Energy efficiency management of lead-acid batteries for micro hybrid vehicles

The increasing number of electrical loads in today's cars poses challenges to the battery. More than half of all vehicle faults caused by the electrical system can be attributed to the lead-acid battery, which could be avoided if the battery status was known. In addition, new functions in micro-hybrid vehicles such as the start-stop system or intelligent control of the alternator also require accurate knowledge of the battery status.

The battery management system (BMS) can quickly and reliably monitor the state of charge (SoC), state of health (SoH) and state of function (SoF) according to the starting ability to provide the necessary information. Therefore, the BMS can minimize the number of vehicle failures caused by unexpected battery failure, thereby maximizing battery life and battery efficiency and achieving CO2 reduction. The key element of the BMS is the intelligent battery sensor (IBS), which can measure the terminal voltage, current and temperature of the battery and calculate the battery status.

This article will describe how to use state-of-the-art algorithms to calculate SoC, SoH and SoF for BMS and their efficient implementation on Freescale lead-acid battery IBS.

Technical Introduction

In the past, the charge level of a car battery has been a poorly understood factor that has led to vehicle failure in many cases. Depending on the life of the car, battery-related failure rates can climb as high as 10,000 ppm.

In addition to the existing severe situation, automotive batteries also face other challenges from the ever-increasing electrical energy and power consumption, while reducing CO2 emissions.

Since electronics plays a very important role in automotive innovation, the demand for energy is increasing as the functions of automobiles such as comfort, electrification of safety-related functions, hybrid power, driver assistance and infotainment continue to develop.

On the other hand, more and more regulations are coming out calling for lower CO2 emissions and fuel consumption.

In order to cope with these conflicting requirements, an advanced energy management system is needed to ensure that the battery can provide sufficient power for engine starting in various working situations.

Power management system

A typical power supply network used to power a start-stop system includes a body control module (BCM), a battery management system (BMS), a generator, and a DC/DC converter (see Figure 1).

The BMS provides the BCM with battery status information using a dedicated load management algorithm, and the BCM stabilizes and manages the power supply network by controlling the generator and DC/DC converter. The DC/DC converter distributes electrical energy to the various electrical components inside the car.

Typically, the BMS for lead-acid batteries is mounted directly in a smart connector on the battery clip. The connector consists of a low-resistance shunt resistor (typically in the 100μΩ range) and a small PCB with highly integrated devices (with accurate measurement and processing capabilities) called intelligent battery sensors (IBS, see Figure 2). The IBS measures the battery voltage, current, and temperature with high resolution and accuracy, even under the harshest conditions and throughout the service life, to correctly predict the battery's state of charge (SoC), state of health (SoH), and state of function (SoF). These parameters are transmitted to the BCM periodically or on request through the automotive-certified in-vehicle network.

In addition to the above functions and parameter performance, other key requirements for IBS include low power consumption, the ability to operate in harsh automotive environments (i.e. EMC, ESD), conformance testing of the in-vehicle communication interface (i.e. LIN) for automotive OEM acceptance, meeting automotive grade test limits (6σ limits for the measured parameters), and compliance with the AEC-Q100 standard requirements.

Freescale has announced a fully integrated LIN battery monitoring device based on Freescale S12 MCU technology that meets all of the above parameters. The device includes three independent measurement channels: current measurement through external shunt resistors; battery voltage measurement through a series resistor directly mounted on the positive electrode of the battery; and temperature measurement through an integrated sensor. An integrated LIN 2.1 interface is used to directly connect the sensor to the LIN bus without the need for additional components. Freescale IBS fully meets the AEC-Q100 standard requirements of the automotive industry.

In the following, we will introduce the BMS implementation scheme using Freescale IBS devices and how to implement an efficient BMS by utilizing the hardware features and fixed-point algorithms of IBS.

Battery Monitoring

As mentioned in the previous paragraph, the main purpose of IBS is to monitor the battery status and transmit the status variables to BCM or other ECUs as needed. The measured battery current, battery voltage and temperature sampling values ​​are used as battery monitoring inputs. The battery monitoring outputs are SoC, SoH and SoF.

1. State of Charge (SoC)

The definition of SoC is very intuitive and is usually expressed as a percentage. A fully charged battery has a SoC of 100%, and a fully discharged battery has a SoC of 0%. The SoC value changes as the battery is charged and discharged.

This leads to formula (1), where Cr is the remaining (dischargeable) capacity of the battery and Ca is the total available battery capacity:

This value is calculated using formula (1), where Cr represents the remaining (dischargeable) capacity of the battery and Ca represents the total available capacity of the battery:

However, there is often a problem that the available battery capacity is different from the nominal capacity of the battery (usually marked on the battery casing). For a new battery, it may be higher than the nominal capacity, and for a battery that has been used for a period of time, the available capacity will be lower. Another problem is that the actual available capacity is difficult to determine based on the input value of the IBS.

Therefore, SoC is usually rated by nominal capacity Cn, which has several advantages:

The available charge of the battery for a specific SoC is known, including old batteries; Cn is measured at a certain current (I=Cn/20h) and temperature (27 °C).

There are two common methods for calculating SoC: coulomb counting, also known as current integration or ampere-hour balance, and open circuit voltage (OCV) measurement.

Coulomb counting is the best algorithm to track fast changes in SoC. It is based on integrating the current flowing into and out of the battery and adjusting the calculated battery SoC accordingly. Formula (2) is used for SoC calculation, where Q(t0) represents the initial charge of the battery, α represents the efficiency factor, i(t) represents the current (forward or reverse), and Cn represents the nominal capacity of the battery.

The parameters in the formula are pretty straightforward, except for the alpha factor. This is a factor that describes the efficiency, also known as Peukert's law. It describes the amount of energy a lead-acid battery can hold at different discharge rates. When the discharge rate increases, the amount of energy available from the battery decreases. Another parameter that affects the amount of energy available is the temperature. The higher the temperature, the more energy is available. Both efficiencies are described by alpha, so the alpha value needs to be a 2D array (temperature and discharge rate). Depending on the measured temperature and discharge rate, the corresponding value is used for each integration step. The alpha value depends strongly on the battery design and chemistry and usually varies even for different models from the same manufacturer. They are usually obtained in the laboratory by charge and discharge tests.

While Peukert's law only applies to discharge, there is also an efficiency factor similar to the α value that is used for the charge cycle. In addition to temperature and charge rate, the actual SoC also needs to be taken into account, because the charging efficiency at a high SoC is less than that at a medium SoC.

Since the current and α values ​​are integrated, the errors caused by changes in the battery condition and the current measurement and quantization errors will increase over time. Therefore, the parameter Q(t0) (the starting point of the current integration) is usually obtained by a different method that can provide higher accuracy: the OCV method. OCV is the voltage between the two terminals of the battery when no electrical device is drawing current from the battery.

Lead-acid batteries show a good linear relationship between OCV and SoC. Therefore, by measuring OCV, SoC can be directly calculated. The exact factor between OCV and SoC must be characterized.

The only drawback of this method is that OCV can only be measured after the car is parked, that is, after (almost) all electrical devices are turned off, and it must be measured tens of minutes or even hours after the car is turned off.

Therefore, the OCV method is often used to recalibrate the coulomb counting method, while the coulomb counting method is run continuously. This combination provides a good method for calculating the SoC, and the self-discharge rate can be used to correct the SoC to make the calculation more accurate during longer parking times.

2. State of Health (SoH)

The various aging effects of lead-acid batteries affect the battery differently. As these aging effects are difficult to monitor and quantify individually via IBS, SoH is usually not rated directly based on these aging effects. Instead, SoH is rated based on the reduction in battery capacity over the service life, which is the main result of aging. Another very important parameter related to battery aging is the starting performance, but it is usually expressed in terms of the state of function (SoF) of the starting capability.

Therefore, SoH can be estimated by formula (3), where Caged represents the aged battery capacity and Cn represents the nominal capacity used as a reference according to the calculation of SoC.

Since Cn is known, the key task in calculating SoH is to find Caged. One possible approach is to track the maximum charge (or SoC) achieved throughout the life of the battery. If after several subsequent full charges, the maximum charge level of the battery is lower than the previously calculated aged capacity, it means that the aged capacity has become smaller. Accordingly, Caged and SoH must be adjusted based on the capacity determined by coulomb counting and OCV methods. The full charge state can be monitored when the charging current drops below a certain threshold.

Another way to determine SoH is to track charge and discharge cycles and evaluate them with the cycle stability provided by the battery manufacturer. Typically, the manufacturer will ensure the total amount of charge and discharge cycles for a certain depth at a specified temperature, for example, 500 cycles at 27°C and 25% discharge depth. By evaluating all cycles with these numbers and applying temperature and state of charge correction factors, tracking of the cage mentioned above can be provided. These correction factors must be determined by characterizing the battery characteristics.

However, both methods are usually combined with other specialized algorithms that are tightly coupled to multiple battery parameters over the battery’s lifetime. These battery parameters are determined in the lab through extensive battery characterization and are usually valid only for one specific battery model.

3. State of Functionality (SoF)

Starting a car engine is a very important function, if not the most important, for a lead-acid battery. Therefore, a very important task of the BMS is to predict whether the car can start under actual conditions. The start prediction is represented by the SoF parameter.

In addition to the "traditional" stop-and-start function, the start prediction function is becoming more important with the introduction of start-stop systems in micro hybrid vehicles. The BMS must decide whether it is possible to start again after the engine is switched off and whether it is safe to enter stop mode and communicate this to the BCM.

A very good way to obtain the SoF parameters is to analyze the recent engine starts, the remaining charge (as a function of SoC and SoH) and the actual temperature. During the start, the battery internal resistance Ri (calculated from the voltage drop and the current) needs to be recorded. Because Ri is relatively consistent during the battery's service life and only increases significantly before the end of the battery's service life, the average Ri value needs to be below a certain threshold to ensure safe starting. Another effect of an aging battery is that during the start phase, the Ri value calculated from the voltage and current samples will tend to be non-linear, that is, different current values ​​will be obtained for the same voltage sample value. For a new battery, Ri is linear. See Figures 3 and 4 for common voltage and current variation trends during the start process.

The combination of Ri (calculated by voltage drop and current), the remaining battery capacity and the actual temperature can well characterize the starting capability. In addition, these thresholds must also be determined by analyzing the characteristics of the battery.

In order to determine the linearity or nonlinearity of Ri with the necessary accuracy, all voltage and current values ​​sampled during the start-up phase need to be filtered using a linear filter, preferably a bandpass filter.

Efficient implementation of BMS hardware and software

Energy efficiency is one of the most important features of new cars, which is achieved by the BMS. In addition to managing some power saving functions, the BMS needs to be highly energy efficient because it is an always-on system that needs to be powered by the lead-acid battery when the generator is not working. To meet this requirement, the power consumption of the IBS must be as low as possible.

To achieve this, Freescale’s IBS implementation uses two low-power models, where the CPU and other unneeded hardware (HW) blocks are turned off. To reduce power consumption in normal operating mode and reduce software (SW) development work on the client side, additional hardware blocks are added to reduce software complexity. This allows the use of smaller, lower-power, more cost-effective 16-bit microcontrollers. Another way to reduce software complexity is to ensure product parameters throughout the lifetime and store factory trim values ​​in non-volatile memory (NVM). As part of the end-of-line testing of the product, these trim values ​​are characterized for each chip individually and stored accordingly. Therefore, there is no need to use complex calibration algorithms in software.

In addition to these three techniques implemented in hardware, this article also presents an efficient software implementation of the battery monitoring algorithm.

1. Low power mode

Implementing low-power modes is a very good way to reduce power consumption. This is achieved by shutting down parts of the SoC (especially the CPU) when they are not needed and switching to normal mode (that is, activating all hardware blocks) only when needed. As mentioned earlier, there are two low-power models that differ only in the program entry point used after the CPU is woken up.

However, the battery status also needs to be monitored in low power (i.e., no software interaction) mode. First, the current needs to be tracked and the SoC needs to be calculated using the coulomb counting method. Accordingly, current measurement in low power mode and automatic summation of current sampling values ​​(i.e., coulomb counting) can be supported.

The IBS must be able to react to changes in the state of the battery and the car, that is, the battery sensor must be awakened when various events occur. Accordingly, it is also necessary to measure the current and temperature in low-power mode. Current changes usually indicate changes in the state of the car (on and off of power devices), while temperature changes sometimes require recalibration of measurement channel parameters. Thresholds for current and temperature sampling values ​​can be configured and wake up if they exceed the threshold. An automatic coulomb counter threshold wake-up mechanism can also be used.

In addition to those wake-up events for measured parameters, other wake-up mechanisms can be implemented, allowing the BCM or other electronics in the car to wake up the IBS (via LIN messages or direct wire connections), as well as timed wake-up mechanisms.

The implementation of the low power mode and wake-up mechanism described above allows the IBS to run in low power mode most of the time (typically around 70%), including when the engine is running. During normal operating mode, the SoC, SoH and SoF parameters are recalculated.

2. Move software tasks to hardware modules

Using dedicated hardware modules to take over software tasks is an effective way to reduce software complexity and save power. Such hardware modules can be very effectively used to pre-process the voltage, current and temperature measurement samples before using them in the battery monitoring algorithm. This is very necessary because the power line of the car is often disturbed and for IBS, the measurement accuracy of the sampled values ​​is very high.

A high-precision 16-bit Σ-Δ ADC with decimation and anti-interference filters is very suitable for this application because of its high measurement accuracy compared to other ADC technologies. In combination with error compensation, this already provides very good accuracy. However, it is often necessary to refilter the sampled values ​​after the signal processing chain. The reason for this is to remove noise from other electronic components in the car, so the filter needs to have a frequency-free switching characteristic. Another reason is that the specific battery parameters that are observed as part of battery monitoring are closely linked to the excitation frequency (determined by the battery's chemical composition). This is the case, for example, with Ri.

Programmable linear filters can meet these requirements: the filter coefficients can be transferred to the hardware filter module through registers. After these registers are programmed, the filtering task no longer needs to be completed in software.

Current measurement faces the challenge of measuring low currents with high accuracy while also supporting a wide measurement range. The required accuracy is better than 10mA, which means a 1μV voltage drop across a 100μΩ shunt. During the start-up process of a car, currents of 1000A or even higher can occur. To support both of these needs while avoiding manual measurement reconfiguration in software, an automatic gain amplifier is introduced. The selectable gain factor is used to adjust the input signal to achieve the best match with the reference voltage of the ADC. The gain factor adjustment can be done automatically, without reconfiguring the software during the entire operation. For testing purposes or when there are special application requirements, a fixed gain factor can also be selected.

3. Simplify calibration work

A very important task to ensure that the device maintains high accuracy throughout its entire service life is adjustment and calibration. For this purpose, correction factors obtained from previous tests are applied to the key device parameters. As part of the testing of the devices in the production line, these factors are tested at various temperatures and stored in the NVM of the IBS. At device startup, the various adjustment parameters need to be written to the device registers by software. The parameters that need to be adjusted are available in the current and voltage measurement chain. In addition, the oscillator, voltage reference and LIN timing also need to be calibrated. Recalibration is required during operation, for example, at regular intervals or when there are drastic changes in temperature. If available, different correction factors need to be written to the respective registers again.

The calibration method mentioned above can avoid the customer from having to perform expensive end-of-line testing of these parameters. In addition, by simply applying the parameters, the complexity of software calibration can also be reduced.

4. Software Implementation

The battery management algorithms mentioned in the previous article require processor-intensive computation and control algorithms. The first implementation of these algorithms is usually done on a PC using model-based simulation tools. These tools usually use floating-point data formats. In the subsequent development process, these algorithms are ported to the IBS. However, due to cost and power reasons, floating-point hardware is not available on the microcontroller used in the IBS. Therefore, in order to obtain a suitable runtime, the data types used in the algorithm must be mapped to fixed-point integer format. There are a variety of data types and corresponding value ranges available. For example, the following table lists the data types provided on Freescale's IBS.

Data Type Range (Unsigned) Range (Signed)

char (8 bits) 0 – 255 -128 – 127

int (16 bits) 0 – 65535 -32768 – 32767

long int (32 bits) 0 – 4294967295 -2147483648 – 2147483647

To represent values ​​less than 1, the LSB is mapped to a specific value. This value is determined by the required resolution. By selecting one of the available data types, you can derive the available value range and virtual fixed decimal point (fixed-point format) of the variable. For example, with a resolution of 1mV, the range that can be represented using an unsigned integer is 0~65.535V.

Since Freescale IBS has a 16-bit S12 CPU, the integer data type can provide 16 bits of precision. This means that 8-bit and 16-bit variables can be processed more efficiently than 32-bit variables. Therefore, 8-bit and 16-bit variables are generally preferred.

The implementation examples of the algorithms used to calculate SoC, SoH, and SoF above show that in many cases, 16-bit variables can provide sufficient numerical accuracy and range. This is because the voltage and temperature input values ​​are 16-bit accurate (by using a 16-bit ADC). Some other values ​​where 16-bit accuracy is sufficient include SoC, SoH, Ri, and the correction factor α. Even if current sample values ​​with 24-bit accuracy are used, they can be mapped to 16-bit data in most cases. For example, at 3mA accuracy, by using a signed 16-bit integer format, current values ​​up to ±98.3A can be represented without further modification of the digital format. This is sufficient to meet the requirements of the car when it is moving and stopped. During the starting process, the current sample values ​​will be out of bounds and a 32-bit data format must be used. Parameters that require a 32-bit format are values ​​related to battery charging (for example, a coulomb counter).

Conclusion

This article describes how to use Freescale IBS to effectively implement the BMS in a micro hybrid vehicle. The state-of-the-art battery state calculation algorithms (SoC, SoH and SoF) are discussed. It is seen which special hardware features can be used to improve the efficiency of the IBS in terms of power consumption. The use of low-power modes with automatic battery state observation (no software interaction) and a sophisticated wake-up mechanism is also presented. As a result, the IBS can be in low-power mode in most cases. In addition, software complexity is reduced by using appropriate hardware signal processing, programmable filtering and simplified calibration. The article also introduces the principles of fixed-point arithmetic and the analysis results show that for the variables in the BMS algorithm, the 16-bit fixed-point data format is generally sufficient and only in a few cases the 32-bit format is required.