LITHIUM-ION BATTERY Performance analysis of equalizer circuits for lithium-ion cells
This paper proposes the analysis of the performance of two different architectures of BMS. The comparison has been performed in terms of equalization speed, power losses, complexity of both hardware and software implementation, size and cost. The functionalities of the equalization process have been verified in charging, discharging and idle state.
The importance of Battery Management System
Nowadays, the adoption of lithium-ion batteries is growing very quickly in numerous fields of application. However, the needed of a large number of series-connected cells depending on the application and the inhomogeneity of cell parameters can lead to unbalanced voltage distribution, which limits the battery pack operation.
Therefore, a Battery Management System (BMS) results fundamental for ensuring the safe operating conditions of the battery pack as well as maximizing the usable capacity by means of an equalization circuit. However, it is still a challenge to identify the optimal architecture for a cell equalizer circuit as well as to define parameters for correctly comparing the performance. So far, the comparison has been operated by describing advantages and disadvantages of the equalizer circuits, without providing a qualitative or quantitative analysis that can fully highlight the effectiveness of the different solutions. This paper reports a quantitative analysis related to the comparison between the passive BMS (P-BMS) and an active BMS (A-BMS) that takes into account different relevant aspects.
Proposed BMS architectures
Both architectures have been designed for six series-connected cells. The design of the P-BMS is optimized for handling the dissipated power due to high balancing current (1A). The A-BMS is based on a Multiple Active Bridge (MAB), in which every cell is connected to a medium-frequency multiwinding transformer by means of an H-bridge converter. This novel architecture allows for achieving a cells-to-cells equalization. The control strategy is based on a phase-shift modulation technique, thus just one common PWM signal is adopted. Figs. 1 and 2 show the generalized scheme for N cells and the prototypes of the P-BMS and the A-BMS, respectively.
The A-BMS results to be more expensive and larger than the P-BMS. Moreover, a higher number of components needs to be considered as well. For what concern the control complexity, despite the hardware complexity, the A-BMS allows for managing all the H-Bridge converters by means of just a single common PWM signal, leading to an easier control algorithm. On the contrary, six control signals are required in the P-BMS, one for each cell that composes the battery pack. From a fault tolerance point of view, fault conditions on switches result in different issues for the P-BMS or the A-BMS. In particular, a fail closed switch leads to an overdischarge of the corresponding cell in the P-BMS. While, a short circuit can potentially occur in the A-BMS in case of the simultaneous closing of both switches in a H-Bridge converter leg. Nevertheless, a short circuit can be interrupted by adopting a fuse, whereas it is more complicated to reduce the risk related to the switch failure in the P-BMS.
All components of both P-BMS and A-BMS have been modeled and implemented in Matlab® using Simulink-Simscape. Accurate models for the battery cells, the multiwinding transformer and the H-Bridge converters have been adopted. A zero order model has been calibrated for the SONY VTC6 3-Ah cell by means of experimental tests considering different C-rate, current direction, state of charge (SoC) and temperature.
The transformer and power switches models have been implemented taking into account the datasheet parameters, including either the conduction or the switching power losses for the H-Bridge converters. The balancing strategy for both P-BMS and A-BMS has been developed for three different operating modes: charging, discharging and idle state, when neither power supply nor load are connected to the battery pack. In idle state, the cells to be balanced are identified and discharged until all cell voltages are within a desired voltage band (Vband). On the other hand, once enabled, the discharging and charging modes are stopped when the minimum cell voltage (Vmin) is lower than the minimum voltage threshold (Vth_down) and the maximum cell voltage (Vmax) is greater than the maximum voltage threshold (Vth_up), respectively.
The idle state mode is performed during charging as well. There are two main differences in the operating modes for the A-BMS with respect to the P-BMS. First, the P-BMS requires a continuous control of the equalization process of each cell, whereas the A-BMS allows for automatically achieving the balancing of the battery pack. Second, the equalization process can be enabled during discharging only for the A-BMS. Fig. 3 reports the numerical results for both BMS architectures in each operating mode. Different voltage imbalances among cells (ΔV) have been considered starting from 50 mV to 350 mV. The SoC of the most charged cell (SoCinit) has been set equal to 100% in idle state and in discharging and equal to 40% in charging.
Therefore, a linear distribution of the initial SoC of the cells has been adopted on the basis of SoCinit and ΔV. The results show that the P-BMS allows for achieving low equalization time (teq), especially for ΔV lower than 150 mV. However, a considerable reduction of the energy involved in the equalization process (Ebal) can be achieved by adopting the A-BMS. In addition,
the A-BMS shows performance improvements either in charging or in discharging with respect to the P-BMS. In detail, more energy can be provided in discharging (Eprov) by the battery pack with an increase of the discharging time (tdisch). While, lower charging time (tch) is achieved and a smaller amount of energy from the charger (Echarger) is required. Moreover, these improvements increase for higher ΔV.
Experimental tests have been performed in idle state for the P-BMS and the A-BMS, as reported in fig. 4. Starting with a voltage imbalance of 350 mV, all the cell voltages result within a desired voltage band of 30 mV at the end of the equalization process, highlighting the correct functionality of both BMS architectures. Moreover, this results are in line with the ones obtained in the numerical analysis. Indeed, the differences in teq between the results achieved for the P-BMS and the A-BMS with respect to the numerical results are equal to +1% and +4%, respectively. In addition, the differences between the teq achieved with the P-BMS and the A-BMS in the same operating conditions are equal to 5% and 11% in the numerical and experimental results, respectively. This fully confirms the validity of the numerical analysis and the goodness of the comparative analysis.