An Experimental System for Battery Management Algorithm evelopment Jonas Hellgren, Lei Feng, Björn Andersson and Ricard Blanc Volvo Technology, Göteborg, Sweden E-mail: {jonas.hellgren, lei.feng, bjorn.bj.andersson, ricard.blanc}@volvo.com Abstract The objective of this paper is to present an experimental battery system. In the development of the system, system engineering principles are utilized. This implies for example that solution independent requirement settings are used. State of charge estimation, maximum power capability estimation and cell balancing are prioritized functionalities. In the paper, system hardware and algorithms are explained and presented. Keywords battery system, battery prototype, battery management, battery management unit, electric energy storage 1. Introduction The Volvo Group has high ambitions regarding heavy hybrid vehicles. From 2009 there are hybrid buses from Volvo on the market. These vehicles utilize a parallel hybrid powertrain with the electric machine placed between the engine and the transmission. Thanks to brake energy regeneration and utilizing engine shuts off when stopping, the fuel consumption is reduced by roughly 30% for city buses. The battery system is essential in any hybrid vehicle. Characteristic for a battery system (BS) is the ability to both store and deliver energy. In hybrid electric vehicles (HEVs), it is essential to design the BS properly since it is normally the major explanation of the higher cost of a hybrid powertrain. ue to the high cost, it is essential that the battery management system has enough accuracy to maximize the usage of the battery while maintaining safe operation without significantly shortening the operational life span. esigning a BS is a difficult task. The reason is that many aspects, for example packaging, performance and safety, must be considered simultaneously. Today the Volvo group BS is provided from a supplier. The objective of this paper is to present an experimental down-scaled BS. The important result for the Volvo group is not the experimental prototype itself, but the experience from developing it. This experience will be fruitful for the development of future hybrid vehicles. Chapter 2 briefly describes the structure of battery systems and their major functionalities. Chapter 3 elaborates the physical configuration, the specifications, and the algorithm design of the proposed experimental BS. Chapter 4 concludes the paper and outlines possible future work. 2. The technology of hybrid vehicle battery systems Figure 1 shows how a BS in general is structured. Typically the load is an electric machine. The voltage of the BS must be within the operating voltage range of the load. Therefore, an appropriate number of cells are connected in series. Figure 1. General system layout of a battery system. This example ESS system has two cells in series. The function of the battery management unit (BMU) in a BS is multifaceted. Examples of important BMU functions are:
Providing information about the state of charge (SoC). Protecting the cells from destructive loading. Such loading can for example mean that some cell voltage is exceeded. Predicting the short and long term power capacity. Normally, the maximum possible charge and arge power depends on SoC and resistance. In addition the BS must be able to compensate for any imbalances in cell voltages. To reach balance, the charge of cells with high voltage should be lowered while cells of low voltage should be charged. 3. Proposed experimental battery system estimated SoC are not allowed. This implies that the absolute value of the time derivate (%/sec) of the SoC always must be less than Value. The maximum short-term arge power ability shall be estimated. The maximum short-term charge power ability shall be estimated. The maximum long-term arge power ability shall be estimated. The maximum long-term charge power ability shall be estimated. 3.2 Hardware description In the project, system engineering principles are utilized. This implies that solution independent requirement setting (defined in the beginning of the project) and systematic concept generation and selection are applied. Another important phase, more relevant in the end of the project, is validation. The process is iterative; one can for example think of adding/removing/changing requirements during the project. Figure 2 shows the BS. It consists of 14 cells, BMU, connector, frame and signal interfaces. Lithium-ion battery cells with properties listed in Table 2 are used. Cell BMU 3.1 Requirements Table 1 presents the functional requirements. These requirements reflect the prioritized functionality of the experimental BS. One can for example see that state of health diagnosis and cell fault detection are not in the scope of the project. An important design requirement is that the BS shall be equipped with an interface for connection to a computer. The reason of this is to enable validation of the functional requirements by data logging. Table 1. Functional requirements. = demand, W = wish. Requirement /W Value Cell balancing is carried out. The cell balancing shall be 2 carried out with power losses smaller than Value % of the loading power An estimation of state of charge (SoC) shall be calculated and available in real-time during operation of the BS. Sudden jumps in the 1 Figure 2. Experimental battery system. The dimensions are 750x250x250 mm. Table 2. Cell properties. Term Value Noal capacity 20 Ah Noal voltage 3.3 V imensions 15x126x220 mm Mass 0.7 kg To achieve a compact design, the cells are placed side-by-side with alternating polarity to the adjacent cell, see Figure 3. Two endplates, to be placed perpendicular to the cells, are made from an isolating and wood-like material. The plates are slitted equidistantly into combs. The cell positions are fixed by sliding the cells into the slits. The
distance between the slits are adapted to allow some space between the cells for the cooling air flow. Every other pair of adjacent cell terals are electrically connected by folding each pair of teral lips over an aluium bar and then clamped together by another bar. The bars were forced and held together using bolts, with spring washers, locking into threaded holes outside of the lips, hence there is no need to drill holes into the fragile lips. This also facilitates replacements of cells. Figure 3. Cell arrangement, series connection (upper picture). Cell connection (lower picture) The BMU hardware is from the Swedish company Effektutveckling[1]. Broadly speaking, the BMU consist of one master and two slave units. The main responsibilities of the slave units are to assemble current and voltage measurements. This system only includes two slave units. In systems with many more cells, several slave units are appropriate. The master unit processes measurement signals provided by the slave unit via an internal signal bus. Thanks to a CAN bus, the master unit can provide the hybrid, plug-in hybrid or electric vehicle with information such as for example SoC. The master includes a digital processor that uses code primary generated from the Real time workshop toolbox in the Matlab/Simulink environment. Other hardware functionalities that characterize the system are: Pre-charge circuit. This circuit prevents the system from high currents when the main contactor is switched on. Air cooling realized by a fan on top of the cells. Because this will work similar to a vacuum cleaner, the solution will only work in a clean environment. Fuse. To protect the system from extreme currents. Temperature measurement. Three temperature sensors are placed at different locations in the BS. 3.3 Algorithm description The main challenge with the battery system presented in this paper is to develop algorithms for cell balancing, SoC estimation and state of power (SoP) estimation. eviations in cell behaviors generally occur because of 1) changes in internal impedance and/or 2) cell capacity reduction due to aging. If a single cell in the BS has deviant behavior, that cell becomes a likely candidate for differing in voltage. Therefore a mechanism to balance the cells must be incorporated. There are two main approaches for balancing a BS, passive and active [2][3]. For passive systems, energy in cells with too high voltage is removed by resistors connected in parallel with the cells, see Figure 4. Active cell balancing methods move energy from high voltage cells to low voltage cells. Normally, active techniques benefit from lower losses but suffer from higher costs. The system presented in this paper utilizes a passive balancing technique. Figure 4 shows a BS with n cells. For each cell a switch controls if current should be dissipated into the leakage resistor or not. The algorithm for cell balancing, in the BS presented in this paper, controls the switches in such a way that cell voltage deviation is imized. Figure 4. Passive cell balancing. SoC estimation is an important problem. If SoC is not accurately estimated, it will for example not be possible to calculate the remaining operating range of an electric vehicle. Incorrect SoC estimation can in worst case cause cell damage/fire because Lithium-ion cells are very sensitive to overcharging. From Figure 5 it is possible to understand why SoC estimation is challenging. For moderate SoC values (0.2<SoC<0.8), the cell voltage has a very small dependency on SoC. In addition, cell voltage varies with load current in a complex manner. Therefore, just measuring cell voltage is far from enough for estimating SoC.
u cell 3. 2. Figure 5. Open circuit voltage curve. The curve represents cell voltage when no load current is applied. Our experimental BS uses an extended Kalman filter for SoC estimation. One estimator for each cell is utilized. It uses measured load current and cell voltage to derive an estimated SoC. Loosely speaking, one can say that the filter uses a state space model of the cell to calculate the cell voltage from the load current. The states, e.g. SoC, of the state space model are affected by the difference between the calculated and measured cell voltage. More about Kalman filter based SoC estimation can be found in [4]-[8]. The purpose of the SoP algorithm is to evaluate the maximum charge and arge capability of the BS. The arge capability is for example crucial to perform maximum acceleration of an electric vehicle. Knowledge of the charge capability is important to maximize regenerative braking energy. In addition, peak power capability estimation is essential to avoid battery over charge or arge. For the specific BS, presented in this paper, SoP for two time horizons (short and long term) is calculated. The online SoP algorithm is based on two input signals, SoC and resistance. In practice, resistance is derived from temperature measurements and prior knowledge of the resistance-temperature relation. Figure 6 shows the in and output signals of the SoP algorithm realization. 0. SoC 1 The look-up tables in Figure 6 are derived from a pre-processing procedure. The procedure for creating maps for arging capability is based on the optimization problem stated in Error! Reference source not found.. Example constraints are stated in (1). (1) P (SoC, R) with respect to c.. c Table 3. SoP constraints for arge case. Constraint Comment i ( > i Cell current limitation. t) cell t) cell, 1 N ischarge implies a negative current; hence the expression can be rewritten as: i ( t) < i. u ( > u Cell voltage restriction. SoC ( t) > SoC Cell charge level restriction. T ( t) < T Cell temperature restriction. max The optimization compares numerous evaluations of P. For each evaluation, a simulation with a length equal to the prescribed time window is performed. If all constraints are valid during the simulation the evaluated is considered as feasible. P Figure 7 shows the look-up table in the case of long term arging. One reflection is that the power decreases with resistance. Because resistance decreases with temperature one can say that the possible arge power is substantially lower at low temperatures. Another reflection is that the arge power capability goes towards zero when SoC approaches zero. This is natural; an empty BS has no power to give. SoC cell resistance Look-up tables Pshort Plong Pchmaxshort Pchmaxlong Figure 7. Look-up table, long term arging. Figure 6. SoP implementation. Figure 8 shows an interesting phenomenon. When the cell voltage deviation is high
(5<std(SoC)<10), the arge power capability is zero for relatively large SoC values (15<SoC<30). The reason is that one single cell limits the SoP when the deviation is high. The conclusion is that a poor cell balancing reduces the practical BS energy capacity. Figure 8. ischarge power as function of SoC and cell voltage deviation. 4. Conclusions and future work Using a downscaled BS for algorithm development is an efficient way to increase the competence within battery technology. In addition there are positive safety aspects when experimenting with a low voltage system. Potential future works are: System validation. For example, it is of interest to verify the accuracy of the SoC algorithm. Further development of SoC, SoP and cell balancing algorithms. Implementation of state of health estimation. Implementation of safety and diagnosis functionality. Full scale BS development. 5. References [1] Effektutveckling, www.effektutveckling.se, 2010. [2] S. Moore and P. Schneider, "A Review of Cell Equalization Methods for Lithium Ion and Lithium Polymer Battery Systems", SAE, 2001. [3] X. Wei and B. Zhu, The Research of Vehicle Power Li-ion Battery Pack Balancing Method, ICEMI, 2009. [4] G. Plett, "Extended Kalman filtering for LiPB-based HEV battery packs. Part 1. Background", Journal of Power Sources, 2004. [5] G. Plett, "Extended Kalman filtering for LiPB-based HEV battery packs. Part 2. Modeling and identification", Journal of Power Sources, 2004. [6] G. Plett, "Extended Kalman filtering for LiPB-based HEV battery packs. Part 3. State and par estimation", Journal of Power Sources, 2004. [7] M. Urbain,etc, State Estimation of a Lithium-Ion Battery Through Kalman Filter, IEEE, 2007. [8] S. Lee, etc, State-of-charge and capacity estimation of lithium-ion battery using a new open-circuit voltage versus state-of-charge, Journal of Power Sources, 2008. 6. Authors The authors are presented in this chapter. r. Jonas Hellgren Hybrid power system specialist. E-mail: jonas.hellgren@volvo.com Jonas has more than ten year experience in r. Lei Feng E-mail: lei.feng@volvo.com Lei s research interests are on control engineering and the model-based development of automotive embedded systems. Björn Andersson Email: bjorn.bj.andersson@volvo. com Björn is working with Ricard Blanc Email: ricard.blanc@volvo.com Ricard is working with