|
Several versions of
the SmoothTalk TM lithium battery management system (BMS) are
in development to address the need for taking care of individual cells in a traction pack
of an electric vehicle. Systems vary in complexity and functionality. There are two main
types of the BMS - pure hardware based for especially harsh operating environments, and
traditional microcontroller-driven type. The hardware based system uses modular charging approach - initially all cells in a pack are charged as a group by a bulk charger while each one is monitored not to exceed max. voltage recommended by the cell manufacturer. Then each cell receives finished equalizing charge from individual isolated mini-chargers set up for the same max allowed finish voltage corresponding to 100% SOC. This way each cell in a traction pack gets what it needs regardless of other cells in the pack. previous cycles as long as charging process is not interrupted until completed. There are common alarm, end of charge circuits and timers as well as supervisors allowing to make sure there are no failed chargers in the chain. The system will take care of proper charging and - by definition - balancing, but will not prevent overdischarge, limit battery current or collect any battery related parameters. Most if this functionality is included in traction controller or inverter. Since there is no software involved, there is no chance for microcontrollers lock-up, wrong code execution, etc. There are no digital circuits in the hardware based BMS - it is all pure analog. The software based BMS takes traditional approach - shunting extra charge away from some cells thus balancing them. Such systems always consist of the main controller (decision maker) and remote controlled switches connecting shunting elements (resistors or FETs) across respective cells thus bypassing charging current or allowing selective discharge. Optionally remote nodes may contain boost circuits - small DC/DC converters that can be remotely turned on and off similar to shunts. All the nodes, main controller and charger typically communicate via CAN bus. Main controller can address each cell's node, measure and store cell's voltages. After processing collected data based on current state of charge, learned previous cycles behavior, and thus expected condition, controller can individually treat each cell by taking out or optionally adding small amounts of charge at the time until cells are balanced according to predefined criteria. For cell chemistries where voltage is fairly good and consistent indication of SOC, such as LiPo, simple voltage balance method is sufficient, whereas for the types which maintain about constant voltage output over wide range of SOC, such as LiFePo, typically coulomb Ah capacity balancing is deployed. The charge is taken out by shunting cells, or more efficiently by patented pseudo-shunting method. Pseudo-shunting allows taking charge out of cells (bypassing charging current) without applying resistive shunt which dissipates energy as heat, e.g. effectively it allows electrically remove cells from series string being charged without physically disconnecting them. Adding the charge is done by individual boosters (see modular charging above). The software based BMS allows collecting and massaging historical data and implement other functions unavailable for hardware based system. Physically the cell electronics PCBs designed by Metric Mind Engineering fit the cells they are installed on. No remote sensing is deployed unless requested by a customer. Currently 3 types of cells are being manufactured: cylindrical (for instance common 18650, A123 Systems, SAFT, GAIA), flat pouch (Kokam, K2) or prismatic (Valence, Thunder-Sky, SAFT and many others) No two cells are created equal. So, connected in series and being cycled as one group, the cells will gradually drift out of SOC balance. Lower capacity cells charge and discharge quicker so their terminal voltage may be higher or lower than the average; the temperature gradient across the battery pack results in further imbalance. Identical initial capacity cells might have different self-discharge rate, and so on. This, however, is expected and does not constitute a problem with the pack, so may not require special powerful balancing actions. Featured BMS systems' smart algorithm anticipates cell behavior learned from previous charge/discharge cycles to avoid pointless activity of trying to keep individual voltages appear the same at all times. A combination of terminal voltage near 0% SOC and 100% SOC and amount of amp-hours stored in a cell (adjusted to actual initial capacity) is used to determine running SOC and required action. Terminal voltage swings during driving or regenerative braking usually do not allow making meaningful measurements. Therefore, during driving the system only tracks energy usage (amount of Ah in and out) and this determines amount of charge needed to refill partially discharged battery to exactly 100% (or other preferred amount) of SOC. Not only manufacturing differences or defects lead to non-uniform cells. If more than one location is used to place all the cells and no active temperature control is deployed, it is practically guaranteed that groups of cells in different locations in a vehicle will have different temperatures (e.g. self-discharge rates). A BMS should take this into account and intelligently compensate for it using manufacturer's (or empirically collected) data. During discharge dynamic and thermal behavior of the battery is very complex and in general unpredictable as depends on the driving pattern, individual cell internal resistance, temp, age, amount of cycles accumulated, etc. It is rare for even cell manufacturers to have adequate cell models allowing to predict its behavior, however, strictly speaking it is not nesessary for successful BMS design. Nevertheless, we prefer to implement custom top performance designs, so when enough data is available, we model a battery behavior as well as vehicle subsystems in matlab environment - that allows optimize the system for intended application and the vehicle. Obvious advantage of this appoach is we can predict how particular battery type will perform *in your vehicle*, not on the lab bench. Thus we avoid blindly building a system and later tweak it into desirable behavior. If any configuration mistakes occur or design assumptions are wrong, we rather prefer to make such mistakes on paper (in simuation software) than witness spectacular battery failure later. Changing configuration and capacity, adding and removing cells and type of chemistry is as simple as editing config file or cut and paste from the cells library which we maintain and expand. Working directly with top OEM cell manufactureres such as EiG, Kokam or Altairnano, we may obtain their proprietary test data that is stored in BMS memory in a form of lookup tables. Should you decide to fit a vehicle with different battery later, changing whole BMS behaviour involves either swapping EEPROM chip or downloading new config file into the BMS' main controller (provided, cell's electronics physically fit and no hardware re-design required). If no data is avaialble from the manufacturer, we can learn the battery by bench testing a sample using cycler and data acquisition system (data logger) and obtaining empirical data for the simulator as well as for real EEPROM later. Customer benefits from shorter development time and the system that works at the optimized point. This approach also allows to prioritize performance, battery life or anywhere in between as requiested, by editing parameters of matlab BMS model, responsible for its behavior. Few prototypes of the software based node construction as well as purpose built hardware based liquid cooled LiP battery module prototype assembly can be seen on the photos below. These are samples of custom BMS systems designed per customer's specifications. |
