Batteries are of vital importance for storing intermittent renewable energy for stationary and mobile applications. In order to charge the battery and maintain its capacity, the states of the battery - such as the current charge, safety and health, but also quantities that cannot be measured directly - need to be known to the battery management system. State estimation estimates the electrical state of a system by eliminating inaccuracies and errors from measurement data. Numerous methods and techniques are used for lithium-ion and other batteries. The various battery models seek to simplify the circuitry used in the battery management system. This concise work captures the methods and techniques for state estimation needed to keep batteries reliable. The book focuses particularly on mechanisms, parameters and influencing factors. Chapters convey equivalent modelling and several Kalman filtering techniques, including adaptive extended Kalman filtering for multiple battery state estimation, dual extended Kalman filtering prediction for complex working conditions, and particle filtering of safety estimation considering the capacity fading effect. This book is necessary reading for researchers in battery research and development, including battery management systems and related power electronics, for battery manufacturers, and for advanced students in power electronics.
Inspec keywords: support vector machines; Kalman filters; equivalent circuits; battery powered vehicles; secondary cells; parameter estimation; nonlinear filters; battery management systems; state estimation
Other keywords: adaptive Kalman filters; support vector machines; battery powered vehicles; parameter estimation; secondary cells; state estimation; battery management systems; nonlinear filters; equivalent circuits
Subjects: Transportation; Secondary cells; Filtering methods in signal processing; Semiconductor device modelling, equivalent circuits, design and testing; General electrical engineering topics
With the development of society, energy security and the environmental pollution caused by it has become a key issue that all sectors of society are concerned about and urgently need to be resolved. The deepening of the world energy crisis has led to the emergence of new energy industries as well as the increasing awareness of environmental protection. Among them, lithium-ion batteries have developed rapidly in the field of new energy due to their higher energy density and longer cyclic life. This chapter briefly introduces the use scenarios and market conditions of lithium-ion batteries, the common methods of lithium-ion battery state estimation, and their research significance. Then, the development status of the battery management system is briefly described. At the same time, the estimation methods of various state parameters are introduced for lithium-ion batteries, which lays a foundation for subsequent research. The research conclusions and further research plans are discussed, the content is reviewed, and the system state estimation is emphasized to achieve the purpose of safety protection and lifetime guarantee.
This chapter introduces the operating mechanism, influencing factors, key indicators, and some mainstream state estimation methods of lithium-ion batteries. First, understanding the main composition and internal working principle of lithium-ion batteries is the prerequisite and basis for other work. Second, internal resistance, open-circuit voltage (OCV), terminal voltage, current thermal energy, capacity variation, and temperature characteristics are the main features of the battery, which can be further used to accurately characterize the battery state. The state of charge, state of health, state of power, depth of discharge, and cyclic life are the key indicators of state estimation, while the discharging test, Ampere-hour integral, OCV, internal resistance, and Kalman filtering are basic state estimation strategies. Based on the Kalman filter, there are many improved algorithms, such as the unscented Kalman filter and adaptive Kalman filter, which have achieved good application effects. Besides, other algorithms such as neural networks, support vector machines, and some improvement strategies are also introduced. The advancement of these algorithms has made an important contribution to improving the whole-life-cycle state estimation effect of lithium-ion batteries.
It is vital to establish an accurate battery model for the characteristic analysis and performance optimization of the batteries. This chapter introduces several popular modeling strategies of batteries, including Rint, partnership new generation of vehicles, and Thevenin modeling methods. According to the different internal characteristics of various batteries, four kinds of battery modeling methods are studied, including the electrochemical model, mathematical model, thermal-based model, and shepherd model. The advantages and disadvantages of these models are discussed in detail through theoretical analysis and open-circuit voltage characteristics. Meanwhile, several improvement measures are investigated for the Thevenin equivalent circuit modeling. Their benefits and limitations are also compared. Through the demonstration of simulated models, the basis for accurate battery state of charge estimation is constructed. Finally, different parameter identification methods for various battery models are introduced in detail and verified with the Beijing bus dynamic stress test. The results show that the model parameters can be well identified.
The Kalman filtering extension strategies for accurate battery state estimation are analyzed, especially the extended Kalman filtering and fractional-order extended Kalman filtering algorithms. The battery equivalent circuit modeling methods include second-order Thevenin modeling and other modelings. The procedure design and verification are introduced in this chapter. The model parameters are identified by the recursive least-square method, the forgetting factor least-square method, and the hybrid pulse-power characteristic (HPPC) experimental tests, in which the results are verified by pulse-current cycling experiments. The battery equivalent modeling strategy is explored to obtain its state-space expression, which is then used to realize its state of health and state of charge estimation. The fractional experiments are then conducted, including real-time platform implementation, HPPC tests, and capacity tracking experiments. The experimental procedure is designed, including the whole experiment structure and the detailed procedure flowchart. After that, the experimental platform is built, and the designed experiments are realized. Consequently, this chapter introduces the extended Kalman filtering algorithm into the state of health and state of charge estimation, which can realize the dynamic parameter estimation of the battery system and improve the state estimation accuracy for lithium-ion batteries under the HPPC, BBDST, and dynamic stress test conditions.
Based on the electrical equivalent modeling, this chapter introduces an adaptive extended Kalman filtering (AEKF) method to estimate the battery state of charge and state of power based on the electrical equivalent modeling procedure design. Accurate online estimation of lithium-ion battery status can effectively extend battery life and improve battery safety, which is essential for the battery management system from the control perspective. Lithium-ion batteries are widely used in control fields such as electric vehicles and drones. The state estimation of the battery has a great influence on its working state, and the highly accurate lithium-ion battery state estimation is more conducive to the controller and the implementation of efficient energy management by the controller. Aiming at the shortcomings of traditional extended Kalman filtering design with insufficient prior information and dynamic environment, a carrier tracking algorithm based on AEKF is proposed. By monitoring the filter innovation or the residual dynamic change in real-time, the algorithm corrects the variance of the state noise and the variance of the observed noise by then adjusting the filter increment. The lithium ion battery state estimation is carried out through online parameter identification and real-time correction of model parameters. Combined with the iterative calculation of the AEKF algorithm, the mean value and variance of the algorithm error are judged. Based on the second-order Thevenin equivalent circuit model, the battery state results are judged and estimated effectively. The experimental results show that the algorithm can significantly improve the convergence speed without reducing the estimation accuracy. It is practical in online state estimation applications.
The Kalman filtering algorithm plays a vital role in battery state estimation. To solve the difficulties in real-time estimation and low precision under various working conditions for the battery state of charge and the state of health (SOH) estimation, the extended Kalman filtering algorithm is improved and a dual extended Kalman filtering method is proposed to realize the joint estimation of the model parameters and state of charge (SOC) in this chapter. Taking the experimental current and voltage as inputs, the model state-space equation is established, and the SOC estimation results are obtained. On this basis, the estimation method of battery internal resistance is established. Through experiment tests, the recursive least-square method is assisted by using the established simulation model to achieve the prediction of model parameters in various working conditions. Through the experimental results under complex operating conditions, the internal resistance of the battery under various operating conditions is estimated, according to which the SOC and SOH can be predicted accurately for lithium-ion batteries.
To get a more accurate state of charge estimation for lithium-ion batteries, this chapter introduces an improved particle filtering algorithm named unscented particle filtering. This algorithm adopts the framework of the unscented Kalman filter. With an appropriate probability density function, the unscented transformation is used in the traditional particle filter algorithm. Through the unscented transformation, the improved algorithm can realize the mean and variance calculation more accurately and solve the shortcoming of particle exhaustion in traditional particle filter algorithms. To verify the effectiveness of the algorithm, the ternary lithiumion battery is selected as the research object, and the Thevenin equivalent circuit model is constructed. Finally, the experimental analysis is carried out under the Beijing bus dynamic stress test condition. The verification results show that there exists a great variation when the improved algorithm is used to estimate the state of charge value for lithium-ion batteries. The effect is very good, and the algorithm shows extremely strong tracking and robustness. The prediction error is stabilized within 1.5%, which brings good performance to the lithium-ion batteries.