Tiza Group

Home / Tools / Plant model validation

Plant model validation

Choose the control and the test, and use an example record or paste your own with time, measured signal and simulated signal. The tool computes the coherence indices of CNO Agreement 1954 in their window, compares them with the thresholds of that test, groups the plant's units into equivalent units and reviews the contents of the report.

Acuerdo CNO 1954§ anexo 3coherence indicesAcuerdo CNO 1954§ anexo 1equivalent unitsAcuerdo CNO 1954§ anexo 2report contentsAcuerdo CNO 1954§ art. 10 y 14tests and follow-upCREG 101 104 de 2026§ art. 4 y 9plant models

Example cases
Plant model validation

Only the basic data; the rest is taken with typical values.

ALearn

E1Plant and test
kV
MW
% CEN

For the load level: low from 10 % to 20 %, medium from 30 % to 70 %, high from 80 % to the CEN.

E2Measured and simulated record

The examples are our own records at 100 samples per second. «My record» uses the table below.

%

In % of the plant's time constant. It changes the simulated signal of the example, as if you were tuning the model.

%

In % of the measured change.

s
s
My record

One row per sample: time in seconds, measured signal and simulated signal on the same time stamp. Paste the columns of your .txt files.

RowTimesMeasuredSimulatedRemove

E3Plant units
Generation blocks

One row per block or collector circuit. Settings set: the same code only if they share protection settings, control settings and plant controller. Deviation: the one your study gives when aggregating that block; leave it empty if you don't have it.

RowBlockInverter or turbineUnitsCapacityMWMounting or typeTilt or trackingSettings setDeviation when aggregating%Remove

E4Comparison criteria
unit
s
s
% CEN
%
%

Use a decimal point. The calculation runs in your browser and updates as you type.

Test result

—

Combined error EC

—

Linear correlation CCL

—

Normalized mean error EAMN

—

Equivalent units

—

Record sampling

—

Plant, controller, recorder and point of connection
The signals the test uses light up: voltage and current at the point of connection, the recorder and, in step or ramp tests, the set point or simulated frequency that goes into the controller. The controller frame turns green if the model passes and red if it does not.
Measured and simulated signal with its bands
  • Measured
  • Simulated
  • ±3 % bands
  • Measurement error
The shaded area is the analysis window of the EC, CCL and EAMN: from the disturbance to the maximum settling time of CREG 060. The two short-dashed lines are the ±3 % bands around the initial and the final value.
Difference between simulated and measured
  • Simulated − measured
  • ± EAMN
  • ± measurement error
The difference is in % of the smaller contribution of the two signals, the same base as the EAMN. The EAMN is the mean of its absolute value inside the window.

Coherence indices of the test

IndexMeasuredSimulatedValueCriterionResult

Equivalent units

UnitBlocksEquipmentSet-upUnitsMWMinimum in serviceImpedance

Report review checklist

PartWhat it must haveStatus

What it is for

A solar or wind plant connected to the STN or STR hands the CND the model of the plant and its controls, and must show that it reproduces what the plant does in the field. This tool compares a test record with the model's simulation and computes the coherence indices of CNO Agreement 1954, with the window and threshold of each test.

It also groups the inverters or wind turbines into equivalent units with the criteria of the agreement, computes how many units must be in service in each test and reviews the contents of the report before you send it.

How it is calculated

Acuerdo CNO 1954§ anexo 3curve shape

Correlation, mean error and combined error

CCL = Σ(Ri − R̄)(Si − S̄) / √(Σ(Ri − R̄)² · Σ(Si − S̄)²) · EAMN = mean |Ri − Si| / min(contribution R, contribution S) · 100

R is the measured signal and S the simulated one. The contribution of each signal is its maximum difference from its initial value. In validation tests they are computed from the disturbance to the maximum settling time of CREG 060 (10 s for voltage, 15 s for frequency); in follow-up, from 1 s before the event to the end of the record. The combined error is EC = EAMN + 100·(1 − CCL)²; with EAMN = 0, the 30 % limit requires a CCL of 0.452 or more.

Acuerdo CNO 1954§ anexo 3times and values

Initial and final values, overshoot and times

SI = (vmax − vf)/(vf − vi)·100 · EVF = |vfR − vfS|/|vfR − viR|·100 · ETE = |TER − TES|/TE060·100

The initial and final values are averages over windows you define. The initial response time is when the signal leaves ±3 % of the step around the initial value; the settling time, when it enters and stays within ±3 % around the final one. Both are checked on the measured and the simulated signal against the CREG 060 maximums. The delay-time error (at 50 % of the change) is computed like the ramp error and shown without a threshold.

Acuerdo CNO 1954§ anexo 3ramp

Ramp error

ER = |RR − RS|/RR·100, with R² > 0.5

The slope of each signal comes from a least-squares line through the points between 20–30 % and 70–80 % of the change. If the coefficient of determination is 0.5 or less, the slope is not used.

Acuerdo CNO 1954§ anexo 1, art. 10grouping

Equivalent units

units in service ≥ ⌈0.1 · n⌉

Blocks that share equipment and rating, mounting (and in solar the tilt or tracking algorithm) and settings set are grouped. A block whose aggregation shifts P, Q or V at the point of connection by more than 1 % stays as a separate unit. In the test of each equivalent unit at least 10 % of its units must be in service, rounded up to the next integer.

Limits of the calculation

  • The 2 in the EC formula is read as a square: EC = EAMN + 100·(1 − CCL)². Confirm it with the official text; the alternative reading, the sum divided by 2, is in Advanced mode.
  • Pending: annex 5 of the agreement (results template, initial and final value windows and test names) is not in the tool. The windows are your inputs.
  • The simulated signal must be on the same time stamps as the measured one (resample it first) and with the same disturbance time.
  • Assumptions: the overshoot is taken from the extreme of the signal in the direction of the change; the ±3 % band is computed with the measured step and applied to both signals; the delay time is the time the signal takes to reach 50 % of the change; TE and TRI are compared as «less than or equal» to the CREG 060 maximums, as in the agreement tables.
  • CREG 101 104 of 2026 requires RMS and EMT models. The tool compares field records with the RMS simulation; the maximum times used by the indices are those of CREG 060, which 101 104 did not change.
  • The tool does not simulate the plant: the model is simulated in the CND tool and the records are compared here. The examples are our own records generated with a second-order response.
  • The deviation when aggregating collectors comes from your study with 10 % voltage and frequency steps; the tool takes it as an input.

Frequently asked questions

Which indices does the CND check to validate a solar plant model?

CNO Agreement 1954 compares the curve measured in the field with the one simulated in the CND tool. In an unlimited set-point step it checks the overshoot error, the final-value error, the initial response and settling times and their errors, the correlation (CCL) and the normalized mean absolute error (EAMN). Limited tests, ramps and events check fewer indices, and events only the combined error (EC).

How are inverters grouped into equivalent units?

Units go together when they share inverter type (maker, model and rating), protection and control settings, the plant controller and, in solar, the tilt and the mounting type; in wind, the turbine type. Collectors are aggregated as long as 10 % voltage and frequency steps give at most 1 % difference in P, Q and V at the point of connection against the aggregated model, with the fewest units.

How many tests must pass to approve the model?

80 % of the tests of each control mode must meet all their indices, and the ones that fail must be due to something outside the control. Voltage control with droop and without droop count as different modes.

Related guides

Other tools

All tools
WhatsApp