> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/dsx/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/dsx/_mcp/server.

# Qualification Tests

> Defines the 12 qualification tests, objectives, required evidence, and pass or disqualification criteria for BESS validation.

## Test 1: Telemetry Verification and Data Integrity

**Objective:** Verify PCS telemetry accuracy, timestamping, and completeness.

### Steps

* Operate PCS at 10%, 50%, 75% and 100% rated MW in grid-connected mode
* Record internal telemetry and compare to external 0.5S (IEC 62053-22) or better revenue-grade meters
* Verify by demonstration reporting of: V, I, P, Q, f, SOC, PCS temperature, current limit flags, mode state
* Event logging shall be 1 ms precision
* Logs shall be held for 7 days minimum
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* P and Q error ≤ ±0.5% or better of reading across points
* Time alignment error ≤ 5 ms
* No missing channels, no "flatlined" sensors

## Test 2: Grid-Adaptive Voltage and Frequency Regulation (Islanded)

**Objective:** Demonstrate stable operation with no external grid reference.

**Setup:** Isolated test bus with configurable resistive/inductance/capacitance (RLC) load bank.

### Steps

* Black start onto dead bus (if Test 10 not yet done, run reduced voltage)
* Hold voltage and frequency within the system's defined stability limits at a steady load
* Apply load steps: +10%, +20%, -20% of PCS rated MW
* Record V, I, P, Q, f, and control mode state throughout the test
* Provide all test data necessary and sufficient to demonstrate compliance with the
  requirements and pass/fail criteria above.

#### Pass Criteria

* Stable V/f with no sustained oscillation
* Voltage and frequency return to their steady-state range in a timely, well-damped manner, with no evidence of control hunting or marginal stability
* No protective trip

## Test 3: Current Limit Characterization (P-first, Q-first, and Mixed Priority)

**Objective:** Demonstrate stable, predictable, and transparent PCS behavior when operating at or beyond current limits under all priority configurations.

### Steps

* Force the PCS into current limit using:
  * Reduced-voltage condition on the test bus
  * Commanded P and/or Q beyond the declared capability envelope
* Repeat for each priority configuration:
  * Active power (P) priority
  * Reactive power (Q) priority
  * Partner default or "grid-support" mode
* Record P, Q, V, I, current-limit flags, and control-mode state variables
* Provide all test data necessary and sufficient to demonstrate compliance with the
  requirements and pass/fail criteria above.

#### Pass Criteria

* PCS remains stable and controllable while in current limit
* PCS explicitly reports entry into and exit from current limit via telemetry
* Control priority and available headroom are observable and consistent with documented behavior

#### Disqualification Criteria

* **Limit-cycle oscillation:** Under steady-state operating conditions, no sustained oscillations shall be present, and PCS performance shall remain within all applicable generator and grid interconnection requirements.
* **Control-mode hunting:** More than one control-mode transition (e.g., P→Q→P, GFM→GFL, or equivalent) during a single current-limit event constitutes test failure.
* **Voltage recovery behavior:** Upon release from current limit, terminal voltage shall recover monotonically to its regulated setpoint. Any overshoot, undershoot, or oscillatory recovery exceeding ±5% of nominal voltage constitutes test failure.

### Deliverables

* Demonstrate through EMT analysis
* P–Q capability map with clearly defined current-limit envelope
* Annotated time-series plots showing:
  * Entry into current limit
  * Behavior while limited
  * Exit from current limit and voltage recovery
* Description of current-limit implementation and priority logic

## Test 4: AI Buffering Proxy Test (Fast Ramp Tracking)

**Objective:** Demonstrate the BESS can buffer AI-induced power ramps at the required rate without inducing control instability.

### Steps

* Operate PCS in grid-connected mode at both weak-grid (SCR ≤ 3) and strong-grid (SCR ≥ 20) configurations
* Apply a time-varying active power command profile that is representative of realistic AI workload ramps, including both short bursts and longer ramp intervals, and reaching ramp rates up to the vendor's declared AI-buffering capability. Partner must state the assumed IT load used for the test and demonstrate correct scaling of results.
* Record P, Q, V, I, and current-limit and control-mode state flags
* Provide all test data necessary and sufficient to demonstrate compliance with the
  requirements and pass/fail criteria above.

#### Pass Criteria

* **Tracking accuracy:** Steady-state ramp tracking error ≤ 2% of commanded ramp magnitude
* **Current limit:** No entry into current limit during execution of required buffering profile
* **Stability:** No sustained oscillatory behavior induced by control interaction

#### Disqualification Criteria

* **Sustained oscillation:** Any clearly observable, sustained oscillatory behavior in measured P, Q, V, I, or frequency that indicates marginal or unstable control response during the ramp profile constitutes test failure, regardless of steady-state tracking accuracy.
* **Current-limit interaction:** Repeated or sustained entry into current limit, as indicated by PCS telemetry or limit-state flags, during execution of the required buffering profile constitutes test failure, unless explicitly declared as outside the partner's stated capability envelope.
* **Ringing tolerance:** Short-duration transient ringing (less than one second) is acceptable only if it is monotonically decaying, does not re-enter current limit, and does not induce control-mode transitions.

#### Note

This test does not, by itself, prove site POI compliance. It verifies that the PCS can execute the required buffering action when properly integrated.

## Test 5: AI Buffering EMT Validation (Model-Based, Required)

**Objective:** Validate PCS stability and control robustness under extreme weak-grid conditions representative of faulted, partially islanded, or switching-dominated AI campus scenarios.

### Partner Provides

* EMT model of PCS + controls (compiled or encrypted acceptable, but must run)
* dq impedance curves vs frequency
* Nyquist or passivity evidence

### Required EMT Case

* **SCR:** 2.0
* **Grid impedance:** Sweep from X/R ratio = 2 to 10, representative of inverter-dominated grids
* **Topology:** Grid-connected with upstream source modeled as weak voltage source
* **Load:** Constant Power Load representative of IT/compute with clearly defined ramp characteristics shall be used

### Disturbances to Apply

* AI buffering proxy ramps
* At least one grid disturbance event (voltage dip, impedance step, or partial islanding)
* Entry into and recovery from PCS current limit

### Steps

* Configure the EMT model with the required case parameters above
* Apply each disturbance scenario sequentially, recording all state variables
* Provide all model outputs and analysis necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* POI ramp compliance for the defined workload profile
* No unstable oscillation modes across operating range
* System remains stable with no loss of synchronism
* No sustained oscillatory modes
* No uncontrolled voltage collapse at the BESS terminals
* PCS transitions into and out of current limit in a controlled and documented manner

#### Failure Criteria

* Any unstable oscillation persisting more than one second
* Voltage collapse or failure to recover to regulated operating point
* Control mode hunting or undefined control state transitions
* Model instability that prevents completion of the test scenario

#### Note

This stress case may be performed model-only.

#### Critical

Failure in this case constitutes non-qualification, regardless of performance at higher SCR values. This test is intended to expose control fragility under realistic worst-case grid conditions expected in large AI campus deployments.

## Test 6: Demand Response Dispatch Performance

**Objective:** Verify DR execution without breaking buffering and ride-through reserves.

### Steps

* Demonstrate dispatch steps at: 10%, 25%, 50%, 100% of DR committed MW
* Demonstrate ramp-limited dispatch (partner must implement ramp limits)
* Demonstrate SOC reserve policy: hold back prescribed SOC capacity reserved for LVRT/buffering
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* Commanded MW achieved within specified response time (default: 2 s for fast DR, 60 s for slow DR)
* No violation of SOC reserve rules
* Clear state machine behavior when DR conflicts with LVRT

**Deliverable:** DR state machine diagram and priority rules

## Test 7: LVRT / HVRT Functional Ride-Through (Hardware or HIL)

**Objective:** Prove ride-through behavior aligned to IEEE 2800 baseline.

### Steps

* Apply voltage sag and swell profiles at AC terminals (grid simulator or HIL) at both weak-grid (SCR ≤ 3) and strong-grid (SCR ≥ 20) configurations:
  * Multiple depth-duration points per IEEE 2800 or utility categories
* Observe whether PCS stays connected and how it prioritizes Q support
* Record V, I, P, Q, f, and protection relay states throughout each ride-through event
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* No trip within IEEE envelope or applicable utility/ISO envelope, whichever governs
* Correct trip outside envelope
* Q injection behavior documented and consistent

## Test 8: Seamless Grid / Island Transition

**Objective:** Validate transition logic and stability.

### Steps

* Start grid-connected in buffering mode
* Execute intentional islanding event (open upstream breaker) while load present
* Maintain V/f and continue supplying load
* Resynchronize and reconnect to grid with controlled closing logic
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* No unstable mode hunting
* No uncontrolled frequency drift
* Reclose without excessive inrush or protective trips
* Event logs show proper sequencing

## Test 9: Islanded Operation with Generator Following (Model-Based)

**Objective:** Validate interaction with turbines or gensets.

### Steps

* EMT model with GFM BESS as voltage master
* Generator modeled as droop-following source with realistic governor time constants
* Apply AI load proxy disturbance and show:
  * BESS absorbs fast dynamics
  * Generator picks up slower energy component
* Include N-1 generation event
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* No sustained oscillation
* SOC drift controlled (see Test 11)
* Frequency and voltage remain within declared bounds

## Test 10: Black Start

**Objective:** Demonstrate ability to energize a dead bus and establish stable V/f.

### Steps

* PCS starts from de-energized condition
* Energize test bus to nominal voltage
* Pick up load in steps: 10%, 25%, 50% rated
* If using auxiliary supply, document requirements
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* Stable energization
* No nuisance trips
* V/f regulation within declared bounds

## Test 11: SOC Drift and Energy Management Under Combined Missions

**Objective:** Demonstrate that the BESS can simultaneously support AI buffering, demand response, and LVRT reserve obligations over time without uncontrolled state-of-charge (SOC) drift.

**Test duration:** Minimum 24-hour continuous simulation or accelerated profile test.

### Steps

* Run a 24-hour simulation or accelerated profile test:
  * AI buffering proxy ramps applied intermittently throughout the test window
  * Demand response dispatch events of varying magnitude and duration
  * At least one LVRT or equivalent grid-disturbance event
  * Declared SOC operating window and reserve thresholds enforced
* Provide all test data necessary and sufficient to demonstrate compliance with the requirements and pass/fail criteria above

#### Pass Criteria

* **Net SOC drift:** ≤ ±5% of usable SOC capacity, excluding declared DR events
* **SOC bounds:** Remain within partner-declared operating window at all times
* **Energy prioritization:** Explicit de-prioritization when approaching reserve thresholds

#### Disqualification Criteria

* **Unbounded SOC drift:** Exceeding ±5% without declared external energy imbalance
* **Undefined reserve behavior:** Loss of capability without explicit state transition
* **Implicit behavior:** SOC management relying on undocumented heuristics

### Required Deliverables

* Time-series plots of SOC, P, Q, service state, and reserve flags
* Energy balance summary table (in/out, losses, DR energy, net drift)
* State machine diagram showing:
  * Normal operation
  * Reserve-threatened operation
  * De-rating or service shedding behavior
* Event log with timestamped state transitions and priority changes

#### Important

Oversizing battery capacity alone does not satisfy this requirement. SOC stability must be achieved through control logic and energy management, not nameplate capacity.

## Test 12: Control Transparency Package Review

**Objective:** Provide the analysis artifacts needed for integration.

### Required Deliverables

* EMT model package with instructions
* dq or sequence (full matrix) impedance vs frequency (magnitude and phase) in key operating points, at least +0.9 p.u. and −0.9 p.u. active power
* Nyquist plot or passivity assessment demonstrating stability margin for extremely weak (SCR = 2, X/R = 10) and strong grid conditions (SCR = 20, X/R = 10)
* Controller description: loops, bandwidths, current limit behavior
* Parameter list: droop coefficients, virtual impedance, inertia, PLL state (if any)

#### Pass Criteria

* Package is complete, runnable, and consistent with measured behavior
* No "black box with no stability evidence" accepted
* Has snapshot feature to start simulations from a specific snapshot
* Model runs with generally accepted Fortran compilers released after 2018