Reliability Assessment of the Configuration of Dynamic Uninterruptible Power Sources: A Case of Data Centers

Overview

The study asks how data-centre power supply configurations can be ranked quickly for reliability at early design stage, when equipment data is incomplete. Six DRUPS configurations are compared: single, parallel (2/3), cross link (3/4), isolated parallel, isolated redundant and distributed redundant. Classical indicators are computed with reliability block diagrams per IEC 61078:2016 using published component data.

The second method is an information-resource analysis that rests only on network structure (an oriented graph and adjacency matrix) and does not need uptime data for each UPS element. The authors find the rankings of the two methods broadly correlated: the correlation coefficient between MTBF and the informational indicator R is 0.793 (0.829), and both methods identify the single system and distributed redundant configurations as the extremes.

Block-diagram results: single system λ = 3.10 × 10^-6 1/h, A = 0.999949, MTBF 306,600 h (35 years); parallel (2/3) λ = 2.59 × 10^-7 1/h, A = 0.999997, MTBF 3,810,600 h (435 years); cross link (3/4) 3,793,080 h (433 years); isolated parallel (2/3) 3,898,200 h (445 years). Isolated redundant (2/3) has λ = 2.25 × 10^-6 1/h, A = 0.999949 and MTBF 420,480 h (48 years), barely better than a single system; distributed redundant (2/3) reaches λ = 4.09 × 10^-9 1/h and MTBF 243,659,400 h (27,815 years).

Limits: the method does not yet include the human factor, recovery time of the power supply or cost indicators, which are listed as future work. Very large MTBF values (hundreds of years) should be read as relative rankings for design comparison, not as real expected lifetimes. The graph method is proposed as a first screening step to find the best configuration, not as a replacement for detailed reliability study.

Key points

1. Single system: λ = 3.10 × 10^-6 1/h, A = 0.999949, MTBF 306,600 h (35 years).
2. Parallel (2/3): λ = 2.59 × 10^-7 1/h, A = 0.999997, MTBF 3,810,600 h (435 years); cross link (3/4) 3,793,080 h (433 years); isolated parallel (2/3) 3,898,200 h (445 years).
3. Isolated redundant (2/3): λ = 2.25 × 10^-6 1/h, A = 0.999949, MTBF 420,480 h (48 years), barely better than a single system.
4. Distributed redundant (2/3): λ = 4.09 × 10^-9 1/h, A = 1.000000 (to six decimals), MTBF 243,659,400 h (27,815 years).
5. MTBF and informational indicator R correlate at 0.793 (0.829), and both identify the single system and distributed redundant configurations as the extremes.
6. MTBF calculations used reliability block diagrams per IEC 61078:2016; the graph method needs only the network structure.

Using it in design

• When choosing among DRUPS or UPS configurations, first screen with the graph-based method to identify the best candidates, then do the detailed RBD calculation for that candidate only.
• Compare configurations by structure to avoid arrangements that look redundant but give near-single-system reliability, such as isolated redundant; mark common failure points explicitly.
• Present MTBF values in the reliability report as relative rankings, and state the source of component data and that the human factor and repair time are not included.

Who should read it

Power-system engineers and reliability analysts comparing DRUPS and UPS configurations at early design stage.

Document contents

1. 1. Introduction
2. 2. State-of-the-Art and Proposed Methods for Reliability Assessment (2.1 Reliability Block Diagram, 2.2 Informational Analysis Concept)
3. 3. Use Cases (3.1 General, six power supply system configurations)
4. 4. Results (reliability indicators and informational indicators, comparison)
5. 5. Discussion
6. 6. Conclusions

Related standards

• IEC 61078:2016

Read the document

Publication details

Authors

Kirill Varnavskiy; Fedor Nepsha; Qingguang Chen; Alexander Ermakov; Sergey Zhironkin

Publisher

Energies (MDPI AG)

Year

2023

Volume

16

Issue

3

Pages

1419

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/en16031419

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation