Lithium battery SOC estimation method and system cooperating with edge end and cloud end

Abstract

The invention discloses a lithium battery SOC estimation method and system cooperating with an edge end and a cloud end, and belongs to the technical field of lithium ion batteries, and the method comprises the steps: obtaining the original data of a lithium battery under a charging and discharging test condition; establishing a second-order equivalent circuit model of the lithium battery, and performing identification and prediction of battery characteristic parameters in the model at an edge end and a cloud end according to original data; on the basis of two groups of battery characteristic parameter values obtained by the edge end and the cloud end, performing joint correction by taking the SOH value as a reference; and according to the corrected battery characteristic parameter value, combining with the reference capacity value, carrying out edge end SOC estimation, and obtaining an SOC estimation value considering the aging state at each moment. According to the method, the influence of the battery state on characteristic parameter identification is considered, and the method has more accurate battery characteristic parameter identification capability, so that more accurate SOC prediction is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium-ion batteries, and in particular, relates to a lithium battery SOC estimation method and system that collaborates with an edge terminal and a cloud terminal. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The state of charge (SOC) of lithium-ion batteries in electric vehicles is an important parameter of the operating state of lithium batteries. SOC reflects the current charge storage state of the battery, usually expressed as the ratio of discharge capacity to current available capacity, and is used to describe the current short-term state change. Efficient and accurate SOC estimation of lithium batteries is crucial. Existing SOC estimation methods include ampere-hour integration method, open circuit voltage method, equivalent circuit model method and data-driven method. Among them, the edge estimation method refers to the SOC estimation method based on Matlab / Simulink, and the cloud estimation method refers to the method of SOC estimation by a data-driven neural network model. Therefore, from the characteristics of the edge and the cloud, the ampere-hour integration method, the open circuit voltage method, and the equivalent circuit model method can be regarded as edge SOC estimation methods, while the data-driven method can be regarded as a cloud SOC estimation method.

[0004] Considering that a single method has a certain SOC estimation error, resulting in low final estimation accuracy, a SOC estimation method based on data fusion is currently proposed. In this method, the SOC estimation values ​​obtained by the neural network and the ampere-hour integration method are fused in a Kalman filter manner to achieve a more accurate estimation result, that is, a neural network for SOC estimation is built, and measurable values ​​such as current and voltage are used as the input of the neural network, and the lithium battery SOC estimation value is used as the output of the neural network. The built neural network is trained using a training data set to learn its internal mapping relationship, and then the trained neural network can be used to perform actual lithium battery SOC estimation, and at the same time, the battery SOC is estimated using the ampere-hour integration method. The ampere-hour integration method integrates the current over time to obtain the SOC change value, and then adds the SOC initial value and the SOC change value to obtain the current SOC estimate. Finally, the SOC estimate obtained by the neural network and the SOC estimate obtained by the ampere-hour integration method are fused using the Kalman filter algorithm, and the fused SOC estimate has higher accuracy.

[0005] However, the existing SOC estimation method still has certain problems, which leads to errors in the final SOC estimation and low accuracy. The problems are as follows:

[0006] (1) The existing technology is still dependent on the quality of directly measured data. For example, when using cloud-based methods to estimate SOC, it is usually necessary to build a machine learning or deep learning model to estimate SOC based on direct measurement data such as current, voltage, and temperature. However, during the actual operation of the vehicle, due to the influence of measurement accuracy and load fluctuations, a large number of outliers will appear in direct measurement data such as current, voltage, and temperature, resulting in poor data quality. There is a complex nonlinear mapping relationship between the above-mentioned direct measurement data and SOC. The existence of outliers will cause the model to overfit when predicting SOC. Therefore, when performing SOC estimation, the existing technology is highly dependent on data quality and is difficult to apply to actual vehicle operation data.

[0007] (2) The existing technology does not take into account the impact of battery aging on the SOC estimation algorithm. The estimation accuracy of the battery SOC is directly related to the maximum available capacity. The maximum available capacity of the battery will decrease as the battery aging increases. Therefore, the capacity of batteries of the same batch and model at different aging levels will cause the SOC estimation value to deviate from the actual SOC value during the operation of the algorithm, resulting in a decrease in the accuracy of the SOC estimation.

[0008] (3) The existing technology does not take into account the error accumulation caused by the calculation itself. For example, when using the ampere-hour integration method to estimate SOC, since it is difficult to directly calculate the integral, the integral calculation is usually discretized. However, since the ampere-hour integration method is an open-loop estimation method, the error caused by current measurement or discretization will continue to accumulate as the integral calculation proceeds, resulting in an increasing error in the SOC estimation result. Summary of the invention

[0009] To address the deficiencies of the above-mentioned prior art, the present invention provides a lithium battery SOC estimation method and system that collaborates with the edge and cloud, wherein the battery status and battery characteristic parameters are predicted by a cloud-side neural network model, and the battery characteristic parameters are identified by an edge-side algorithm. The two sets of battery characteristic parameters are jointly corrected in combination with the battery status obtained by the neural network model, and the edge-side SOC estimation is performed using the corrected battery characteristic parameters, thereby achieving more accurate battery characteristic parameter identification, and thus achieving more accurate SOC prediction and estimation.

[0010] In a first aspect, the present invention provides a lithium battery SOC estimation method that collaborates with the edge and the cloud.

[0011] A lithium battery SOC estimation method that collaborates with an edge end and a cloud end, comprising:

[0012] Obtaining raw data of lithium batteries under charge and discharge test conditions; wherein the raw data includes the measured values ​​of multiple parameters at each moment;

[0013] A second-order equivalent circuit model of a lithium battery is established, and the characteristic parameters of the battery in the model are identified at the edge and in the cloud based on the original data. The original data is used to identify the characteristic parameters of the battery at the edge using the least square method with a forgetting factor to obtain the identification value of the characteristic parameters of the battery at each moment. In the cloud, the time series characteristics of each parameter in the set time period in the original data are extracted and input into the pre-trained battery parameter prediction model, and the predicted value of the characteristic parameters of the battery at each corresponding moment is output, and the SOH value and the reference capacity value at each moment are calculated.

[0014] Based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, the battery characteristic parameter values ​​are jointly corrected with the SOH value as a reference;

[0015] According to the corrected battery characteristic parameter value and the reference capacity value, the edge SOC is estimated to obtain the SOC estimation value considering the aging state at each moment.

[0016] In a second aspect, the present invention provides a lithium battery SOC estimation system that collaborates with the edge and the cloud.

[0017] A lithium battery SOC estimation system that cooperates with an edge terminal and a cloud terminal, comprising:

[0018] A data acquisition module is used to acquire raw data of the lithium battery under charge and discharge test conditions; wherein the raw data includes the measured values ​​of multiple parameters at each moment;

[0019] The battery characteristic parameter identification module is used to establish a second-order equivalent circuit model of the lithium battery, and identify the battery characteristic parameters in the model at the edge and cloud according to the original data; wherein, the battery characteristic parameters are identified at the edge using the original data using the least squares method with a forgetting factor to obtain the battery characteristic parameter identification value at each moment; at the cloud, the time series characteristics of each parameter within a set time period in the original data are extracted and input into the pre-trained battery parameter prediction model, and the battery characteristic parameter prediction value at each corresponding moment is output, and the SOH value and the reference capacity value at each moment are calculated;

[0020] A battery characteristic parameter correction module is used to perform joint correction of the battery characteristic parameter values ​​based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, with the SOH value as a reference;

[0021] The SOC estimation module is used to perform edge SOC estimation based on the corrected battery characteristic parameter value in combination with the reference capacity value, and obtain the SOC estimation value considering the aging state at each moment.

[0022] In a third aspect, the present invention also provides an electronic device, comprising: a memory for storing executable instructions; and a processor for implementing the above-mentioned collaborative edge and cloud-based lithium battery SOC estimation method when executing the executable instructions stored in the memory.

[0023] In a fourth aspect, the present invention further provides a computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the above-mentioned collaborative edge and cloud-based lithium battery SOC estimation method.

[0024] In a fifth aspect, the present invention also provides a computer program product, which includes executable instructions, and the executable instructions are stored in a computer-readable storage medium; wherein, when the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the above-mentioned collaborative edge and cloud-based lithium battery SOC estimation method is implemented.

[0025] One or more of the above technical solutions have the following beneficial effects:

[0026] 1. The present invention provides a lithium battery SOC estimation method and system that coordinates the edge and cloud. The battery state and battery characteristic parameters are predicted by a cloud neural network model, and the battery characteristic parameters are identified by an edge algorithm. The two sets of battery characteristic parameters are jointly corrected in combination with the battery state obtained by the neural network model, and the edge SOC is estimated with the corrected battery characteristic parameters. Compared with other existing SOC estimation methods (such as the ampere-hour integration method, the open circuit voltage method, the equivalent circuit model method, and the data-driven method), the present invention takes into account the influence of the battery state on the characteristic parameter identification, and has a more accurate battery characteristic parameter identification capability, thereby achieving a more accurate SOC prediction.

[0027] 2. The present invention integrates multiple SOC estimation methods and adopts the method of edge-end and cloud-end fusion to achieve more accurate SOC estimation. First of all, considering the practical application level and computing power issues, the present invention generally follows the idea of ​​the edge-end SOC estimation method and selects a more effective method based on an equivalent circuit model. In order to solve the problem of large SOC estimation errors due to inaccurate parameter identification, the present invention adopts a cloud-end method for parameter identification and realizes battery characteristic parameter prediction through a neural network model to solve the problem of inaccurate parameter identification of the edge-end algorithm. In addition, since the neural network model requires complex model algorithms and massive and high-quality training data to determine the complex mapping relationship between the input and SOC, the cloud-end method used in the present invention is not directly used for SOC estimation, but only for parameter identification and battery status prediction, that is, the neural network model is not directly used to learn the complex mapping relationship between the input and SOC, but it is used to learn the relatively simple mapping relationship between the input and the battery characteristic parameters. Due to the relationship between the input such as current, voltage, temperature and the characteristics such as the battery ohmic internal resistance, the mapping relationship between the input and SOC is not directly used for SOC estimation. The parameters are in the same equivalent circuit model, and there is a simpler mapping relationship between the two. Therefore, predicting the characteristic parameters instead of directly predicting SOC can effectively reduce the complexity of the neural network model and its dependence on data, while improving the accuracy and efficiency of the model. In addition, the combination of cloud and edge fully combines the advantages of edge and cloud, and reduces the computational complexity and complexity of the cloud model on the basis of more accurate battery status and parameter identification, thereby achieving more efficient and accurate SOC estimation. Considering that the ohmic internal resistance in the battery characteristic parameters is directly related to the battery aging state, on the basis of characteristic parameter identification, the battery aging state is obtained by predicting the ohmic internal resistance, thereby providing an accurate benchmark capacity, and then the battery state and characteristic parameters predicted by the cloud are synchronized to the edge model, which effectively improves the accuracy of edge parameter identification while realizing capacity correction considering the battery aging state, thereby achieving more accurate SOC estimation.

[0028] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0030] Figure 1 This is an overall flow chart of the lithium battery SOC estimation method for coordinating edge and cloud according to an embodiment of the present invention;

[0031] Figure 2Schematic diagram of an equivalent circuit model of a second-order RC network of a battery in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] It should be noted that the following detailed descriptions are exemplary only, are intended to describe specific embodiments, are intended to provide further explanation of the present invention, and are not intended to limit exemplary embodiments according to the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art to which the present invention belongs. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0033] As mentioned in the background technology, existing SOC estimation methods all have certain SOC estimation errors, resulting in low final estimation accuracy, including:

[0034] (1) The ampere-hour integration method is based on the principle of calculating the change value of SOC and then adding the change value to the initial value of SOC to obtain the estimated SOC value at the current moment. The change value of SOC is obtained by integrating the current over time. The ampere-hour integration method is an open-loop estimation method, which means that the ampere-hour integration method cannot correct the estimation error, and the estimation error will accumulate with the increase of time. In addition, since it is difficult to perform accurate integration operations at the edge, the integration operation is usually discretized, which will introduce errors caused by discretization, which will also accumulate with the increase of time.

[0035] (2) The open circuit voltage method, similar to the ampere-hour integration method, is also a traditional SOC estimation method commonly used at the edge. The basic idea of ​​the open circuit voltage method is to find the mapping relationship between the SOC characterization parameters and the SOC, and then directly obtain the current SOC value by comparing these acquired characterization parameters when estimating. The selected characterization parameters include open circuit voltage, impedance spectrum and remaining capacity, but among the three main characterization parameters mentioned above, the open circuit voltage cannot be measured directly, and usually needs to be left to stand for several hours until the polarization reaction disappears before it can be measured. The impedance spectrum can usually only be obtained in a laboratory environment because it needs to measure the response at different frequencies. This makes it difficult to obtain parameters in the open circuit voltage method, and its real-time performance is poor. In addition, when this method is in the platform area of ​​the SOC-OCV curve, a small open circuit voltage measurement error will lead to a large SOC estimation error.

[0036] (3) The method based on the equivalent circuit model usually uses the Kalman filter and observer method. First, the system state space equation is established according to the battery circuit model, and the SOC is used as the system state for estimation. Then, the state of the next moment is predicted through the system state and the model, and the system output is obtained according to the predicted state. Then, the output is compared with the actual system output to form a negative feedback. Different from the above two methods, the method based on the equivalent circuit model is a negative feedback closed-loop estimation method, which can realize error correction. Compared with the first two methods, the method based on the equivalent circuit model has higher real-time and accuracy, and can be well applied to the actual environment. Since this method is based on the equivalent circuit model for prediction, it means that the accuracy of the model is the most important for this method. The key to establishing an accurate equivalent circuit model lies in the accurate identification of the battery characteristic parameters, which is also the disadvantage of the method based on the equivalent circuit model. The traditional edge parameter identification method usually assumes that the battery is in a certain state for characteristic parameter identification, and does not consider the influence of the battery state on the parameter identification result. This leads to a large deviation between the parameter identification result and the actual result, making the SOC estimation value obtained by the Kalman filter have a large error.

[0037] (4) Data-driven methods: Data-driven methods refer to methods such as neural networks, extreme learning machines, and deep learning, where the algorithm learns the relationship between input data and output SOC by inputting current, voltage, temperature, and other data. This method avoids the problem of building a battery model and can directly learn the complex mapping relationship between input and output. However, this complex mapping relationship requires a large amount of data training to obtain, which makes the method more dependent on data. If the amount of data is insufficient, the training effect will be extremely poor and the SOC estimation will be inaccurate. Secondly, the mapping relationship between direct measurement quantities such as current, voltage, and temperature and SOC is too complex, and usually requires complex model algorithms and massive training data to obtain, which makes the method more time-consuming and difficult to apply in practice.

[0038] In addition, the existing SOC estimation methods based on data fusion also have problems such as not being able to break away from the dependence on the quality of direct measurement data, not considering the impact of battery aging on the SOC estimation algorithm, and not considering the error accumulation caused by the calculation itself, resulting in low accuracy of the final SOC estimation.

[0039] To this end, the present invention is based on the above-mentioned existing technical solutions and proposes a lithium battery SOC estimation method and system that coordinates the edge and cloud. The cloud-side neural network model is used to predict the battery status and characteristic parameters, and the battery characteristic parameters are identified through the edge-side algorithm. The two sets of characteristic parameters are jointly corrected in combination with the battery status obtained by the neural network, and the edge-side SOC estimation is then performed using the corrected battery characteristic parameters to achieve more accurate SOC estimation.

[0040] Embodiment 1

[0041] This embodiment provides a lithium battery SOC estimation method that cooperates with the edge and the cloud. Figure 1 As shown, the specific steps include:

[0042] Step S1, obtaining raw data of a lithium battery under a charge and discharge test condition; wherein the raw data includes measurement values ​​of multiple parameters at each moment, and the multiple parameters include current, voltage, charge and discharge capacity, temperature, power and operating time.

[0043] Specifically, the lithium battery is charged and discharged to obtain the corresponding working condition test data, and the working condition test data is grouped and the parameter data is screened, and each parameter data is divided into segments to obtain the original data, including:

[0044] Step S1.1, dividing the working condition test data into different groups according to different test temperatures and test conditions;

[0045] Step S1.2, for each group of data, divide it into charging segment data and discharging segment data;

[0046] Step S1.3, filter each group of data after division, remove irrelevant parameter data such as test steps and relative time, and retain parameter data such as current, voltage, charge and discharge capacity, temperature, power, and operating time;

[0047] Step S1.4: Calculate the actual SOC value at each moment based on the total capacity of the lithium battery and the discharge capacity at each moment, and then obtain the open circuit voltage-state of charge (OCV-SOC) curve of the lithium battery.

[0048] Step S2: Establish a second-order equivalent circuit model of the lithium battery, and identify the battery characteristic parameters in the model at the edge and in the cloud based on the original data.

[0049] Among them, the battery characteristic parameters are identified at the edge using the original data and the least squares method with a forgetting factor to obtain the battery characteristic parameter identification value at each moment; at the same time, in the cloud, the time series characteristics of each parameter in the set time period in the original data are extracted and input into the pre-trained battery parameter prediction model, and the battery characteristic parameter prediction value at each corresponding moment is output, and the SOH value and benchmark capacity value at each moment are calculated.

[0050] Specifically, first, an equivalent circuit model of a lithium iron phosphate battery is established. Considering the accuracy and complexity of the battery model, two sets of RC circuits are used to describe the polarization characteristics inside the battery, and an equivalent circuit model of the battery second-order RC network (referred to as the second-order equivalent circuit model) is established, such as Figure 2 As shown, R0, R s , R p , C s , C p is the battery internal resistance parameter in the model.

[0051] Secondly, based on the acquired raw data, two different methods, edge and cloud, are used to identify the battery status and battery characteristic parameters respectively.

[0052] Step S2.1: Using the original data, the least square method with a forgetting factor is used at the edge end to identify the characteristic parameters of the battery, and the identification value of the characteristic parameters of the battery at each moment is obtained. Specifically, according to the original data in step S1, the open circuit voltage-state of charge (OCV-SOC) curve of the battery is obtained, and the least square method with a forgetting factor is used at the edge end to identify the characteristic parameters of the battery. The specific steps of the least square method parameter identification with a forgetting factor are as follows:

[0053] Step S2.1.1: According to the second-order RC equivalent circuit model and Kirchhoff's law established above, the system state equation of the battery is obtained as follows:

[0054]

[0055]

[0056] U t =U OC -U P -U S -IR0

[0057] Then the system equation of the battery can be established as follows:

[0058] y k =Φ k θ k +e k

[0059] In the above formula, yk Indicates terminal voltage, Φ k represents the observed variables, including the terminal voltage and current at the previous moment, θ k is the unknown parameter matrix to be solved, e k represents the system white noise.

[0060] Step S2.1.2, perform Laplace transformation on the system state equation of the battery to obtain the transfer function, and use the bilinear transformation method to convert the transfer function into the discrete domain, which can be expressed as:

[0061]

[0062]

[0063]

[0064] U t,k =(1-c1-c2)U oc +c1U t,k-1 +c2U t,k-2 +c3I k +c4I k-1 +c5I k-2

[0065] Therefore, the observed variable Φ k and the unknown parameter matrix θ to be solved k They are:

[0066] Φ(k)=[1U t,k-1 U t,k-2 I k I k-1 I k-2 ]

[0067] θ(k)=[(1-c1-c2)U oc,k c1 c2 c3 c4 c5] T

[0068] In the above formula, U oc is the open circuit voltage, which can be obtained by interpolation of the SOC-OCV curve; c1, c2, c3, c4, and c5 represent the unknown parameters that need to be solved.

[0069] Step S2.1.3, after determining the parameter matrix and the observation matrix, introduce the forgetting factor, and use the least squares method to solve it by iterative calculation based on the original data, which can be expressed as:

[0070]

[0071] Among them, u represents the forgetting factor, K k is the algorithm gain, Pk is the error covariance matrix.

[0072] The parameter matrix obtained by the above calculation is only an intermediate variable, which needs to be solved further to obtain the battery characteristic parameter identification result. The formula is:

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] Furthermore, the above battery characteristic parameter identification and solution process is specifically as follows:

[0081] (1) Parameter initialization. Assign initial values ​​to the error covariance matrix, that is, let the initial value of the error covariance matrix be P0 = αI, α = 10 5 ; Where I is the unit matrix; take the forgetting factor u = 0.99; take the initial value of the parameter matrix θ0 as the zero vector;

[0082] (2) Based on the original data, obtain the corresponding data. Specifically, obtain the SOC value, voltage value, and current value, and assign them to the SOC k ,U t,k-1 ,U t,k-2 ,I k ,I k-1 ,I k-2 ;

[0083] (3) Iterative calculation. According to the above algorithm, the gain K k , observed variable P k and the parameter matrix The calculation formula is iterated continuously to obtain P at the current k moment. k , K k and

[0084] (4) According to the parameter matrix obtained by calculation The parameters (including c1, c2, c3, c4, c5) in the calculation are used to calculate the battery characteristic parameters (including R0, R s , R p , C s , C p)value;

[0085] (5) Repeat the above steps (2) to (4) to calculate the identification results of the battery characteristic parameters at each moment.

[0086] Step S2.2: In the cloud, extract the time series characteristics of each parameter in the set time period in the original data, input them into the pre-trained battery parameter prediction model, output the predicted value of the battery characteristic parameter at each corresponding moment, and calculate the SOH value and benchmark capacity value at each moment.

[0087] SOH reflects the aging state of the battery over its entire life cycle, and is usually expressed as the ratio of the current available capacity of the battery to the initial capacity. It is used to describe the degree of degradation of the battery under different cycles. That is, in this embodiment, while the battery characteristic parameters are identified at the edge, a cloud-based neural network model is built as a battery parameter prediction model. The model is used to identify the battery characteristic parameters and obtain the battery aging state.

[0088] Step S2.2.1: Construct a neural network model in the cloud as a battery parameter prediction model, using the mean, maximum, variance, range, change rate and other time series characteristics of the voltage, current, temperature, power, operating time and other measurement information (i.e., raw data) within a set time period as the model input, and the internal resistance and other battery characteristic parameters as the model output, and pre-train the model using the raw data. This training process includes:

[0089] First, the original data is reordered and arranged in the order of input and output of the neural network;

[0090] Secondly, the original data is divided into training set and test set, and then the training set and test set data are divided into input and output matrices according to a certain step size to facilitate feature extraction;

[0091] Then, the divided data is subjected to time series feature extraction, and time series features including mean, maximum, variance, range, and rate of change are extracted as inputs to the model.

[0092] After that, build a neural network model, including but not limited to a fully connected neural network model, a convolutional neural network model, a recurrent neural network model, etc.;

[0093] Again, the constructed model is trained on the training set to obtain the mapping relationship between the input timing features and the output battery characteristic parameters, and a trained battery characteristic parameter prediction model is obtained.

[0094] Step S2.2.2: While performing parameter identification at the edge, the original data is input into a pre-trained battery parameter prediction model. The model identifies the battery characteristic parameters corresponding to the battery at each moment based on the original data.

[0095] Step S2.2.3, after obtaining the battery characteristic parameters, calculate the corresponding SOH value, that is, the battery aging state, according to the battery ohmic internal resistance (i.e. resistance R0) in the battery characteristic parameters. Specifically, firstly, use the box plot to screen and eliminate the abnormal values ​​appearing in the ohmic internal resistance prediction, and use the data after eliminating the abnormal values ​​as the data for calculating the battery aging state; secondly, for the data, extract the mean of the stable data of the ohmic internal resistance before 4000s as the initial ohmic internal resistance value R new , take 1.5 times of the initial value as the ohmic internal resistance value R when the battery is scrapped end , combined with the current predicted ohmic internal resistance R cur , the SOH value at the current moment is calculated by the following formula:

[0096]

[0097] Step S2.2.4, after obtaining the SOH value at each moment, i.e., the battery aging status, calculate the accurate reference capacity at the current moment according to the product of the SOH value and the initial capacity, and store the SOH value and the reference capacity in the cloud database. In this embodiment, the frequency of SOH estimation (i.e., battery aging estimation) is much lower than that of SOC estimation. Here, the SOH at each moment is estimated, which is actually the estimation of the end moment of each charge and discharge cycle or each equivalent charge and discharge cycle. The charging segment data and the discharging segment data are divided in the above raw data in order to clarify each charge and discharge cycle or equivalent charge and discharge cycle for SOH estimation.

[0098] Step S3: Based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, the battery characteristic parameter values ​​are jointly corrected with the SOH value as a reference.

[0099] Specifically, based on the two groups of battery characteristic parameters obtained by the edge and cloud methods in the above step S2, the parameters are jointly corrected by a weighted average method to determine appropriate battery characteristic parameters, including:

[0100] Step S3.1: Compare the battery characteristic parameters obtained by the edge algorithm and the cloud model respectively, and use the battery aging state (i.e., SOH value) predicted by the cloud model as a benchmark to correct the battery characteristic parameters obtained by the edge algorithm and the cloud model by weighted average. The formula is:

[0101]

[0102] Among them, X cloud represents the battery characteristic parameters estimated by the cloud, X edge represents the battery characteristic parameter obtained at the edge, α represents the weight, Indicates the corrected battery characteristic parameter value. In specific implementation, it can be based on X cloud and X edge The value of the weight α is adjusted according to the error. Here, the characteristic parameters estimated in the cloud are used as the benchmark. Therefore, when adjusting the weight α, α is made greater than 0.5. The accurate battery characteristic parameters can be obtained through the above weighted average method.

[0103] Step S3.2, verify the battery characteristic parameters calculated in the above step S3.1, and judge them with the terminal voltage error. Specifically, input the accurate battery characteristic parameters obtained in step S3.1 into the battery system state equation, and obtain the battery terminal voltage calculated by the identification parameters, and compare it with the terminal voltage of the actual charge and discharge data of the battery. The error between the two is required to be within 3%. If the verification result meets the requirements, proceed to the next step. Otherwise, if the verification result does not meet the requirements, return to step S3.1 and re-correct the parameters until the verification result meets the requirements.

[0104] Step S4: perform edge SOC estimation based on the corrected battery characteristic parameter value in combination with the reference capacity value to obtain an SOC estimation value that takes the aging state into consideration at each moment.

[0105] Specifically, the battery second-order equivalent circuit model established in step S2 and the corrected battery characteristic parameter values ​​finally obtained in step S3 are used to perform SOC estimation using an edge algorithm, such as an extended Kalman filter. In this embodiment, the battery characteristic parameter values ​​obtained above are substituted into the extended Kalman filter estimation, and the SOC estimation value is obtained by iterative solution. The estimation process is:

[0106] First, SOC, R p C p Branch voltage U p , R s C s Branch voltage U s As the state quantity, the battery system state equation is established; the current I(t) is used as the input quantity, and the terminal voltage U t (t) is used as the output, and the state equation of the battery model is obtained as follows:

[0107]

[0108]

[0109]

[0110] Among them, Q ν is the rated capacity of the battery, and its output equation is:

[0111] U t (t) = U ocv SOC(t)-U P (t)-U S (t)-I(t)R0

[0112] The above battery model is discretized, and the discretized state equation is:

[0113]

[0114] Among them, w k is the process noise, and the discretized output equation is:

[0115] U t,k =U ocv SOC k -U P,k -U S,k -I k R0

[0116] Using Taylor's formula to perform a first-order series expansion, we can get the parameter A after the model is linearized. k , B k , and D k , respectively expressed as:

[0117] D k =-R0

[0118] Then the discrete model after linearization is:

[0119] x k+1 =A k x k +B k u k +w k

[0120]

[0121] In the above formula, x k is the state variable of the system, u k is the control input of the system, y k is the observed variable, A k is the system transfer matrix, B k is the input matrix, is the output matrix, D k is the feedforward matrix, w k is the process noise, v kis the system observation noise. Among them, the process noise and observation noise are uncorrelated Gaussian white noise, and their covariances are Q k and R k , the formula is:

[0122]

[0123] Furthermore, the SOC value is solved recursively by the extended Kalman filter, and the steps are as follows:

[0124] (1) Initialize the mean and covariance P0;

[0125] (2) Update state estimation time, that is:

[0126] (3) Update the error covariance time, that is:

[0127] (4) Calculate the Kalman gain, that is:

[0128] (5) Update the state estimation measurement, that is:

[0129] (6) Update the error covariance measurement, that is:

[0130] In the above process, is the predicted value of the current state, is the optimal estimate at the previous moment, It is the optimal estimated value of the system state variable at the next moment. The SOC value is estimated through continuous iteration.

[0131] Furthermore, while estimating the characteristic parameters, the cloud also estimates the battery aging state (i.e., SOH). The characteristic parameter estimation on the cloud is performed synchronously with the edge, and the characteristic parameters are transmitted to the edge in real time to realize the joint correction of the battery characteristic parameters. Since the battery aging time period is long, it does not need to be estimated all the time. Therefore, a complete charge and discharge cycle or an equivalent charge and discharge cycle is considered to estimate the aging state, and the aging state (i.e., SOH value) is transmitted to the edge at the end of the cycle to provide an accurate capacity benchmark, that is, the capacity correction is performed based on a complete charge and discharge cycle or an equivalent complete charge and discharge cycle as the time benchmark, so as to eliminate the problem of inconsistent time scales of SOC and SOH. At each time benchmark, the cloud database sends the benchmark capacity value to the edge, and replaces the current maximum available capacity in the SOC estimation with the benchmark capacity value, so as to obtain the final accurate SOC estimation value considering the aging state.

[0132] Based on the above-mentioned collaborative edge and cloud lithium battery SOC estimation method proposed in this embodiment, the method uses a cloud method to identify battery characteristic parameters to achieve accurate characteristic parameter identification; on the basis of characteristic parameter identification, the battery aging state is calculated, and the obtained characteristic parameters and aging state are synchronized to the edge, and the cloud characteristic parameters are used as a reference, combined with the parameters obtained at the edge for joint correction to achieve more accurate parameter identification; parameter identification is performed by the cloud method, which solves the problem of inaccurate parameter identification of the edge algorithm, so that the edge can perform more accurate SOC estimation; the cloud method used is not directly used for SOC estimation, but only for parameter identification and battery state prediction. The method greatly reduces the complexity and computational complexity of the cloud model while improving the recognition accuracy, and reduces the dependence of the cloud model on data; the method combines the cloud and edge, fully combines the advantages of the edge and cloud, and reduces the computational complexity and computational complexity of the cloud model on the basis of achieving more accurate battery status and parameter identification, achieving more efficient and accurate SOC estimation; in addition, the impact of battery aging is also considered, and the battery characteristic parameters are predicted by the neural network model while calculating the battery aging status and benchmark capacity, and the capacity value in the edge SOC estimation is corrected on a unified time scale with the benchmark capacity, further improving the accuracy of SOC estimation. Compared with the traditional SOC estimation method, which has problems such as large error accumulation, poor practicality, inaccurate parameter identification, high computational cost, and excessive dependence on data, the method proposed in this embodiment can achieve more efficient and accurate SOC estimation.

[0133] Embodiment 2

[0134] This embodiment provides a lithium battery SOC estimation system that cooperates with an edge terminal and a cloud, including:

[0135] A data acquisition module is used to acquire raw data of the lithium battery under the charging and discharging test conditions; wherein the raw data includes the measured values ​​of multiple parameters at each moment;

[0136] The battery characteristic parameter identification module is used to establish a second-order equivalent circuit model of the lithium battery, and identify the battery characteristic parameters in the model at the edge and cloud according to the original data; wherein, the battery characteristic parameters are identified at the edge using the original data using the least squares method with a forgetting factor to obtain the battery characteristic parameter identification value at each moment; at the cloud, the time series characteristics of each parameter within a set time period in the original data are extracted and input into the pre-trained battery parameter prediction model, and the battery characteristic parameter prediction value at each corresponding moment is output, and the SOH value and the reference capacity value at each moment are calculated;

[0137] A battery characteristic parameter correction module is used to perform joint correction of the battery characteristic parameter values ​​based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, with the SOH value as a reference;

[0138] The SOC estimation module is used to perform edge SOC estimation based on the corrected battery characteristic parameter value in combination with the reference capacity value, and obtain the SOC estimation value considering the aging state at each moment.

[0139] Embodiment 3

[0140] This embodiment provides an electronic device, including: a memory, used to store executable instructions; and a processor, used to implement the above method provided in this embodiment when executing the executable instructions stored in the memory.

[0141] Embodiment 4

[0142] This embodiment also provides a computer-readable storage medium storing executable instructions. When the executable instructions are executed by a processor, the processor will be caused to execute the above method provided in this embodiment.

[0143] Embodiment 5

[0144] This embodiment provides a computer program product, which includes an executable instruction, which is a computer instruction; the executable instruction is stored in a computer-readable storage medium. When a processor of an electronic device reads the executable instruction from the computer-readable storage medium and the processor executes the executable instruction, the electronic device executes the above method provided in this embodiment.

[0145] The steps involved in the above embodiments 2 to 5 correspond to the method embodiment 1. For the specific implementation, please refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0146] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0147] The above description is only a preferred embodiment of the present invention. Although the specific implementation mode of the present invention is described in conjunction with the accompanying drawings, it is not a limitation of the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the protection scope of the present invention.

Claims

1. A lithium battery SOC estimation method for coordinating edge and cloud terminals, characterized in that: include: Obtaining raw data of lithium batteries under charge and discharge test conditions; wherein the raw data includes the measured values ​​of multiple parameters at each moment; A second-order equivalent circuit model of a lithium battery is established, and the characteristic parameters of the battery in the model are identified at the edge and in the cloud based on the original data. The original data is used to identify the characteristic parameters of the battery at the edge using the least square method with a forgetting factor to obtain the identification value of the characteristic parameters of the battery at each moment. In the cloud, the time series characteristics of each parameter in the set time period in the original data are extracted and input into the pre-trained battery parameter prediction model, and the predicted value of the characteristic parameters of the battery at each corresponding moment is output, and the SOH value and the reference capacity value at each moment are calculated. Based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, the battery characteristic parameter values ​​are jointly corrected with the SOH value as a reference; According to the corrected battery characteristic parameter value and the reference capacity value, the edge SOC is estimated to obtain the SOC estimation value considering the aging state at each moment.

2. The method for estimating SOC of a lithium battery in a coordinated edge and cloud environment according to claim 1, characterized in that: The various parameters include current, voltage, charge and discharge capacity, temperature, power and operating time.

3. The method for estimating SOC of a lithium battery in a coordinated edge and cloud according to claim 1, characterized in that: Using the original data, the least square method with forgetting factor is used at the edge to identify the battery characteristic parameters, and the identification value of the battery characteristic parameters at each moment is obtained, including: Establish a second-order equivalent circuit model of lithium iron phosphate battery, obtain the system state equation of the battery based on the second-order equivalent circuit model and Kirchhoff's law, and then establish the system equation of the battery; Perform Laplace transformation on the system state equation of the battery to obtain the transfer function, and use the bilinear transformation method to transform the transfer function into the discrete domain to determine the observation matrix and parameter matrix in the battery system equation; Based on the determined parameter matrix and observation matrix, the forgetting factor is introduced, and the least square method is used to solve the original data through iterative calculation to obtain the parameter values ​​in the parameter matrix; According to the parameter values ​​in the parameter matrix, the battery characteristic parameter identification value is calculated.

4. The method for estimating SOC of a lithium battery in a coordinated edge and cloud environment according to claim 1, wherein: In the cloud, extract the time series characteristics of each parameter in the set time period in the original data, input it into the pre-trained battery parameter prediction model, output the predicted value of the battery characteristic parameter at each corresponding moment, and calculate the SOH value and benchmark capacity value at each moment, including: Construct a neural network model in the cloud as a battery parameter prediction model, using the mean, maximum, variance, range, and change rate time series features of voltage, current, temperature, power, and operating time in the original data within a set time period as the model input, and the battery characteristic parameters as the model output, and pre-train the model using the original data; The raw data is input into the pre-trained battery parameter prediction model, and the model identifies the battery characteristic parameters corresponding to the battery at each moment based on the raw data; Calculate the corresponding SOH value, i.e. the battery aging state, based on the battery ohmic internal resistance in the battery characteristic parameters; According to the SOH value obtained at each moment, the product of the SOH value and the initial capacity is calculated to obtain the reference capacity at each moment.

5. The method for estimating SOC of a lithium battery in collaboration between an edge terminal and a cloud as claimed in claim 1, characterized in that: Based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, the battery characteristic parameter values ​​are jointly corrected with the SOH value as the benchmark, including: Compare the battery characteristic parameter values ​​obtained from the edge and cloud, respectively, and use the battery aging status predicted by the cloud model as a benchmark to correct the battery characteristic parameter values ​​obtained from the edge and cloud by weighted average to obtain the corrected battery characteristic parameter values; The corrected battery characteristic parameter values ​​are verified and judged by the terminal voltage error. If the verification result meets the requirements, proceed to the next step. Otherwise, if the verification result does not meet the requirements, re-correct the parameters until the verification result meets the requirements.

6. The method for estimating SOC of a lithium battery in collaboration between an edge terminal and a cloud as claimed in claim 1, characterized in that: According to the corrected battery characteristic parameter value and the reference capacity value, the edge SOC is estimated to obtain the SOC estimation value considering the aging state at each moment, including: Based on the second-order equivalent circuit model of the battery and the corrected battery characteristic parameter values ​​finally obtained, an edge algorithm is used to estimate the SOC; wherein, the edge algorithm is an extended Kalman filter algorithm, and in the SOC estimation process, the capacity is corrected according to the reference capacity value at the current moment and with a complete charge and discharge cycle or an equivalent complete charge and discharge cycle as the time reference, and the current maximum available capacity in the SOC estimation is replaced by the reference capacity value to obtain the final SOC estimation value taking into account the aging state.

7. A lithium battery SOC estimation system that collaborates with edge and cloud, characterized in that: include: A data acquisition module is used to acquire raw data of the lithium battery under charge and discharge test conditions; wherein the raw data includes the measured values ​​of multiple parameters at each moment; The battery characteristic parameter identification module is used to establish a second-order equivalent circuit model of the lithium battery, and identify the battery characteristic parameters in the model at the edge and cloud according to the original data; wherein, the battery characteristic parameters are identified at the edge using the original data using the least squares method with a forgetting factor to obtain the battery characteristic parameter identification value at each moment; at the cloud, the time series characteristics of each parameter within a set time period in the original data are extracted and input into the pre-trained battery parameter prediction model, and the battery characteristic parameter prediction value at each corresponding moment is output, and the SOH value and the reference capacity value at each moment are calculated; A battery characteristic parameter correction module is used to perform joint correction of the battery characteristic parameter values ​​based on the two sets of battery characteristic parameter values ​​obtained from the edge and the cloud, with the SOH value as a reference; The SOC estimation module is used to perform edge SOC estimation based on the corrected battery characteristic parameter value in combination with the reference capacity value, and obtain the SOC estimation value considering the aging state at each moment.

8. An electronic device, characterized in that: include: A memory for storing executable instructions; The processor is used to implement the lithium battery SOC estimation method for the collaborative edge and cloud as described in any one of claims 1-6 when executing the executable instructions stored in the memory.

9. A computer-readable storage medium, characterized in that: Executable instructions are stored, which are used to cause the processor to execute the executable instructions to implement the lithium battery SOC estimation method of the collaborative edge and cloud as described in any one of claims 1-6.

10. A computer program product, characterized in that The computer program product includes executable instructions stored in a computer-readable storage medium; When the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the collaborative edge and cloud-based lithium battery SOC estimation method described in any one of claims 1-6 is implemented.

Images