The Frequency-Dependent Network Equivalent (FDNE) is an advanced tool to model system characteristics. However, this model should be used cautiously for accurate and meaningful results.
Please consider the following points:
The FDNE assumes that the input parameters, such as impedance and admittances, are passive. For example, in the case of impedance data, the resistive part should be always positive at each frequency. A non-passive model may give unstable simulations.
Input data, such as impedance or admittance, is defined for several frequencies (i.e. discrete data). The FDNE approximates the samples with a continuous impedance function, ensuring accuracy at each sample frequency. This continuous function is defined from zero to infinity. So at any other frequencies, FDNE represents an impedance value defined by the continuous function. This will effect accuracy of the simulation, when energized at that frequency. For example, let us assume that the input data file contains impedance parameters for 100 frequency samples, ranging from 120 Hz to 5 kHz. The data does not contain a power frequency sample (60 Hz). When the circuit (including the FDNE) is energized with 60 Hz, the FDNE represents an impedance (at 60 Hz) defined by the continuous function. This will affect the accuracy of the simulation. To overcome this, it is better to add additional impedance data sample at 60 Hz (or in general any interested frequency).
Ideally it is better if the data samples include a few low frequencies (ex. 1 Hz) and power frequency as well. This will also improve stability of the simulation (less likely to have passivity violations).
It is assumed that the frequency-dependent parameters are smooth in magnitude, as well as angle. The non-smooth parameters can lead to poor curve-fitting results and hence inaccurate simulation (ex. adding artificial impedances to the existing smooth data may lead to poor curve-fitting results).
The parameters are assumed to be close to the minimum phase function. A non-minimum phase function may require a very high-order transfer function (the transfer impedance/admittance of a long distributed parameter transmission line may significantly deviate from the minimum phase function).
Impedance data is converted in to admittance data and it is assumed that inverse of the impedance matrix is possible.
It is assumed that the parameter matrix is symmetrical. (ex. Z(i,j) = Z(j,i) ).
The following are some questions that get asked frequently on our Support Desk:
In the Frequency-Dependent Network Equivalent (FDNE) component parameters, set Detailed log file to Create.
In the PSCAD temporary folder for the project, open the log file (FDNE*.log). Check the maximum fitting error as a percentage. If the fitting error is too high, the simulation may not be accurate and there can be increased chance of simulation instability. To reduce fitting error, increase the maximum order of the function and check the input data.
You can also check the _Y_MAG.out and _Y_ANG.out output files. These files contain the actual and fitted data (magnitude and angle). It is better to plot the actual and fitted admittance functions as a function of frequency and see if the fitted function is in close agreement with the actual function (based on given data).
You can also connect the Harmonic Impedance component to compare impedance parameters for a defined frequency range.
It depends on the fastest transient in the simulation. For example, if you are studying lightning surges, the highest frequency can be up to 1 MHz, and the minimum frequency can be 0.5 Hz or less. The algorithm checks for passivity violations at a number of frequency samples within the lowest and highest frequency range.
If the time domain simulation is unstable, you should enable passivity enforcement.
The algorithm is based on linearization and constrained optimization. These algorithms work fine if the passivity violations are small.
Weighting factors are used to improve accuracy at a steady-state frequency. If the weighting factor is increased (above 1) for a frequency or frequency range, accuracy for that range is increased. However this may lead to relatively poor accuracy in other frequencies. It is recommended to set weighting factors equal to 1.0 for all frequencies.
This is only important to define weighting factors.
If the parameters (ex. impedance/ admittance) are de-coupled (ex. there is no mutual between phases, or only positive sequence is available), it is better to use three independent FDNEs (one for each phase instead of one FDNE (dimension 3) to represent the three phases).