Lithium battery SOH estimation method based on charging parameters and electrochemical characteristics

Abstract

The invention discloses a method for estimating SOH (state of health) of a lithium battery based on charging parameters and electrochemical characteristics. The method comprises the following steps: collecting a sequence of charging current, charging time, temperature change, battery charging and discharging times, charging start SOC (state of charge) and charging end SOC of the battery; elements in the sequence of the charging start SOC and the sequence of the charging end SOC are calibrated through a voltage differential capacity curve constructed by voltage differential capacities corresponding to a plurality of SOCs in the charging process; performing standardization operation on elements in the sequence of the charging current, the charging time, the temperature change, the charging and discharging times of the battery, the calibrated charging start SOC and the calibrated charging end SOC of the battery, and constructing a characteristic matrix X; constructing a CNN-LSTM model for estimating the SOH of the lithium battery, and inputting historical data of the battery into the CNN-LSTM model to train the model; and inputting the feature matrix X as features into the trained CNN-LSTM model to obtain the accurate SOH of the battery.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics. Background Art

[0002] With the widespread use of lithium-ion batteries, accurate assessment of their health status is crucial to ensure the safety of equipment operation and extend battery life. Battery health (SOH) is a key indicator for measuring battery performance, usually evaluated by the remaining capacity of the battery. However, existing SOH assessment methods often require complex testing equipment and a long testing process, and cannot be monitored in real time and quickly evaluated. Summary of the invention

[0003] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a lithium-ion battery health estimation method based on charging parameters and chemical properties, which can quickly and accurately evaluate the health status of the battery.

[0004] Technical solution: To achieve the above-mentioned purpose, a method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics of the present invention comprises collecting a sequence of a battery's charging current, charging time, temperature change, battery charge and discharge times, a charging start SOC, and a charging end SOC; the elements in the sequence of the charging start SOC and the charging end SOC are calibrated by a voltage differential capacity curve constructed by voltage differential capacities corresponding to several SOCs during the charging process; the elements in the sequence of the battery's charging current, charging time, temperature change, battery charge and discharge times, a calibrated charging start SOC, and a calibrated charging end SOC are standardized, and a feature matrix X is constructed; a CNN-LSTM model for estimating the SOH of a lithium battery is constructed, and the battery's historical data is input into the CNN-LSTM model to train the model; the feature matrix X is input as a feature into the trained CNN-LSTM model to obtain the battery health SOH of the battery.

[0005] Further, the sequence I of the charging current of the battery, the sequence S of the charging time, the sequence T of the temperature change, the sequence C of the number of times of battery charge and discharge, the sequence SOCA of the charging start SOC and the sequence SOCB of the charging end SOC all contain at least two elements;

[0006] I=[I1,I2,...,I n ]

[0007] S=[S1,S2,...,S n ]

[0008] ΔT=[ΔT1, ΔT2,..., ΔT n ]

[0009] C=[C1,C2,...,C n ]

[0010] SOCA=[SOCA1, SOCA2,..., SOCA n ]

[0011] SOCB=[SOCB1,SOCB2,...,SOCB n ].

[0012] Furthermore, during the charging process, the voltages corresponding to several SOCs of the battery and the amount of charge charged are collected, the voltage differential capacity corresponding to each SOC is calculated, and a voltage differential capacity curve is constructed; a standard voltage differential capacity curve is selected, and the SOC corresponding to the peak position of the standard voltage differential capacity curve is compared with the SOC corresponding to the peak position of the constructed voltage differential capacity curve and calibrated, and the charging start SOC and charging end SOC of the calibrated battery are obtained based on the calibrated voltage differential capacity curve.

[0013] Furthermore, the elements in the battery charging current sequence, charging time sequence, temperature change sequence, battery charge and discharge times sequence, calibrated charging start SOC sequence and calibrated charging end SOC sequence are all standardized, and the calculation process is as follows:

[0014] X std =(x-μ) / σ

[0015] In the formula, Xstd is the standardized data, x is the original data, μ is the mean of all features in the sequence, and σ is the standard deviation of the feature.

[0016] Further, construct a feature matrix X based on all elements in the sequence of charging current, the sequence of charging time, the sequence of temperature change, the sequence of battery charge and discharge times, the sequence of calibrated charging start SOC, and the sequence of calibrated charging end SOC of the battery after the standardized operation;

[0017]

[0018] Furthermore, the change in battery temperature is the difference between the highest temperature and the lowest temperature during the charging process, and the calculation process is as follows:

[0019] ΔT n =T H -T L

[0020] Where, T His the highest temperature of the battery during the same charge, T L It is the lowest temperature of the battery during the same charge.

[0021] Furthermore, the CNN-LSTM model includes a CNN layer, an LSTM layer and a fully connected layer; the CNN layer includes multiple convolutional layers and a pooling layer; the CNN layer extracts local features of the charging current, charging time, temperature change, battery charge and discharge times, calibrated charging start SOC and calibrated charging end SOC in the feature matrix X through multiple convolutional layers, and performs a pooling operation on the extracted local features; the LSTM layer analyzes the multiple local features extracted by the CNN layer and outputs a hidden state, the fully connected layer integrates the features, and outputs a predicted battery health SOH; after comparing and calibrating the calculated battery health SOH with the predicted battery health SOH, the final battery health SOH is output.

[0022] Furthermore, the calculated battery health SOH is calculated as follows:

[0023] Q charge =∫I charge *T charge

[0024]

[0025] SOH=C real / C nominal

[0026] In the formula, Q charge The charge capacity when charging the battery, I charge The current used to charge the battery, C real is the actual capacity of the battery, SOC end1 is the charging starting SOC, SOC start2 To end the charging SOC, C nominal is the rated capacity of the battery, T charge For charging time.

[0027] Beneficial effects: The method of estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics of the present invention can quickly respond to changes in the battery state through real-time monitoring of charging parameters; it is simple to operate, does not require complex testing equipment, and is easy to implement in various environments; it combines the two key parameters of charging start SOC and charging end SOC to more accurately evaluate the health status of the battery with higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a curve diagram of voltage differential capacity;

[0029] Figure 2 CNN-LSTM model for estimating SOH of lithium batteries. DETAILED DESCRIPTION

[0030] The present invention will be further described below in conjunction with the accompanying drawings.

[0031] like Figure 1 As shown, a method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics is provided. The battery management system BMS collects in real time the charging current, charging time, temperature change, battery charge and discharge times, charging start SOC and charging end SOC of the battery during the charging process; the elements in the charging start SOC sequence and the charging end SOC sequence are calibrated by a voltage differential capacity curve constructed by voltage differential capacities corresponding to several SOCs during the charging process; the elements in the battery charging current, charging time, temperature change, battery charge and discharge times, calibrated charging start SOC and calibrated charging end SOC sequence are standardized, and a feature matrix X is constructed; a CNN-LSTM model for estimating SOH of a lithium battery is constructed, and the historical data of the battery is input into the CNN-LSTM model to train the model; the feature matrix X is input as a feature into the trained CNN-LSTM model to obtain the battery health SOH of the battery.

[0032] The method has a wide range of applicability and is applicable to various types of lithium-ion batteries, including but not limited to electric two-wheeled vehicles, electric vehicles, energy storage batteries and portable electronic devices.

[0033] The sequence I of the battery charging current, the sequence S of the charging time, the sequence T of the temperature change, the sequence C of the battery charge and discharge times, the sequence SOCA of the charging start SOC and the sequence SOCB of the charging end SOC all contain no less than two elements; however, the elements in each sequence are generally 2 to 4, and the number of collected charge and discharge times cannot be too small, otherwise the accuracy of estimating the battery health will be reduced; the number of collected charge and discharge times cannot be too large, otherwise the estimated battery health will not be the battery health of the battery at the time of detection, and the obtained battery health will be too large, which will also reduce the accuracy of the estimated battery health.

[0034] I=[I1,I2,...,I n ]

[0035] S=[S1,S2,...,S n ]

[0036] ΔT=[ΔT1, ΔT2,..., ΔT n ]

[0037] C=[C1,C2,...,C n ]

[0038] SOCA=[SOCA1, SOCA2,..., SOCA n ]

[0039] SOCB=[SOCB1,SOCB2,...,SOCB n ].

[0040] During the charging process, the voltages corresponding to several SOCs of the battery and the amount of charge charged are collected, the voltage differential capacity corresponding to each SOC is calculated, and a voltage differential capacity curve is constructed; a standard voltage differential capacity curve is selected, and the SOC corresponding to the peak position of the standard voltage differential capacity curve is compared with the SOC corresponding to the peak position of the constructed voltage differential capacity curve and calibrated, and the charging start SOC and charging end SOC of the calibrated battery are obtained based on the calibrated voltage differential capacity curve.

[0041] In a charging cycle, during the charging process, the voltage and the amount of charge charged at different SOCs are measured from the start SOC to the end SOC of the battery; the start SOC and the end SOC of the battery are obtained, and the coulomb counting method is first used to make a preliminary estimate of the start SOC and the end SOC of the battery. The voltage differential capacity corresponding to each SOC is calculated by the voltage and the amount of charge charged, and the voltage differential capacity curve at this time, the dQ / dV curve, is constructed; the voltage differential capacity curve constructed in this charging cycle is calibrated by the SOC corresponding to each peak position of the standard voltage differential capacity curve; and the charge start SOC and charge end SOC of the battery in this charging cycle after calibration are obtained based on the calibrated voltage differential capacity curve. The above steps are used to monitor the charging start SOC and charging end SOC in each charging cycle in real time. When it is necessary to estimate the battery health at this time, in addition to the charging start SOC and charging end SOC of the current charging cycle, the charging start SOC and charging end SOC of the previous one or two charging cycles are also required; the charging start SOC and charging end SOC are added to estimate the battery health of the battery to eliminate the influence of factors such as battery aging and temperature changes.

[0042] The battery SOH is estimated by combining the calibrated charge start SOC, the calibrated charge end SOC, the battery charging data, and the rated capacity of the battery to construct a CNN-LSTM model for estimating the battery health SOH; the calculation process of the battery health SOH in the battery using the formula is as follows:

[0043] Q charge =∫I charge *T charge

[0044]

[0045] SOH=C real / C nominal

[0046] In the formula, Q charge The charge capacity when charging the battery, I charge The current used to charge the battery, C real is the actual capacity of the battery, SOC end1 is the charging starting SOC, SOC start2 To end the charging SOC, C nominal is the rated capacity of the battery, T charge For charging time.

[0047] The calibrated SOC data and battery charging data are used as features to construct a CNN-LSTM model for estimating the SOH of lithium batteries. The estimation of SOH is achieved through a machine learning algorithm. The historical usage data of the battery is collected, and features that are helpful in predicting SOH are extracted from the historical data, including the battery charging time, number of charge and discharge times, and temperature changes. The CNN-LSTM model for estimating the SOH of lithium batteries is trained using historical data. The SOC change trend of the battery throughout its service life and how these changes affect the health status of the battery can be learned. The battery aging process can be understood by analyzing the SOC data of multiple charging cycles, and the future SOH can be predicted accordingly. The SOC at the beginning and end of charging can help LSTM better understand the performance of the battery under different usage conditions, because they reflect the state of the battery before and after a complete charging and discharging cycle.

[0048] The elements in the battery charging current sequence, charging time sequence, temperature change sequence, battery charge and discharge times sequence, calibrated charging start SOC sequence and calibrated charging end SOC sequence are normalized as features to ensure the stability and efficiency of model training. The previous charging power, actual capacity and SOH value are included; then the elements in the battery charging current sequence, charging time sequence, temperature change sequence, battery charge and discharge times sequence, calibrated charging start SOC sequence and calibrated charging end SOC sequence are standardized for the subsequent construction of the feature matrix to improve the accuracy of estimating the battery SOH. The calculation process is as follows:

[0049] X std =(x-μ) / σ

[0050] In the formula, Xstd is the standardized data, x is the original data, μ is the mean of all features in the sequence, and σ is the standard deviation of the feature.

[0051] Constructing a feature matrix X based on all elements in the sequence of charging current, the sequence of charging time, the sequence of temperature changes, the sequence of battery charge and discharge times, the sequence of calibrated charging start SOC, and the sequence of calibrated charging end SOC of the battery after the standardized operation;

[0052]

[0053] The battery temperature change is the difference between the highest temperature and the lowest temperature during the charging process, and the calculation process is as follows:

[0054] ΔT n =T H -T L

[0055] Where, T H is the highest temperature of the battery during the same charging process, T L It is the lowest temperature of the battery during the same charging process.

[0056] The CNN-LSTM model includes a CNN layer, an LSTM layer and a fully connected layer; the CNN layer includes multiple convolutional layers and a pooling layer; the CNN layer extracts local features in the battery charge and discharge cycle from the charging current, charging time, temperature change, battery charge and discharge times, calibrated charging start SOC and calibrated charging end SOC in the feature matrix X through multiple convolutional layers, and performs a pooling operation on the extracted local features; the LSTM layer analyzes the multiple local features extracted by the CNN layer and outputs a hidden state, the fully connected layer adjusts the weight of the features and integrates the features, and outputs the predicted battery health SOH; after comparing and calibrating the calculated battery health SOH with the predicted battery health SOH, the final battery health SOH is output.

[0057] The CNN layer extracts useful features from the original charging cycle data through convolution and pooling operations, and uses a one-dimensional convolution layer to extract local features in the battery charging and discharging cycle; the extracted local features are pooled so that these features become a higher-dimensional data representation, which is then passed to the LSTM layer. The features extracted by CNN are input into the LSTM network. The LSTM layer can process time series data and capture the long-term dependencies of changes in the battery's health status; the LSTM layer can learn the SOC change trends of the battery throughout its service life and how these changes affect the battery's health status;

[0058] The LSTM unit contains four main parts: forget gate, input gate, unit state and output gate. The forget gate determines which information needs to be discarded, which helps the model forget unimportant historical information.

[0059] f t =σ(W f ·[h t-1 , x t ]+b f )

[0060] Where σ is the sigmoid activation function, W f and b f are the weight matrix and bias term, respectively, x t is the input of the current time step, h t-1 is the hidden state at the previous moment;

[0061] The input gate determines what new information will be added. It consists of two parts: a sigmoid layer that determines what to update and a tanh layer that creates a new candidate value vector.

[0062] Update ratio:

[0063] i t =σ(W i ·[h t-1 , x t ]+b i )

[0064] New candidate values:

[0065] hatC t =tanh(W C ·[h t-1 , x t ]+b C )

[0066] Where W C and W i are weight matrices, b C and b i All are bias terms, tanh is the hyperbolic tangent function;

[0067] The cell state, which is the core of LSTM, is a memory storage throughout the entire link. The forget gate and the input gate work together to update the cell state.

[0068] C t =f t *C t-1 +i t *hatC t

[0069] In the formula, f t is the output of the forget gate, C t-1 is the updated status of the previous moment, i t is the input gate value, hatCt is a candidate value;

[0070] The output gate determines what to output based on the cell state and the current input. First, a sigmoid layer determines which part of the cell state will be output; then, the cell state passes through the tanh activation function and is multiplied by the result of the sigmoid layer to obtain the hidden state h. t .

[0071] Output decision:

[0072] o t =σ(W o ·[h t-1 , x t ]+b o )

[0073] Hidden state:

[0074] h t =o t *tanh(C t )

[0075] Where Wx and b o are the weight matrix and bias term respectively;

[0076] The LSTM layer can understand the battery aging process by analyzing the SOC data of multiple charging cycles and predict the future SOH accordingly; the calculation of the fully connected layer converts the output h of the LTSM layer into t Through the weight matrix W out Perform a linear transformation and add the bias term b out , the predicted output can be obtained That is to predict the SOH value of the battery. Since it is a regression task, there is no activation function in the last layer, so the result of the linear transformation is directly output.

[0077]

[0078] The formula is used to calculate the battery health SOH of the most recent charge and discharge before the detection battery SOH; the trained CNN-LSTM model is used to compare and calibrate the SOH estimation result with the SOH calculated by the formula to obtain the calibrated battery SOH value and improve the accuracy of the battery SOH; the output layer outputs the calibrated battery health, that is, the SOH value, as the final estimation result of the battery health status.

[0079] The above is only a description of the preferred embodiments of the present invention. Ordinary technicians in this technical field can make several modifications and optimizations based on the above disclosure without departing from the above basic principles. These improvements and optimizations should be regarded as the understood protection scope of the present invention.

Claims

1. A method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics, characterized in that: The method includes collecting a sequence of a battery's charging current, charging time, temperature change, battery charge and discharge times, a charging start SOC, and a charging end SOC; the elements in the charging start SOC sequence and the charging end SOC sequence are calibrated by a voltage differential capacity curve constructed by voltage differential capacities corresponding to a plurality of SOCs during the charging process; the elements in the sequence of the battery's charging current, charging time, temperature change, battery charge and discharge times, a calibrated charging start SOC, and a calibrated charging end SOC are standardized, and a feature matrix X is constructed; Build a CNN-LSTM model for estimating lithium battery SOH, and input the battery's historical data into the CNN-LSTM model to train the model; The feature matrix X is input as features into the trained CNN-LSTM model to obtain the battery health SOH.

2. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 1, characterized in that: The sequence I of the battery charging current, the sequence S of the charging time, the sequence T of the temperature change, the sequence C of the battery charge and discharge times, the sequence SOCA of the charging start SOC and the sequence SOCB of the charging end SOC all contain at least two elements; I=[I1,I2,...,I n ] S=[S1,S2,...,S n ] ΔT=[ΔT1,ΔT2,...,ΔT n ] C=[C1,C2,...,C n ] SOCA=[SOCA1, SOCA2, ..., SOCA n , SOCB=[SOCB1,SOCB2,...,SOCB n ]。 3. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 1, characterized in that: During the charging process, the voltage corresponding to several SOCs of the battery and the amount of charge charged are collected, the voltage differential capacity corresponding to each SOC is calculated, and a voltage differential capacity curve is constructed; A standard voltage differential capacity curve is selected, and the SOC corresponding to the peak position of the standard voltage differential capacity curve is compared with the SOC corresponding to the peak position of the constructed voltage differential capacity curve, and calibration is performed. Based on the calibrated voltage differential capacity curve, the charging start SOC and charging end SOC of the calibrated battery are obtained.

4. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 1, characterized in that: The elements in the battery charging current sequence, charging time sequence, temperature change sequence, battery charge and discharge times sequence, calibrated charging start SOC sequence and calibrated charging end SOC sequence are all standardized, and the calculation process is as follows: X std =(x-μ) / σ In the formula, Xstd is the standardized data, x is the original data, μ is the mean of all features in the sequence, and σ is the standard deviation of the feature.

5. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 1, characterized in that: Constructing a feature matrix X based on all elements in the sequence of charging current, the sequence of charging time, the sequence of temperature changes, the sequence of battery charge and discharge times, the sequence of calibrated charging start SOC, and the sequence of calibrated charging end SOC of the battery after the standardized operation; 6. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 1, characterized in that: The battery temperature change is the difference between the highest temperature and the lowest temperature during the charging process, and the calculation process is as follows: ΔT n =T H -T L Where, T H is the highest temperature of the battery during the same charge, T L It is the lowest temperature of the battery during the same charge.

7. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 1, characterized in that: The CNN-LSTM model includes a CNN layer, an LSTM layer and a fully connected layer; the CNN layer includes multiple convolutional layers and a pooling layer; the CNN layer extracts local features of the charging current, charging time, temperature change, battery charge and discharge times, calibrated charging start SOC and calibrated charging end SOC in the feature matrix X through multiple convolutional layers, and performs a pooling operation on the extracted local features; the LSTM layer analyzes the multiple local features extracted by the CNN layer and outputs a hidden state, the fully connected layer integrates the features, and outputs a predicted battery health SOH; after comparing and calibrating the calculated battery health SOH with the predicted battery health SOH, the final battery health SOH is output.

8. The method for estimating SOH of a lithium battery based on charging parameters and electrochemical characteristics according to claim 7, characterized in that: The calculated battery health SOH is calculated as follows: Q charge =∫I charge *T charge SOH=C real / C nominal In the formula, Q charge The charge capacity when charging the battery, I charge The current used to charge the battery, C real is the actual capacity of the battery, SOC end1 is the charging starting SOC, SOC start2 To end the charging SOC, C nominal is the rated capacity of the battery, T charge For charging time.

Images