A method for estimating branch current of battery system based on machine learning
Through machine learning methods, the branch current of the battery system is estimated using BP neural network, which solves the problem of battery pack current estimation dependence parameters in the prior art, realizes real-time and accurate battery pack monitoring, and improves the service life of the battery pack and Security.
Patent Information
- Application Number
- CN202210254508.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-15
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2042-03-15
AI Technical Summary
In the prior art, the battery parallel branch current estimation method requires the battery cell parameters to be determined in advance, and it is difficult to obtain the internal parameters of the parallel cell in real time, resulting in a shortening of the battery life and safety hazards.
Using the branch current estimation method of battery system based on machine learning, the branch current and branch voltage of the parallel battery pack under different operating conditions is used to estimate the branch current without determining the battery cell parameters.
Real-time branch current estimation without dependent on battery cell parameters is realized, the battery system monitoring is simplified, and the service life and safety of the battery pack are improved.
Smart Images

Figure CN114624603B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of battery technology, and in particular to a battery system branch current estimation method based on machine learning. Background Art
[0002] Lithium-ion batteries are widely used in electric vehicles due to their high energy density, long cycle life and light weight. In the use of automotive lithium-ion batteries, hundreds or thousands of cells are usually used in series and parallel to meet the needs of high power output. However, due to the inevitable differences in the manufacturing process of each battery cell, the capacity, temperature, aging degree and other characteristics of each cell will vary during use. These differences will lead to different branch currents of parallel battery packs, which will in turn cause differences in the depth of charge and discharge of different battery cells. As time goes by, the differences between parallel batteries will further increase, thereby affecting the service life of the battery pack and posing safety risks.
[0003] At present, there are few studies on the estimation of parallel branch current of batteries. In the prior art, the branch current is estimated by the equivalent circuit model of the parallel battery pack. The equivalent circuit model of the parallel battery pack is established by the equivalent circuit model of the battery cell and Kirchhoff's current-voltage law. The parameters of the battery cell model can be obtained by experimental calibration. The parallel battery pack model is based on the state space theory and selects the open circuit voltage V oc and polarization voltage drop V p As a state variable, this model can be used to analyze the unbalanced current distribution inside the parallel battery pack in real time. However, the branch current estimation method based on the equivalent circuit model has disadvantages. For example, whether it is a simple or complex equivalent circuit model, the parameters of the battery cell need to be determined in advance. But in practical applications, it is difficult to obtain the real-time parameters inside the parallel single battery from external observable parameters. Moreover, various inconsistencies also increase the uncertainty of battery parameters. Summary of the invention
[0004] In order to address the deficiencies in the prior art, the present invention provides a battery system branch current estimation method based on machine learning. This method does not require determining the parameters of the battery cells, and can estimate the branch current only by relying on the main current and branch voltage of the parallel battery pack under different operating conditions.
[0005] In order to achieve the above object, the present invention is implemented through the following technical solutions:
[0006] A method for estimating branch current of a battery system based on machine learning comprises the following steps:
[0007] S1. Obtain the main circuit current and branch circuit voltage and current of the parallel battery pack under four dynamic working conditions: DST, FUDS, UDDS, and HPPC;
[0008] S2, integrating the main current, two branch voltages and two branch currents under three of the working conditions into a main current data set I, two branch voltage data sets V1 and V2, and two branch current data sets I1 and I2, and normalizing the integrated data sets;
[0009] S3, training BP neural network;
[0010] S4. The main current I' and the voltages V'1 and V'2 of the two branches under the remaining working condition are input as input features into the trained BP neural network. After the estimation results are obtained, the results are denormalized to obtain the estimated branch current values I'1 and I'2.
[0011] Furthermore, the following method is used in S3 to train the BP neural network:
[0012] S3.1 Setting the parameters of BP neural network
[0013] S3.2 takes the main current data set I and the two branch voltage data sets V1 and V2 as inputs, and the branch current data sets I1 and I2 as outputs, trains the BP neural network, and obtains a trained BP neural network.
[0014] Furthermore, the data recorded in step S1 is obtained by the following method: current sensors are connected in series to the main circuit and two branches of the parallel battery pack, and voltage sensors are connected in parallel to the two branches, so as to obtain the main circuit current of the parallel battery pack and the voltage and current of the two branches.
[0015] Furthermore, the Min-Max normalization method is used in S2 to normalize the data set.
[0016] Compared with the prior art, the beneficial effects of the present invention are:
[0017] 1. The present invention uses a BP neural network to estimate branch current. The BP neural network algorithm has a simpler structure, faster training and learning, and occupies less memory than the deep learning algorithm, and is more suitable for mobile transport equipment.
[0018] 2. Compared with the analytical model, the present invention does not need to determine the parameters of the battery cells, and the branch current of the battery system can be estimated only by the main current and branch voltage of the parallel battery group under different working conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Attached Figure 1 It is a parallel battery pack model of the present invention;
[0020] Attached Figure 2 is a structural diagram of a neural network of the present invention;
[0021] Attached Figure 3 is a branch current estimation flow chart of the present invention;
[0022] Attached Figure 4 It is a diagram of the experimental results of the present invention. DETAILED DESCRIPTION
[0023] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the application equally.
[0024] The present invention discloses a method for estimating and correcting the parallel branch current of an energy storage battery system, comprising the following steps:
[0025] S1. Obtain the main circuit current and branch circuit voltage and current of the parallel battery pack under four dynamic working conditions: DST, FUDS, UDDS, and HPPC;
[0026] S2, integrating the main current, two branch voltages and two branch currents under three of the working conditions into a main current data set I, two branch voltage data sets V1 and V2, and two branch current data sets I1 and I2, and normalizing the integrated data sets;
[0027] S3, setting the parameters of the BP neural network, taking the trunk current data set I and the two branch voltage data sets V1 and V2 as inputs, and the branch current data sets I1 and I2 as outputs, training the BP neural network, and obtaining a trained BP neural network;
[0028] S4. The main current I' and the voltages V'1 and V'2 of the two branches under the remaining working condition are input as input features into the trained BP neural network. After the estimation results are obtained, the results are denormalized to obtain the estimated branch current values I'1 and I'2.
[0029] The data recorded in step S1 is obtained by the following method: current sensors are connected in series to the main circuit and two branches of the parallel battery pack, and voltage sensors are connected in parallel to the two branches, so as to obtain the main circuit current of the parallel battery pack and the voltage and current of the two branches.
[0030] In order to understand the present method more clearly, the present method is further described below with reference to the accompanying drawings:
[0031] The present invention discloses a battery system branch current estimation method based on machine learning, such as Figure 1As shown, the parallel battery pack uses two lithium-ion batteries Cell1 with a nominal capacity of 30.244Ah and Cell2 with a nominal capacity of 29.927Ah. In the experiment, a current sensor is connected in series in the main circuit, and voltage sensors are connected in parallel in the two branches to obtain the main circuit current of the parallel battery pack and the voltages of the two branches. The Arbin sampling time is fixed at 1s, and the dspace sampling time is fixed at 50ms.
[0032] like Figure 3 The branch current estimation flow chart is obtained: S1, obtaining the main circuit current and branch voltage and current of the parallel battery pack under four dynamic working conditions of DST, FUDS, UDDS and HPPC;
[0033] S2. Integrate the main current and the voltage and current of the two branches under the three working conditions of FUDS, UDDS, and HPPC into a main current data set I, two branch voltage data sets V1, V2, and two branch current data sets I1, I2, and normalize the integrated data sets. The Min-Max normalization method is used to achieve data set normalization.
[0034] S3, setting the parameters of the BP neural network, taking the trunk current data set I and the two branch voltage data sets V1 and V2 as inputs, and the branch current data sets I1 and I2 as outputs, training the BP neural network, and obtaining a trained BP neural network;
[0035] S4, the main circuit current I under DST condition DST And the voltage of the two branches V 1-DST 、V 2-DST As input features, it is input into the trained BP neural network. After the estimation result is obtained, the result is denormalized to obtain the estimated value of branch current I under DST condition. 1-DST , I 2-DST ;
[0036] like Figure 2 As shown in Figure 1, the structure of the BP neural network is divided into three parts: an input layer, one or more hidden layers, and an output layer. The number of nodes in different types of layers can be set as needed. i , x k , x m represents input features. f1, f2 represent activation functions. y1, y k ,y l Represents output parameters. i,j , w k,i Represents the weights between different nodes. 1,1 ,θ 1,2 ,θ 1,i ,θ 1,j ,θ1,q ,θ 2,1 ,θ 2,k ,θ 2,l Represents the threshold of different nodes. Set the parameters of the BP neural network, the number of input layer nodes is 3, the number of hidden layer nodes is 10, the number of output layer nodes is 2, the activation function between the input layer and the hidden layer is logsig, the activation function between the hidden layer and the output layer is purelin, the training algorithm is trainlm, the number of network iterations is set to 1000, the network training accuracy is 0.000027, the learning rate is set to 0.01, and the training ends when the error of the verification sample does not decrease for 20 consecutive times. Then take the main current dataset I and the two branch voltage datasets V1 and V2 as input, and the branch current datasets I1 and I2 as output, and train the BP neural network to obtain the trained BP neural network. The main current I under DST conditions -DST And the voltage of the two branches V 1-DST 、V 2-DST Input as input features into the trained BP neural network.
Claims
1. A battery system branch current estimation method based on machine learning, characterized in that: The following steps are involved: S1. Obtain the main circuit current and branch circuit voltage and current of the parallel battery pack under four dynamic working conditions: DST, FUDS, UDDS, and HPPC; S2, integrating the main current, two branch voltages and two branch currents under three of the working conditions into a main current data set I, two branch voltage data sets V1 and V2, and two branch current data sets I1 and I2, and normalizing the integrated data sets; S3, training BP neural network; S4, the remaining main circuit current I ’ And the voltage of the two branches V ’ 1. V ’ 2 is input as the input feature into the trained BP neural network. After obtaining the estimated result, the result is denormalized to obtain the estimated value of branch current I ’ 1.I ’ 2; The following method is used in S3 to train the BP neural network: S3.1 set the parameters of BP neural network; S3.2 takes the main current data set I and the two branch voltage data sets V1 and V2 as inputs, and the branch current data sets I1 and I2 as outputs, trains the BP neural network, and obtains a trained BP neural network.
2. The method for estimating branch current of a battery system based on machine learning according to claim 1, characterized in that: The data recorded in step S1 is obtained by the following method: current sensors are connected in series to the main circuit and two branches of the parallel battery pack, and voltage sensors are connected in parallel to the two branches, so as to obtain the main circuit current of the parallel battery pack and the voltage and current of the two branches.
3. The method for estimating branch current of a battery system based on machine learning according to claim 1, characterized in that: In S2, the Min-Max normalization method is used to normalize the data set.
Citation Information
Patent Citations
Sub-current calculation method and simulation method and simulation system of parallel storage battery pack
CN107609259A
Parallel battery pack internal current distribution estimation method and system based on LSTM
CN112083346A
Method for estimating and correcting temperature field of battery pack by using virtual temperature sensor
CN112092676A
Parallel battery pack modeling method and device
CN114154327A
Parallelly connected battery branch circuit current detection device
CN205898880U
