МОДЕЛИРОВАНИЕ ОТКАЗОВ ЭЛЕМЕНТОВ ГОРОДСКОЙ ЭЛЕКТРИЧЕСКОЙ СЕТИ (10 КВ) НАПРИМЕРЕПРАВОБЕРЕЖНОГООКРУГАГОРОДАИРКУТСКА

Abstract

The paper presents mathematical models for probabil-istic evaluation and forecasting of emergency shutdowns in electric networks by the example of the Pravoberezhniy District of the Irkutsk City from 2008 through 2017. At the first stage, the autocorrelation function of the parameter series is determined to estimate its randomness. According to the calculated statistical parameters and the consent criteria, a number of equipmentfailures may be described by a three-parameter gamma distribution. A method of two-level identification of extreme (maximum and minimum) values of the parameter under study is proposed; accord-ing to this method, a significant polynomial trend is ob-tained for predicting the largest number of failures. The evaluation of the presence of trends based on monthly data showed that polynomial and power trends may be used to predict failures on electric networks. At the same time, sig-nificant trends were identifiedonly for January, February, May and December. At the next stage, trend-seasonal models are constructed; the least squares method is used to calculate their components. According to the obtained seasonality indices, the greatest increase in emergency shutdowns takes place in April and July, and a decrease -in February and March. On the basis of the correlation-regression model, factor models of failures of electrical network elements and the accumulated average daily tem-peratures for months and time are constructed. Linear and nonlinear models with and without trends are obtained. To evaluate the accuracy of the forecasts of the obtained models, the results of the retrospective forecast for 2017 were compared with the actual values. According to the resultsobtained for predicting failures on electric networks in February, June, July and September, the best result is shown by a trend-seasonal model, in May -a polynomial trend, in November -a factorial one taking into account time, in March, May and October-a nonlinear regression equation, and in December -a power trend. There are no qualitative forecast models for the months of January, April, and August. In this regard, the values of emergency shut-downs on these months may be estimated using a proba-bilistic model.