XL BESS Projects in Germany – Article 3: Grid Studies Part 1

Prologue — Two Sentences That Should Govern Every XL-BESS Project

Before any switchgear closes, before any cell ships, before any controller is loaded, there are two sentences that should be hung on the wall of every XL-BESS project meeting room. They represent the constitution of the power system and BESS -regardless of its size, is just a part of it.

The grid is more important than the project any project and all projects.

The grid is not the laboratory.

The first sentence says: when the grid and the project disagree, the project bends. The schedule, the milestones, the OEM's eagerness to energize — all of them are negotiable. The grid is not.

The second sentence says: you do not validate an XL-BESS with GFM by switching it onto a live network and observing what happens. The validation happens elsewhere — in models, in benchmarks, in replicas, in real-time hardware-in-the-loop rooms. The field is the confirmation of what was already proven, not the experiment . If the field surprises you, you did the wrong work before it.

Everything that follows in this article — every gate, every study, every test, every iteration — is a consequence of these two sentences. The grid studies a serious XL-BESS program runs are not paperwork for a regulator. They are the way an engineer earns the right to close a breaker on a national transmission system.

This is the story of those studies. What they are. Why they exist. Why they cannot be ordered as separate deliverables from a catalogue. I will do my best to keep it simple and avoid equations and diagrams.

Let’s start by defining what distinguishes an XL-BESS project from a small BESS projects connected to the MV level. A 5 MWh commercial BESS behind a factory meter and a 1,500 MWh grid-scale BESS feeding a 400 kV substation are not the same machine scaled up. They are different categories of risk.

At XL scale — call it anything above roughly 500 MWh, definitively above 1 GWh — three things change at once.

First, the BESS becomes a stability actor, not a stability follower. A small BESS rides the grid. An XL-BESS shapes it. Its inverters, when configured as grid-forming (GFM), present an internal voltage phasor reference and can dictate frequency and voltage in their electrical neighborhood. When configured as grid-following (GFL), they still inject hundreds of megawatts of fast active and reactive current that can excite or damp oscillation modes nobody knew existed. The plant is no longer a passive customer of the grid — it is a participant in its dynamics.

Second, the BESS is asked to do many jobs at once. A modern XL-BESS specification typically asks for grid-forming behavior, Fast Frequency Response (FFR), reactive power / voltage control (VAR), synthetic inertia, RoCoF withstand, black-start capability, peak shaving and energy shifting — sometimes all on the same hardware. These services are not an "add-on toolbox" of independent functions; they share controller loops, share SoC, share headroom, and share the EMT-level transient behaviour of the converters. A change to the FFR ramp affects the inertial measurement. A change to the voltage loop affects small-signal damping. Anyone who treats the services as separable line items in a Bill of Quantities is building a liability.

Third, XL-BESS projects increasingly sit in electrically weak parts of the network. They are placed where the grid needs them — where renewables overwhelm conventional generation, where new load is being added faster than transmission, where Short Circuit Ratio (SCR) at the Point of Common Coupling is low and falling. (Convention: SCR below roughly 3 is treated as weak; below 2 as very weak. XL-BESS projects increasingly sit in the second band.) Weak grids are exactly the conditions where inverter-control phenomena — sub-synchronous control interaction (SSCI), harmonic resonance, control instability, low-frequency oscillation — show up. And these phenomena, critically, are often invisible in classical positive-sequence (RMS) simulations. You can run thousands of PSS®E cases and miss them entirely. They only appear in EMT (PSCAD-class) studies and in hardware-in-the-loop testing.

This is the technical reality that the two constitutional sentences address. If the grid is more important than the project, then the project must build the evidence base the grid requires. If the grid is not the laboratory, then that evidence must be produced before energization, in environments that genuinely represent the system.

That evidence base is the grid studies program.

Seventeen Studies, Four Evidence Levels, One Truth (17–4–1)

A typical XL-BESS interconnection package runs to around seventeen formal studies in the practice I run; the exact count varies by jurisdiction. Grouped by their role in the story, they fall into four acts.

Foundation Studies establish the world the BESS lives in: the Steady-State Analysis (load flow, voltage band, short-circuit levels), the Insulation Coordination Studies Part 1 and Part 2, the Power Quality concept, the initial PSS®E DMAT and PSCAD DMAT setups, the Benchmarking between RMS and EMT models, the Control Replica Study, and the Harmonic Measurement program. I keep the Control Replica in the foundational tier deliberately: no dynamic study downstream is robust without a documented, version-controlled representation of the actual controller, so it belongs with the artefacts that hold the rest up.

Dynamic Risk Reduction Studies look for what can go wrong when the inverters speak to the grid: Small-Signal Stability, the SSCI Study, the SSTI (Sub-Synchronous Torsional Interaction) Study, the Interaction Study (IBR-IBR, IBR-grid, weak-grid resonance), the RTDS / Hardware-in-the-Loop Testing Report, and the broader Dynamic Grid Impact Report.

As-Built Real-World Validation confirms the world matches the model after the equipment is in place: field harmonic measurements, the as-built update of Insulation Coordination Part 2, and the staged commissioning tests.

Final Synthesis is the Black Start Study and the Grid Code Compliance Report — the integration document that the TSO signs.

Across all of these, four evidence levels feed the studies and are fed by them. Offline model-quality tests on a Single-Machine-Infinite-Bus (SMIB) representation — voltage steps, frequency steps, HVRT/LVRT, reference step tests, SCR sweeps. Real-time HIL / PHIL tests where the actual plant controller (or a reproducible replica of it) is exercised against a real-time grid model. Staged field tests during commissioning. And, after energization, real operational event data captured by high-resolution recorders and replayed against the models to keep them honest.

This is the cast. Now the central plot point: these studies are not independent deliverables. They share a controller parameter state, a grid equivalent, a model build, and a test definition. Touch one, and several others are forced to re-run. That is not a defect of the process — it is the process. It is what integration means at XL scale.

Gate 1: Data and Power-Flow Freeze

Every story has a setting. For grid studies, the setting is the steady-state case: a single, agreed, version-controlled representation of the topology, the operating points, the transformer impedances, the cable parameters, the short-circuit levels at the PCC, and the assumed neighboring generation. Until this case is frozen, nothing else is real.

The most expensive mistake in XL-BESS projects is not a technical one. It is when three engineering teams start the dynamic work using three different steady-state cases, because the developer has not yet declared which one is canonical. Three months later, the small-signal study is conditioned on a 2027 dispatch, the PQ study is conditioned on a 2030 dispatch, and the SSCI study is using yesterday's network impedance from the TSO. Every result is invalid. Every report has to be re-run. I have lived this.

The constitutional principle "the grid is more important than the project" is operationalized here in a single act: the developer accepts that the TSO's network data, with its constraints and limitations, is the single source of truth. The project does not get to pick its preferred grid; it gets the grid the TSO publishes. From that, the Steady-State Analysis becomes the foundation everything else stands on.

Once Gate 1 is closed, three workstreams can begin in parallel under the same data regime: Insulation Coordination Part 1 (definitions, principles, rules for over-voltage withstand levels) starts. The Power Quality concept — what will be measured, against which IEC compatibility and planning levels — starts. And the DMAT setup for both PSS®E and PSCAD begins.

Gate 1 looks administrative. It is not. It is the moment the project agrees what kind of grid it is being built into.

Gate 2: Model Freeze and Benchmark Sign-Off

This is the gate where most XL-BESS projects bleed schedule, and the root cause is almost always the same: the interface between the converter OEM and the grid studies contractor is treated as a file transfer, not as the months-long collaboration it actually is.

A point worth saying plainly: "grid-forming" on the datasheet is true about the hardware. It says nothing about the model. A DMAT-grade GFM model is a separate engineering product, often years behind the hardware in maturity — built loop by loop, parameter by parameter, and matured against your benchmark events. Sometimes that work is already done from a previous project; sometimes yours is the first. I have been lucky to witness both situations, and there is a lot to learn in either. Pretending you will always inherit the mature case is a mistake.

The studies sit on two models — PSS®E for positive-sequence (RMS) work and PSCAD for EMT — and each one must carry its own DMAT (Dynamic Model Acceptance Test) report: block diagrams of every control loop, parameter ranges, protection and limiter logic, mode-transition documentation. On top of the two DMATs sits a Benchmark Report that places the PSS®E and PSCAD models side by side and demonstrates they give consistent results across faults, voltage and frequency steps, ramps, and phase jumps. The TSO — or its modelling authority — must accept all three: the PSS®E DMAT, the PSCAD DMAT, and the cross-model Benchmark.

So a “.dll file” plus a one-page parameter sheet is not a model. The TSO will not accept it, the small-signal study cannot linearize it, the interaction study cannot diagnose it, and the Replica downstream is impossible. Worse, the studies contractor cannot defend the report — every TSO question bounces back to an OEM on a different continent, and the benchmark cycle drags across the whole project as email.

And it is not only the TSO who is reading. On any bankable XL-BESS, the Lender's Technical Advisor opens the same DMAT artefact pack during due diligence; an incomplete or black-box model is not just a compliance problem with the grid operator, it is a financing problem with the lender. The artefact list is, in practice, a procurement clause, a compliance gate, and a bankability gate at the same time. I know because I also do this . My advise here Design the interface, do not discover it.

While the model is being matured, the rest of the project does not have to sit idle. Several studies are genuinely independent of the dynamic model and should run in parallel with the OEM benchmark cycle: the first part of Insulation Coordination, the Power Quality concept, the pre-energisation Harmonic Background Measurement campaign at the PCC, and the Black Start system-sequencing framework (if requested).

Everything else depends on the frozen model and must wait. The trap is starting any of those before the freeze: the work is real, but it is conditioned on a model that will change, and every model revision retroactively invalidates it.

A word of caution to the project director, though, because the word "parallel" misleads more projects than it helps. Parallelism is a scheduling benefit, not a redistribution of attention, and not an engineering license to treat these studies as independent.

Gate 2 closes when PSS®E DMAT, PSCAD DMAT, and the cross-benchmarking report are signed off by the TSO or its modelling authority. From that moment, the model is frozen under configuration management — every subsequent study cites the model version, and any model change retrospectively invalidates downstream work.

End of Part 1

To keep the story readable, I have split it in two. Today we close at Gate 2, with the model frozen and the benchmark signed off. Next week, Part 2 picks up where the grid actually starts pushing back — and where most of the questions a serious XL-BESS engineer loses sleep over finally get answered.

A few of those questions, to set the stage:

What actually happens, when small-signal stability, SSCI, SSTI, and interaction studies meet a weak grid — and why is iteration here a feature of the schedule rather than a failure of it?

What is the real difference between a Replica and a HIL/PHIL room, and why does that difference decide whether your field test confirms what you already know or discovers something you wish you had found earlier?

What does it mean to stress the "constitution" of the plant under real voltage — and which tests are worth the risk of doing live versus the cost of not doing them at all?

Once the Grid Code Compliance Report is signed and the plant is energised, why does the model refuse to retire — and what is the long-tail obligation that quietly outlives the commissioning team?

Part 2 is the answer to those questions. I hope Part 1 was useful, and I am looking forward to the discussion.

Part 2 follows.

Originally published on LinkedIn Pulse. Open the original →