Online Estimation of Internal States Distributions in Li-ion Batteries
鋰電池內部狀態分佈的在線預測方法
Student thesis: Doctoral Thesis
Author(s)
Detail(s)
| Awarding Institution | |
|---|---|
| Supervisors/Advisors |
|
| Award date | 15 Sep 2016 |
Link(s)
| Permanent Link | https://scholars.cityu.edu.hk/en/theses/theses(f2a50f27-775d-44ec-b893-3ca6abf987be).html |
|---|---|
| Other link(s) | Links |
Abstract
Hybrid electrical vehicles (HEV) are prevailing worldwide nowadays due to their environmental friendliness. Lithium-ion batteries are suitable for powering the HEV given their advantages in high energy density, weak memory effect and low environmental pollution. Battery management system (BMS) aims to carry out the real-time monitoring of the battery status to ensure safety and health of batteries.
A BMS monitors the internal states distribution of the battery, including (1) Current: complex internal current profile caused by real-road operation, (2) Voltage: total voltage, voltages of individual cells, minimum and maximum cell voltage or voltage of periodic taps, (3) Temperature: temperature distribution under various cooling, (4) State of charge (SOC) or depth of discharge (DOD): to indicate the charge level of the battery. The internal states can be sorted into two groups, electrical states and thermal states.
The online monitoring of internal processes is a difficult task. Main difficulties lie in the following aspects:
1. Spatiotemporal coupling: Electrochemical process and thermal process in the battery cell are nonlinear PDEs systems. They can be solved by finite element method which unfortunately cannot be used online for its high computational cost.
2. Boundary cooling effects: to keep the batteries in good working conditions, the various cooling methods may be adopted by the BMS which makes the online modeling of temperature distribution difficult.
3. Spatial nonlinearity: the internal states governed by partial differential equations have spatial nonlinearity. Traditional linear methods may not be able to model the nonlinearity.
4. Physical measurement limitations: very few sensors can be used for online measurement which causes difficulties in the online monitoring of the processes. Internal states, such as local current distributions, are not measurable in the HEV application.
5. The electro-chemical behaviors in batteries are sensitive to the external disturbances and variation of operational conditions. The online adaptive algorithms should be developed for real time estimation.
There is still an open area for online estimation of internal state distributions in lithium-ion battery systems. The main objective of this research is to develop a systematic estimation method for internal states distributions in vehicle battery systems, which is compatible to the existing management system. The estimated distributions can be used for control, health monitoring and design purposes.
The electrical states are spatiotemporal coupled and difficult to measure in the real-time operation. Therefore, a space-time separation method is applied to model the electrochemical properties of the battery with the help of the extended Kalman filter. The model is efficiently optimized by using least absolute shrinkage and selection operator based adaptation method and can be updated through data-based learning. The derived analytical model is able to offer a fast estimation of internal states of the battery. Thus, it has potential to become a prediction model for battery management system.
When considering the cooling effects, the estimation of the internal temperature of cylindrical battery is very challenging. The Karhunen-Loève (K-L) decomposition method is capable of dealing with constant boundary. However, the time-varying cooling condition should be considered in the real operation. An explicit compensation algorithm has been proposed to model the boundary cooling effects on the truncated models. With the compensated model, the dual extended kalman filter can be then applied for the adaptive online estimation.
In addition, the thermal process of the embedded batteries cannot be precisely modeled by traditional methods such as proper orthogonal decomposition. Since it is spatiotemporal process governed by partial differential equations with spatial nonlinearity. A deep model reduction method is proposed for the complex temperature distribution. The proposed method also can be applied to other distributed parameter system such as hyperbolic PDEs and shown better modeling efficiency and accuracy.
The effectiveness of the modeling method is validated through simulation experiments, and its high modeling accuracy has also been demonstrated in the comparison studies. The developed method is simple and can be adaptive for online application. It can also be easily integrated into existing battery management systems for estimation.
A BMS monitors the internal states distribution of the battery, including (1) Current: complex internal current profile caused by real-road operation, (2) Voltage: total voltage, voltages of individual cells, minimum and maximum cell voltage or voltage of periodic taps, (3) Temperature: temperature distribution under various cooling, (4) State of charge (SOC) or depth of discharge (DOD): to indicate the charge level of the battery. The internal states can be sorted into two groups, electrical states and thermal states.
The online monitoring of internal processes is a difficult task. Main difficulties lie in the following aspects:
1. Spatiotemporal coupling: Electrochemical process and thermal process in the battery cell are nonlinear PDEs systems. They can be solved by finite element method which unfortunately cannot be used online for its high computational cost.
2. Boundary cooling effects: to keep the batteries in good working conditions, the various cooling methods may be adopted by the BMS which makes the online modeling of temperature distribution difficult.
3. Spatial nonlinearity: the internal states governed by partial differential equations have spatial nonlinearity. Traditional linear methods may not be able to model the nonlinearity.
4. Physical measurement limitations: very few sensors can be used for online measurement which causes difficulties in the online monitoring of the processes. Internal states, such as local current distributions, are not measurable in the HEV application.
5. The electro-chemical behaviors in batteries are sensitive to the external disturbances and variation of operational conditions. The online adaptive algorithms should be developed for real time estimation.
There is still an open area for online estimation of internal state distributions in lithium-ion battery systems. The main objective of this research is to develop a systematic estimation method for internal states distributions in vehicle battery systems, which is compatible to the existing management system. The estimated distributions can be used for control, health monitoring and design purposes.
The electrical states are spatiotemporal coupled and difficult to measure in the real-time operation. Therefore, a space-time separation method is applied to model the electrochemical properties of the battery with the help of the extended Kalman filter. The model is efficiently optimized by using least absolute shrinkage and selection operator based adaptation method and can be updated through data-based learning. The derived analytical model is able to offer a fast estimation of internal states of the battery. Thus, it has potential to become a prediction model for battery management system.
When considering the cooling effects, the estimation of the internal temperature of cylindrical battery is very challenging. The Karhunen-Loève (K-L) decomposition method is capable of dealing with constant boundary. However, the time-varying cooling condition should be considered in the real operation. An explicit compensation algorithm has been proposed to model the boundary cooling effects on the truncated models. With the compensated model, the dual extended kalman filter can be then applied for the adaptive online estimation.
In addition, the thermal process of the embedded batteries cannot be precisely modeled by traditional methods such as proper orthogonal decomposition. Since it is spatiotemporal process governed by partial differential equations with spatial nonlinearity. A deep model reduction method is proposed for the complex temperature distribution. The proposed method also can be applied to other distributed parameter system such as hyperbolic PDEs and shown better modeling efficiency and accuracy.
The effectiveness of the modeling method is validated through simulation experiments, and its high modeling accuracy has also been demonstrated in the comparison studies. The developed method is simple and can be adaptive for online application. It can also be easily integrated into existing battery management systems for estimation.
