Predictive models of the dependability of electric power transmission (using the example of AO Tyvaenergo)

Abstract

Aim. The paper analysed the damage rate of the electric power networks of AO Tyvaenergosbyt between 2014 and 2022 and preventively evaluates emergency shutdowns in the company’s networks in 2023. The author provides a brief historical reference regarding the company’s foundation, a description of the structure of its power supply network. The damage rate of elements, as well as main electrical equipment of the company’s power supply network were examined. The quantity of electrical power undersupplied to customers as the result of emergency shutdowns was analysed. The number of failures was classified depending on the amount of undersupplied electrical power. The paper studied the applicability of certain methods of prediction of the number of emergency shutdowns for the obtained interpolated series of preceding failures. The most appropriate method was established that was used for preventively evaluating the damage rate of the examined power networks for each month of 2023.Methods. The paper uses methods of mathematical analysis, general scientific methods of research, properties and capabilities of the MATLAB graphic interface and Excel tables. Statistical and cybernetical methods of prediction were considered for the purpose of predictive model development. The probability of failures within the examined power networks was evaluated. Equations were obtained for autoregression failure models that enable predictive evaluation of failure by months.The Results of the study may prove to be useful to the experts of Tyvaenergosbyt, as well as other power grid companies that are developing advanced emergency prevention activities aimed at improving the reliability of power supply.Conclusion. The approach to predictive evaluation of emergency shutdowns proposed in the paper showed that the most efficient method of predicting failures in a company’s power networks is the method of autoregression models using interpolation of data on preceding failures. The probability of prediction fulfilment is 95%.