Tag Archive for: BESS
Application NotesFAESR and the New PCS Design Metric
FAESR and the New PCS Design Metric
Grid inertia, NC RfG 2.0 and what changes operationally for European Power Conversion System manufacturers
Executive summary
On 24 April 2026, Sustainable Energy, Grids and Networks published a paper on grid inertia estimation and FAESR — Fast-Acting Energy Storage Resource — requirements for the Indian power system. The work is relevant not for its geography but for its methodology: it offers a reproducible framework to quantify how much fast-response storage is needed for a given drop in rotational inertia level.
The quantitative question, formulated for India, is the qualitative question that Europe is already answering through regulation. ENTSO-E has closed the Phase II report of the new Network Code RfG 2.0 with binding grid-forming requirements for all new BESS or renewables above 1 MW. Germany published the first price for the inertia services market on 28 January 2026. The UK Pathfinder already has operational commercial projects. And the Iberian Peninsula experienced, one year earlier, a major blackout that the European technical literature interprets as the case study the Indian paper models from theory.
The industrial consequence is direct. The design metric for a BESS PCS is no longer stored kWh or round-trip efficiency at steady state. It is the capacity to sustain a sub-100-millisecond response to a grid event over a transient of a few seconds, maintaining a minimum damping ratio of 5 %, full ride-through of phase-angle jumps, and conformity with the ENTSO-E mechanical starting time method.
The FAESR concept — and why the design metric changes
The term FAESR — Fast-Acting Energy Storage Resource — designates a storage resource that responds to a grid event on a sub-second timescale, with the explicit purpose of compensating for the mechanical inertia lost when a conventional synchronous machine is replaced by a renewable source connected via inverter. The rotational inertia of a thermal alternator is not programmed: it exists physically as kinetic energy stored in rotating parts, and is released automatically upon a frequency drop.
The synthetic equivalent — the virtual synchronous machine — emulates that behaviour from a bidirectional power converter with downstream storage, but only if the inverter-plus-battery assembly can inject the necessary power within the correct timeframe. That timeframe is demanding: ENTSO-E places the current response time below 10 milliseconds in the Phase II report of NC RfG 2.0.
A battery with slow chemistry, or a PCS with a classical PI controller tuned for steady state, may have perfectly adequate rated capacity and still not qualify. What determines qualification is the capacity to sustain the response during the characteristic dt of the event — typically between 100 ms and 10 s — without the regulation becoming unstable.
This shifts the relative weighting of five design metrics:
| Metric | Traditional PCS | FAESR-capable PCS |
|---|---|---|
| Rated kWh | Determining | Necessary, not sufficient |
| Round-trip efficiency | Primary KPI | Secondary KPI |
| Current response time | <100 ms acceptable | <10 ms required (NC RfG 2.0) |
| Damping ratio | Not specified | ≥5 % on electromechanical oscillations |
| Phase-angle ride-through | Optional | Mandatory |
The paper’s contribution — and its limits
The paper proposes a quantitative estimation methodology for the residual inertia level of a power system under different renewable penetration scenarios, and from that level derives the minimum FAESR needed to keep frequency within the acceptable operating range during a generation-loss transient. The valuable contribution is the methodology, not the Indian case: the work explicitly articulates the relationship between the admissible RoCoF — Rate of Change of Frequency — the available equivalent inertia, and the required FAESR power with its temporal profile.
Our critique is threefold. First, the Indian case is not linearly transferable to Europe: the energy mix, grid age, geographic dispersion and quantity of residual synchronous reserves are different. Second, the methodology does not cover the actual worst case observed — the Iberian event of 28 April 2025 showed that cascades can be qualitatively different from academic scenarios. Third, the paper models the FAESR response as if it came from a single ideal source; in practice, it comes from hundreds of distributed PCS units, each with their own computation times, communication delays and potential fleet-level degradations.
These three caveats do not invalidate the contribution — they consolidate it as a starting point for a project-specific engineering analysis.
Transposing to Europe — what changes operationally in 2026
European regulation has closed, within six months, the technical requirements boundary that the paper models theoretically:
- ENTSO-E NC RfG 2.0 Phase II: binding grid-forming requirements for all new BESS or renewables >1 MW. Current response time <10 ms. Damping ratio ≥5 % on electromechanical oscillations. Full ride-through. Effective application expected 2028–2029 with national transitional periods.
- Germany: inertia services market launched on 28 January 2026 with four product categories (positive/negative response × base 30 % / premium 90 % availability). Mandatory VDE FNN certification. Multi-year contracts with remuneration indexed to verified availability.
- United Kingdom: Pathfinder programme with its first commercial operational plant (Blackhillock), demonstrating system-strengthening performance under real conditions.
- Iberian Peninsula: the event of 28 April 2025 — the only documented cascade in the European continental grid — remains the most-read operational reference for FAESR sizing assumptions across the European technical community.
Implications for PCS design
Three families of power conversion assets in the Premium PSU portfolio map directly onto the FAESR requirement space:
| Asset | FAESR mapping | Primary adaptation |
|---|---|---|
| ODX/OVX-6000 series (critical microgrid) | Bidirectional PCS with VSM and sub-second response | Control migration to NC RfG 2.0 with auditable dossier |
| EPS series sub-millisecond DC UPS | Edge FAESR node in an industrial microgrid | Minor engineering changes building on railway fast-transfer know-how |
| CHB topology (Cascaded H-Bridge) | Grid-scale BESS where the inertia contract becomes bankable | Pure modularity, cell bypass, natural isolation — canonical topology |
The catalogue alone is not enough. Technical rules have tightened within six months and TSOs now demand compliance traceability. A Design for Compliance (D4Co) approach articulates, from the design phase, the dossier that the operator will present to the regulator to make the inertia service contract remunerable. That discipline turns a technically good PCS into a contractually bankable PCS.
About Premium PSU
Premium PSU (Premium SA) designs and manufactures custom power supplies and power converters in Barcelona, with over 40 years of experience in railway, energy, industrial, data centre and defence applications.
Talk to our engineering team
Discuss your power conversion requirements with Premium PSU.
Application Notes, Data CenterThe Hyperscale Data Centre as a New MV Load
The Hyperscale Data Centre as a New MV Load
What changes for the DSO when AI connects to the distribution grid: primary AC/DC conversion, grid-support services and mitigation assets at the point of connection are redefining the quality contract on the European distribution network.

Born in Barcelona, Powering the World
AI load no longer resembles a classic industrial load
Three structural differentiators separate the load of an AI campus from that of a traditional data centre.
Power density per rack
NVIDIA Blackwell Ultra dissipates 1.4 kW per GPU. A GB300 NVL72 rack dissipates between 125 and 140 kW. The 2026–2028 frontier is 200 kW per rack, with a 1 MW horizon. This density has no precedent in classic industrial grid connection.
Dynamic time profile
AI training cycles aggregate and release tens of MW on sub-second time scales. Distributed inference modulates the load in the 0.1–10 Hz range, within the flicker and power-quality disturbance band monitored by DSO analysers.
Geographical concentration
An AI campus aggregates 50–200 MW in a footprint of a few hectares, something historically seen only in heavy industrial plants with a dedicated HV connection. The 2025–2026 transition to bipolar 800 V DC architectures, driven by NVIDIA and leading power-infrastructure manufacturers, concentrates primary AC/DC conversion at a single point.
Frankfurt, London, Amsterdam, Dublin, Madrid, Paris: hyperscale hubs are redefining MV feeder planning and investment cycles on the distribution network.
Four measurable effects at the point of common coupling
These are the effects captured by power-quality analysers at the head of an MV feeder with an active AI campus downstream.
| Effect at the PCC | Physical origin in the AI load | Implication for the DSO |
|---|---|---|
| Voltage dips and ride-through | Sub-second power ramps at training start-up; synchronised start-up of GPU banks generating abrupt load steps. | HV→MV→LV propagation in adjacent Dy networks; industrial customers sensitive to Type I and II dips. A 0.5 p.u. dip at MV propagates as 0.73/0.48 p.u. on the LV side. |
| Harmonic distortion (THD) and interharmonics | AC/DC rectifiers in the 800 V DC front end; SiC switching with sub-5 ns rise times generating broadband harmonic content. | Risk of EN 50160 non-compliance on shared feeders; active filtering or reinforcement required at the MV substation. |
| Flicker and low-frequency oscillations | Load modulation by inference/training cycles in the 0.1–10 Hz range. Systematic and semi-periodic, unlike classic industrial process flicker. | Pst and Plt in the residential-complaint range for customers connected to the same MV/LV transformer. DSO responsibility under EN 50160. |
| Unbalance and sequence components | Non-homogeneous load distribution per phase, with single-phase sidecar racks in three-phase campus installations. | Neutral overload, additional transformer heating and degradation of feeder equipment. Accelerated ageing of distribution assets. |
The mitigation chain: from the rack to the PCC
Four coordinated layers with distributed allocation of the dynamic response. A poorly sized layer degrades the technical contract of the whole.
| Layer | Function | Topology | Premium SA platform |
|---|---|---|---|
| 01 · In-rack CBU | Absorb sub-100 ms transients and hold the DC bus during grid events. | EDLC supercapacitor bank + bidirectional DC/DC converter. | Industrial DC/DC platforms with telemetry and BMS, derived from regenerative traction. |
| 02 · Sidecar BBU | Hold-up from seconds to minutes. Staggered restart after a dip or transfer. | Li-ion module + bidirectional DC/DC in a cabinet outside the compute rack. | EPS line (sub-ms DC UPS) rescaled to an industrial sidecar · EN 50600 + OCP. |
| 03 · Campus BESS + STATCOM | Power-quality management at the PCC · reactive voltage support · peak shaving to reduce contracted MW. | Multilevel Cascaded H-Bridge · three-phase AC bus at MV · Li-ion batteries + grid-forming control. | CHB topology (grid-scale BESS and STATCOM, NC RfG 2.0 ready). |
| 04 · Grid-forming PCS at the PCC | Meet the connection contract: ride-through, damping ≥5%, current response <10 ms. The critical layer for the DSO. | Multilevel inverter with certifiable Virtual Synchronous Machine (VDE FNN where applicable). | ODX/OVX-6000 platform: MV grid-forming adaptation with D4Co dossier. |
Regulatory framework: the DSO perspective
European grid codes are tightening the real-time technical connection contract.
ENTSO-E NC RfG 2.0, Phase II
Current response <10 ms · damping ≥5% on electromechanical oscillations · full ride-through of phase-angle jumps · compliance with the mechanical starting time method for virtual synchronous systems. Threshold: any BESS or renewable plant >1 MW. Its transposition to the data centre as a grid-interactive load is an open area of regulatory discussion; CIRED 2026 (Dublin, June 2026) was one of the forums where that discussion was being formalised among European operators.
German inertia market (28 January 2026)
First published European inertia price, with four product categories and mandatory VDE FNN certification. A data centre with a bidirectional BESS and a compliant PCS is technically in a position to bid for those contracts, turning mitigation infrastructure into a monetisable asset that improves the business case of the connection.
UK Pathfinder: Blackhillock
First commercial Virtual Synchronous Machine plant in operation, demonstrating grid-forming performance under real grid conditions. It is the benchmark for what an NC RfG 2.0-class system must deliver commercially, not only in laboratory certification.
Iberian blackout (28 April 2025)
The most widely read case study in Europe on cascades caused by low residual inertia. Direct evidence of the consequences of an under-specified dynamic response at points of connection, including large loads with passive PCS. The event has accelerated DSO requirements for documentation of dynamic response in new connection agreements.
The Premium SA proposition
Over 40 years of mission-critical industrial power electronics, mapped to the grid interface of the hyperscale market.
ODX/OVX-6000
Rugged inverters for critical microgrids: the base on which bidirectional PCS with Virtual Synchronous Machine (VSM) and sub-second response are developed. Main adaptation: migration of the control to an NC RfG 2.0-compliant architecture, with an auditable dossier for the DSO generated from the D2W gate.
EPS line
DC UPS with sub-millisecond response and fast transfer certified in critical railway applications. Rescaled to the industrial data centre sidecar, it covers the BBU/CBU layer without structural platform modification; it only requires additional qualification gates for EN 50600 and, where applicable, OCP.
CHB topology
Cascaded H-Bridge: pure modularity · no capacitor-balancing issues · fault tolerance through cell bypass · natural isolation per cell. It is the canonical topology for the grid-scale BESS + STATCOM segment on data centre campuses, where the inertia or voltage-support contract becomes bankable before the DSO or the ancillary services market.
D4Co methodology
Design for Compliance: from the first D2W gate, the dossier that the operator will submit to the DSO or the regulator so that the connection is approved and the grid-support services provided are remunerable. Derived from over 40 years of railway RAMS, derating to 80% of Tj,max and post-deployment FRACAS. It turns a technically good PCS into a contractually bankable PCS.
Technical excellence is not enough: bankable compliance is what makes the difference.
The boundary between data centre and grid is not a cable: it is a well-designed PCS
When a European hyperscaler requests a new 50 or 100 MW connection from a DSO, the technical debate is no longer settled by sizing an additional transformer. It is settled by articulating a quality contract at the PCC: THD limits, ride-through profile, dynamic reactive power response and behaviour during phase-angle jumps. That contract requires equipment at the point of connection that behaves like the grid, not like a load.
Premium SA provides the industrial asset — ODX/OVX-6000, EPS, CHB topology — and the D4Co methodology that turns the operator’s technical specification into an auditable compliance dossier. When the difference between award and rejection is measured by the quality of the technical dossier, that discipline is decisive.
Talk to our engineering team
Defining the grid interface of a data centre campus? Our engineering team can help you size the conversion and mitigation chain from the rack to the point of connection.
About Premium SA
Premium SA (Premium PSU) is a Barcelona-based designer and manufacturer of custom power supplies and power converters for railway, energy, industrial, data centre and defence applications, with over 40 years of experience since its foundation in 1981. Born in Barcelona, Powering the World.
