It is accepted that battery monitoring is vital for safety and best lifetime performance, particularly in EVs, but current monitoring methods are an evolution from cumbersome old techniques. A new, truly wireless approach is presented which leverages the advantages of fast and flexible edge computing.
Battery monitoring – a necessary but high-value overhead
Battery monitoring is like a dental check-up: it’s expensive and you really only want to know there’s nothing bad to report. Worse, you have to keep doing it, to get any level of confidence that things are not deteriorating. In an EV, you have to carry the dentist’s chair – the monitoring hardware – round with you all the time, occupying valuable space that could have been more battery cells. The hardware and cabling also adds cost and range-reducing weight.
But what price insurance? A prime function of a battery monitoring system (BMS) is to maintain safe charge and discharge, reducing risk of cell degradation, damage and even fire. The benefits go beyond that though, accurately knowing the state-of-charge (SoC) of a battery pack enables the vehicle range to be determined, reducing ‘range anxiety’ and charge times to be reduced. At the same time, accumulation of monitored information such as temperature, voltage and charge/discharge cycles over time can indicate battery state-of-health (SoH). As a fleet of EVs ages, SoH of the battery can become a deal-breaker for ‘second life’ use, either in resale of the car, or re-purposing of the battery pack in another less onerous application such as grid energy storage. Maximising battery longevity reduces the lifetime cost of the battery, and minimises the frequency and cost of recycling, reducing environmental impact of transport generally.
Cable shorts are a main cause of battery failure
So, battery monitoring is a good thing, and that has been recognised for decades in industry and telecomms, where back-up for critical systems is important. A 48V lead-acid battery array feeding a UPS in a server farm can afford wiring harnesses connecting the bulky centralised monitoring hardware to each cell, but carrying the principles over to EVs with up to 800V strings of cells in a highly contained and harsh environment is not an ideal solution. A BMS is currently typically implemented this way though and because of the high voltages and risk of wire abrasion with vibration, the cable connections to cells in strings have to be oversized for the signals they are carrying, with the associated weight and space penalties, not to mention installation costs.
‘Wireless’ is the go-to technology for many applications and EV battery monitoring is an obvious candidate. Solutions exist which have evolved from older modular architectures, where the voltages of a number of cells in a string are monitored. The resulting analogue values are multiplexed in one of a number of modules built-in to the battery pack, ‘digitised’ and then passed over an RF link to a central processor. The number of cells monitored is typically 12 or 14, limited by the voltage rating of the multiplexer, with each cell adding around 3.7V. The number of cells monitored is set to increase to 16 or higher, to reduce the number of multiplexers needed, but this only amplifies the need to use a high voltage technology in the IC fabrication process. This precludes the easy incorporation of local data aggregation and processing which therefore must be done centrally, creating a bottleneck in the RF connection. More significant disadvantages though are that measurement accuracy of each multiplexed cell voltage degrades up the string and longer physical wire connections to each cell are needed – hardly a ‘wireless’ solution. Noise pick-up is an additional concern. Close attention has to be made to the location of RF antennas, to ensure every module has ‘line-of-sight’ to the central receiver, or complex and unpredictable mesh networks have to be built, making data rates and latency unpredictable.
Dukosi solution leverages ‘edge’ computing
A new solution, proposed by Dukosi, is to adopt the modern idea of ‘edge’ computing – monitoring cells individually with local processing to interpret readings and wirelessly transmit instantaneous and aggregated data over time in the form of histograms created by proprietary embedded software. The ultra low-power hardware is a tiny CMOS chip powered by the monitored battery cell, so the IC technology is compatible with common processor cores and memory. No analog signal multiplexing is necessary, so precision is optimised and the chip is fitted directly at the cell for maximum measurement accuracy of both voltage and local temperature. The problem of connection to an antenna is solved by the use of patented NFC technology. Similar to the inductive loops for ‘contactless’ payment, a thin, low voltage, single wire loop is routed around the battery pack, close to each Dukosi monitor, loosely coupling into a loop on the sensor with a few millimeters of physical separation. This ensures a fast and robust data connection, but is enough to easily provide the electrical isolation needed for the highest battery pack voltage. Each IC has a unique identifier and is polled via the NFC connection by a radio manager which controls the communication process and passes data to the vehicle management electronics. The whole system is designed to be safe, as an ASIL C component of an ASIL D-rated battery pack.
Putting intelligence which is ‘always-on’ at the battery pack, even when the EV is not in use, opens opportunities for long-term logging of usage and performance data which can be interpreted as state-of-health and even maintained as provenance of the battery at any point in its life. With reduced hardware, cabling and installation costs, the lifetime benefit of the Dukosi system is compelling and can read across to all electric vehicle types, as well as to wider energy storage applications.