Page 214 - Decision Making Applications in Modern Power Systems

Adaptive estimation and tracking of power quality disturbances Chapter | 6 177

first part of the chapter that adaptive filters play an important role in design-
ing the PQ estimation models. To judge the tracking and estimation accuracy
of different models, simulations have been carried out by using MATLAB/
SIMULINK before testing the model in FPGA platform. Simulated compari-
son results show a comparison between LMS, NLMS, and RLS algorithms
in the presence of swell and momentary interruptions, which clearly indicates
that RLS has better estimation accuracy than the other two algorithms. The
second part of the chapter deals with the classification of seven PQ events,
that is, sag, swell, harmonics, sag with harmonics, swell with harmonics,
notch, and spikes. The seven PQ event signals (sag, swell, harmonics, sag
with harmonics, swell with harmonics, notch, and spikes) are generated by
using MATLAB/SIMULINK environment by considering a system having
two generators on both sides feeding a long transmission line under different
abnormal conditions such as symmetrical fault and sudden loading of large
load at different distances. It was concluded from the simulation results of
the second part of this chapter that the EMD HT SVM technique gives
better results (94.4%) as compared to EMD HT ANN (65.8%) and
EMD HT PNN (80.9%) techniques.

Appendix

Parameters of ANN

TABLE A1 Details of the ANN parameters.

Network type Feed-forward back propagation network
Training function Levenberg Marquardt
Size of first hidden layer 20
Size of second hidden layer 05
Train parameter goal 7 3 10 29
Performance function MSE

No. of epochs 1000
MSE, Mean squared error.

Parameters of probabilistic neural network

Kernel function used in PNN: RBF
Spread factor (σ) 5 0.10.