Initial Situation at Forschungszentrum Jülich

Even before JUPITER, more than 120 Power Quality Analysers were deployed campus-wide, including in JUWELS.
These included PQI-DA smart and PQI-DE devices with central evaluation and low-latency data access for simulation servers to build a digital twin of the campus MV network.
With JUPITER, the monitoring was expanded by 17 additional measuring points.

Initial situation - summary

• Exascale operation with very high requirements for voltage quality and availability
• Campus grid with mixed load and feed-in profiles
• 120 measuring points as foundation, expandable for JUPITER
• Low-latency data access for research and model validation

Technical key data - grid setup & measurement architecture

Feature	Value
Total measurement points	> 140 (PQI-DE & PQI-DA smart)
Time synchronisation	GPS / NTP (≤ 1 ms Genauigkeit)
Standards	IEC-61000-4-30 Klasse A / EN 50160
Data formats	OPC UA / CSV / REST API

These data form the basis for continuous voltage-quality acquisition in the Jülich campus grid.

Typical Challenges in High-Performance Data Centres

Data centres are sensitive to voltage dips, flicker, transient events, harmonics and supraharmonics.
GPU-intensive jobs and cooling cause rapid load steps.
The integration of CHP, PV, storage and V2G charging alters load flows and spectra up to 20 kHz.
Beyond technical control, IEC-61000-4-30 Class A compliant measurement and EN 50160 documentation are required – and the monitoring must scale economically.

In HPC infrastructures like JUPITER, high-frequency current changes arise from GPU cluster clocks and load transitions.
These can cause short-term harmonic peaks and frequency-selective coupling between MV and LV levels.
Continuous, high-resolution Power Quality Monitoring detects and assesses such transients early – crucial for energy supply and operational security in data centres and industrial campus grids.

Challenges - summary

• High load dynamics from HPC jobs and cooling
• PQ effects up to 20 kHz due to power electronics
• Obligation for standard-compliant measurement & documentation
• Scalable Power Quality Monitoring with acceptable OPEX

Measurement Data - Hpc-Specific Power Quality Effects

Influencing factor	Description
GPU load steps	Current change > 200 A / ms
Supraharmonics	Frequency ranges up to 20 kHz
PQI-LV, PQI-DA smart & PQI-DE	40,96 kHz
Standards reference	EN 50160 / IEC-61000-4-7

These parameters characterise typical grid phenomena in exascale operation.

Project Objectives

Key objectives: full transparency of load flows, validation of simulation & grid models, effective early-warning for critical events, and flexibility for future expansions.

Objectives - summary:

• Transparency, validation, early warning, flexibility, compliance
• 10-minute averages + event-based disturbance records
• Precise time base and central, role-based evaluation

The measurement data serve not only for operations but also for validating digital twins of the campus grid – enabling realistic simulation of voltage stability and response times compared with live PQ data.

Technical Solution by A. Eberle

FZJ employs a tiered architecture with PQI-DE, PQI-DA smart, and WebPQ®.
PQI-DE units are installed at key nodes with large displays for direct live data access.
In deeper network layers, PQI-DA smart units are DIN-rail mounted – compact, cost-efficient and optimised for large-scale monitoring.
WebPQ® consolidates all measuring points, reports and alarms in a central browser interface.

The system is time-synchronised via GPS or NTP servers, allowing event correlation across voltage levels – vital for analysing cause-and-effect between MV and LV grids.

IT Security in WebPQ®

In high-performance environments, IT security is crucial. WebPQ® ensures data protection and cyber-resilience through:

• Encrypted HTTPS / TLS communication
• Role-based user & access management
• Secure authentication (2FA optional)
• Audit trails & tamper-proof logging
• Optional network segmentation

WebPQ® also integrates with SIEM systems so that security-related events can be part of the overall cyber-security strategy.

It Security and Integration Paths in Webpq®

Aspect	Implementation
Encryption	TLS 1.3 / HTTPS
Authentication	Two-factor (2FA / token)
Role model	Multi-user permission management
SIEM integration	Syslog / SNMP trap
Data APIs	OPC UA / REST / Modbus TCP/IP

The implementation ensures secure data paths and compatibility with existing IT-security concepts.

Example: Time-synchronised query (PostgreSQL)

-- pgsql
SELECT timestamp, voltage, frequency
FROM pq_measurements
WHERE device = 'PQI-DE'
AND frequency BETWEEN 49.9 AND 50.1
ORDER BY timestamp DESC;

Purpose: Quick retrieval of frequency values near EN 50160 limits from a PQI-DE for situational assessment.

Example: WebPQ® API behind NGINX reverse proxy

# nginx
location /webpq/api {
    proxy_pass http://localhost:8080/api/measurements;
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header Host $host;
    proxy_set_header X-Forwarded-Proto $scheme;
}

Purpose: Secure, centralised access path to measurement data (reverse proxy).

Example: Programmatic fetch (Java)

/* java */
HttpResponse<String> response = Unirest.get("https://webpq.example.org/api/v1/measurements")
  .header("Authorization", "Bearer <API_TOKEN>")
  .asString();
System.out.println(response.getBody());

Purpose: Automated retrieval for reporting / research / Digital Twin.

Example: Visual style in dashboard (CSS)

/* css */
.graph-voltage {
  stroke: #0073e6;
  stroke-width: 2;
  fill: none;
}

Purpose: Readable voltage traces in SVG/Canvas charts.

Measurement Results and Findings

Load flows and trend analyses

Long-term trends reveal changes in load behaviour (e.g., software updates, server expansions, cooling strategies) and enable early actions to keep EN 50160-compliant voltage quality.

Transients and response times

Event-triggered recordings up to 200 kHz per channel enable precise analysis of millisecond-range voltage dips and quantification of effects on UPS systems or other sensitive loads.

Harmonics and Supraharmonics

The analysis reveals spectral clusters and resonances tied to specific PSU switching frequencies – insights that directly support Power Quality optimisation, filter dimensioning and EMC assessment.

Example: Quick trend fit (Mathematica)

(* mathematica *)
Fit[data, {1, x, x^2}, x]

Purpose: Fast curve fit (trend, curvature) for time series.

Benefits for the Operator

The combined PQI-DE / PQI-DA smart / WebPQ® concept delivers end-to-end grid transparency from the main feed to sub-distribution, proves standards compliance and provides robust decision support. It also supplies KPIs for energy management, voltage-quality trends and capacity planning. Continuous analysis of reactive power, voltage asymmetry and grid interactions reduces operating costs and optimises components.

Conclusion and Outlook

The project shows how Exascale requirements and the energy transition can be mastered together – with Class A Power Quality Monitoring, a tiered architecture and central WebPQ® evaluation. The approach is transferable to other data centres, utilities and campus grids. Next steps include predictive analytics, automation and machine learning in WebPQ® to detect recurring patterns and generate early warnings – a step towards autonomous Power Quality Monitoring in HPC and industrial grids.

Standards, Measurement Methods and Integration

Measurements in line with IEC-61000-4-30 Ed. 4 Class A; harmonics per IEC-61000-4-7; flicker per IEC-61000-4-15; supraharmonics 2–20 kHz (FGW TR3). Documentation per EN 50160. Via CSV, OPC UA, REST API, voltage-quality data flow into MATLAB/Simulink or DIgSILENT PowerFactory to compare real vs. modelled behaviour and calibrate load-flow and stability models.

Standards and Methods at a Glance

Parameter	Standard / method
Voltage quality	IEC-61000-4-30 Class A
Harmonics	IEC-61000-4-7
Flicker	IEC-61000-4-15
Supraharmonics	2–20 kHz (FGW TR3)

Example: SQL export to CSV (PostgreSQL)

-- sql
COPY pq_measurements
TO '/var/data/pq_exports/pq_jupiter.csv'
WITH CSV HEADER;

Purpose: Standards-compliant data extract for audit/analysis.

Product Innovation: Pqi-LV

The PQI-LV extends the portfolio with a cost-efficient solution for applications requiring high measurement-point density in LV grids. Ideal for utilities, grid operators and data-centre operators needing continuous, accurate voltage-quality monitoring with standards-compliant reporting.

Authors

Fabian Leppich - Product Manager PQSys
Maximilian Sefz - Head of Marketing
Wolfgang Reitmeier - Sales