State of charge (SOC) meaning

SOC is StateofCharge, which refers to the state of charge of the battery. From different perspectives such as power and energy, SOC has many different ways of meaning. The United States Advanced Battery Consortium (USABC) meaning of SOC is widely used, that is, the battery in a certain discharge rate, the remaining power and the ratio of rated capacity under the same conditions.

Li-ion battery charge state prediction method

The charge state of Li-ion battery is one of the important parameters of the battery management system, and is also the basis for the whole charge/discharge control strategy and battery equalization work. However, due to the complexity of the structure of the lithium-ion battery itself, its charge state cannot be obtained by direct measurement, and the prediction of the charge state can only be done based on certain external characteristics of the battery, such as the internal resistance, open circuit voltage, temperature, current and other related parameters, using the relevant characteristic curves or calculation formulas.

The charge state estimation of lithium-ion battery is nonlinear, and the commonly used methods are discharge experiment method, open-circuit voltage method, ampere-time integration method, Kalman filter method, neural network method, etc.

1、Discharge experiment method

The principle of the discharge experiment method is that the battery is discharged at a constant current without interruption, and the amount of discharge is calculated when the discharge reaches the cut-off voltage. Discharge power value is the product of the constant current value and discharge time used in the discharge. The discharge test method is often used to estimate the state of charge of a battery under laboratory conditions, and many battery manufacturers are now using the discharge method for battery testing.

Its significant advantage is the simplicity of the method and the relatively high accuracy of the estimation. Its disadvantages are also prominent: it cannot be measured with a load, it takes up a lot of measurement time, and during the discharge measurement, the work previously performed on the battery must be interrupted so that the battery is placed in an offline state, and therefore cannot be measured online. Driving electric car battery has been in working condition, its discharge current is not constant, this method is not applicable. However, the discharge experiment method can be used in the battery overhaul and parameter model determination.

2、Open circuit voltage method

The parameters of the battery are relatively stable after a long time of full resting, and the functional relationship between the open circuit voltage and the battery charge state is relatively stable at this time. If you want to get the battery charge state value, you only need to measure the open circuit voltage at both ends of the battery, and compare the OCV-SOC curve to get the corresponding information.

The advantage of the open-circuit voltage method is that it is simple to operate, and the charge state value can be obtained by simply measuring the open-circuit voltage value against the characteristic curve. However, there are many disadvantages: first of all, in order to obtain accurate values, the battery voltage must be in a relatively stable state, but the battery often has to be left for a long time before it can be in this state, so it cannot meet the real-time monitoring requirements, and it is often applied to electric vehicles when they are parked for a long time.

When the battery charge/discharge ratio is different, the current fluctuation will cause the battery open circuit voltage to change, thus leading to inconsistent open circuit voltage of the battery pack, which makes the predicted remaining power and the actual remaining battery power deviate significantly.

3、Amperometric integration method

The ampere-time integration method does not consider the internal usage mechanism of the battery, and calculates the total power flowing into and out of the battery by integrating the time and current and sometimes adding certain compensation coefficients based on certain external characteristics of the system, such as current, time, temperature compensation, etc., so as to estimate the charge state of the battery. At present, the ampere-time integration method is widely used in battery management systems.

The advantage of the ampere-time integration method is that it is relatively less restricted by the battery’s own condition, the calculation method is simple and reliable, and it can make real-time estimation of the battery’s charge state. Its disadvantage is that because the ampere-time metering method is an open-loop detection in the control, if the current collection accuracy is not high, the given initial charge state has a certain error, along with the extension of the system operation time, the previous error will gradually accumulate, thus affecting the prediction result of the charge state. And because the ATS method only analyzes the charge state from the external characteristics, there are certain errors in multiple links. It can be seen from the calculation formula of the ampere-time integration method that the initial charge of the battery has a large impact on the accuracy of the calculation results.

In order to enable the accuracy of current measurement to be improved, high-performance current sensors are usually used to measure the current, but this increases the cost. For this reason, many scholars apply the open-circuit voltage method along with the ampere-time integration method, combining the two. The open-circuit voltage method is used to estimate the initial state of charge of the battery, and the ampere-time integration method is used for real-time estimation, and relevant correction factors are added to the equation to improve the accuracy of the calculation.

4、Kalman filter method

The Kalman filter algorithm is a kind of minimum variance estimation using time-domain state space theory, which belongs to the category of statistical estimation, macroscopically it is to minimize and eliminate the influence of noise on the observed signal, and its core is the optimal estimation, that is, the input quantity of the system is effectively corrected on the basis of the prediction of the state variables.

The basic principle of the algorithm is that the state-space model of noise and signal is used as the model of the algorithm, and the estimation of the state variables is updated at the time of measurement by applying the observed value of the current moment and the estimated value of the previous moment. The essence of the Kalman filter algorithm for predicting the state of charge of a lithium-ion battery is the ampere-time integration method, while the measured voltage value is used to correct the value obtained from the initial prediction.

The advantages of Kalman filter method are that it is suitable for computer to process the data in real time, has a wide range of applications, can be used for non-linear systems, and has a good effect on predicting the charge state of electric vehicles during the driving process. The disadvantage of Kalman filter method is that it depends on the accuracy of the battery model, and in order to improve the accuracy and precision of the prediction results of this algorithm, a reliable battery model should be established. In addition, the algorithm of Kalman filter method is relatively complex, so its computation is also relatively large, and has high requirements on the performance of the operator.

5、Neural network method

The purpose of neural network is to imitate the intelligent behavior of human beings, to obtain the ability of data expression through parallel structure and its own strong learning ability, to be able to give the corresponding output response in the presence of external excitation, and to make a good nonlinear mapping ability.

The principle of neural network method applied to lithium-ion battery charge state detection is that a large amount of corresponding external data such as voltage and current as well as the battery charge state data are used as training samples, which are repeatedly trained and modified by the forward propagation of input information and backward propagation of error transmission during the learning process of the neural network itself, and the predicted charge state of the battery is obtained by inputting new data when the predicted charge state reaches within the error range of the design requirements.

The advantages of the neural network method are that it can estimate the charge state of various batteries with a wide range of application; do not establish a specific mathematical model, do not need to consider the complex chemical change process inside the battery, only need to select suitable samples, as well as establish a better neural network model, and the more sample data, the higher the accuracy of its estimation; can determine the charge state of the battery at any time. The disadvantages of the neural network method are the high hardware requirements, the accuracy of the data samples used in training, the sample capacity and sample distribution, and the training method all have a great impact on the battery charge state prediction.

Summary

In this paper, several important methods for predicting the charge state of Li-ion batteries have been briefly introduced and their respective advantages and disadvantages have been analyzed in detail. At present, the ATS method is still the most applied charge state prediction method, but due to the limitations of the ATS method itself, it is often combined with other methods such as the open circuit voltage method to complete the detection of the initial charge state of Li-ion batteries.

From the development trend, the factors considered in the prediction of the charge state of lithium-ion batteries are becoming more and more comprehensive, and the prediction methods used are often a combination of several aforementioned methods, making the prediction results more accurate. Moreover, the equivalent circuit model of lithium-ion battery has been developed continuously and is closer to the reality, which makes the accuracy of charge state prediction further improved.