC system sags, and real breaker timing all differ from the test floor — and integrated systems testing is the only stage that exercises them together.

9. Design Checklist for Data Center Owners and Engineers

• Architecture — Define the availability objective first (concurrent maintainability vs. fault tolerance) and derive the MV architecture from it — not the reverse.
• Ratings — Rate bus continuous current and short-circuit duty for the ultimate build-out and for paralleled-source transfer conditions.
• Arc safety — Specify arc-resistant (Type 2B / IAC AFLR) construction and layer protective arc-energy reduction: AFD relays, maintenance mode, 87B or ZSI, remote racking.
• Protection — Require a complete coordination and arc-flash study before settings are energized, and put settings under configuration control from day one.
• Automation — Document the transfer scheme in a cause-and-effect matrix and test every row at commissioning, including failure modes.
• Separation — Eliminate hidden common modes: independent DC systems per lineup, separated raceways, independent automation and relay networks for A and B systems.
• Maintainability — Design for maintainability: drawout breakers, IR windows, online PD provisions, spare cubicles, and a critical-spares strategy including a spare breaker per frame size.
• O&M by design — Plan the O&M program during design — monitoring points, test access, and switching procedures are far cheaper to provide on paper than to retrofit.
  

10. Standards and References Landscape

11. How Keentel Engineering Supports Data Center MV Systems

Keentel Engineering LLC provides end-to-end medium-voltage engineering for data centers and other mission-critical facilities: MV system architecture and single-line development; short-circuit, coordination, and arc-flash studies; switchgear specification and procurement support; protection settings development and relay programming; transfer-scheme logic design and cause-and-effect documentation; commissioning and integrated systems testing support as owner's engineer; and O&M program development including condition-monitoring strategy and switching procedures. Our team also supports utility interconnection for data center campuses — from point-of-interconnection studies through NERC compliance — so the MV design is coordinated with the transmission-side realities that shape it.



Whether you are planning a new campus, expanding an operating facility, or hardening an existing MV lineup against arc-flash and reliability risk, Keentel brings licensed professional engineering judgment and current, standards-based practice to the highest-consequence layer of your power system.

12. Key Takeaway

An MV switchgear lineup is an integrated system of switching, protection, control, and monitoring engineered to deliver safe, reliable, and uninterrupted power to mission-critical facilities. Know your MV switchgear — its architecture, its protection zones, its transfer logic, and its condition — and operate it with discipline, and your data center stays on. Neglect it, and no quantity of downstream redundancy will save you. The medium-voltage layer is where data center reliability is truly won or lost.

Case Studies

The following case studies are drawn from representative mission-critical engineering engagements. All client, project, location, and personnel identifiers have been removed or generalized to protect confidentiality.

Case Study 1: Main-Tie-Main MV Design for a Hyperscale Campus Expansion

Challenge. The owner's phase-one design had sized bus continuous current and short-circuit duty for the initial 20 MW deployment. Mid-design, planned rack densities more than doubled. Preliminary short-circuit analysis showed that closed-transition transfers at ultimate build-out would exceed the specified 25 kA switchgear rating during the paralleled interval, and the two-step coordination between feeder and incoming relays left no margin against the utility's upstream clearing requirements.

Approach. The MV single-line was re-architected around 40 kA, 3000 A arc-resistant lineups rated for the ultimate phase, with bus differential (87B) protection added to each lineup to decouple bus-fault clearing from the time-overcurrent ladder. The transfer scheme was formalized in a cause-and-effect matrix covering source loss, bus-fault lockout, VT failure, and DC-loss scenarios. Feeder settings were coordinated against transformer damage and inrush curves, and maintenance-mode arc-energy reduction settings were engineered for every bus.

Outcome. The final design achieved full selectivity with utility approval on the incoming settings, closed-transition transfer capability within equipment ratings at ultimate build-out, and calculated incident-energy reductions of more than 80 percent at the lineups in maintenance mode. The owner avoided a mid-life switchgear replacement that the phase-one sizing would have forced.

Case Study 2: Arc-Flash Mitigation Retrofit on an Operating Colocation Facility

Challenge. An updated IEEE 1584 study showed incident energy at the main lineup exceeding the practical limits of available PPE, driven by slow electromechanical relay clearing and settings that had drifted from any study of record. The facility could not accept extended outages, and full switchgear replacement was not fundable in the near term.

Approach. A staged retrofit replaced feeder and incoming relays with multifunction microprocessor relays incorporating arc-flash detection (light plus current) and a supervised maintenance-mode instantaneous group, executed bus section by bus section using the tie to keep tenants energized. CT circuits were verified and settings were rebuilt from a new coordination study. Remote racking equipment and rewritten switching orders completed the personnel-safety layer.

Outcome. Calculated incident energy at the lineup dropped from beyond PPE limits to within standard arc-rated PPE levels in maintenance mode, with normal-mode selectivity fully preserved. The retrofit was completed without any tenant load interruption, and the facility adopted a configuration-controlled settings and testing program aligned with ANSI/NETA MTS.



Case Study 3: Owner's Engineer for MV Commissioning with Generator Paralleling

Challenge. Factory testing had validated the transfer scheme against simulated inputs only. During site review, the cause-and-effect matrix was found to omit several credible failure modes — VT fuse loss during transfer, DC system sag under simultaneous breaker operations, and GOOSE network degradation affecting interlocking — and generator-mode ground-fault protection had not been verified against the plant's actual grounding configuration.

Approach. The commissioning plan was expanded to test every matrix row at the switchgear with primary sources, including deliberately induced failure cases. Secondary injection proved each protection element in both utility and generator source modes; GOOSE message supervision and fail-safe states were configured and demonstrated. Integrated systems testing exercised pull-the-plug transfers and generator paralleling at full load-bank load, with relay event records analyzed after every evolution.



Outcome. Testing surfaced and resolved a transfer misoperation path (a security timer set below the utility's reclose interval) and a generator-mode ground-fault sensitivity gap — both before the facility carried live load. The owner accepted the system with a fully executed cause-and-effect record, as-left settings archive, and a baseline breaker-timing and PD dataset for the future condition-based maintenance program.

FAQ

About Keentel Engineering

Keentel Engineering is a power systems and grid interconnection consulting firm headquartered in Tampa, Florida, with offices in Austin, Sacramento, and Baltimore. Our service lines span grid interconnection engineering, substation and transmission design, power system studies NERC compliance, renewables and BESS engineering, EMT modeling, and owner's engineer services. We serve developers, independent power producers, utilities, and large-load customers across North American RTO/ISO footprints.