https://doi.org/10.36719/2789-6919/42/157-164
Ilkhombek Kholiddinov
Ferghana Polytechnic Institute
Doctor of Technical Sciences
https://orcid.org/0000-0002-0120-4043
i.xoliddinov@ferpi.uz
Mashkhurakhon Holiddinova
Ferghana Polytechnic Institute
doctoral student
https://orcid.org/0009-0006-2430-5708
mashhuraxonxoliddinova@gmail.com
Modeling a Block to Determine the Type of Damage to a Power Line Using Matlab
Abstract
This article discusses the modeling of a block for determining the type and location of damage to a power line based on artificial neural networks. A description of the working scheme, the principles of the model and the key stages of neural network training is introduced. The simulation was performed in the MATLAB environment, including Simulink and Neural Network Toolbox, which allowed us to build an accurate and scalable model. Phase voltages and currents, as well as their symmetrical components, were used as input parameters for the analysis.
As part of the research, the architecture of a two-layer neural network with one hidden layer containing 10 neurons using a sigmoidal activation function has been developed. The linear output layer provided prediction of the type of damage and its location. The Levenberg-Marquardt algorithm was used to train the network, which provided a fast and stable solution to the problem. The training was conducted on a large data set containing more than two million observations, divided into training, validation and test samples.
The simulation results demonstrated high accuracy in damage classification, minimal standard deviation, and a stable correlation between model predictions and real data. Graphs and error histograms confirm the stability of the network. The use of artificial neural networks has significantly reduced the diagnostic time and improved the quality of the analysis of emergency modes in 10 kV power lines. The results obtained open up prospects for further implementation of such approaches in the energy industry.
Keywords: power lines, reliability, modeling, neural networks, type of damage, short circuit