Collection

Power and UPS

Power distribution, UPS, batteries, protection and redundant supply.

Sources
418
Downloadable PDFs
95
Years
1984–2027

All sources

217–240 / 418

Technical guidePDF

Open Compute Project Foundation 2026

Open Rack Wide (ORW) Base Specification, Revision V1.0.1 (effective May 5, 2026)

Add an ORW rack family whose footprint, height and payload come from the chosen design specification rather than fixed defaults, prompt users for those values, and warn that floor-loading and aisle-pitch checks depend on them.

  • The frame accepts air-cooled or liquid-cooled 48 V busbars; busbar mounting interfaces are identical to ORV3 HPR.
White paperFree to read

Eaton 2020

Parallel UPS Systems

Add maxParallelUnits, paralleling architecture (distributed or centralised bypass) and static switch rating to the UPS catalogue.

  • N+1 alarm: five 100 kVA modules alarm when load exceeds 400 kVA.
Technical guideFree to read

Caterpillar 2018

Paralleling Generator Set Systems and Design

Build an architecture selector with the three options and show consequences: LV-only limit for on-set, required breaker closing speed, and master-controller need for complex sequences.

  • The connecting device needs about 5-cycle close/open speed to avoid out-of-phase paralleling.
Research paperFree to read

arXiv preprint (Microsoft Azure) 2023

POLCA: Power Oversubscription in LLM Cloud Providers

In the power planner, apply different diversity and oversubscription factors by workload: inference rows can be provisioned below sum-of-nameplate (for example peak about 80% of provisioned), while training rows should assume near-coincident peaks with only about 10% headroom.

  • POLCA allows 30% more servers in the same inference GPU cluster while maintaining SLOs.
White paperFree to read

Schneider Electric 2024

Power and Cooling Capacity Management for Data Centers

For every rack, row and upstream device store design maximum supply, design maximum demand (nameplate or as-configured) and measured actual where available; enforce supply >= demand at each hierarchy level including under N-1.

  • Safety margins are typically 10-20%, up to 30% or more with weak change control.
Research paperFree to read

arXiv preprint (Microsoft / OpenAI / NVIDIA) 2025

Power Stabilization for AI Training Datacenters

For AI-training halls, carry a 'synchronous load' flag and capture utility time-domain (MW/s ramp up/down, dynamic range) and frequency-domain (critical band, spectral cap) limits as interconnection inputs.

  • GPU power smoothing with the floor at 90% of TDP gave a total energy overhead of 10.5% in simulation.
Research paperFree to read

arXiv preprint (Emerald AI / EPRI / National Grid / NVIDIA / Oracle; submitted to IEEE) 2026

Power-Flexible AI Data Centers: A New Paradigm for Grid-Responsive Compute

Add a 'flexible load' profile in the grid-connection module: committed reduction depth (%), response time (s), duration (h) and recovery, derived from the workload priority mix.

  • Achieved 100% compliance across 200+ distinct National Grid power events on a 130 kW cluster of NVIDIA Blackwell Ultra GPUs.
Research paperPDF

IEEE Transactions on Industry Applications (Institute of Electrical and Electronics Engineers (IEEE)) 2026

Practical Battery Arc Flash Models

In battery-room arc-flash labelling, calculate the self-extinguishing model alongside the maximum power method rather than using the latter as the only reference, and check cases where excessive PPE categories make maintenance impractical.

  • An arc-in-a-box coefficient of 1.52, based on the value calculated for panelboards, was used for enclosed-space experiments.
Technical guideFree to read

Consulting-Specifying Engineer 2016

Preventing arc flash in mission critical facilities

Run a preliminary arc-flash estimate at design stage and flag white-space panels above about 1.2 cal/cm2 or above PPE category 2.

  • NEC 240.87 (2014) requires arc energy reduction on overcurrent devices rated 1,200 A or higher.
Technical guideFree to read

Consulting-Specifying Engineer 2012

Preventing arc flash incidents by design

Arc-flash outputs should label both line and load sides of mains and compare normal versus maintenance-mode settings.

  • Incident energy above 1.2 cal/cm2 can cause second-degree burns; IEEE 1584 suggests a 2 s maximum arcing exposure time.
Research paperPDF

Fire (MDPI AG) 2026

Probabilistic Risk Assessment of Grid-Scale Lithium-Ion Battery Energy Storage System Fire Hazards: Hydrogen Fluoride (HF) Toxicity, Suppression Effectiveness, and Comparative Compartment Design Analysis

Compare BESS compartmentation quantitatively: two compartments halve HF dose and IDLH clearance time, but IDLH is still exceeded for anyone in the room in full thermal runaway, so runaway prevention (BMS integrity, UL 9540A containment, emergency power off) must stay the primary control.

  • Mean IDLH (25 mg/m³) clearance time with 9 ACH was 599 min for one compartment and 301 min for two; the probability of clearing within 5 min was 0% for both.
Technical guideFree to read

Uptime Institute Journal 2020

Rack Density is Rising (Andy Lawrence)

Ask users for modal and peak rack density separately, not just the average, and size cooling and power for the peak cluster.

  • Mean density was calculated after excluding respondents above 30 kW as outliers.

All content and rights belong to the respective authors and publishers. SoSo CAD provides only the citation, its own summary and a link to the source, and is not responsible for the sources' content.

Monthly newsletter · free

Keep your reading list current

Once a month: new library sources, changes to Turkish regulations and upcoming data centre events.