Deliverable 3.2 (June 2020)

Refined Tools to fit electrical models according to experimental data

It consists on the last Deliverable of the AEMS-IdFit project. It presents the system identification of the devices by means of a black-box approach. It also shows the final version of the graphical user interfaces (GUIs) of the algorithms presented in the previous deliverables.

Two black-box identification methods are described and tested using experimental data of DC-DC power converters and six-pulse diode rectifier. The methods are the Nonlinear Autoregressive Exogenous Model (NARX) and Long-Short Term Memory (LSTM) neural networks. The results presented show an accurate representation of the behavior of the different devices.

Four different GUI were delivered, which are the following:

  • Parameter estimation of power electronics devices using the Non-linear least squares algorithm
  • Parameter identification of synchronous machine
  • Parameter estimation of synchronous machine
  • Black-box identification