Electrochemical model online parameter identification method and device

Abstract

The application provides an online parameter identification method of an electrochemical model, comprising: simulating historical electrochemical model parameters of a battery by an electrochemical model to obtain a historical discharge capacity estimation value of the battery; adjusting the historical electrochemical model parameters to generate a plurality of electrochemical model parameter sets and obtain corresponding discharge capacity estimation values through simulation; simulating the electrochemical model parameters under different working conditions by the electrochemical model to obtain corresponding voltage and current data; taking the voltage and current data as a training set, and taking the electrochemical model parameter sets and the corresponding discharge capacity estimation values as labels to train the electrochemical model; identifying battery data to be identified by the trained electrochemical model to output initial electrochemical model parameters and an initial discharge capacity estimation value corresponding to the battery data to be identified; fitting a comprehensive loss of the initial discharge capacity estimation value and the voltage, and identifying parameters through a heuristic algorithm to obtain target electrochemical model parameters.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and more particularly to a method and apparatus for online parameter identification using an electrochemical model. Background Technology

[0002] Due to the instability of new energy systems, energy storage systems are needed, and lithium-ion batteries have been widely used. To ensure the safety and reliability of lithium-ion batteries during long-term use, a battery management system (BMS) composed of hardware and software is required to manage them. Currently, widely used BMSs are developed based on equivalent circuit models (ECMs). Due to the limited predictive power of ECMs, battery operation strategies are designed based on simple safety constraints, such as charging cutoff voltage, discharging cutoff voltage, and maximum current.

[0003] However, terminal voltage cannot fully reflect the internal state of the battery, especially under high current conditions. Due to the large overpotential, this can significantly increase or decrease the battery's terminal voltage during charging and discharging. With the improvement of hardware computing power, newer, more intelligent and advanced BMS based on electrochemical models (EM) will soon be applied. Because EM can fully reflect the internal state of the battery, such as the lithium-ion concentration distribution at the positive and negative electrodes, potential distribution, and overpotential, it can greatly improve the management capability of lithium batteries. Electrochemical models involve a large number of coupled partial differential equations, especially dozens of physical parameters, which limits the practical application of EM models. With the improvement of hardware capabilities, model parameters can be obtained in a data-driven manner through methods such as heuristic algorithms (genetic algorithms, particle swarm optimization, cuckoo algorithm, etc.), neural networks, and Kalman filtering.

[0004] These methods typically divide battery data into constant current and dynamic operating conditions, and perform step-by-step identification. They also require that the initial SOC (state of charge) of the positive and negative electrode materials be consistent across different operating conditions. However, in practical applications, it is almost impossible for batteries to operate at a constant current continuously, and the initial state of the battery often varies significantly depending on the different operating conditions obtained after cleaning. Therefore, this method has significant limitations. Summary of the Invention

[0005] The purpose of this invention is to provide an online parameter identification method for electrochemical models, which can be used to solve the above-mentioned problems.

[0006] The technical solution provided by this invention is as follows:

[0007] An online parameter identification method for electrochemical models includes:

[0008] The historical electrochemical model parameters of the battery are simulated using an electrochemical model to obtain an estimated value of the battery's historical discharge capacity.

[0009] Adjust the parameters of the historical electrochemical model to generate multiple electrochemical model parameter sets and simulate to obtain the corresponding discharge capacity estimates;

[0010] The electrochemical model parameters were simulated under different operating conditions to obtain the corresponding voltage and current data.

[0011] The electrochemical model is trained using the voltage and current data as the training set and the electrochemical model parameter set and the corresponding discharge capacity estimate as labels.

[0012] The trained electrochemical model identifies the battery data to be identified and outputs the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data to be identified.

[0013] The combined loss of the initial discharge capacity estimate and voltage is fitted, and the parameters are identified using a heuristic algorithm to obtain the target electrochemical model parameters.

[0014] In some embodiments, before simulating the electrochemical model parameters of the battery using an electrochemical model to obtain the discharge capacity distribution corresponding to the electrochemical model parameters, the method further includes:

[0015] Clean the historical data of the battery used for parameter identification; the historical data used for parameter identification includes data of the battery being stationary for a preset time to ensure that the battery is in a balanced state when it starts to work, and the SOC change range exceeding a preset range threshold.

[0016] The historical electrochemical model parameters of the battery are obtained by using the electrochemical model to identify parameters from the historical data used for parameter identification after cleaning.

[0017] In some embodiments, the step of simulating the historical electrochemical model parameters of the battery using an electrochemical model to obtain an estimated historical discharge capacity of the battery includes:

[0018] The historical electrochemical model parameters are charged to the charging cutoff voltage under a constant current state at a first preset current, and then discharged under a constant voltage state until the current is less than the second preset current; the first preset current is greater than the second preset current.

[0019] The historical electrochemical model parameters are simulated under the first preset current using the electrochemical model until the voltage drops to the discharge cutoff voltage, and the estimated historical discharge capacity of the battery is output.

[0020] In some implementations, the overall voltage loss includes:

[0021] Loss = α*Loss_Q + β*Loss_V;

[0022] Loss_Q = |Q_iden - Q_model|;

[0023]

[0024] Where loss is the overall voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; Loss_V is the voltage loss; Q_iden is the capacity corresponding to the identified parameter set; Q_model is the capacity corresponding to the identified parameter set; MSE is the root mean square error; N is the total number of data points; i is the current; v cell The voltage of the battery; v sim This is the simulation voltage for the electrochemical model.

[0025] In some implementations, the method further includes selecting a loss function for training the electrochemical model based on the voltage fitting results.

[0026] In some embodiments, an online parameter identification device for an electrochemical model includes:

[0027] The simulation module is used to simulate the historical electrochemical model parameters of the battery through an electrochemical model, and obtain the estimated historical discharge capacity of the battery.

[0028] The filtering module is used to adjust the parameters of the historical electrochemical model, generate multiple sets of electrochemical model parameters, and simulate to obtain the corresponding discharge capacity estimates.

[0029] The simulation module is also used to simulate the electrochemical model parameters under different operating conditions using the electrochemical model to obtain corresponding voltage and current data;

[0030] The training module is used to train the electrochemical model using the voltage and current data as the training set and the electrochemical model parameter set and the corresponding discharge capacity estimate as labels.

[0031] The output module is used to identify the battery data to be identified through the trained electrochemical model, and output the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data to be identified.

[0032] The identification module is used to fit the combined loss of the initial discharge capacity estimate and voltage, and to identify the parameters through a heuristic algorithm to obtain the target electrochemical model parameters.

[0033] In some implementations, it further includes: a cleaning module, used for:

[0034] Clean the historical data of the battery used for parameter identification; the historical data used for parameter identification includes data of the battery being stationary for a preset time to ensure that the battery is in a balanced state when it starts to work, and the SOC change range exceeding a preset range threshold.

[0035] The historical electrochemical model parameters of the battery are obtained by using the electrochemical model to identify parameters from the historical data used for parameter identification after cleaning.

[0036] In some implementations, the simulation module is further used for:

[0037] The historical electrochemical model parameters are charged to the charging cutoff voltage under a constant current state at a first preset current, and then discharged under a constant voltage state until the current is less than the second preset current; the first preset current is greater than the second preset current.

[0038] The historical electrochemical model parameters are simulated under the first preset current using the electrochemical model until the voltage drops to the discharge cutoff voltage, and the estimated historical discharge capacity of the battery is output.

[0039] In some implementations, the overall voltage loss includes:

[0040] Loss = α*Loss_Q + β*Loss_V;

[0041] Loss_Q = |Q_iden - Q_model|;

[0042]

[0043] Where loss is the overall voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; Loss_V is the voltage loss; Q_iden is the capacity corresponding to the identified parameter set; Q_model is the capacity corresponding to the identified parameter set; MSE is the root mean square error; N is the total number of data points; i is the current; v cell The voltage of the battery; v sim This is the simulation voltage for the electrochemical model.

[0044] In some implementations, it also includes:

[0045] Based on the voltage fitting results, the loss function for training the electrochemical model is selected.

[0046] Compared with existing technologies, the online parameter identification method for electrochemical models provided by this invention can bring the following beneficial effects:

[0047] This invention provides a method for online electrochemical parameter identification. First, parameters are initially identified by analyzing battery data. Then, the dataset is simulated by an electrochemical model by adjusting sensitive parameters, and a neural network is trained. During online identification, the neural network first outputs the initial values ​​of capacity and identified parameters. The parameters are further identified by the combined loss of capacity and fitted voltage, which solves the problem of weak generalization ability for parameters under limited operating conditions. Attached Figure Description

[0048] The preferred embodiments will be described below in a clear and easy-to-understand manner, with reference to the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of the online parameter identification method for electrochemical models.

[0049] Figure 1 This is a flowchart of an embodiment of the online parameter identification method for electrochemical models of the present invention;

[0050] Figure 2 This is a histogram showing the distribution of a certain parameter in 1000 identifications of this invention;

[0051] Figure 3 This is the capacity distribution diagram obtained from 100 parameter identifications of the present invention;

[0052] Figure 4 This is the voltage diagram obtained by fitting the parameter set selected based on the discharge capacity according to the present invention. Detailed Implementation

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.

[0054] To keep the drawings concise, only the parts relevant to the invention are shown schematically in each figure, and they do not represent the actual structure of the product. Furthermore, for ease of understanding, in some figures, only one of components with the same structure or function is shown schematically, or only one is labeled. In this document, "one" can mean not only "only one" but also "more than one".

[0055] In one embodiment, such as Figure 1 As shown, this invention provides a method for online parameter identification of an electrochemical model, comprising:

[0056] S101 uses an electrochemical model to simulate the historical electrochemical model parameters of the battery, thereby obtaining an estimate of the battery's historical discharge capacity.

[0057] In this embodiment, the historical electrochemical model parameters include high-sensitivity parameters and low-sensitivity parameters.

[0058] S102 Adjust the parameters of the historical electrochemical model, generate multiple electrochemical model parameter sets, and simulate to obtain the corresponding discharge capacity estimate.

[0059] Specifically, a new set of electrochemical model parameters is obtained by adjusting the values ​​of highly sensitive reading parameters.

[0060] The electrochemical model of a lithium battery involves dozens of physical parameters. Under constant current conditions, only some of these parameters affect the output voltage of the model. These parameters are defined as high-sensitivity parameters, while the remaining parameters that affect the output voltage of the model under high-rate or dynamic conditions are defined as low-sensitivity parameters.

[0061] For example, such as Figure 2 The battery discharge capacity distribution diagram shown illustrates that when there is limited battery data available for parameter identification, the identified parameters are unstable.

[0062] S103 simulates the electrochemical model parameters under different operating conditions using the electrochemical model to obtain the corresponding voltage and current data.

[0063] Therefore, this invention generates multiple electrochemical model parameter sets by adjusting the parameters of each electrochemical model within a reasonable range, thus generating the capacity of each set. Furthermore, it extracts data (U, I, t) from different power plant operating conditions (i.e., current I and time t data) and performs simulations on each set of electrochemical model parameters to obtain data. Here, U represents voltage.

[0064] S104 uses the voltage and current data as the training set and the electrochemical model as the label, and trains the electrochemical model with the electrochemical model parameter set and the corresponding discharge capacity estimate.

[0065] Using the aforementioned U, I, and t data as the training set, and the electrochemical model parameter set and its corresponding capacity as labels, a neural network is trained.

[0066] S105 identifies the battery data to be identified using the trained electrochemical model and outputs the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data to be identified.

[0067] Since the electrochemical model has been trained, the battery data can be identified online through the electrochemical model to obtain the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data.

[0068] S106 fits the combined loss of the initial discharge capacity estimate and voltage, and uses a heuristic algorithm to identify parameters to obtain the target electrochemical model parameters.

[0069] The combined loss of the initial discharge capacity estimate and voltage mentioned here refers to the combined loss of the initial discharge estimate and voltage. The specific fitting process is as follows: the parameter set and current are input into the electrochemical model, and the electrochemical model will simulate the voltage.

[0070] Among them, heuristic algorithms include genetic algorithms, particle swarm optimization, and cuckoo algorithms.

[0071] This invention provides a method for online electrochemical parameter identification. First, parameters are initially identified by analyzing battery data. Then, the sensitive parameters are adjusted to simulate the dataset using an electrochemical model, and a neural network is trained. During online identification, the neural network outputs the initial values ​​of capacity and identified parameters. The parameters are further identified by the combined loss of capacity and fitted voltage. The problem of weak generalization ability for parameters under limited operating conditions is solved by increasing the capacity loss.

[0072] In this embodiment, since the actual capacity of the battery cell to be identified is unknown, parameters are first identified from the actual battery cell data. Then, highly sensitive parameters are adjusted to generate a large amount of parameter data, which is used to train the neural network. The neural network is then used to estimate the capacity value from the data to be identified.

[0073] Because there are many parameters in the electrochemical model of a battery, in order to ensure the accuracy of parameter identification and the generalization ability of the parameter set, traditional parameter identification methods based on experimental data need to obtain full charge and discharge data of the battery under different operating conditions, such as data from SOC=100% to SOC0% under 1C conditions, and data from SOC=100% to SOC=0% under dynamic conditions.

[0074] When batteries are actually working in a power station, it is difficult to obtain data on the change from SOC=100% to SOC=0% under constant current conditions. Moreover, the concentrations of the positive and negative electrodes in the electrochemical model are also important parameters directly related to the battery. Furthermore, since the calculation of SOC in a power station is always subject to error, it is difficult to establish a direct correlation between the two data points regarding concentration.

[0075] Therefore, before simulating the electrochemical model parameters of the battery using an electrochemical model to obtain the discharge capacity distribution corresponding to the electrochemical model parameters, the present invention further includes:

[0076] Clean the historical data of the battery used for parameter identification; the historical data used for parameter identification includes data of the battery being stationary for a preset time to ensure that the battery is in a balanced state when it starts to work, and the SOC change range exceeding a preset range threshold.

[0077] The historical electrochemical model parameters of the battery are obtained by using the electrochemical model to identify parameters from the historical data used for parameter identification after cleaning.

[0078] In practical applications, a data point for parameter identification is extracted from each battery cell in the power plant. This data point needs to meet the following requirements:

[0079] 1.1 Before starting work, the battery cell should be left to stand still for a sufficient period of time (generally more than 1 hour) to ensure that the internal structure of the battery cell is in a balanced state.

[0080] 1.2 It should contain a sufficiently long data set, and the range of SOC variation should be sufficiently large, for example, it should cover more than 70% of the SOC fluctuations.

[0081] SOC, or State of Charge, refers to the battery's state of charge. The state of charge of a lithium battery is a crucial parameter of the battery management system and serves as the basis for the vehicle's charge and discharge control strategy and battery balancing. SOC can be defined in various ways from different perspectives, such as by charge capacity or energy level. The SOC defined by the US Advanced Battery Association (USABC) is widely adopted: it is the ratio of the battery's remaining charge to its rated capacity under the same conditions at a given discharge rate. Operating condition refers to the battery's working state.

[0082] In this embodiment, by cleaning the battery data, the accuracy of parameter identification and the generalization ability of the parameter set are further ensured.

[0083] In some embodiments, the step of simulating the historical electrochemical model parameters of the battery using an electrochemical model to obtain an estimated historical discharge capacity of the battery includes:

[0084] The historical electrochemical model parameters are charged to the charging cutoff voltage under a constant current state at a first preset current, and then discharged under a constant voltage state until the current is less than the second preset current; the first preset current is greater than the second preset current.

[0085] The historical electrochemical model parameters are simulated under the first preset current using the electrochemical model until the voltage drops to the discharge cutoff voltage, and the estimated historical discharge capacity of the battery is output.

[0086] The first preset current can be 1C, and the second preset current can be 0.05C. It should be noted that the charging and discharging current during the simulation does not necessarily have to be 1C, but it is necessary to first charge to the cutoff voltage and then discharge to the discharge cutoff voltage.

[0087] After the parameters are identified, each set of parameters is charged at a current of 1C, first in the CC mode to the charging cutoff voltage, and then in the CV mode until the current is less than 0.05C.

[0088] Then, the cell capacity is obtained by simulating the process using an electrochemical model at a 1C current (or according to the capacity measurement method in the cell datasheet) until the voltage drops to the discharge cutoff voltage.

[0089] The capacity calculation formula is as follows:

[0090] The discharge capacity Q of each set of parameters was calculated using the ampere-hour integration method. i It conforms to the following formula:

[0091] Q i =∫Idt

[0092] Where I is the current in the discharge simulation, t is the time of the discharge simulation process, and i represents the i-th set of parameters.

[0093] Discharge capacity distribution is Figure 3 This indicates that when there is limited battery data available for parameter identification, the identified parameters are unstable.

[0094] In some implementations, the overall voltage loss includes:

[0095] Loss = α*Loss_Q + β*Loss_V;

[0096] Loss_Q = |Q_iden - Q_model|;

[0097]

[0098] Where loss is the overall voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; Loss_V is the voltage loss; Q_iden is the capacity corresponding to the identified parameter set; Q_model is the capacity corresponding to the identified parameter set; MSE is the root mean square error; N is the total number of data points; i is the current; v cell The voltage of the battery; v sim This is the simulation voltage for the electrochemical model.

[0099] Where α and β are coefficients, for example, they can be set to 0.2 and 0.8.

[0100] In some implementations, the method further includes selecting a loss function for training the electrochemical model based on the voltage fitting results.

[0101] Currently, data-driven parameter identification methods primarily use loss functions to measure the simulation results of the model.

[0102] The voltage value is compared with the actual voltage value of the battery to optimize the electrochemical parameters of the model. The loss functions mainly include root mean square error (MSE) and mean absolute error (MAE).

[0103] The loss function for training the electrochemical model is:

[0104]

[0105] Where N is the number of the first type of sensitivity parameters; v cell This is the actual voltage of the lithium battery; v sim These are the parameters of the electrochemical model; i is the parameter number.

[0106] For example, the specific implementation method of the present invention is as follows:

[0107] Step 1: Clean the battery data to obtain the actual data used for parameter identification.

[0108] Because there are many parameters in the electrochemical model of a battery, in order to ensure the accuracy of parameter identification and the generalization ability of the parameter set, traditional parameter identification methods based on experimental data need to obtain full charge and discharge data of the battery under different operating conditions, such as data from SOC=100% to SOC=0% under 1C conditions, and data from SOC=100% to SOC=0% under dynamic conditions.

[0109] When batteries are actually operating in a power station, it's difficult to obtain data on the state of charge (SOC) from 100% to 0% under constant current conditions. Furthermore, the concentrations of the positive and negative electrodes in the electrochemical model are crucial parameters directly related to the battery. Since SOC calculations in power stations often contain errors, it's difficult to establish a direct correlation between the concentrations of the two data points. Therefore, in practical applications, only one data point for parameter identification can be obtained from each cell in the power station. This data point needs to meet the following requirements:

[0110] 1.1 Before starting work, the battery cell should be left to stand still for a sufficient period of time (generally more than 1 hour) to ensure that the internal structure of the battery cell is in a balanced state.

[0111] 1.2 It should contain a sufficiently long data set, and the range of SOC variation should be sufficiently large, for example, it should cover more than 70% of SOC fluctuations.

[0112] Step 2: Identify parameters using a data-driven approach.

[0113] The cleaned data is used as the data for parameter identification, and the loss function is the difference between the actual operating voltage and the model output voltage:

[0114] The specific loss function can be selected based on the voltage fitting effect, choosing MSE.

[0115] The formula for calculating MSE is as follows:

[0116]

[0117] The data for this battery cell were subjected to multiple parameter identifications using heuristic algorithms, typically no fewer than 100 times. Due to the randomness of heuristic algorithms and the limited amount of data available for parameter identification, the parameters exhibit significant fluctuations. Taking a lithium iron phosphate battery as an example, the distribution of one of the identified parameters is as follows: Figure 2 .

[0118] Step 3: Simulate the parameters obtained from each identification using an electrochemical model to obtain the discharge capacity distribution corresponding to each set of parameters. This discharge capacity distribution is used to generate the dataset.

[0119] After the parameters are identified, each set of parameters is charged at 1C current, first in CC mode to the charging cutoff voltage, and then in CV mode until the current is less than 0.05C.

[0120] Then, the cell capacity is obtained by simulating the process using an electrochemical model at a 1C current (or according to the capacity measurement method in the cell datasheet) until the voltage drops to the discharge cutoff voltage.

[0121] The formula for calculating discharge capacity is as follows:

[0122] The discharge capacity Q of each set of parameters was calculated using the ampere-hour integration method. i It conforms to the following formula:

[0123] Q i =∫Idt

[0124] I represents the current in the discharge simulation, t represents the time of the discharge simulation process, and i represents the i-th set of parameters.

[0125] It should be noted that the charging and discharging current during simulation does not necessarily have to be 1C current, but it is necessary to first charge to the cutoff voltage and then discharge to the discharge cutoff voltage.

[0126] The battery discharge capacity distribution is as follows: Figure 3 This indicates that when there is limited battery data available for parameter identification, the identified parameters are unstable.

[0127] Step 4: Generate a battery dataset by adjusting the electrochemical model parameters.

[0128] By adjusting the parameters of each electrochemical model within a reasonable range, multiple electrochemical model parameter sets are generated. The capacity of each electrochemical model parameter set is then generated according to step 3. For example, such as... Figure 4 The voltage diagram shown is obtained by fitting the parameter set selected based on the capacity (the voltage fitting effect diagram obtained by identification).

[0129] The system extracts data for different operating conditions of the power plant (i.e., current I and time t), and performs simulations on each set of electrochemical model parameters to obtain data (U, I, t). Here, U represents voltage.

[0130] Using the aforementioned U, I, and t data as the training set, and the electrochemical model parameter set and its corresponding capacity as labels, a neural network is trained.

[0131] The training of a parameter identification neural network includes the following steps: Voltage, current, and other data from the training dataset are used as feature inputs to train the neural network. High-sensitivity parameters and discharge capacity estimates are used as labels. A suitable deep learning method is selected to train the neural network, where the neural network can be CNN, LSTM, etc.

[0132] C refers to the charge / discharge rate of the battery.

[0133] Step 5: Identify cell parameters.

[0134] During online identification, the data to be identified is first cleaned according to step 1. Then, the initial values ​​of capacity and identification parameters are output through a neural network. The identification parameters are further identified through a heuristic algorithm by using the combined loss of capacity and fitted voltage.

[0135] The overall loss is as follows:

[0136] Loss = α*Loss_Q + β*Loss_V;

[0137] Loss_Q = |Q_iden - Q_model|;

[0138]

[0139] Where loss is the overall voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; Loss_V is the voltage loss; Q_iden is the capacity corresponding to the identified parameter set; Q_model is the capacity corresponding to the identified parameter set; MSE is the root mean square error; N is the total number of data points; i is the current; v cell The voltage of the battery; v sim This is the simulation voltage for the electrochemical model.

[0140] In some embodiments, an online parameter identification device for an electrochemical model includes:

[0141] The simulation module is used to simulate the historical electrochemical model parameters of the battery through an electrochemical model, and obtain the estimated historical discharge capacity of the battery.

[0142] The filtering module is used to adjust the parameters of the historical electrochemical model, generate multiple sets of electrochemical model parameters, and simulate to obtain the corresponding discharge capacity estimates.

[0143] The simulation module is also used to simulate the electrochemical model parameters under different operating conditions using the electrochemical model to obtain corresponding voltage and current data;

[0144] The training module is used to train the electrochemical model using the voltage and current data as the training set and the electrochemical model parameter set and the corresponding discharge capacity estimate as labels.

[0145] The output module is used to identify the battery data to be identified through the trained electrochemical model, and output the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data to be identified.

[0146] The identification module is used to fit the combined loss of the initial discharge capacity estimate and voltage, and to identify the parameters through a heuristic algorithm to obtain the target electrochemical model parameters.

[0147] In some implementations, it further includes: a cleaning module, used for:

[0148] Clean the historical data of the battery used for parameter identification; the historical data used for parameter identification includes data of the battery being stationary for a preset time to ensure that the battery is in a balanced state when it starts to work, and the SOC change range exceeding a preset range threshold.

[0149] The historical electrochemical model parameters of the battery are obtained by using the electrochemical model to identify parameters from the historical data used for parameter identification after cleaning.

[0150] In some implementations, the simulation module is further used for:

[0151] The historical electrochemical model parameters are charged to the charging cutoff voltage under a constant current state at a first preset current, and then discharged under a constant voltage state until the current is less than the second preset current; the first preset current is greater than the second preset current.

[0152] The historical electrochemical model parameters are simulated under the first preset current using the electrochemical model until the voltage drops to the discharge cutoff voltage, and the estimated historical discharge capacity of the battery is output.

[0153] In some implementations, the overall voltage loss includes:

[0154] Loss = α*Loss_Q + β*Loss_V;

[0155] Loss_Q = |Q_iden - Q_model|;

[0156]

[0157] Where loss is the overall voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; Loss_V is the voltage loss; Q_iden is the capacity corresponding to the identified parameter set; Q_model is the capacity corresponding to the identified parameter set; MSE is the root mean square error; N is the total number of data points; i is the current; v cell The voltage of the battery; v sim This is the simulation voltage for the electrochemical model.

[0158] In some implementations, it also includes:

[0159] Based on the voltage fitting results, the loss function for training the electrochemical model is selected.

[0160] It should be noted that the above embodiments can be freely combined as needed. The above description is only a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for online parameter identification of an electrochemical model, characterized in that, include: The historical electrochemical model parameters of the battery are simulated using an electrochemical model to obtain an estimated value of the battery's historical discharge capacity. Adjust the parameters of the historical electrochemical model to generate multiple electrochemical model parameter sets and simulate to obtain the corresponding discharge capacity estimates; The electrochemical model parameters were simulated under different operating conditions to obtain the corresponding voltage and current data. The electrochemical model is trained using the voltage and current data as the training set and the electrochemical model parameter set and the corresponding discharge capacity estimate as labels. The trained electrochemical model identifies the battery data to be identified and outputs the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data to be identified. The initial estimated value of the initial discharge capacity and the combined loss of voltage are fitted, and the parameters are identified by a heuristic algorithm to obtain the target electrochemical model parameters; Before simulating the historical electrochemical model parameters of the battery using an electrochemical model to obtain the estimated historical discharge capacity of the battery, the process further includes: cleaning the historical data of the battery used for parameter identification; the historical data used for parameter identification includes data where the battery has been stationary for a preset time to ensure that the battery is in a balanced state when it starts working, and where the SOC change range exceeds a preset range threshold; and using the electrochemical model to identify the parameters of the cleaned historical data used for parameter identification to obtain the historical electrochemical model parameters of the battery. The step of simulating the historical electrochemical model parameters of the battery using an electrochemical model to obtain an estimated historical discharge capacity of the battery includes: charging the historical electrochemical model parameters at a first preset current under constant current conditions to the charging cutoff voltage, and discharging them under constant voltage conditions until the current is less than a second preset current; the first preset current is greater than the second preset current; simulating the historical electrochemical model parameters using the electrochemical model at the first preset current until the voltage drops to the discharge cutoff voltage, and outputting an estimated historical discharge capacity of the battery. The overall voltage loss includes: ； ； ； Where loss is the total voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; and Loss_V is the voltage loss. The corresponding capacity of the identified parameter set; The capacity predicted by the trained model; is the root mean square error; N is the total number of data points; i is the current; This refers to the battery voltage. This is the simulation voltage for the electrochemical model.

2. The method for online parameter identification of an electrochemical model according to claim 1, characterized in that, Also includes: Based on the voltage fitting results, the loss function for training the electrochemical model is selected.

3. An online parameter identification device for an electrochemical model, characterized in that, include: The simulation module is used to simulate the historical electrochemical model parameters of the battery through an electrochemical model, and obtain the estimated historical discharge capacity of the battery. The filtering module is used to adjust the parameters of the historical electrochemical model, generate multiple sets of electrochemical model parameters, and simulate to obtain the corresponding discharge capacity estimates. The simulation module is also used to simulate the electrochemical model parameters under different operating conditions using the electrochemical model to obtain corresponding voltage and current data; The training module is used to train the electrochemical model using the voltage and current data as the training set and the electrochemical model parameter set and the corresponding discharge capacity estimate as labels. The output module is used to identify the battery data to be identified through the trained electrochemical model, and output the initial electrochemical model parameters and the initial estimated value of the initial discharge capacity corresponding to the battery data to be identified. The identification module is used to fit the initial estimated value of the initial discharge capacity and the combined loss of voltage, and to identify parameters through a heuristic algorithm to obtain the target electrochemical model parameters. The cleaning module is used to: clean the historical data of the battery used for parameter identification; the historical data used for parameter identification includes data on the battery being stationary for a preset time to ensure that the battery is in a balanced state when it starts working, and the SOC change range exceeding a preset range threshold; and to perform parameter identification on the cleaned historical data of the battery using the electrochemical model to obtain the historical electrochemical model parameters of the battery. The simulation module is also used to: charge the historical electrochemical model parameters under a first preset current in a constant current state to the charging cutoff voltage, and discharge them under a constant voltage state until the current is less than the second preset current. The first preset current is greater than the second preset current; The historical electrochemical model parameters are simulated under the first preset current using the electrochemical model until the voltage drops to the discharge cutoff voltage, and the estimated historical discharge capacity of the battery is output. The overall voltage loss includes: ； ； ； Where loss is the total voltage loss, α is a coefficient; Loss_Q is the capacity loss; β is a coefficient; and Loss_V is the voltage loss. The corresponding capacity of the identified parameter set; The capacity predicted by the trained model; is the root mean square error; N is the total number of data points; i is the current; This refers to the battery voltage. This is the simulation voltage for the electrochemical model.

4. The online parameter identification device for electrochemical models according to claim 3, characterized in that, Also includes: Based on the voltage fitting results, the loss function for training the electrochemical model is selected.

Images