METHOD AND ALGORITHMS OF CALCULATION OF PARAMETERS OF RELIABILITY IN ACCORDANCE WITH MULTIVARIATE DATA

Abstract

Assessments of the relevance of the kinds of attributes regulated by classifiers, evaluation of parameters of individual reliability, of parameters of reliability of multivariate data clusters relate to the primary problems the decision of which would make it possible to reduce operational expenses of electric power system objects. The lack of specialized software systems, the knowledge-intensiveness of algorithms of calculation needed for the solution of operational tasks determine the predominant application of the intuitive approach. Some factors, such as lack of operating experience and subjectivity of the intuitive approach lead to the essential risk of making the incorrect decision. The article presents analytical methods that make it possible to consider the expediency of classification of multivariate statistical data, to identify the relevant types of signs, to rank the objects in power systems and to form clusters of homogeneous data. The methods function with the aid of calculations of indicators of reliability. The originality of methods consists in application of an fiducial approach. Experimental data are compared to boundary values of a fiducial interval. Statistical functions of fiducial distributions are obtained by imitating modeling of samples of random variables for a series of numbers of random variables samples. The analysis of dependency of boundary values on a number of realizations made it possible to obtain a fairly simple equation of regress with the accuracy of no less than 0.995. Formation of a set of relevant kinds of attributes is demonstrated on behalf of an estimation of average duration of idle time during emergency repair made for the analysis of the law of alterations by months of year when the calculation of a parameter of individual reliability for the third power unit in case of failure of boiler facility, representation of set of realizations of duration of idle time in emergency repair clusters. As a result of the repairs, the aging of equipment and devices, alterations of important types of signs that cause the change of the results of the ranking of objects of power system and their technical condition, are observed.

Keywords

• многомерные данные
• разновидность признаков
• функции распределения
• показатели надежности