KR20220163736A - Method for detecting abnomal cells and estimating SOH in Lithium-Ion battery pack - Google Patents

Abstract

The present invention relates to a technique for providing a more accurate state of health (SOH) prediction value. A method for detecting abnormal cells and predicting the SOH of a lithium-ion battery pack according to the present invention may include: a first step of acquiring actual measured values for four parameters of voltage, capacity, temperature, and humidity for each battery cell; a second step of calculating a threshold value for each parameter by inputting the measured values of voltage, capacity, temperature, and humidity of each battery cell together with battery cell location information into a learning model for each parameter; a third step of determining whether each battery cell is in an abnormal state; and a fourth step of calculating the first weight for the electrochemical factor, calculating the second weight for the external environment factor, and calculating the sum of the values acquired by performing the convolution operation on the previous SOH prediction value as the SOH prediction value for the corresponding battery pack and each battery cell, respectively.

Description

Method for detecting abnomal cells and estimating SOH in Lithium-Ion battery pack}

The present invention detects a battery cell operating abnormally in a lithium ion battery pack using threshold values for each of a number of parameters related to performance changes according to electrochemical characteristics and environmental factors of the battery cell, and more accurately SOH (State of Health) relates to technology that provides predictive values.

Lithium-ion batteries are one of the most widely used secondary batteries today.

Lithium-ion batteries have an electromotive force of 3.6V, are very light, have a high energy density compared to other batteries, and have very low power loss due to self-discharge. It is being used.

Such a lithium ion battery is mainly used in the form of a battery pack composed of one or more electrochemical cells capable of charging and discharging when a certain amount or more is required.

Recently, 25% of new car sales are vehicles equipped with electric engines, and the demand for electric vehicles is rapidly increasing as the operation of electric vehicles is actively encouraged in response to environmental pollution and climate change.

However, lithium ion batteries used as a main power source for electric vehicles, etc., have disadvantages in that, due to their temperature-sensitive nature, the aging progresses faster and the risk of explosion increases as the temperature increases.

In addition, as a lithium ion battery is used and repeatedly charged and discharged, a voltage difference between battery cells increases, resulting in unbalance. In this case, the lifetime of the entire battery is shortened by some of the deteriorating battery cells, and even the risk of fire or explosion increases.

Accordingly, lithium-ion batteries used in electric vehicles must operate stably in various harsh environments and must be managed precisely. To this end, internal and external factors that threaten the stability of lithium batteries, including the identification of the cause of thermal instability, Research on this is actively progressing.

In general, SOH (State of Health), which numerically presents the deterioration state of lithium-ion batteries, is modeled to determine the current state of lithium-ion batteries. Existing methods use internal resistance/impedance, charge capacity, and accumulated charge/discharge cycles. A method of calculating approximate SOH using numbers or estimating SOH using a Kalman filter, fuzzy logic, or SOC variation between successive cycles is used.

However, these techniques have a fatal disadvantage in that the parameters used for SOH estimation are limited, and the accuracy of estimation and prediction eventually deteriorates when operating in long cycles and various environments because various factors that cause the overall performance degradation and explosion of the LIB are not considered. has

In addition, battery cells have different deterioration characteristics depending on their placement, and SOH can be changed not only by deterioration characteristics but also by other environmental factors, so it is difficult to determine the current state of each battery cell only by estimating SOH. .

1. Korean Patent Publication No. 10-2020-0097170 (Name: Battery management device, battery management method and battery pack) 2. Korean Patent Publication No. 10-2020-0056716 (Name: Battery SOH Calculation System)

Therefore, the present invention was created in view of the above circumstances, and a number of parameters corresponding to the charging and discharging states of lithium ion battery packs and battery cells were applied to a deep learning model through an RNN-based structure to reflect the criticality of deterioration characteristics. By calculating the value, determining the battery cell state based on the threshold value for each parameter, and calculating the current SOH prediction value, it is possible to more accurately monitor the battery cell state and provide the SOH prediction value. Its technical purpose is to provide a cell detection method.

According to one aspect of the present invention for achieving the above object, a lithium ion battery pack analyzes a signal applied from a lithium ion battery pack composed of a plurality of battery cells in a battery state monitoring device to monitor the current state of the lithium ion battery pack. In the method of detecting abnormal cells and predicting SOH of, the battery condition monitoring device obtains actual values for four parameters of voltage, capacity, temperature, and humidity for a lithium-ion battery pack and each battery cell in the lithium-ion battery pack In the first step, in the battery condition monitoring device, the measured values of the voltage, capacity, temperature, and humidity for each battery cell in the lithium-ion battery pack and the corresponding lithium-ion battery pack are converted to the learning model for each parameter together with the battery cell location information. A second step of outputting a threshold to which different deterioration characteristics are applied according to the position of the battery cell, the difference between the threshold of the battery pack and the actual value in the battery condition monitoring device. A third step of determining whether or not each battery cell is in an abnormal state based on the difference between the measured value of each battery cell and a threshold value while performing a state detection operation for each battery cell in the battery pack when is greater than a preset reference value, and battery state monitoring In the device, a first weight for an electrochemical element is calculated based on a voltage threshold and a capacity threshold for each battery pack and each battery cell, and a second weight for an external environment element is calculated based on a temperature threshold and a humidity threshold. and a fourth step of calculating the sum of the values calculated by convolution with the previous predicted SOH values as SOH predicted values for the corresponding battery pack and each battery cell, respectively. A prediction method is provided.

여기서, Vt, Ct, Tt, Ht 는 각각 전압 임계값, 캐패시티 임계값, 온도 임계값, 습도 임계값이고, σ는 시그모이드 함수, W는 열화 가중치, X는 입력값, B는 바이어스 값, * 는 컨볼루션 연산자이며, 상기 열화 가중치(W)는 배터리셀의 위치가 배터리팩의 중심에서 멀어질수록 보다 큰 값으로 설정됨.In addition, in the second step, the learning model to which the deterioration characteristics according to the battery cell position are applied is a model using a signoid function, and a threshold value for each parameter is output according to the following equation. A cell detection and SOH prediction method is provided.

Here, Vt, Ct, Tt, and Ht are the voltage threshold, capacity threshold, temperature threshold, and humidity threshold, respectively, σ is the sigmoid function, W is the degradation weight, X is the input value, and B is the bias value. , * is a convolution operator, and the deterioration weight (W) is set to a larger value as the location of the battery cell is farther from the center of the battery pack.

In addition, the humidity learning model calculates only the humidity threshold for the battery pack, and the battery condition monitoring device uses the humidity threshold for the battery pack as the humidity threshold for all battery cells in the battery pack. A method for detecting abnormal cells in a battery pack and predicting SOH is provided.

Further, in the third step, at least one of a first difference between the measured voltage of the battery pack and the voltage threshold and the second difference between the measured battery pack capacity and the capacity threshold in the battery state monitoring device is a predetermined criterion. Provided is a method for detecting abnormal cells in a lithium ion battery pack and predicting SOH, characterized in that performing a state detection operation on all battery cells in a corresponding battery pack when the difference is greater than or equal to the difference value.

In addition, in the battery cell state detection operation in the third step, a first condition in which both the difference value between the measured voltage value of the battery cell and the voltage threshold value and the difference value between the measured capacity value and the capacity threshold value are less than a preset reference value. If the battery cell is satisfied, determining that the battery cell is in a stable state, and if the first condition is not satisfied, the average temperature change amount and average humidity change amount of the battery cell compared to a predetermined reference cycle are less than the temperature threshold value and the humidity threshold value. If two conditions are satisfied, determining that the battery cell is in a degraded state, and if the first and second conditions are not satisfied, determining that the battery cell is in an abnormal state. A method for detecting an abnormal cell of a lithium ion battery pack and predicting SOH is provided.

여기서, Pt는 제1 가중치, Qt 는 제2 가중치, ReLU 은 ReLU 함수, h(t-1)은 이전 SOH 예측값이고, * 는 컨볼루션 연산자임.In addition, in the fourth step, the first weight and the second weight are calculated by the ReLU function, and a method for detecting an abnormal cell of a lithium ion battery pack and predicting SOH is provided, characterized in that they are calculated by the following equation.

Here, Pt is the first weight, Qt is the second weight, ReLU is the ReLU function, h(t-1) is the previous SOH prediction value, and * is the convolution operator.

여기서, h(t)는 SOH 예측값이고, K 는 운용 가중치임.In addition, a method for detecting an abnormal cell of a lithium ion battery pack and predicting SOH is provided, wherein the SOH prediction value in the fourth step is calculated by the following equation.

Here, h(t) is the predicted value of SOH, and K is the operating weight.

In addition, in the fourth step, the battery condition monitoring device collects charge/discharge cycle data and extracts the effective charge start point, charge peak point, and effective discharge end point of the cycle, and the extracted effective charge start point and charge peak. Calculating the slope of the straight line generated by connecting the points as the charging voltage slope, and calculating the slope of the straight line generated by connecting the charging peak point and the effective discharge end point as the discharge voltage slope, the charging voltage and the discharging voltage of the cycle Calculating an SOH weight of a smaller value as the difference between the slope and the reference charge voltage and discharge voltage slope period in the preset reference cycle increases, and correcting and outputting the predicted SOH value by multiplying the SOH weight by the predicted SOH value A method for detecting an abnormal cell of a lithium ion battery pack and predicting SOH is provided.

According to the present invention, a threshold value reflecting deterioration characteristics is calculated based on various parameters such as voltage, capacity, temperature, and humidity corresponding to the charging and discharging state of a lithium ion battery pack and battery cell, and using this, the battery cell state and By simultaneously calculating the current predicted SOH value, a more accurate battery cell state and predicted SOH value can be provided.

In addition, in the present invention, the SOH prediction value is calculated using the threshold value for each parameter in consideration of the electrochemical characteristics by the voltage and capacity of the battery cell and the external environmental characteristics by temperature and humidity, but the change in the charge/discharge cycle It is possible to provide a more reliable SOH prediction value considering various environmental conditions by correcting the SOH prediction value by additionally considering .

In addition, in the present invention, waste of resources due to unnecessary battery cell state detection can be prevented by performing the battery cell state detection operation in the battery pack only when the threshold value of the battery pack is equal to or greater than the reference value.

1 is a diagram showing a schematic configuration of a battery state monitoring device having an abnormal cell detection and SOH prediction function of a lithium ion battery pack to which the present invention is applied.
FIG. 2 is a diagram for explaining a learning model of the threshold calculation unit 200 shown in FIG. 1;
FIG. 3 is a diagram showing activation function characteristics applied to each learning model of the threshold value calculator 200 and the SOH predictor 400 shown in FIG. 1;
FIG. 4 is a diagram for explaining a method of correcting an SOH predicted value of the SOH predictor 400 shown in FIG. 1;
5 is a view for explaining an operation of detecting an abnormal cell of a lithium ion battery pack and predicting SOH of the battery state monitoring device shown in FIG. 1;
FIG. 6 is a diagram for explaining a battery cell state determination process (ST300) shown in FIG. 5 in more detail;
FIG. 7 is a diagram for explaining a battery SOH prediction process (ST400) shown in FIG. 5 in more detail;

The embodiments described in the present invention and the configurations shown in the drawings are only preferred embodiments of the present invention, and do not represent all the technical ideas of the present invention, so the scope of the present invention is limited to the embodiments and drawings described in the text. should not be construed as being limited by That is, since the embodiment can be changed in various ways and can have various forms, it should be understood that the scope of the present invention includes equivalents capable of realizing the technical idea. In addition, since the object or effect presented in the present invention does not mean that a specific embodiment should include all of them or only such effects, the scope of the present invention should not be construed as being limited thereto.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs, unless defined otherwise. Terms defined in commonly used dictionaries should be interpreted as being consistent with meanings in the context of related art, and cannot be interpreted as having ideal or excessively formal meanings that are not clearly defined in the present invention.

1 is a diagram showing a schematic configuration of a battery state monitoring device having an abnormal cell detection and SOH prediction function of a lithium ion battery pack to which the present invention is applied.

Referring to FIG. 1, the battery state monitoring device applied to the present invention includes a battery information collection unit 100, a threshold value calculation unit 200, a battery state determination unit 300, and an SOH prediction unit 400. do. The battery condition monitoring device may be provided and operated in a battery management system (BMS) or may be an independent device from the battery management system.

In addition, the battery pack 10 applied to the present invention includes a plurality of battery cells S, and one or more battery packs 10 may be connected to the battery state monitoring device. In addition, each battery cell (S) has different deterioration characteristics depending on its location, and the battery cell disposed outside the battery pack 10 has a relatively higher deterioration rate than the cell located in the center of the battery pack 10. It has fast-paced characteristics.

In addition, a battery controller is provided inside or outside the battery pack 10 to collect and output charge/discharge cycle information of each battery cell S included in the corresponding battery pack 10, through which the battery information collection unit 100 ) can communicate with.

The battery information collection unit 100 collects battery pack charge/discharge cycle data from each battery cell S, and extracts four parameters of voltage, capacity, temperature, and humidity from the charge/discharge data. The battery information collection unit 100 extracts cell state information including voltage, capacity, and temperature information for each battery cell S, and pack state information including voltage, capacity, temperature, and humidity information for each battery pack. At this time, the voltage, capacity, and temperature information of the battery pack is calculated as an average value for each parameter of all battery cells S constituting the battery pack. Also, the humidity information may be collected for each battery pack or may be calculated as an average value of humidity values for each cell collected for each battery cell.

The threshold value calculation unit 200 calculates a threshold value for each parameter corresponding to the battery state value acquired by the battery information collection unit 100 using the learning model.

As shown in FIG. 2, the learning model is provided for each battery state parameter, and includes a voltage learning model, a capacity learning model, a temperature learning model, and a humidity learning model.

In this case, the voltage learning model, the capacity learning model, and the temperature learning model learn battery pack state values and battery cell state values in parallel within one learning model, and output threshold values for each battery cell and battery pack threshold values, respectively. And, the learning model that outputs the threshold value for each cell is configured to output the threshold value to which different degradation characteristics are applied according to cell positions. That is, the above-described learning model uses the battery cell location value as an input value, and is configured so that cells disposed at both ends of the battery pack 10 output relatively lower threshold values than cells disposed in the middle of the battery pack 10 for the same state value. do.

In addition, the humidity learning model is trained as a single parameter for the battery pack in order to reduce the amount of calculation based on the experimental result that the humidity of the entire battery pack and the humidity of each cell inside the pack have almost the same characteristics. That is, the humidity learning model calculates only the humidity threshold value for the battery pack and uses the humidity threshold value for the battery pack as the humidity threshold value for all battery cells in the corresponding battery pack.

Each learning model described above uses a sigmoid function of an RNN structure to calculate a threshold value for each parameter in units of a certain time. As shown in FIG. 3(A), the sigmoid function has a characteristic of outputting a value of “0 to 1” for an input value. That is, the threshold value for each parameter is calculated by Equation 1 below.

Here, Vt, Ct, Tt, and Ht are the voltage threshold, capacity threshold, temperature threshold, and humidity threshold, respectively, σ is the sigmoid function, W is the degradation weight, X is the input value, and B is the bias value. , and * is the convolution operator. Here, the deterioration weight (W) and bias value (B) for each parameter are set in advance, and the deterioration weight (W) for each cell is set differently according to the location of the corresponding cell. set to a larger value.

That is, the threshold values (Vt, Ct, Tt, Ht) for each parameter are calculated by performing a convolution operation with a preset weight on the input value and applying a value calculated by summing the preset bias value to the result to the sigmoid function. do.

The threshold calculation unit 200 outputs voltage thresholds, capacity thresholds, temperature thresholds, and humidity thresholds for each battery pack, as well as voltage thresholds and capacity thresholds for all battery cells of each battery pack 10 . value, the temperature threshold is output.

The battery state determination unit 300 determines the current state of the battery cell based on the difference between the measured value for each parameter collected by the battery information collection unit 100 and the threshold value calculated by the threshold value calculation unit 200. . At this time, the battery state determining unit 300 performs a state detection operation for each battery cell in the corresponding battery pack 10 when the difference between the threshold value and the measured value of the battery pack 10 is equal to or greater than a predetermined reference value, but the measured value for each battery cell It is determined whether or not each battery cell is in an abnormal state based on the difference between ? and a threshold value.

The SOH prediction unit 400 applies the threshold value for each parameter calculated by the threshold value calculation unit 200 to a learning model based on a Rectified Linear Unit (ReLU) function to calculate SOH prediction values for the battery pack and battery cell. .

As shown in FIG. 3(B), the ReLU function returns “0” when a value smaller than “0” is obtained, and outputs the corresponding output value as it is when a value greater than “0” is obtained. That is, the predicted SOH value of the battery pack and each battery cell is calculated by Equation 2 below.

Here, Pt is a first weight based on an electrochemical factor, Qt is a second weight based on an external environment factor, ReLU is a ReLU function, h is an output value (SOH prediction value), and K is an operating weight. Here, K is a weight considering the battery pack operating environment.

That is, the SOH predictor 400 calculates the first weight for the electrochemical factor based on the voltage threshold and the capacity threshold for each battery pack and each battery cell, and the external environment based on the temperature threshold and humidity threshold. A second weight for each element is calculated, and a sum of values obtained by performing a convolution operation on previous SOH predicted values is calculated as an SOH predicted value for each battery pack and each battery cell.

As a result, in the present invention, the SOH predictor 400 outputs an SOH value based on electrochemical factors and external environmental factors based on the threshold value calculated by reflecting the deterioration weight for each battery cell position, so that a more accurate SOH value can be obtained. it is possible to predict

In addition, the SOH predictor 400 extracts a specific point for each cycle starting from a preset reference cycle, and connects the extracted feature points to determine the charging voltage slope and the discharge voltage slope in the corresponding cycle. Each is calculated, and a final SOH value can be output by additionally applying a weight to the SOH output value based on the slope of the charging voltage and the slope of the discharging voltage.

At this time, as shown in FIG. 4, the SOH predictor 400 extracts the effective charge start point P1, charge peak point P2, and effective discharge end point P3 as feature points in one charge/discharge cycle. And, using a mathematical algorithm, the slope of the straight line connecting the effective charging start point (P1) and the charging peak point (P2) is calculated as the charging voltage slope, and the charging peak point (P2) and the effective discharge end point (P3 ) is calculated as the slope of the discharge voltage. 4 illustrates a charge/discharge voltage slope for a reference cycle, for example, a third cycle, and a charge/discharge voltage slope for a current cycle (nth cycle).

The SOH prediction unit 400 increases the difference between the reference charging voltage slope in the reference cycle and the charging voltage slope of the current cycle and the absolute value of the reference discharge voltage slope and the absolute value of the discharge voltage slope in the current cycle. The smaller the ratio of the current cycle's charging voltage slope to the slope and the absolute value of the current cycle's discharge voltage slope to the absolute value of the reference discharge voltage slope, the smaller the weighted value is applied to the SOH prediction value so that the smaller SOH prediction value is finally output. It can be.

Subsequently, the operation of detecting abnormal cells of the lithium ion battery pack and predicting SOH of the battery state monitoring device configured as described above will be described with reference to flowcharts shown in FIGS. 5 to 7 .

Referring to FIG. 5, the battery state monitoring device acquires state values for four parameters of voltage, capacity, temperature, and humidity for each battery cell (S) in a charge/discharge cycle applied from the battery pack 10, A battery state value is obtained by calculating an average value of state values for each battery cell S for each parameter as a state value for each parameter for a corresponding battery pack (ST100). At this time, the charge/discharge cycle is composed of data constituting one complete cycle including all feature points, and if charge/discharge of the current cycle is in progress, 4 values of voltage, capacity, temperature and humidity are used using the entire data of the previous cycle. State values for each parameter can be obtained.

Subsequently, the battery state monitoring device calculates a threshold value for each parameter using a learning model pre-learned for the battery state value (ST200). At this time, the learning model is a model in which deterioration characteristics of each battery cell location are reflected, and the voltage threshold value (Vt), capacity threshold value (Ct), temperature threshold value (Tt) and humidity of the battery pack according to the algorithm of Equation 1 A threshold value (Ht) is output, and a voltage threshold value (Vt), a capacity threshold value (Ct), and a temperature threshold value (Tt) for each battery cell are output.

Thereafter, the battery state monitoring apparatus determines the battery cell state by performing a battery cell state detection operation based on the threshold value for each battery cell in the battery pack based on the threshold value for each parameter calculated in step ST200 (ST300).

In addition, the battery state monitoring device calculates first and second weights using different threshold values and applies them to the SOH prediction value of the previous time to calculate the current SOH prediction value (ST400).

Next, referring to FIG. 6, the battery cell state determination process (ST300) shown in FIG. 5 will be described in more detail.

Referring to FIG. 6, the battery state monitoring apparatus calculates a first difference between the measured voltage of the battery pack 10 and the voltage threshold and a second difference between the measured capacity and the threshold, respectively (ST310). ). At this time, the measured voltage value and the measured capacity value are values obtained from a signal applied from the battery pack 10, and the voltage threshold value and the capacity threshold value are values output by the learning model in step ST200 of FIG. 5.

The battery state monitoring device determines whether both the first and second difference values are less than a predetermined reference difference value in step ST310 (ST320). At this time, the reference value is set to an error value considering the number of cells and operating environment variables, and is set to a value in the range of "0.001 to 0.01".

In the step ST320, when the difference between the measured values of the voltage and the capacity and the threshold is less than the reference value, the battery state monitoring apparatus determines the battery pack 10 to be in a stable operating state (ST330).

Meanwhile, in the step ST320, when at least one of the first and second difference values is equal to or greater than the reference difference value, the battery state monitoring device sequentially performs state detection operations for all battery cells included in the corresponding battery pack 10. do it with

That is, the battery state monitoring apparatus calculates the difference between the measured voltage value of the battery cell and the voltage threshold value, and the measured capacity value and the capacity threshold value (ST340). In this case, the measured voltage value and the measured capacity value are values obtained from a signal applied from the corresponding battery cell, and the voltage threshold value and the capacity threshold value are values output by the learning model in step ST200 of FIG. 5 .

Then, the battery state monitoring device determines whether the first condition of whether the difference between the measured value of the voltage or capacity calculated in step ST340 and the threshold value is less than a preset reference value is satisfied (ST350).

In the step ST350, when the difference between the measured values of the voltage and the capacity and the threshold value are all less than the reference value, the battery state monitoring apparatus determines the battery cell to be in a stable operating state (ST330).

Meanwhile, in the step ST350, if the difference between the measured value of at least one of the voltage and the capacity and the threshold value is higher than the reference value and does not satisfy the first condition, the battery condition monitoring device determines the average temperature of the corresponding battery cell compared to the preset reference cycle. And the amount of change in humidity is calculated (ST360). At this time, the reference cycle is set as the third cycle of the corresponding battery cell, and the difference between the average temperature and average humidity up to the current cycle is calculated by setting the temperature and humidity during the third cycle operation as the reference temperature and reference humidity.

Then, the battery state monitoring device determines whether the second condition of whether the average temperature and humidity change amount compared to the reference cycle calculated in the step ST360 is less than the temperature threshold value and the humidity threshold value is satisfied (ST370). Here, the temperature threshold and humidity threshold are values output by the learning model in step ST200 of FIG. 5 for the corresponding battery cell.

In the step ST370, if both the average temperature and humidity change compared to the reference cycle are less than the reference value and satisfy the second condition, the battery state monitoring device determines that the battery cell is in a stable temperature and humidity state, but has deteriorated performance (ST380). This is a state in which the battery cell is normally operated, but the electrochemical performance of a specific battery cell is degraded, and measures can be taken to operate stably in connection with a cell balancing module of a battery management system (not shown).

On the other hand, in the step ST370, when at least one of the average temperature and humidity change compared to the reference cycle does not satisfy the second condition by a reference value or more, the battery state monitoring device determines the battery cell to be in an abnormal state (ST390). The abnormal state is a state in which there is a problem in both electrochemical performance and the external environment, and the system can be operated more stably through active measures such as replacing the battery cell.

At this time, the above-described operations of ST340 to ST390 are equally performed for all battery cells provided in the corresponding battery pack 10 .

Next, with reference to FIG. 7, the battery SOH prediction process (ST400) shown in FIG. 5 will be described in more detail.

Referring to FIG. 7, the battery state monitoring apparatus calculates a first weight according to an electrochemical factor by applying the voltage threshold and the capacity threshold calculated in step ST200 of FIG. 5 to a learning model based on the ReLU function (ST410). ). At this time, the first weight is calculated as the Pt value of Equation 2 above.

In addition, the battery state monitoring apparatus applies the temperature threshold and humidity threshold calculated in step ST200 of FIG. 5 to a learning model based on the ReLU function to calculate a second weight according to external environmental factors (ST420). At this time, the second weight is calculated as the Qt value of Equation 2 above.

Subsequently, the battery state monitoring apparatus calculates an SOH prediction value for the corresponding battery pack or the corresponding battery cell using the first and second weights (ST430). At this time, the SOH predicted value is calculated as the h(t) value of Equation 2 above.

In addition, the battery condition monitoring device extracts the effective charge start point, charge peak point, and effective discharge end point of the cycle based on the charge and discharge cycle data applied from the battery pack 10, and the effective charge start point and charge peak point. The slope of the straight line created by connecting the points is calculated as the charging voltage slope, and the slope of the straight line created by connecting the charging peak point and the effective discharge end point is calculated as the discharge voltage slope (ST440).

Subsequently, the battery state monitoring device calculates an SOH weight corresponding to a difference between the charge voltage/discharge voltage slope and the reference charge/discharge voltage slope period in a preset reference cycle (ST450). In this case, the SOH weight may be determined through a separate learning model that outputs an SOH weight for outputting the SOH as an optimal predicted value by taking the charging voltage slope and the discharge voltage slope as inputs. That is, as the difference between the charge voltage and discharge voltage slopes of the corresponding cycle and the reference charge voltage and discharge voltage slopes in the preset reference cycle increases, a smaller SOH weight is calculated.

Then, the battery state monitoring device corrects the predicted SOH value calculated in the step ST430 using the SOH weight calculated in the step ST450, and outputs an optimal predicted SOH value (ST460). That is, the battery state monitoring device multiplies the SOH weight by the SOH prediction value, and additionally corrects and outputs the SOH prediction value based on the threshold value calculated in the step ST430 through the weight according to the charge/discharge cycle.

100: battery information collection unit, 200: threshold value calculation unit,
300: battery state detection unit, 400: SOH prediction unit,
10: battery pack, S: battery cell.

Claims (8)

A method for detecting abnormal cells in a lithium ion battery pack and predicting SOH by analyzing a signal applied from a lithium ion battery pack composed of a plurality of battery cells in a battery state monitoring device to monitor the current state of the lithium ion battery pack,
A first step of acquiring actual measured values for four parameters of voltage, capacity, temperature, and humidity for a lithium ion battery pack and each battery cell in the lithium ion battery pack in a battery condition monitoring device;
In the battery condition monitoring device, the measured values of the voltage, capacity, temperature, and humidity of each battery cell in the lithium-ion battery pack and the corresponding lithium-ion battery pack are entered into the learning model for each parameter along with the battery cell location information, and the threshold for each parameter is entered. A second step of calculating a value, but the learning model outputs a threshold value to which different deterioration characteristics are applied according to the location of the battery cell;
In the battery condition monitoring device, if the difference between the threshold value and the measured value of the battery pack is greater than the preset reference value, the state detection operation is performed for each battery cell in the battery pack, and each battery cell is based on the difference between the measured value and the threshold value for each battery cell. A third step of determining whether or not is in an abnormal state, and
In the battery condition monitoring device, a first weight value for an electrochemical element is calculated based on a voltage threshold value and a capacity threshold value for each battery pack and each battery cell, and a control value for external environmental factors is determined based on a temperature threshold value and a humidity threshold value. 2. An abnormal cell of a lithium-ion battery pack, characterized in that it comprises a fourth step of calculating the weight and calculating the sum of the values obtained by convolution with the previous SOH predicted value as the SOH predicted value for the corresponding battery pack and each battery cell, respectively. Detection and SOH prediction methods.
According to claim 1,
In the second step, the learning model to which the deterioration characteristics according to the battery cell position are applied is a model using a signoid function, and a threshold value for each parameter is output according to the following equation. and SOH prediction methods.

Here, Vt, Ct, Tt, and Ht are the voltage threshold, capacity threshold, temperature threshold, and humidity threshold, respectively, σ is the sigmoid function, W is the degradation weight, X is the input value, and B is the bias value. , * is a convolution operator, and the deterioration weight (W) is set to a larger value as the location of the battery cell is farther from the center of the battery pack.
According to claim 1 or 2
The humidity learning model calculates only the humidity threshold for the battery pack,
A method for detecting abnormal cells in a lithium ion battery pack and predicting SOH, characterized in that the battery condition monitoring device uses a humidity threshold for the battery pack as a humidity threshold for all battery cells in the battery pack.
According to claim 1,
In the third step, at least one of the first difference between the measured battery pack voltage and the voltage threshold and the second difference between the measured battery pack capacity and the capacity threshold in the battery state monitoring device is a preset reference difference value. A method for detecting abnormal cells in a lithium-ion battery pack and predicting SOH, characterized in that in the case of an abnormality, a state detection operation is performed for all battery cells in the corresponding battery pack.
According to claim 4,
The battery cell state detection operation in the third step,
Determining that the battery cell is in a stable state when a difference value between the measured voltage value of the battery cell and the voltage threshold value and the difference value between the measured capacity value and the capacity threshold value both satisfy a first condition that is less than a preset reference value;
If the first condition is not satisfied and the average temperature change and average humidity change of the battery cell compared to the preset reference cycle satisfy the second condition that is less than the temperature threshold and humidity threshold, the battery cell is determined to be in a performance deterioration state Steps to do and
A method for detecting an abnormal cell in a lithium ion battery pack and predicting SOH, characterized in that it comprises the step of determining that the battery cell is in an abnormal state when the first condition and the second condition are not satisfied.
According to claim 1,
In the fourth step, the first weight and the second weight are calculated by the ReLU function, and the abnormal cell detection and SOH prediction method of the lithium ion battery pack, characterized in that calculated by the following equation.

Here, Pt is the first weight, Qt is the second weight, ReLU is the ReLU function, h(t-1) is the previous SOH prediction value, and * is the convolution operator.
According to claim 6,
In the fourth step, the SOH prediction value is calculated by the following equation.

Here, h(t) is the predicted value of SOH, and K is the operating weight.
The method of any one of claims 1, 6, and 7,
In the fourth step, the battery state monitoring device collects charge/discharge cycle data and extracts an effective charge start point, a charge peak point, and an effective discharge end point of the cycle;
Calculating the slope of the straight line generated by connecting the extracted effective charge start point and the charging peak point as the charging voltage slope, and calculating the slope of the straight line generated by connecting the charging peak point and the effective discharge end point as the discharge voltage slope ,
Calculating an SOH weight of a smaller value as the difference between the slope of the charge voltage and discharge voltage of the corresponding cycle and the slope of the reference charge voltage and discharge voltage in the preset reference cycle increases;
A method for detecting an abnormal cell in a lithium ion battery pack and predicting SOH, further comprising the step of correcting and outputting the predicted SOH value by multiplying the SOH weight by the predicted SOH value.

Images