ПРИМЕНЕНИЕ ИНТЕЛЛЕКТУАЛЬНОГО АНАЛИЗА ДАННЫХ ДЛЯ ОЦЕНКИ РАСХОДА ЗАРЯДА АККУМУЛЯТОРА БЕСПИЛОТНОГО ЛЕТАТЕЛЬНОГО АППАРАТА

Abstract

Assessing and predicting the dynamic flight performance of an unmanned aerial vehicle is associated with the complexity of constructing and applying strict mathematical models that take into account many heterogeneous parameters. The paper examines the problem of estimating the battery consumption of an unmanned aerial vehicle when flying along a given route, considering weather conditions and geospatial characteristics. An approach to its solution is proposed, based on clarifying the estimate obtained using an approximate dependence by finding the interval with the highest probability containing the actual flow value. To find the specified interval, it is proposed to use the binary classification method applying logistic regression models based on the training data set. A pipeline is built that combines the mechanisms of constructing, training, and applying the assessment model. Implementing this pipeline using the Loginom analytical platform, which is included in the Russian software register, is described. The features of preparing training data, as well as the results of teaching the model and checking it on the test data set are considered.