Title :
A Probabilistic Approach for Animal-Caused Outages in Overhead Distribution Systems
Author :
Sahai, Swati ; Pahwa, Anil
Author_Institution :
New England ISO, Holyoke, MA
Abstract :
Faults caused by animals are a significant factor in the reliability of distribution systems. Examination of historical data on outages recorded by a utility in Kansas showed that the occurrences of animal-caused faults, most of which are caused by squirrels, are dependent on weather conditions and the time of the year. It was observed that the majority of animal-caused outages take place on fair weather days, which are the days with temperature between 40 degrees F and 85 degrees F with no other weather activity. Also, the breeding cycle and behavioral patterns of squirrels over the year has a strong correlation with the animal-caused outages in different months of the year. Population density of squirrels will have an influence on the number of outages caused by animals in overhead distribution systems, but reliable data on animal population density could not be found, thus an adequate direct relationship could not be established. In this paper, a Bayesian model is constructed for prediction of animal-caused outages in overhead distribution systems given the type of month and the number of fair weather days per week. The model uses historical data for training purposes and predicts the number of animal-caused outages per week given a combination of month type and number of fair weather days per week. A confidence interval is built around the predicted values of animal-caused outages, and the percentage of observed values lying within the confidence limits are computed in order to obtain accuracy of the predictions. Also, the time-sequences of monthly averaged predictions are compared with the observed values. The model is illustrated with seven years of data obtained for a service territory of a utility in Kansas
Keywords :
Bayes methods; power distribution lines; power distribution reliability; power overhead lines; Bayesian model; Kansas; animal-caused outages; overhead distribution systems; population density; reliability; squirrels; weather; Accidents; Animals; Bayesian methods; Distributed computing; Lightning; Power distribution; Power system modeling; Power system reliability; Predictive models; Weather forecasting; Bayesian network model; failure rate prediction; overhead power distribution feeders; power distribution reliability;
Conference_Titel :
Probabilistic Methods Applied to Power Systems, 2006. PMAPS 2006. International Conference on
Conference_Location :
Stockholm
Print_ISBN :
978-91-7178-585-5
DOI :
10.1109/PMAPS.2006.360321