First I will talk about the entire BMS Road, and the second is software architecture and development. We had the first-generation BMS in August 2017, and the second-generation in May this year. The design of our BMS is distributed, but there will be a little difference. The acquisition board will have two configurations. The functional safety level is ASIL C. In fact, everyone knows that President Liu just said that domestic requirements for functional safety are ASIL C, and foreign BMS requirements. ASIL D is required, for example: many foreign houses It is a wooden house. If a fire occurs during the charging process, it will affect the safety of human life. When we do HARA analysis, the safety level must be from C to D in this case. In the second half of this year, we We will upgrade our second-generation BMS with a small change, which includes upgrading the functional safety level from ASIL C to ASIL D.
The planning of the third generation BMS considers the introduction of the concept of a domain controller. When we do a dynamic domain control analysis, the introduction of a domain controller, the VCU will not exist and will be replaced by a domain controller. In addition, the domain controller will migrate some functions of the Cover BMS, such as the SOC algorithm, from the current control version of the BMU to the domain controller.
To add a little bit, whether it is high-voltage BMS or other, it must be inseparable from semiconductor technology, whether high-voltage BMS or 48V BMS, our design is based on the NXP solution, because we and NXP are strategic partners. CMU we use a centralized, two configuration methods, the first configuration has four MC33771, the second configuration we do two MC33771 plus one MC33772. The CMU acquisition board can support CAN2.0 and CAN FD communication methods. The reason for designing CAN FD is that after 2021, some BMS functions will be migrated to the domain controller, and the voltage and temperature data of the CMU will be uploaded to the domain. The controller takes into account the requirements of real-time performance and communication bandwidth, so ordinary CAN2.0 cannot meet the needs, and CAN FD needs to be upgraded.
At the same time, our BMS voltage can support 800V platforms. One of the main reasons for the possibility that the vehicle voltage will rise in the later period is the requirement for fast charging. Because the biggest bottleneck in the development of new energy is the problem of charging, fast charging can solve the problem of increasing the current or increasing the voltage. For some questions, the currently more feasible method is to upgrade the voltage from the current 400V platform to the 800V platform. We support ordinary AC charging and DC charging. At the same time, we also support European standard charging. Everyone knows that European standard charging is different from ordinary domestic standards. European standard is PLC communication.
This is the functional indicator of our main design. We use a distributed solution for the high-voltage BMS. This is the architecture that controls the motherboard and the BMU framework of the motherboard. We use 5774. The motherboard has low-voltage and high-voltage parts. Standard, the insulation voltage rating must be above 3500V, which is our approximate architecture.
This is our acquisition board, CMU framework, which supports two methods. The first method is to support common solutions. At the same time, if cost is to be considered, it can be daisy chained. According to our experience, the current 33771B version, It is best to control it below 10 nodes. If it exceeds 10 nodes, everyone knows that the 33771B attenuation is quite strong. If it exceeds 10 nodes, the overall vehicle performance cannot be guaranteed.
The second acquisition version supports sampling of 34 strings of cells. Why do we take this approach? In fact, each power demand of the vehicle OEM is different, and the solutions are mainly the BMU and CMU. At present, we and many OEMs can basically meet the power configuration requirements of more than 85% of OEMs in the future. When designing, we must consider frequency performance, flexible configuration, and customized products with low cost performance. Now the threshold for BMS will be very high, such as the required functional safety. If it is customized for him, each cost will be very high. Initially, our company invested about 30 million yuan for BMS. This investment is still very high. huge. All our development, whether hardware or software, is based on modular development. The modules of the application layer, including the main control, high-voltage loop, SOC, safety monitoring, etc. are basically designed by the module.
48V BMS originated in Europe. New energy still has some safety and charging problems. Because 48V just makes up for this space, it only takes about 5,000 yuan in traditional cars to ok, so we developed 48V BMS. Now it is developed to the 1.0 generation, and the pre-researched 2.0 generation, including the later 3.0 generation, Mercedes-Benz is currently in the 3.0V 48V system. At present, the fuel-saving rate has reached 15%, and it is still relatively good.
This is a functional indicator. The SOC is nominally 5%. In fact, we have measured that the data in our laboratory will gradually decrease with aging. It is no problem to achieve 5% throughout the life cycle.
This is our entire 48V design architecture. We use the S32K1XX + MC33664 + MC33771 solution, which is not really defined as a high-voltage product. Considering the reliability of the product, avoid high-voltage products coming in. The design has high and low voltage. Isolation. This is what we do with high-voltage architecture. This is a list of functions of the 48V application layer.
The last piece, about the algorithm of our SOC, we apply it to the 48V BMS. The traditional BMS uses the AH integral as the basis. On this basis, we have added an extended Kalman filter algorithm. SOS performs calibration to achieve the purpose of preparing and trimming the SOC in real time, ensuring the accuracy of the SOC.
Going a bit deeper, how do we do it? In the case of low current, the extended Kalman filter algorithm can be used to eliminate the integral error. When the current is greater than 5A, we will switch the algorithm to the traditional integration, which can ensure that the algorithm has both the AH integration algorithm and the extended Kalman algorithm. Filter algorithm advantages. The traditional second-order model used by the extended Kalman filter. The EKF algorithm takes current and voltage as input, R1, R2, and other data. At that time, we measured the battery for four months. At different temperatures and at different rates To measure these data, it took about 500,000 or more difficult algorithms, the workload of development, including how you deal with these calibrated models.
This is our algorithm at that time. The accuracy of the simulation data is controlled within 0.5%. This is our entire simulation data. The actual measured data will be displayed on the next page. This is the data tested after the entire product is assembled. The accuracy of the SOC of the battery pack is verified. The initial error of the test data is 5%. After 10 pulse cycles, the SOC error is within 1%.
That ’s all, thank you!