The objective of the Distribution Fault Cause Identification project is to investigate and develop appropriate technologies to reliably identify the cause of faults in power distribution systems. This can lead to economical improvements by (a) reducing operation costs such as money spent for non-experts to perform unnecessary and exhaustive search for the causes of the faults, (b) reducing the waiting time to find an expert to solve the problem, since artificial neural networks can be trained to possess the knowledge of an expert, (c) increasing the security of the system operations (through early stage fault detection), and (d) increasing the life time of power components by repairing early stage faults of the components, if any.

This project has investigated the general and local properties of distribution faults. The likelihood measure, based on the local region properties, provides valuable information to estimate the distribution fault cause when the fault cause is not known. (The likelihood measure concept has been used to train artificial neural networks to identify animal-caused distribution faults, yielding highly satisfactory results.)

Power Distribution Systems Outage Analysis

This milestone addresses the analysis of distribution faults from a quantitative approach using historical data, from a qualitative approach based on experience, and from a heuristic point of view. Four different measures are used in this project for thedistribution fault analysis : (1) actual number of faults; (2) normalized number of faults; (3) relative number of faults; and (4) likelihood of faults.

Anamial Fault Cause Identification Artificial Neural Network