Dynamic model of distribution network cell using system identification approach

Abstract

The centralised generation is generally a passive network. The interconnection of distributed generation (DG) to this centralised generation have changed this passive network perspective to become an active power network. This DG interconnected to distribution network is also called active Distribution Network Cell (DNC). However, this active DNC normally has limitations because of computational time constrains and huge dimensions of the network systems. The dynamic model of this active DNC provides a remedy for these limitation since it offers a simple representation of the system without effecting the DNC dynamic characteristics and behavior. Thus, this research aims to develop an active DNC model that represents the dynamic characteristics of the distribution network. The model development deployed the System Identification approach. The model used in this research is a transfer function model which has eighteen parameters. The transfer function model is comprised of double-fed induction generator as the generator model and for the load part, the composite load model is used which contains the static constant impedance, constant current and constant power (ZIP). This ZIP is combined with the induction motor as dynamic load. The developed model then formulated under system identification framework before the parameter estimation procedure is conducted. The estimation procedure used is the nonlinear least square optimisation and was conducted in MATLAB software which considered the input ( ) and output ( ). The parameter estimation procedure evaluation is considered by the best fit values of the transfer function model. Lastly, the performance of developed equivalent model is evaluated under three phase to ground fault at different fault such as Bus 1, 2, 3, 4 and Bus 5 for small and large disturbance studies. The graphical comparison of the estimated responses and measured responses are done using the best fit values. The original transfer function model has eighteen parameters. The results indicated there are four parameters that have zero values for all cases studies. From investigation, it is proven that these four parameters are not involved in parameter estimation procedure. Thus, these four zeroes parameter can be ignored and the original transfer function model can be reconstructed to a new reduced transfer function model which has only fourteen parameters.