Unmanned Aerial Vehicles (UAVs) Battery Power Anomaly Detection Using Temporal Convolutional Network with Simple Moving Average Algorithm

Abstract

Nowadays, the widespread use of UAVs in various fields has led to studies on many issues related to UAVs in the literature. In particular, any deterioration that may occur in UAV systems and their inability to detect them, cause uncontrolled accidents and financial losses. In this study, temporal convolutional network and moving average are used to identify anomalies in instantaneous battery power consumption of UAVs. In time series prediction, usually preprocessed data with various statistical methods are used as inputs to the neural networks. In this study, both the power consumption data obtained from the UAV battery sensors and the calculated simple moving average data are given as multi-variate inputs to the temporal convolutional network. In the simulation results, the effect of using the simple moving average data together with the data from the UAVs sensor, the instantaneous battery power consumption prediction and the detection of anomalies have been examined. The simulation results showed that the combined use of simple moving average data and data from UAVs sensors helps to achieve more successful results in instantaneous power consumption prediction and anomaly detection with temporal convolutional network.