A "filling-inspection" integrated electric vehicle battery state detection and evaluation method

Abstract

The application discloses a kind of " fills - detects " integrated electric vehicle battery state detection and evaluation method, its steps include:1 to electric vehicle battery pack is carried out routine charging, gathers the current, voltage, temperature of module in certain voltage interval range in constant current charging stage, calculates charging capacity and capacity increment;2 when charging to battery pack state of charge value is S0, to module is applied 1-20Hz frequency sine wave excitation current, gathers the module response voltage under different frequencies, calculates ac impedance;3 to battery pack is applied pulse charging current, root gathers module terminal voltage information, calculates the ohmic resistance of each module;4 the current, voltage, incremental capacity, temperature, ac impedance, ohmic resistance etc. parameter gathered above is input to corresponding battery model, after analyzing data, the performance evaluation of battery whole and module is completed in charging stage, to realize the function of battery online state estimation and consistency evaluation.

Description

Technical Field

[0001] This invention is applied to the field of electric vehicles, specifically a "charging-inspection" integrated method for detecting and evaluating the state of electric vehicle batteries, which is applicable to the state estimation and fault early warning of electric vehicle battery management systems. Background Technology

[0002] As a secondary energy source, electricity can replace fossil fuels in transportation, significantly alleviating carbon emission pressures. Consequently, the electric vehicle industry has entered a period of rapid development, and batteries, as energy storage devices, are being widely used in electric vehicles. As one of the core subsystems in electric vehicle technology, the Battery Management System (BMS) plays a crucial role. The BMS is responsible for monitoring the operating status of each battery within the battery storage unit, ensuring the safe and reliable operation of the energy storage unit.

[0003] Traditional BMS can monitor and collect the state parameters of energy storage batteries (single cell voltage, battery terminal temperature, battery circuit current, battery pack terminal voltage, battery system insulation resistance, etc.) and perform necessary analysis and calculation on the relevant state parameters. However, it cannot perform performance evaluation of the battery during the charging and replenishment stage, and the types of parameters acquired are limited, and it cannot obtain the battery impedance spectrum information online. Summary of the Invention

[0004] The present invention addresses the shortcomings of the prior art by proposing an integrated "charging-inspection" method for estimating and diagnosing the state of electric vehicle batteries. This method aims to enable battery state estimation and fault diagnosis during the charging and recharging phase of electric vehicles, thereby ensuring safer and more reliable use of electric vehicles.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] The present invention provides an integrated charging-inspection method for detecting and evaluating the state of electric vehicle batteries. This method is characterized by its application in a scenario consisting of a battery management system and a charger. The battery management system includes a control module, a sampling circuit, a calculation module, and a thermal management module. The detection and diagnosis process comprises the following steps:

[0007] Step 1: Extract the capacity increment curve during the constant current charging stage:

[0008] During the initial charging phase of the electric vehicle battery, the control module sends a charging command to the charger. After receiving the charging command, the charger performs constant current charging on the electric vehicle battery pack. When the voltage of the module in the electric vehicle battery pack increases within the range [U1, U2], the sampling circuit collects the voltage, current, and temperature of the module and sends them to the calculation module. The calculation module calculates the capacity increment dQ / dV of the module within the voltage range [U1, U2] based on the collected results, thereby obtaining the capacity increment curve within the voltage range [U1, U2].

[0009] Step 2: Measurement of electrochemical impedance spectroscopy in the low-to-mid frequency range using AC impedance spectroscopy:

[0010] Step 2.1: When the electric vehicle battery pack is charged at a constant current until the state of charge is the set value S0, charging is paused; the control module sends a control command to the charger, so that the charger starts to apply a sinusoidal AC excitation current I to the battery module and uses the sampling circuit to collect the response AC voltage U;

[0011] Step 2.2: The calculation unit calculates the AC impedance Z based on the input sinusoidal AC excitation current I and the acquired response AC voltage U, and calculates the impedance Z using the real part Zi. Re Z is the horizontal axis and the imaginary part. Im Electrochemical impedance spectroscopy was plotted with the vertical axis as the ordinate to obtain the Nyquist curves for the mid-to-low frequency range.

[0012] Step 3: Measuring battery internal resistance using the hybrid pulse power characterization method:

[0013] Step 3.1: When the state of charge of the battery pack is the set value S0, and the battery thermal management controls the temperature of the battery within the temperature range [T1, T2], the control module sends a control command to the charger, so that the charger applies a pulse charging current I to the battery module and uses the sampling circuit to measure the response voltage value U of each module.

[0014] Step 3.2: The calculation unit calculates the ohmic internal resistance R of the battery based on the response voltage value U. ∞ and polarization internal resistance R p Thus, due to the ohmic internal resistance R ∞ and polarization internal resistance R p The sum of these values ​​represents the total internal resistance R of a single battery cell.

[0015] Step 4: State estimation and performance evaluation:

[0016] Step 4.1, Joint State Estimation:

[0017] Step 4.1.1: Normalize the capacity increment within the voltage range [U1, U2], the AC impedance Z when the battery state of charge is the set value S0, and the charging amount Q within the voltage range [U1, U2] into a three-dimensional matrix, and input it into a pruned convolutional neural network model with transfer learning for processing, and output the estimated capacity C of the battery.

[0018] Step 4.1.2: Input the estimated capacity C, the normalized response voltage U' from step 3.1, and the normalized total internal resistance R′ of the battery from step 3.2 into the Gaussian process regression model to obtain the online estimated state of charge of the battery;

[0019] Step 4.1.3: Obtain the charge amount Q` based on the online estimated state of charge of the battery, and use it to update the three-dimensional matrix before inputting it into the pruned convolutional neural network model to obtain the battery health state estimation result;

[0020] Step 4.2, Multi-parameter fusion consistency evaluation:

[0021] Based on the battery module temperature and calculated capacity increment in step one, the AC impedance Z in step two, the total internal resistance R of the battery measured in step three, and the estimated battery health and state of charge in step four, the Pearson correlation coefficient matrix between the characteristics of each module is calculated. The main characteristic parameters of consistency are obtained through principal component analysis. Then, a multi-parameter evaluation method is used to evaluate the consistency of the battery modules. Based on the evaluation results, battery modules with internal short circuit faults and fault risks are screened out to provide repair and replacement prompts.

[0022] The "charging-inspection" integrated electric vehicle battery state estimation and diagnosis method of the present invention is characterized in that: the set value S0 in step 2.1 is 50% state of charge, and the amplitude of the applied sinusoidal AC excitation current I is one-tenth of the nominal capacity, the frequency range is 1-20Hz, and 60 points are sampled every ten signal frequencies.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] 1. In the electric vehicle charging phase, the BMS, in addition to measuring and analyzing conventional parameters such as battery voltage U, current I, and temperature T, also pauses charging when the battery pack is in a specific state of charge (SOC). The charger applies multiple modes of self-test current to each battery module in a short period of time, and collects parameters such as IC curve, electrochemical impedance Z, and ohmic internal resistance R, thereby performing a more comprehensive analysis of battery performance and realizing the function of real-time and accurate estimation of battery status and fault diagnosis.

[0025] 2. The joint state estimation method used in this invention establishes a pruned convolutional neural network model with transfer learning and a Gaussian process regression model. By utilizing the coupling relationship between battery state of health (SOH) and state of charge (SOC), it achieves a high-precision joint estimation of SOH and SOC, thus correcting and supplementing the traditional ampere-hour integral method for estimating SOH and the open-circuit voltage method for estimating SOC.

[0026] 3. The method of the present invention analyzes parameters related to battery safety, such as temperature, voltage and internal resistance, collected during the battery charging stage, and then combines the estimation results of state parameters to perform multi-parameter performance evaluation and consistency evaluation of each module in the battery pack, thereby realizing the functions of fault diagnosis and early warning.

[0027] 4. The method of the present invention achieves non-destructive online detection of the state parameters and safety assessment of the power battery through the coordinated cooperation between the BMS system and the charger. It does not require professional personnel or equipment and site, and does not damage the original battery module structure. It has the advantages of being convenient, efficient and highly feasible. Attached Figure Description

[0028] Figure 1 This is a block diagram illustrating the technical principle structure of the present invention;

[0029] Figure 2 This is a flowchart illustrating the functional implementation of the present invention;

[0030] Figure 3 This is a schematic diagram of the CC-CV battery charging curve shown in this invention;

[0031] Figure 4 This is a schematic diagram of the electrochemical impedance spectroscopy measured in this invention;

[0032] Figure 5 This is a schematic diagram of the HPPC method used in this invention to measure the internal resistance of a battery.

[0033] Figure 6 This is a flowchart illustrating the framework of the joint state estimation method of the present invention.

[0034] Figure 7 This invention relates to a pruned convolutional neural network model with transfer learning. Detailed Implementation

[0035] In this embodiment, a "charge-inspection" integrated electric vehicle battery state detection and evaluation method involves sending different commands to the charger via the BMS system, causing it to apply different current modes to the battery, and then measuring the corresponding response parameters to obtain characteristic parameters of the battery charging stage. Based on these characteristic parameters, the data is input into a pruned convolutional neural network model with transfer learning to estimate the battery's state of equilibrium (SOH). Then, based on the SOH estimation results and other characteristic parameters, the data is input into a Gaussian process regression model to estimate the state of charge (SOC). The method analyzes the inconsistencies in battery cell voltage, internal resistance, temperature, SOH, and SOC parameters to complete the battery fault diagnosis function. Specifically, the technical structure is as follows: Figure 1 As shown, the functional implementation process is as follows: Figure 2 The method is performed according to the following steps:

[0036] Step 1: Extract the capacity increment curve during the constant current charging stage:

[0037] The basic charging method for electric vehicles is the constant current and constant voltage (CC-CV) charging mode, such as... Figure 3 As shown, the charging process includes two stages: constant current charging until the voltage stops increasing and constant voltage charging until the current reaches its minimum value. During the initial charging stage of the electric vehicle battery, the BMS control module sends a charging command to the charger. Upon receiving the command, the charger performs constant current (CC) charging on the electric vehicle battery pack. When the voltage of the modules in the battery pack increases within the range [U1, U2], the sampling circuit collects the voltage, current, and temperature of the modules and sends them to the calculation module. The calculation module then calculates the increment capacity (IC) of the modules within the voltage range [U1, U2] based on the collected data, thus obtaining the capacity increment curve within the voltage range [U1, U2].

[0038]

[0039] In equation (1), t1 and t2 are the times when adjacent currents are collected, and V1 and V2 are the battery voltages at times t1 and t2, respectively.

[0040] Step 2: Measurement of electrochemical impedance spectroscopy in the low-to-mid frequency range using AC impedance spectroscopy:

[0041] Step 2.1: When the electric vehicle battery pack is charged at a constant current until the state of charge is the set value S0, the charging is paused; the control module sends a control command to the charger, so that the charger starts to apply a sinusoidal AC excitation current I to the battery module and uses the sampling circuit to collect the response AC voltage U;

[0042] Step 2.2: The calculation unit calculates the AC impedance Z based on the input sinusoidal AC excitation current I and the acquired response AC voltage U.

[0043]

[0044] Because the excitation current and feedback voltage have different amplitudes and phases, the impedance Z is a complex number:

[0045] Z W =Z Re +jZ Im (3)

[0046] in, Z Re Z Im These represent the real and imaginary parts of the impedance, respectively. Let Z be the real part of the impedance Z. Re Z is the horizontal axis and the imaginary part. Im Plotting an electrochemical impedance spectroscopy diagram with the vertical axis as the ordinate yields the mid-to-low frequency Nyquist curve, as shown below. Figure 4 As shown;

[0047] Step 3: Measure the battery internal resistance using the Hybrid Pulse Power Characterization (HPPC) method.

[0048] Step 3.1: When the battery pack's state of charge is the set value S0, and the battery thermal management controls the battery temperature within the temperature range [T1, T2], the control module sends a control command to the charger, causing the charger to apply a pulse charging current I to the battery module, and uses a sampling circuit to measure the response voltage U of each module. The module's terminal voltage will immediately respond with a step change (U). A ~U B This voltage step change is caused by the battery's ohmic internal resistance; subsequently, the voltage across the battery will undergo a relatively slow change (U). B ~U C This voltage change is caused by the battery's polarization resistance, and the current and voltage change over time during the response process is as follows: Figure 5 As shown;

[0049] Step 3.2: According to Ohm's law, the ohmic internal resistance R of the battery can be calculated. ∞ and polarization internal resistance R p The calculation formulas are shown in equations (4) and (5); the sum of the measured ohmic internal resistance and polarization internal resistance is the total internal resistance R of the battery, and the calculation formula is shown in equation (6):

[0050]

[0051]

[0052] R = R ∞ +R p (6)

[0053] Step 4: State estimation and performance evaluation:

[0054] Step 4.1, Joint State Estimation:

[0055] Joint state estimation includes the estimation of SOC and SOH, and the framework flowchart is as follows. Figure 6 As shown, the specific details are as follows:

[0056] Step 4.1.1: Normalize the capacity increment IC in the voltage range [U1,U2], the AC impedance Z when the battery state of charge is the set value S0, and the charging amount Q in the voltage range [U1,U2] into a three-dimensional matrix, and input it into a pruned convolutional neural network model with transfer learning for processing, and output the estimated capacity C of the battery.

[0057] Step 4.1.2: Input the estimated capacity C, the normalized response voltage U' from step 3.1, and the normalized total internal resistance R′ from step 3.2 into the Gaussian process regression model to obtain the online estimated battery SOC;

[0058] Step 4.1.3: Obtain the charge quantity Q` based on the online estimated SOC, and use it to update the three-dimensional matrix before inputting it into the trimmed convolutional neural network model to obtain the SOH estimation result of the battery;

[0059] The structure of the pruned convolutional neural network model with transfer learning is shown below. Figure 7 The method for establishing this is as follows:

[0060] First, a CNN model is constructed, consisting of two sets of convolutional layers and a max-pooling layer, followed by two consecutive convolutional layers, a flattening layer, and finally two fully connected layers. Second, transfer learning techniques are applied to the CNN to reduce the required dataset size by leveraging knowledge learned from source tasks with large datasets, thus enabling the application of different but related tasks on smaller datasets to different but related tasks on smaller datasets. The CNN model is first pre-trained on a large source dataset collected from the same type of power batteries, with each battery undergoing approximately 1000 charge-discharge cycles. The learned model structure and training parameters are then transferred to a relatively small target dataset collected from real vehicles, with each unit tested for 30 reference cycles. The two convolutional layers and two fully connected layers are then fine-tuned on the target dataset to ensure model performance. Finally, a network pruning technique based on a fast recursive algorithm is used in the two fully connected layers to remove redundant neurons, significantly reducing model size and computational complexity, allowing the resulting model to be implemented in an onboard BMS.

[0061] Step 4.2, Multi-parameter fusion consistency evaluation:

[0062] 4.2.1 First, the temperature and calculated capacity increment of the battery module in step one, the AC impedance Z in step two, the total internal resistance R of the battery measured in step three, and the estimated SOH and SOC of the battery in step four are standardized, and the Pearson correlation coefficient matrix P between each module for a certain characteristic is calculated:

[0063]

[0064] In equation (7), m is the number of different battery modules, and n is the number of times the feature parameter is sampled in each module. The correlation coefficient is calculated as follows:

[0065]

[0066] In equation (8), r is the correlation coefficient between module x and module y for the same discrete feature variable, and its value ranges from [-1, 1]. cov(x, y) is the covariance between different batteries, and δx and δy are the variances of x and y, respectively. The mean of variable x;

[0067] 4.2.2 Calculate the eigenvalues ​​and eigenvectors using |λI-P|=0, and sort them by size: λ1≥λ2≥…λ n ≥0. The eigenvalues ​​corresponding to the eigenvectors satisfy the condition: |e ij |=1 and

[0068] 4.2.3 Calculate the contribution rate and cumulative contribution rate:

[0069] Calculate the contribution rate using equation (9):

[0070]

[0071] Calculate the cumulative contribution rate using equation (10):

[0072]

[0073] 4.2.4 Calculate the principal component loadings using equation (11):

[0074]

[0075] 4.2.5 Calculate the principal component scores using equation (12):

[0076] z i =l i1 x i1 +l i2 x i2 +…+l in x in (12)

[0077] 4.2.6 Principal component analysis is used to obtain the principal component scores of different characteristic parameters of each battery module; then, a multi-parameter evaluation method is used to evaluate the consistency of the battery modules; based on the evaluation results, battery modules with internal short circuit faults and fault risks can be screened out, and repair and replacement prompts can be made.

Claims

1. A method for integrated charging-inspection of electric vehicle battery status detection and evaluation, characterized in that, This is applied in scenarios consisting of a battery management system and a charger. The battery management system includes: a control module, a sampling circuit, a computing module, and a thermal management module. The detection and diagnosis include the following steps: Step 1: Extract the capacity increment curve during the constant current charging stage: During the initial charging phase of the electric vehicle battery, the control module sends a charging command to the charger. Upon receiving the charging command, the charger performs constant current charging on the electric vehicle battery pack. When the voltage of the modules in the electric vehicle battery pack is within the range […], , During the growth within the voltage range, the sampling circuit collects the voltage, current, and temperature of the module and sends them to the calculation module. The calculation module then calculates the voltage range of the module based on the collected data. , Capacity increment within ] Thus, the voltage range is obtained. , The capacity increment curve within ]; Step 2: Measurement of electrochemical impedance spectroscopy in the low-to-mid frequency range using AC impedance spectroscopy: Step 2.1: When the electric vehicle battery pack is charged at a constant current until the state of charge reaches the set value S0, charging is paused; the control module sends a control command to the charger, causing the charger to start applying a sinusoidal AC excitation current to the battery module. The sampling circuit is used to acquire the response AC voltage. ; Step 2.2: The calculation module calculates based on the input sinusoidal AC excitation current. and the collected response AC voltage Calculate AC impedance and respectively with impedance real part x-axis, imaginary part Electrochemical impedance spectroscopy was plotted with the vertical axis as the ordinate to obtain the Nyquist curves for the mid-to-low frequency range. Step 3: Measuring battery internal resistance using the hybrid pulse power characterization method: Step 3.1: When the battery pack's state of charge is the set value S0, and the battery thermal management controls the battery temperature within the specified temperature range […]. , During this time, the control module sends a control command to the charger, causing the charger to apply a pulse charging current I to the battery module and use the sampling circuit to measure the response voltage value U of each module; Step 3.2: The calculation module calculates the ohmic internal resistance of the battery based on the response voltage value U. and polarization resistance Thus, due to the ohmic internal resistance and polarization resistance The sum of these values ​​represents the total internal resistance of a single battery cell. ; Step 4: State estimation and performance evaluation: Step 4.1, Joint State Estimation: Step 4.1.1: Set the voltage range [ , The capacity increment within [ ], the AC impedance Z when the battery state of charge is the set value S0, and the voltage range [ ] , The charge amount Q within the battery is normalized into a three-dimensional matrix and then input into a pruned convolutional neural network model with transfer learning for processing, outputting the estimated battery capacity C. Step 4.1.2: Combine the estimated capacity C with the normalized response voltage value U' from Step 3.1 and the normalized total internal resistance of the battery from Step 3.

2. Inputting the data into a Gaussian process regression model yields an online estimate of the battery's state of charge. Step 4.1.3: Obtain the charge amount Q` based on the online estimated state of charge of the battery, and use it to update the three-dimensional matrix before inputting it into the pruned convolutional neural network model to obtain the battery health state estimation result; Step 4.2, Multi-parameter fusion consistency evaluation: Based on the battery module temperature and calculated capacity increment in step one, and the AC impedance in step two... Step 3: Measure the total internal resistance of the battery Step 4 estimates the battery health status and state of charge, calculates the Pearson correlation coefficient matrix between the characteristics of each module, and obtains the main characteristic parameters of consistency through principal component analysis. Then, a multi-parameter evaluation method is used to evaluate the consistency of the battery modules. Based on the evaluation results, battery modules with internal short circuit faults and fault risks are screened out to provide repair and replacement prompts.

2. The "charging-inspection" integrated electric vehicle battery state detection and evaluation method according to claim 1, characterized in that: In step 2.1, the set value S0 is 50% of the state of charge, and the applied sinusoidal AC excitation current is... The amplitude is one-tenth of the nominal capacity, the frequency range is 1-20Hz, and 60 points are sampled every ten signal frequencies.

Images