A Predictive Energy Management Strategy for Hybrid UAV Based on Flight Condition Recognition
Year: 2026
Type: conference-paper
Abstract: Efficient energy management is critical to improving the endurance and economic performance of hybrid unmanned aerial vehicles (UAVs). This paper proposes a predictive energy management strategy for hybrid UAVs based on flight condition recognition. First, statistical features of UAV state variables are extracted and reduced by principal component analysis, and K-means clustering is used to identify typical flight conditions, which are then recognized online using a K-nearest neighbors classifier. Second, a bidirectional long short-term memory (BiLSTM) network integrating flight condition information and historical power demand is constructed to predict future power demand, and a hyperparameter sensitivity analysis is conducted to determine an appropriate network structure. Finally, a model predictive control strategy is developed based on the prediction results to optimize engine power allocation while maintaining battery state of charge within a desired range. Simulation results show that the proposed strategy improves future power-demand prediction accuracy and achieves near-optimal economic performance, with equivalent fuel consumption reaching 98.65% of the dynamic programming benchmark and 1.24% lower than that of the strategy using only historical power demand for prediction.
Authors: Aoxuan Li, Yansong Liu, Weiqing Huang
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204179957