Machine Learning-Driven Prediction of Lithium-Ion Battery Power Capability for eVTOL Aircraft

TL;DR

This paper tackles the challenge of predicting lithium-ion battery power capability for eVTOLs under high C-rate discharges and safety constraints. It introduces a novel two-horizon, emergency-aware problem formulation and proposes a hybrid physics-ML dynamic model (NDCTNet) to accurately predict voltage and temperature, together with an ML-based remaining discharge time (RDT) predictor to accelerate the search for the maximum allowable current. The solution yields the power limit $P_{ ext{max}} = i_{ ext{max}} V(t+H)$ by guiding a computationally efficient search, significantly outperforming a traditional shortcut approach in compute time while maintaining safety. Validation on a high-fidelity cell model demonstrates accurate predictions across a broad range of C-rates and horizon lengths, with substantial real-time speedups suitable for safety-critical eVTOL operations.

Abstract

Electric vertical take-off and landing (eVTOL) aircraft have emerged as a promising solution to transform urban transportation. They present a few technical challenges for battery management, a prominent one of which is the prediction of the power capability of their lithium-ion battery systems. The challenge originates from the high C-rate discharging conditions required during eVTOL flights as well as the complexity of lithium-ion batteries' electro-thermal dynamics. This paper, for the first time, formulates a power limit prediction problem for eVTOL which explicitly considers long prediction horizons and the possible occurrence of emergency landings. We then harness machine learning to solve this problem in two intertwined ways. First, we adopt a dynamic model that integrates physics with machine learning to predict a lithium-ion battery's voltage and temperature behaviors with high accuracy. Second, while performing search for the maximum power, we leverage machine learning to predict the remaining discharge time and use the prediction to accelerate the search with fast computation. Our validation results show the effectiveness of the proposed study for eVTOL operations.

Figures (6)

• Figure 1: Proposed LiB power prediction problem formulation for eVTOL. The emergency landing is specifically considered in the predictions to account for intricate dynamics of the LiB at high C-rates.
• Figure 2: The NDCTNet model in Tu:AE:2023Tu:AE:2024. It combines equivalent circuit models with FNNs to predict a LiB's terminal voltage and temperature behaviors. The model has been shown to have high accuracy and computational efficiency over broad C-rate ranges.
• Figure 3: ML-based fast RDT prediction model, denoted as $\Psi_{\textrm{ML}}$, which is used to accelerate the search for $i_\mathrm{max}$ in the proposed power prediction approach.
• Figure 4: The current, voltage and temperature of the LiB under a notional eVTOL mission profile at $T_{\mathrm{amb}}=25$$^{\circ}\mathrm{C}$. 0$\sim$75 s: take-off; 75$\sim$975 s: cruise; 975$\sim$1080 s: landing.
• Figure 5: LiB power capability prediction under a notional eVTOL profile using the proposed method. (a) The predicted $i_\textrm{max}$. (b) The resultant $P_\textrm{max}$. The $P_\textrm{max}$ reflects the maximum power load of the LiB for a cruise time of length $H$ before safely completing the landing of an eVTOL aircraft. It provides important information for the pilot to decide the landing location in an emergency.