Прогнозування вторинних оцінок параметрів сигналу тестування обмотки статора трифазного асинхронного електродвигуна

Abstract

UA: Для визначення поточних величин активного та реактивного опорів обмотки статора асинхронного електродвигуна вимірюють параметри пропущеного крізь неї тестового струму. Описано процедуру прогнозування вторинних оцінок величин цих параметрів при нерівноточному та нееквідистантному в часі їх вимірюванні. Отримано вираз для розрахунку дисперсії спрогнозованої величини при згладжуванні послідовності вторинних оцінок довільними детермінованими функціями часу. Цей вираз конкретизовано для випадку поліноміального згладжування.

EN: Forecasting the technical condition of a threephase asynchronous electric motor is the foundation of a cost-effective maintenance strategy for both the motor itself and the it is incorporated into. The stator winding of such a motor is one of the most unreliable elements of its design. The values of its active and reactive resistances are the main indicators of its technical condition. A known method of measuring these quantities during normal operation of the motor is to introduce a small sinusoidal test voltage with all known parameters into it and then measure the amplitude and initial phase of the current in the winding caused by this voltage. The results of direct measurements of the specified current parameters (primary estimates) are distorted by the additive effect of stationary Gaussian noise and shortterm impulse interference. This results in two different dispersion values of the primary estimates. To reduce the influence of highvariance estimates, smoothing of their time sequence is performed. As a result, a time sequence of secondary estimates is obtained, which has a smooth character. Due to this property, the sequence can be used to predict the numerical value of the monitored parameter at a specific time horizon. The paper examines the transformation of the existing smoothing procedure for primary estimates into a forecasting procedure for secondary estimates. An analytical expression is derived for calculating the variance of the predicted value in the general case of smoothing by deterministic functions of time. This expression is specified for polynomial smoothing both for arbitrary time intervals between the obtained primary estimates and for time-equidistant primary estimates. Computer simulation revealed that as the majority of lowvariance primary estimates shift toward the end of the observation interval, the time horizon of the most accurate forecast increases. Therefore, a theoretically substantiated forecasting procedure has been developed for predicting the sequence of secondary estimates of the test current parameters of the stator winding of a three-phase asynchronous electric motor under conditions of nonuniform-accuracy primary measurement, as well as under conditions of possible non-equidistant arrangement of primary measurements in time.