CN108693478A - A kind of method for detecting leakage of lithium-ion-power cell - Google Patents

Abstract

The invention provides a liquid leakage detection method of a lithium-ion power battery, which can accurately determine whether a short-circuit fault has caused liquid leakage of the battery without opening the battery case for observation. Joint leakage identification was achieved by building a battery model of external short-circuit faults and running a classifier based on a random forest method. The method is suitable for a battery fault diagnosis system, can provide a basis for prediction and diagnosis of the fault degree after a short circuit of the battery, and has many beneficial effects such as simple operation and easy realization.

Description

A liquid leakage detection method for lithium ion power battery

technical field

The invention relates to the technical field of battery safety, in particular to a liquid leakage detection method of a lithium-ion power battery.

Background technique

External short circuit is the most common type of sudden failure in vehicle lithium-ion power battery faults. Within tens of seconds after the external short circuit occurs, the high temperature and large current generated by the power battery are very likely to cause serious liquid leakage. , The leaked electrolyte will further cause more serious consequences including fire, so it is very important to distinguish the leakage of the power battery. However, since the lithium-ion power battery for vehicles is usually placed in the battery box, it is neither convenient nor safe to check whether the battery is leaking through opening the box, and the lithium-ion battery for vehicles is often composed of hundreds of batteries connected in series and in parallel to form a battery pack. The batteries are densely arranged, and even if the leakage is observed, it is difficult to locate the location of the specific leakage battery. It can be seen that the existing liquid leakage detection methods of lithium-ion power batteries still have certain limitations, and there is an urgent need in this field for a method that can realize liquid leakage detection without opening the battery box.

Contents of the invention

Aiming at the above-mentioned technical problems in this field, the present invention provides a liquid leakage detection method for a lithium-ion power battery, which specifically includes the following steps:

Step 1. Conduct multiple sets of external short-circuit tests on different lithium-ion power batteries, and record current, voltage and temperature data;

Step 2, establishing a battery model for the lithium-ion power battery, and performing parameter identification on the battery model based on the data recorded in the step 1, and determining the short-circuit fault state threshold and the liquid leakage state threshold;

Step 3. Use the battery model established in step 2 to diagnose the fault online, and judge its coincidence with the short-circuit fault state and the liquid leakage state according to the power battery data;

Step 4. Establish a classifier based on the random forest algorithm, and train the classifier according to the data recorded in the step 1, so as to realize the monitoring of the short-circuit fault state and the liquid leakage state.

Step 5, combining the battery model and the classifier based on the random forest algorithm to detect whether liquid leakage occurs.

Further, said step two specifically includes:

The established battery model adopts the fractional impedance model, and uses the current, voltage and temperature data when no liquid leakage occurs to train the model, so as to establish a battery model capable of simulating the change of the external short-circuit electrical characteristics in the non-leakage state; Parameter identification is performed on the fractional impedance model based on a genetic algorithm.

Further, the online diagnosis of the fault described in the step 3 specifically includes:

The actual power battery voltage is compared with the voltage output by the fractional-order impedance model in the short-circuit fault state, and the coincidence degree χ in the time period from the beginning of the short circuit to the current moment is calculated:

V _m is the terminal voltage predicted by the model, V is the actual terminal voltage of the battery, and n is the data length.

Comparing the coincidence degree χ with the short-circuit fault state threshold δ1 and the liquid leakage state threshold _{δ2 determined in step 2 .}

Further, the classifier and its training process based on the random forest algorithm established in the step 4 specifically include:

4.1. Collect the statistical data of the maximum temperature rise and discharge capacity of the leaking battery and the non-leaking battery in the M lithium-ion power battery samples after the short circuit occurs, and input it into the random forest algorithm as the total training data set;

4.2. Use the Bootstrap method for resampling, and randomly generate N training data sets S ₁ , S ₂ ,...,S _N . Among them, S _i (1≤i≤N) is the data set formed by the i-th extraction from the M battery sample data with replacement;

4.3. Based on the generated N training data sets, N decision trees are randomly generated. After each decision tree is trained, it has the classification ability to judge whether leakage occurs through the maximum temperature rise and discharge capacity data.

Further, the joint detection of the fractional impedance model and the classifier based on the random forest algorithm described in step five specifically includes:

5.1. Judgment is made according to the comparison results of the coincidence degree χ, the short-circuit fault state threshold δ ₁ and the liquid leakage state threshold δ ₂ , if χ>δ ₂ , it is considered that no liquid leakage occurs, and the value K1=0 is assigned; if δ ₁ <χ<δ ₂ , it is considered that leakage occurs, and the value K=1;

5.2. Run the classifier based on the random forest algorithm, input the collected maximum temperature rise and discharge capacity of the lithium-ion power battery as the test data set X into each trained decision tree, and output the leakage monitoring status K2, when Assign K2=1 when leakage occurs;

5.3. When K1 and K2 are 1 at the same time, assign K=1, indicating that the battery has leaked, otherwise K=0, indicating that no leak has occurred.

According to the above-mentioned method provided by the present invention, it is possible to accurately determine whether the short-circuit fault has caused battery liquid leakage without opening the battery box for observation. Joint leakage identification was achieved by building a battery model of external short-circuit faults and running a classifier based on a random forest method. The method is suitable for a battery fault diagnosis system, can provide a basis for prediction and diagnosis of the fault degree after a short circuit of the battery, and has many beneficial effects such as simple operation and easy realization.

Description of drawings

Fig. 1 is a schematic flow chart of the method provided according to the present invention

Figure 2 is a schematic diagram of a fractional impedance model of a lithium-ion power battery

Fig. 3 is the relationship between the predicted value and the measured value of the power battery terminal voltage according to a specific example of the present invention (SoC=20%)

Fig. 4 is the relationship between the predicted value and the measured value of the power battery terminal voltage according to a specific example of the present invention (SoC=60%)

Fig. 5 is the relationship between the predicted value and the measured value of the power battery terminal voltage according to a specific example of the present invention (SoC=100%)

Figure 6 is a schematic diagram of a classifier based on the random forest algorithm

Figure 7 is the result of leakage detection and discrimination

Detailed ways

The technical solution of the liquid leakage detection method of the lithium-ion power battery provided by the present invention will be further explained in detail below in conjunction with the accompanying drawings.

As shown in Figure 1, the method provided by the present invention specifically includes the following steps:

Step 1. Conduct multiple sets of external short-circuit tests on different lithium-ion power batteries, and record current, voltage and temperature data;

Step 2. Establishing a fractional impedance model for the lithium-ion power battery, and performing parameter identification on the fractional impedance model based on the data recorded in step 1, and determining the short-circuit fault state threshold and the liquid leakage state threshold ;

Step 3. Use the fractional impedance model established in step 2 to diagnose the fault online, and judge its consistency with the short-circuit fault state and the liquid leakage state according to the data of the power battery;

Step 4. Establish a classifier based on the random forest algorithm, and train the classifier according to the data recorded in the step 1, so as to realize the monitoring of the short-circuit fault state and the liquid leakage state.

Step 5, combining the fractional impedance model and the classifier based on the random forest algorithm to detect whether liquid leakage occurs.

In a preferred embodiment of the present application, the second step specifically includes:

The established fractional-order impedance model adopts the fractional-order impedance model, as shown in Figure 2, in which V _ocv is the open circuit voltage, R _o is the ohmic internal resistance, R _ct is the charge transfer internal resistance, CPE is the constant phase angle element, and W is the Weber impedance. The model is trained using current, voltage and temperature data when no liquid leakage occurs, at room temperature (20°C) for low, medium and high initial SoC (20%, 60% and 100%), under the same conditions Next, test 1 battery cell separately, and record the battery voltage and current changes during the experiment of 3 battery cells. None of the 3 cells leaked. Therefore, a battery model capable of simulating the change of the external short-circuit electrical characteristics in a non-leakage state is established; the parameters of the fractional impedance model are identified based on a genetic algorithm. The predicted results of the power battery terminal voltage and the measured terminal voltage results are shown in Figure 3-5. The root mean square value of the error is listed in Table 1.

Table 1

In a preferred embodiment of the present application, the online diagnosis of the fault described in step 3 specifically includes:

The actual power battery voltage is compared with the voltage output by the fractional-order impedance model in the short-circuit fault state, and the coincidence degree χ in the time period from the beginning of the short circuit to the current moment is calculated:

V _m is the terminal voltage predicted by the model, V is the actual terminal voltage of the battery, and n is the data length.

Comparing the coincidence degree χ with the short-circuit fault state threshold δ1 and the liquid leakage state threshold _{δ2 determined in step 2 .} Another three batteries with initial SoC of 20%, 60% and 100% were selected to carry out external short-circuit experiments, and the current, voltage and temperature data were recorded. The model trained in the first step is used to predict the external short-circuit fault behavior, and the root mean square error between the predicted value and the measured value of the internal terminal voltage in the first 2s is obtained, as shown in Table 2.

Table 2

In a preferred embodiment of the present application, as shown in Figure 6, the classifier based on the random forest algorithm and its training process established in step 4 specifically include:

4.1. Collect the statistical data of the maximum temperature rise and discharge capacity of the leaking battery and the non-leaking battery in the M lithium-ion power battery samples after the short circuit occurs, and input it into the random forest algorithm as the total training data set;

4.2. Use the Bootstrap method for resampling, and randomly generate N training data sets S ₁ , S ₂ ,...,S _N . Among them, S _i (1≤i≤N) is the data set formed by the i-th extraction from the M battery sample data with replacement;

4.3. Based on the generated N training data sets, N decision trees are randomly generated. After each decision tree is trained, it has the classification ability to judge whether leakage occurs through the maximum temperature rise and discharge capacity data.

In a preferred embodiment of the present application, the joint detection of the fractional impedance model and the classifier based on the random forest algorithm described in step five specifically includes:

5.1. Judgment is made according to the comparison results of the coincidence degree χ, the short-circuit fault state threshold δ ₁ and the liquid leakage state threshold δ ₂ , if χ>δ ₂ , it is considered that no liquid leakage occurs, and the value K1=0 is assigned; if δ ₁ <χ<δ ₂ , it is considered that leakage occurs, and the value K=1;

5.2. Run the classifier based on the random forest algorithm, input the collected maximum temperature rise and discharge capacity of the lithium-ion power battery as the test data set X into each trained decision tree, and output the leakage monitoring status K2, when Assign K2=1 when leakage occurs;

5.3. When K1 and K2 are 1 at the same time, assign K=1, indicating that the battery has leaked, otherwise K=0, indicating that no leak has occurred.

It is known from this example verification experiment that when the SoC is 100%, the battery leaks, and the result of the final leakage judgment using the patent of the present invention is also consistent with the actual situation. The number of classified trees is shown in Figure 7.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications and substitutions can be made to these embodiments without departing from the principle and spirit of the present invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (5)

1. A liquid leakage detection method of a lithium-ion power battery, characterized in that: specifically comprise the following steps: Step 1. Conduct multiple sets of external short-circuit tests on different lithium-ion power batteries, and record current, voltage and temperature data; Step 2, establishing a battery model for the lithium-ion power battery, and performing parameter identification on the battery model based on the data recorded in the step 1, and determining the short-circuit fault state threshold and the liquid leakage state threshold; Step 3. Use the battery model established in step 2 to diagnose the fault online, and judge its coincidence with the short-circuit fault state and the liquid leakage state according to the power battery data; Step 4. Establish a classifier based on the random forest algorithm, and train the classifier according to the data recorded in the step 1, so as to realize the monitoring of the short-circuit fault state and the liquid leakage state. Step 5, combining the battery model and the classifier based on the random forest algorithm to detect whether liquid leakage occurs. 2. The method according to claim 1, characterized in that: said step 2 specifically comprises: The established battery model adopts the fractional impedance model, and uses the current, voltage and temperature data when no liquid leakage occurs to train the model, so as to establish a battery model capable of simulating the change of the external short-circuit electrical characteristics in the non-leakage state; Parameter identification is performed on the fractional impedance model based on a genetic algorithm. 3. The method according to claim 2, characterized in that: the online diagnosis of the fault in the step 3 specifically includes: The actual power battery voltage is compared with the voltage output by the fractional-order impedance model in the short-circuit fault state, and the coincidence degree χ in the time period from the beginning of the short circuit to the current moment is calculated: V _m is the terminal voltage predicted by the model, V is the actual terminal voltage of the battery, and n is the data length. Comparing the coincidence degree χ with the short-circuit fault state threshold δ1 and the liquid leakage state threshold _{δ2 determined in step 2 .} 4. the method for claim 1 is characterized in that: the classifier based on random forest algorithm and training process thereof set up in described step 4 specifically include: 4.1. Collect the statistical data of the maximum temperature rise and discharge capacity of the leaking battery and the non-leaking battery in the M lithium-ion power battery samples after the short circuit occurs, and input it into the random forest algorithm as the total training data set; 4.2. Use the Bootstrap method for resampling, and randomly generate N training data sets S ₁ , S ₂ ,...,S _N . Among them, S _i (1≤i≤N) is the data set formed by the i-th extraction from the M battery sample data with replacement; 4.3. Based on the generated N training data sets, N decision trees are randomly generated. After each decision tree is trained, it has the classification ability to judge whether leakage occurs through the maximum temperature rise and discharge capacity data. 5. The method according to claim 3, characterized in that: the combined fractional impedance model described in the step 5 and the classifier based on the random forest algorithm are detected, specifically comprising: 5.1. Judgment is made according to the comparison results of the coincidence degree χ, the short-circuit fault state threshold δ ₁ and the liquid leakage state threshold δ ₂ , if χ>δ ₂ , it is considered that no liquid leakage occurs, and the value K1=0 is assigned; if δ ₁ <χ<δ ₂ , it is considered that leakage occurs, and the value K=1; 5.2. Run the classifier based on the random forest algorithm, input the collected maximum temperature rise and discharge capacity of the lithium-ion power battery as the test data set X into each trained decision tree, and output the leakage monitoring status K2, when Assign K2=1 when leakage occurs; 5.3. When K1 and K2 are 1 at the same time, assign K=1, indicating that the battery has leaked, otherwise K=0, indicating that no leak has occurred.