Analysis of Animal-Related Outages in Overhead Distribution Systems With Wavelet Decomposition and Immune Systems-Based Neural Networks

Abstract

Outages in overhead distribution systems caused by different factors significantly impact their reliability. Since animals cause large number of outages in overhead distribution systems, analysis of these outages has a practical value as it allows utilities to keep track of historical trends. This paper presents a methodology for yearend analysis of animal-caused outages in the past year. Models to estimate weekly animal-caused outages in overhead distribution systems using combination of wavelet transform techniques and neural networks are presented. Discrete wavelet transform is applied to decompose the time series of weekly animal-caused outages into two components and separate neural networks are constructed for each decomposed coefficient series. The outputs of neural networks are combined according to wavelet reconstruction techniques to get estimated results for the weekly animal-caused outages. Artificial immune system (AIS) is used to overcome the overtraining problem associated with neural networks. Results obtained for four districts in Kansas of different sizes are compared with observed outages to evaluate performance of three different models for estimating these outages.