US-12625184-B2 - System and method for estimation of battery state and health

US12625184B2US 12625184 B2US12625184 B2US 12625184B2US-12625184-B2

Abstract

Disclosed herein is a method for predicting the performance characteristics of a Li-ion battery over the course of the lifetime of the battery. A model based on integration of a set of ordinary differential equations is used to predict the voltage and thermal characteristics of the battery in a short period of time and a universal ordinary differential equation is used to predict the degradation of the battery by changing the parameters of the system of ordinary differential equations. The degradation of the battery is predicted in terms of change in various parameters of the battery (e.g., capacity, resistance, thermal behavior).

Inventors

Venkatasubramanian Viswanathan
Alexander Bills
Shashank Sripad

Assignees

CARNEGIE MELLON UNIVERSITY

Dates

Publication Date: 20260512
Application Date: 20210819

Claims (10)

1 . A method for estimation of the state and health of a battery comprising: predicting, using a performance model based on a set of performance parameters, a voltage, a temperature and a state of charge of the battery during a given duty cycle of the battery and estimating, using a degradation model, implemented as a neural network, changes in the performance parameters over a lifetime of the battery, the performance parameters including at least residual loss of lithium inventory and ohmic resistance gain, based on an input of a state vector generated by a model of the battery; wherein the performance parameters include one or more aging parameters evolved by the degradation model over the lifetime of the battery by a process comprising: choosing an initial set of aging parameters; evaluating a loss function for each duty cycle of the battery; and altering the selection of the aging parameters until the loss function is minimized.
2 . The method of claim 1 wherein the parameters of the performance model are estimated to predict performance of the battery.
3 . The method of claim 2 wherein the physics-based degradation model models mass, charge and potential redistribution over domains including bulk and surface regions of a cathode and an anode of the battery using ordinary differential equations.
4 . The method of claim 1 wherein the aging parameters are those performance parameters for which an estimation changes over the lifetime of the battery.
5 . The method of claim 1 wherein the degradation model comprises: a physics-based degradation model; and a universal function approximator, including, but not limited to a neural network and random forest, implemented using a set of neural differential equations and universal ordinary differential equations for estimating the degradation parameters used by the physics-based degradation model.
6 . The method of claim 5 wherein the physics-based degradation model is specific to a particular model of the battery.
7 . The method of claim 6 wherein the physics-based degradation model further includes a battery resistance model and a thermal model.
8 . The method of claim 5 wherein the physics-based degradation model comprises: a charge-loss due to SEI formation component; an active material isolation component; a lithium plating component; and a resistance increase component; wherein each of the components are implemented using ordinary differential equations.
9 . The method of claim 1 wherein the battery is used in an electric vertical take-off and landing (eVTOL) aircraft and electric short take-off and landing (eSTOL) aircraft.
10 . The method of claim 1 , implemented in an electrically-powered aircraft, wherein the duty cycle comprises: a take-off phase; a cruise phase; a landing phase; a first rest and cooling phase; a charging phase; and a second rest and cooling phase.

Description

RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Patent Application No. 63/068,398, filed Aug. 21, 2020, the contents of which are incorporated herein in their entirety. BACKGROUND The invention described herein pertains to the field of lithium-ion (Li-ion) batteries. Li-ion batteries are widely used in various industries due to their high energy density. Monitoring of the available energy of a Li-ion battery is important for iterative design procedures, safety considerations, and other applications in any industry which widely uses Li-ion batteries, including, for example, electrified mobility, electric grid and personal electronics. The improving performance, increasing cycle life, and decreasing cost of Li-ion batteries spurred by the mass market adoption of personal electronics and electric vehicles (EV) has recently enabled the development of electric aircraft. Electric aircraft convert energy stored in on-board batteries to propulsive thrust through a high voltage distribution bus, an electric motor, and a power inverter. This mode of propulsive energy transfer eliminates the complexity typically associated with gearboxes or mechanical transmissions, affords relatively low unit costs of propulsors (inverter, motor, propeller), and increases overall powertrain efficiency compared to internal combustion or turbine engine systems. This paradigm shift in the aircraft propulsion system has enabled a vast array of new hybrid and electric aircraft configurations. Many electric aircraft designs utilize distributed electric propulsion to realize novel configurations which can achieve a significant safety and efficiency advantage over conventional single or multi-engine aircraft. Notably, the developments in battery technology and distributed electric propulsion have opened the door to urban air mobility (UAM) by enabling the development of electric vertical takeoff and landing (eVTOL) aircraft. As outlined by the National Aeronautics and Space Administration (NASA), UAM aims to safely and efficiently transport passengers and cargo in an urban area. UAM has obvious benefits of convenience and speed for mobility in some markets and may also have an energy-efficiency advantage over ground transport. The design trade-offs that arise from using Li-ion batteries for electric aircraft designs are distinct from those with combustion engines, in large part due to the orders of magnitude difference in specific energy between Li-ion batteries and jet-fuel. Compared to terrestrial electric vehicles, the performance of aircraft is much more sensitive to battery weight. Thus, electric aircraft require careful integration and use of Li-ion battery systems. Most eVTOL aircraft are designed for the critical case: cruise to maximum range into a headwind, followed by a redirect reserve segment and a subsequent contingency landing such as a single propulsor failure. This mission profile is especially challenging to achieve close to the battery retirement state-of-health (SOH, retirement SOH is typically 80-85%), as maximum power output is demanded at a low state-of-charge (SOC). To ensure that the co-design of electric propulsion sub-systems is consistent with the sized vehicle geometry and weights, a rapid battery performance estimation method is required in the sizing loop. Importantly, the model must be supplemented with a degradation model to account for changes in performance over the lifetime of the battery. Estimating the state-of-charge over a duty cycle and the state-of-health over the lifetime of a cell has historically been performed using physics based models or empirical and data-driven machine learning models. While physical models are typically interpretable and accurate, they can often be computationally expensive. On the other hand, empirical and machine learning approaches trade interpretability for speed, while retaining accuracy. The physics-informed machine learning approach can help overcome this trade-off. In fields such as atomistic simulations and fluid mechanics, encoding physical principles for data efficiency and extrapolation in machine learning methods such as neural networks has shown promising results. Machine learning has been used extensively in the field of energy storage, including being used to optimize charging protocols with closed loop and reinforcement learning based methods, to predict the cycle life of batteries based on early cycle features, and to predict and forecast a battery's state of health. However, there have been few works which closely integrate electrochemical battery models and constraints with machine learning methods to improve performance predictions over the full life of a battery. SUMMARY OF THE INVENTION The invention described herein comprises a method for predicting the performance characteristics of a Li-ion battery over the course of the lifetime of the battery. The lifetime of the battery includes a large number of charge, discharge and