DMAT, AEMO's Dynamic Model Acceptance Test, is a standard battery of simulations that a plant's PSS/E and PSCAD models must pass before AEMO and the network service provider will use them in connection studies. It checks that the models initialise cleanly, stay numerically stable and respond credibly to faults, steps and setpoint changes. Most failures come from three causes: poor initialisation, numerical instability, and PSS/E and PSCAD models that do not agree with each other.
DMAT is not a compliance test of the plant against its Generator Performance Standards. It comes first. It is a test of whether the models are fit to be trusted, and only once they are does the GPS assessment begin.
The DMAT Guideline: Purpose and Version in Force
AEMO publishes the Dynamic Model Acceptance Test Guideline alongside its Power System Model Guidelines. The model guidelines say what a model package must contain. The DMAT guideline says how the models must behave under a defined set of test conditions, run on a simple test network rather than the full NEM model.
The purpose is screening. AEMO and NSPs run detailed studies on large network models, and a plant model that will not initialise, drifts on a flat run or oscillates after a fault wastes everyone's time inside those studies. DMAT catches those problems on a small test system where they are cheap to fix.
Version 3.0 of the guideline took effect on 3 June 2024, and as far as we are aware it remains the version in force as at November 2026. AEMO updates its connection guidelines from time to time, so confirm the current version on AEMO's network connections pages before you start a model package, and check which version your NSP's study requirements letter refers to.
DMAT applies to the models of generating systems and integrated resource systems, including solar, wind and BESS. Our overview of PSS/E and PSCAD modelling for renewables in Australia explains where model acceptance sits in the wider GPS process.
Test Categories
The guideline sets out a test matrix that is repeated across operating points, typically different active power levels, reactive power conditions and network strengths at the test connection point. For PSCAD, that matrix runs to hundreds of cases for a single plant. The table below summarises the families of test that a DMAT package typically covers. The exact cases, values and pass criteria are defined in the guideline itself, and that document governs.
| Test family | What it checks | Typical failure signs |
|---|---|---|
| Flat run | The model sits at its initial operating point with no disturbance applied | Drift in P, Q or voltage; slow oscillation; controllers that never settle |
| Balanced and unbalanced faults | Fault ride-through behaviour for three-phase and single- or two-phase faults of varying depth and duration | Trips that should not happen; non-physical current spikes; slow or no post-fault recovery |
| Multiple or sequential faults | Behaviour when faults occur in quick succession | Protection or ride-through logic that locks out; energy or thermal limits mis-modelled |
| Voltage and frequency disturbances | Response to grid voltage steps, frequency steps or ramps, and phase angle changes | Wrong direction or size of response; oscillation; loss of synchronism in the PLL |
| Setpoint changes | Steps in active power, reactive power, voltage or power factor setpoints at the plant controller | Rise and settling times that do not match the claimed performance; overshoot; steady-state error |
| Network strength changes | Performance across a range of short-circuit ratios, including weak-grid cases | Instability at low SCR; control interactions that appear only in EMT |
PSS/E and PSCAD cases are run in parallel for many of these tests, so that the two models can be compared. That benchmarking is where many packages come unstuck.
Common Failures
On projects we have supported, DMAT failures cluster into four groups, and most could have been caught before submission.
Initialisation
The model must start at the operating point specified for each case, with every internal state in equilibrium. Common faults:
- Plant controller and inverter reactive power references that are inconsistent, so the model starts pulling against itself.
- Transformer tap positions, collector cable data or auxiliary loads that differ between the load flow and the dynamic model.
- PSCAD models that start from a hard-coded snapshot rather than the case's actual load flow.
- Flat runs that look settled over a few seconds but drift visibly over the full run time.
If the flat run is not flat, nothing after it can be trusted.
Numerical Stability
PSCAD in particular is sensitive to time step, solution order and interface choices.
- Models validated by the OEM at one time step that misbehave at the time step the test system uses.
- Chatter or spurious oscillation caused by discontinuous logic, such as hard switching between control modes without hysteresis.
- PSS/E models with time constants too small for the integration step, producing numerical rather than physical oscillation.
- Case-to-case sensitivity, where a model passes at one SCR and goes unstable at a slightly different one.
PSS/E and PSCAD Mismatch
The two models describe the same plant, so their responses must agree within the tolerances the guideline sets. Mismatches usually come from:
- Different firmware or parameter versions in the two models. The OEM updated one and not the other.
- Control functions, such as reactive current injection during faults or active power recovery ramps, implemented in PSCAD but simplified away or mis-parameterised in PSS/E.
- Plant controller models that were built separately for each tool, by different people, from different documents.
- Fault level and impedance data entered differently in the two test setups.
PSS/E is a positive-sequence RMS tool and PSCAD is an EMT tool, so some differences are expected, particularly in the first cycles of a fault. The question is whether the differences are explainable and within tolerance, and the applicant should explain them before AEMO has to ask.
Controller Logic
The last group is the plant behaving as modelled, but not as claimed.
- Fault ride-through entry and exit thresholds that do not match the settings in the GPS proposal or the inverter data sheet.
- Plant controller droop, deadband or ramp settings in the model that differ from the settings that will be commissioned.
- Mode-switching logic, for example between voltage control and power factor control, that does not respond to a setpoint change as described in the model documentation.
- Frequency response that is present in the inverter model but blocked by the plant controller model, or the reverse.
These failures matter beyond DMAT. The models are later checked against site behaviour during R2 testing, and a setting that differs between the model and the plant will surface again then.
A Workflow That Passes First Time
A first-time pass is not luck. It comes from treating DMAT as an internal quality gate, run by the applicant before AEMO ever sees the package.
DMAT Pre-Submission Workflow
- Confirm the guideline version: check AEMO's current DMAT guideline and the NSP's study requirements letter before building the test cases
- Lock the model versions: one firmware and parameter set, recorded, used in both PSS/E and PSCAD
- Reconcile the data: load flow, transformer, cable and auxiliary data identical across both tools and matching the design drawings
- Run long flat runs first: every operating point, full duration, before any disturbance case
- Check time-step robustness: rerun representative PSCAD cases at the test system time step and a smaller one
- Benchmark early: overlay PSS/E and PSCAD results for faults and setpoint steps, and resolve or document every material difference
- Cross-check against the GPS proposal: ride-through thresholds, control settings and response times must match what the GPS will claim
- Review every plot: automated pass/fail checks are useful, but an engineer should look at each case for non-physical behaviour
- Document the results: a clear report with plots, deviations explained and the model versions stated
Two process points carry most of the weight. First, engage the OEM early. Many problems sit inside the vendor's inverter model, and OEM model updates can take weeks. Asking for a model that has already passed DMAT on another NEM project, at the same firmware, saves real time. Second, run DMAT before the full connection studies, not after. Finding an initialisation fault after hundreds of network study cases have been run means running them again.
The same discipline helps with the rest of the GPS application. Our article on common pitfalls in AEMO GPS applications covers the wider process and where modelling problems usually cause delays.
Get DMAT-Ready Models
DMAT is a fixed cost in every NEM connection. Done badly, it becomes repeated review rounds and a GPS assessment that cannot start.
ElectraGlobe's power systems team builds, tunes and tests PSS/E and PSCAD models for solar, wind and BESS connections, runs DMAT packages ahead of submission, and resolves model issues raised in AEMO and NSP review. See our steady-state and dynamic studies service, or contact us to discuss your model package.
Frequently Asked Questions
What is DMAT?
DMAT is AEMO's Dynamic Model Acceptance Test, a standard set of simulations that a plant's dynamic models must pass before they are used in connection studies. It checks that the models initialise correctly, remain numerically stable and respond credibly to faults, disturbances and setpoint changes. It tests the models, not the plant's compliance with its performance standards.
Which DMAT guideline version applies?
Version 3.0 of AEMO's DMAT guideline took effect on 3 June 2024, and we understand it to be the current version as at November 2026. AEMO revises its connection guidelines periodically, so check AEMO's website and your NSP's study requirements before you begin, and confirm which version your application will be assessed against.
Why do models fail DMAT?
The most common causes are poor initialisation, where the model drifts on a flat run, numerical instability in PSCAD, and PSS/E and PSCAD models that do not agree. Controller settings that differ from the GPS proposal or the inverter data are another frequent cause. Most of these can be found by running the tests internally before submission.
Does DMAT apply to PSS/E and PSCAD models?
Yes. DMAT covers both the PSS/E (RMS) model and the PSCAD (EMT) model, and many tests are run in both tools so that the results can be compared. The PSCAD test set is usually the larger and more time-consuming of the two.