Introduction to the estimation method of the remaining capacity SOC of lithium batteries

qwqwqw2088

Introduction to the estimation method of the remaining capacity SOC of lithium batteries [Copy link]

The state of charge, also known as the remaining power, SOC, is a parameter that reflects the percentage of the current power in the battery pack to the total available capacity. Electric vehicle drivers can infer the current battery life based on the total mileage of the fully charged state, and some models directly display the range.

Electric vehicles with inaccurate battery estimation often cause some troubles to their owners.

The battery level jumped. When I was about to park, I checked the battery level and it was only 50%, which was probably enough for the return trip. After a while, I started the car and found that the battery level was only 40%, which seemed like it would never go back...

Suddenly the battery was cut off. When the battery was still 30%, I accelerated and suddenly the alarm sounded that the battery was too low, so I stopped the car...

The accuracy of SOC has always been a point of criticism for electric car users. In online forums, you can often see complaints from electric car owners who were put off track. After years of development, electric cars still don’t have a particularly good way of accurately estimating SOC.

1. What is SOC estimation used for?

Informing the driver of the remaining mileage is the value of SOC to the driver. At the same time, the management and control process of electric vehicles and battery packs also requires accurate SOC data.

Prevent overcharge and over discharge

As an important threshold for charging and discharging, SOC regulates and protects the battery pack. During the charging process, if the battery pack power is too low, such as less than 10%, the charging current cannot be fully charged, but current-limited charging is performed until the normal power range is reached, and the restriction is released. During the discharge stage, when the power is already relatively low, but still above the discharge cut-off power, such as as low as 20% SOC, the power output is generally limited to prevent large currents from causing the system to reach the parking voltage, and it is expected to run in the most energy-efficient way and travel a longer distance.

As low battery current limit threshold

Why do we need to set a current limiting strategy when the battery pack has low power? When the power is low, the battery terminal voltage is low. If a large current discharge occurs suddenly, the battery polarization internal resistance will increase rapidly, causing the internal resistance voltage to rise. The battery terminal voltage after the battery potential minus the internal resistance voltage will decrease accordingly. If the current is large enough, the terminal voltage may be pulled down below the stop power supply voltage. If the low voltage lasts longer than the delay time, the battery management system determines that the battery voltage is too low and cannot continue to work, and needs to be powered off. So the battery pack main circuit contactor is actively disconnected. A sudden power outage occurs. Of course, the control authority of the main circuit contactor of some models is on the vehicle controller.

As the threshold of vehicle control strategy

There are many electrical appliances on electric vehicles, such as air conditioner, audio system, power steering, power brake, etc. When the SOC drops to a certain level, the permissions of the electrical appliances need to be sorted. For example, the power brake must be powered regardless of the power level, and the air conditioner and audio system must stop powering when the power level is less than a certain value.

Second, the algorithm

The algorithms that are already well known include the open circuit voltage method, the ampere-hour integration method, and the internal resistance method. The methods that have been studied more recently include the Kalman filter method and the neural network method. New methods are constantly emerging, but most of them are in the research and discussion stage, and the algorithms in practical applications are still mainly based on the old methods.

Open circuit voltage method

The open circuit voltage of a lithium battery has a clear monotonic correspondence with the battery charge. As long as the accurate open circuit voltage is obtained, the battery charge can be calculated. The corresponding curves of the open circuit voltage and SOC of several battery cells are shown in the figure below.

First, measure the open circuit voltage values at different temperatures and SOCs offline and form a table. After the battery system is installed on the vehicle, whenever the power supply stops, the table data can be called to determine the battery charge state based on the measured open circuit voltage.

The open circuit voltage method can accurately determine the battery charge, but it has many restrictions. The battery must be left alone for a period of time with the circuit disconnected. This requirement makes online measurement impossible.

Some people have conducted research on the above problem and found that the battery terminal voltage and open circuit voltage maintain a stable relationship after the circuit stops supplying power for a certain period of time. This method can avoid the long-term static process, which expands the application scope of the open circuit voltage method.

Ampere-hour integration method

The main circuit current of the battery pack is measured in real time and integrated over time, with charging being negative and discharging being positive. During the discharge process, the initial power is subtracted from the integral result to obtain the current power; during the charging process, the initial power is added to the integral result to obtain the current power.

One problem with the ampere-hour integration method is that the initial power cannot be directly determined. In addition, since the system current fluctuates greatly, and current sampling is performed at regular intervals, the sampled value is not necessarily close to the average value over a period of time. Over a long period of time, this results in a significant error, and the error cannot be eliminated by the ampere-hour integration method itself. Therefore, the practical application of the ampere-hour integration must be combined with other methods to solve the problems of initial value and cumulative error.

Internal resistance method

There is a corresponding relationship between the SOC and internal resistance of the battery cell. It is theoretically feasible to measure the internal resistance in real time and then obtain the SOC of the battery.

However, it can be found from the above figure that the trend of internal resistance following SOC change is very gentle. A small change in internal resistance or measurement error can cause a large error in the SOC value. In addition, there are various possibilities for errors in the internal resistance measurement process. The measured contact resistance is too large; the battery current is large, resulting in a large interference of polarization internal resistance; the battery temperature rise is inconsistent, resulting in the temperature of the temperature monitoring point being inconsistent with the battery body temperature, causing temperature compensation deviations, etc. There are relatively few actual application cases of the internal resistance method to estimate SOC.

Extended Kalman Filter Algorithm

The Kalman filter method introduces the concept of time domain, which is to regard a process as a set of countless states that are continuously played on the time axis. The state equation is used to describe the dynamic process, the measurement equation is used to describe the observation information, and the estimated value at the previous moment and the observed value at the current moment are iterated to update the estimate of the state variable. Some people describe the Kalman filter algorithm as: using the ampere-hour integration algorithm to calculate the SOC, and using the open circuit voltage method to verify the result of the ampere-hour integration. The ampere-hour integration method corresponds to the state, and the open circuit voltage method corresponds to the measurement. If you want to introduce the temperature variable, you must choose a circuit model that expresses the relationship between internal resistance and voltage, because among the battery parameters, the internal resistance is a relatively easy to measure quantity that is closely related to temperature.

Various algorithms derived from Kalman filtering have been applied in multiple computing fields and have become the most mainstream new method for estimating the remaining power of power batteries.

Neural Network Algorithms

In the context of the AI pandemic, the neural network method is also applied to SOC estimation. The input parameters are loop current and battery terminal voltage, and the output result is SOC. The rest is a general neural network model. As long as a sufficient amount of data with the correct correspondence between voltage, current and SOC can be provided, the output accuracy of the model can be continuously improved.

Theoretically, the neural network algorithm is a very ideal estimation method, but the time for its application is not yet mature. The accumulation of data and the improvement of computing power will promote the practical application of the neural network algorithm in the field of power battery SOC calculation.

3. Factors affecting SOC accuracy

From the definition, the SOC value is directly determined by two parameters: the maximum available capacity of the battery and the current charge. From the parameters used by various algorithms, the battery terminal voltage and loop current are the most important measurement parameters.

Maximum available capacity

As the number of battery cycles increases, the battery ages. The maximum available capacity of the battery decreases as the aging degree deepens. The maximum available capacity used to calculate the SOC in the system must be adjusted in real time to obtain an approximately accurate SOC.

Current measurement

Common factors that affect current measurement accuracy include: the accuracy of the current sensor, whether the range is appropriate, the presence of electromagnetic interference, and whether the sampling algorithm is reasonable.

If the sampling algorithm samples uniformly along the time axis, there will be a problem that the current value at the sampling point deviates seriously from the average value of this stage. A more reasonable way is to increase the sampling density during the period of drastic current changes and reduce the sampling density during the period of smooth current changes, so that the sampled value is closer to the true value.

Voltage measurement

Voltage measurement can be divided into two situations: real-time working end voltage measurement and open circuit voltage measurement.

The real-time measurement of the working voltage is mainly affected by the real-time current. On the one hand, the current affects the polarization internal resistance of the battery, and on the other hand, it determines the proportion of the internal resistance to the voltage. Another accidental factor is the resistance between the voltage measurement point and the battery ear. It may be a poor contact or a long wire. In short, if this resistance is too large and reaches the order of magnitude of the internal resistance of the battery cell, the voltage measurement result will be affected.

Open circuit voltage measurement is a relatively simple situation with relatively few influencing factors. What needs to be pointed out is that the battery needs to be left standing for a long enough time.

toothache

This technology is really useful for electric vehicles