A lithium battery health state estimation method based on improved long short-term memory neural network
By improving the long short-term memory neural network using particle swarm optimization algorithm and grey relational analysis, the problem of key hyperparameters depending on experience is solved, and the estimation accuracy of the health status of lithium-ion batteries is improved.
Patent Information
- Application Number
- CN202210094201.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-26
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2042-01-26
AI Technical Summary
In existing methods for estimating the health status of lithium-ion batteries based on long short-term memory neural networks, the selection of key hyperparameters relies excessively on researchers' experience, leading to unstable estimation results.
The parameters of the long short-term memory neural network are automatically tuned using the particle swarm optimization algorithm, and aging characteristics are extracted by combining the grey relational analysis method to establish an improved lithium-ion battery health status estimation model.
This improves the accuracy of lithium-ion battery health state estimation, reduces reliance on researcher experience, and enhances the accuracy of the estimation.
Smart Images

Figure CN114487890B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of lithium ion batteries, and in particular to a method for estimating the health state of a lithium ion battery by improving a long short-term memory neural network. Background Art
[0002] Nowadays, lithium-ion batteries are widely used as power sources for various types of devices, including electric vehicles, portable devices, and spacecraft, due to their significant advantages such as high energy density, long cycle life, and low environmental pollution. However, the performance of the battery will decrease with continuous charge and discharge cycles. For many devices, when the battery capacity drops below 80% of the rated capacity, the battery needs to be replaced in time, otherwise it may cause serious safety accidents. The capacity change of the battery is strongly correlated with the health status, so it is necessary to know the current health status of the battery.
[0003] There are many methods for estimating the health status of lithium-ion batteries, which can be summarized into three categories: direct measurement method, model-based method, and data-driven method. The direct measurement method is to obtain the health status of the battery by designing relevant experiments. This type of method has a long test time and high equipment cost. It is not suitable for actual engineering applications and is more suitable for use in laboratory research. The model-based method usually requires the establishment of an electrochemical model or equivalent circuit model of the battery, and then uses advanced estimation algorithms on this basis to track the decline of the aging factor. Commonly used estimation methods include sliding mode observer, H infinite filter, particle filter, etc. The disadvantage of this type of method is that the accuracy of the estimation is easily affected by noise and actual working conditions. Due to the low reliance on the physical model of the battery, the data-driven health status estimation method has attracted the attention of many researchers. This type of method is mainly based on historical monitoring data and uses machine learning algorithms to describe the mapping relationship between the voltage, current and other data measured during the charging or discharging process of the lithium-ion battery and the health status. Long Short-term Memory (LSTM) neural network is one of the commonly used methods to estimate health status. Although it has achieved good estimation results, it also has certain problems. The key hyperparameters of the corresponding network model are usually difficult to determine. The selection of these parameters mostly needs to be based on the researcher's experience, which to a certain extent reduces the estimation effect of the model. Summary of the invention
[0004] In order to solve the above problems existing in the prior art, the present invention provides a method for estimating the health status of lithium batteries by improving the long short-term memory neural network. The present invention can significantly improve the estimation accuracy and has high practical application value.
[0005] The technical solution adopted by the present invention is: a method for estimating the health status of a lithium battery using an improved long short-term memory neural network, comprising the following steps:
[0006] Step 1: Perform charge and discharge experiments on each battery according to given parameter settings to obtain a lithium-ion battery experimental data set;
[0007] Step 2: Calculate the actual health status of the battery based on the capacity, extract four aging features that can characterize the health status of the battery, and standardize the feature data;
[0008] Step 3, initialize parameters and establish an improved long short-term memory neural network model, determine the parameters that need to be optimized in the long short-term memory neural network, and use the particle swarm algorithm to optimize;
[0009] Step 4: Train the improved long short-term memory neural network estimation model: use the standardized feature data as the input of the neural network model, the actual health status of the battery as the output of the network model, and the mean absolute error of the network model output value as the fitness value of the particles in the particle swarm optimization algorithm.
[0010] Step 5: Use the optimal parameter values obtained through training as the corresponding values in the long short-term memory neural network model to estimate the health status of the lithium-ion battery.
[0011] The calculation formula for the health status of the lithium-ion battery in step 2 is as follows:
[0012]
[0013] In the above formula, C current is the current capacity of the lithium battery, and C0 is the rated capacity of the lithium battery.
[0014] The step 2 requires the use of grey correlation analysis to quantitatively demonstrate the correlation between the extracted aging characteristics and the battery health status. The specific calculation process is:
[0015] First, for a given data set, determine the comparison sequence X = {x i (t)} and the reference sequence Y = {SOH t , t=1,2,...n}, where x i (t) represents the extracted features, and n represents the length of the sequence;
[0016] Secondly, each sequence is standardized. Here, the min-max method is used, and the calculation formula is as follows:
[0017]
[0018] Calculate the grey correlation coefficient:
[0019]
[0020] In the above formula, μ is the resolution coefficient, and its value range is (0, 1), and here it is 0.5.
[0021] The grey correlation degree is obtained as:
[0022]
[0023] The parameters that need to be initialized in step 3 include the population size, number of iterations, initial position and velocity of particles in the particle swarm optimization algorithm, initial weights and biases in the long short-term memory neural network, and the number of network layers. Here, the number of network layers is 3.
[0024] The mathematical formula representing the information flow in the long short-term memory neural network model in step 3 is as follows:
[0025] f t =σ(W f ·[h t-1 ,x t ]+b f );
[0026] i t =σ(W i ·[h t-1 ,x t ]+b i );
[0027] C t ′=tanh(W c ·[h t-1 ,x t ]+b c );
[0028] C t =f t *C t-1 +i t *C t ′;
[0029] o t =σ(W o ·[h t-1 ,x t ]+b o );
[0030] h t =o t *tanh(C t );
[0031] In the above formula, f t ,i t , ot are the outputs of the forget gate, input gate, and output gate at time t, σ represents the sigmoid function, W f , W i , W o are the weights corresponding to the forget gate, input gate and output gate, respectively, f , b i , b o They are the deviations corresponding to the forget gate, input gate and output gate, respectively, x t is the input at time t, h t-1 is the output at time t-1.
[0032] In step 3, the number of neurons in each layer is optimized by using a particle swarm optimization algorithm, specifically:
[0033] Step 3.1, taking the mean absolute error output by the long short-term memory neural network model as the fitness value of the particle, calculating the initial fitness value according to the initial position of each particle and determining the initial individual optimal position and the global optimal position;
[0034] Step 3.2: Update the position and speed of each particle, calculate the new fitness value based on the new position, and update the individual optimal position and global optimal position of the particle;
[0035] Step 3.3, when the number of iterations reaches the maximum or the mean absolute error reaches the minimum, the value of the optimal particle at this time is the optimal value of the parameter to be optimized in the long short-term memory neural network;
[0036] The step 5 is to estimate the health status of the lithium-ion battery using the optimal network parameter value obtained in the step 4.
[0037] Beneficial effects of the present invention: By introducing the particle swarm optimization algorithm, the present invention can avoid some key hyperparameters in the long short-term memory neural network from being overly dependent on the subjective parameter adjustment experience of researchers, and can automatically tune the parameters, thereby further improving the estimation accuracy of the health status of lithium-ion batteries. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 This is a flow chart of the estimation method proposed by the present invention;
[0039] Figure 2 The internal structure diagram of the long short-term memory neural network of the present invention;
[0040] Figure 3 It is a flow chart of the particle swarm optimization algorithm in the present invention;
[0041] Figure 4This is a result diagram of estimating the B0005 battery using the algorithm of the present invention and three other algorithms provided in the method of the present invention;
[0042] Figure 5 This is a result diagram of estimating the B0007 battery using the algorithm of the present invention and three other algorithms provided in the method of the present invention.
[0043] Figure 4 and Figure 5 Curve 1 in the middle represents the actual value of the battery health status, curve 2 adopts the BP neural network algorithm, curve 3 adopts the support vector regression algorithm, and curve 4 adopts the algorithm based on the improved long short-term memory neural network proposed by the present invention. DETAILED DESCRIPTION
[0044] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.
[0045] like Figure 1 As shown, a method for estimating the health status of a lithium-ion battery based on an improved long short-term memory neural network comprises the following steps:
[0046] Step 1: Perform charge and discharge experiments on each battery according to the given parameter settings to obtain a lithium-ion battery experimental data set: the data set includes the time, current, voltage, battery temperature during charging, and the time, current, voltage, capacity, etc. during discharging; the battery charge and discharge experiment plan is to first perform a 1.5A constant current charge until the battery voltage reaches 4.2V, then a constant voltage charge until the current drops to 20mA, and finally a 2A constant current discharge until the battery reaches the corresponding lower cut-off voltage. The entire charge and discharge experiment is carried out at room temperature.
[0047] Step 2: Calculate the actual health status of the battery based on the capacity, extract four aging features that can characterize the health status of the battery and standardize the feature data: constant current charging time, discharge duration, battery temperature peak, and battery incremental capacity curve peak horizontal and vertical coordinates;
[0048] The calculation formula for the health status of lithium-ion batteries is as follows:
[0049]
[0050] In the above formula, C current is the current capacity of the lithium battery, and C0 is the rated capacity of the lithium battery.
[0051] In addition, the grey correlation analysis method is needed to quantitatively demonstrate the correlation between the extracted aging features and the battery health status, so as to prove the effectiveness of the selected features. The specific calculation process is:
[0052] First, for a given data set, determine the comparison sequence X = {x i (t)} and the reference sequence Y = {SOH t , t=1,2,...n}, where x i (t) represents the extracted features, and n represents the length of the sequence.
[0053] Secondly, each sequence is standardized. Here, the min-max method is used, and the calculation formula is as follows:
[0054]
[0055] Then, the grey correlation coefficient is calculated:
[0056]
[0057] In the above formula, μ is the resolution coefficient, and its value range is (0, 1), and here it is 0.5.
[0058] Finally, the grey relational degree is obtained as:
[0059]
[0060] Step 3, initialize relevant parameters and establish an improved long short-term memory neural network model, and determine the parameters that need to be optimized in the long short-term memory neural network: the number of neurons in the first layer of the network, the number of neurons in the second layer of the network, and the number of neurons in the third layer of the network;
[0061] Initialization related parameters include the population size, number of iterations, initial position and velocity of particles in the initial particle swarm optimization algorithm, initial weights and biases in the long short-term memory neural network, and the number of network layers. Here, the number of network layers is 3;
[0062] like Figure 2 As shown in the figure, the mathematical formula for expressing information flow in the long short-term memory neural network model is as follows:
[0063] f t =σ(W f ·[h t-1 , x t ]+b f );
[0064] i t =σ(W i ·[h t-1 , x t ]+b i );
[0065] C t ′=tanh(W c ·[h t-1, x t ]+b c );
[0066] C t =f t *C t-1 +i t *C t ′;
[0067] o t =σ(W o ·[h t-1 , x t ]+b o );
[0068] h t =o t *tanh(C t );
[0069] In the above formula, f t ,i t , o t are the outputs of the forget gate, input gate, and output gate at time t, σ represents the sigmoid function, W f , W i , W o are the weights corresponding to the forget gate, input gate and output gate, respectively, f , b i , b o They are the deviations corresponding to the forget gate, input gate and output gate, respectively, x t is the input at time t, h t-1 is the output at time t-1.
[0070] like Figure 3 As shown in the figure, the number of neurons in each layer is optimized by particle swarm optimization algorithm. The specific steps are as follows:
[0071] Step 3.1, taking the mean absolute error output by the long short-term memory neural network model as the fitness value of the particle, calculating the initial fitness value according to the initial position of each particle and determining the initial individual optimal position and the global optimal position;
[0072] Step 3.2: Update the position and speed of each particle, calculate the new fitness value based on the new position, and update the individual optimal position and global optimal position of the particle;
[0073] The update formula for the particle's velocity and position is:
[0074]
[0075]
[0076] In the above formula, represents the velocity of the ith particle at the tth iteration, represents the position of the i-th particle at the t-th iteration, ω represents the inertia weight, represents the optimal position of the particle in the past, represents the optimal position of the entire particle swarm, c1 and c2 represent learning factors, and r1 and r2 represent random numbers between 0 and 1. Among them, c1 and c2 are both set to 1.5, and ω is set to 0.8.
[0077] Step 3.3, when the number of iterations reaches the maximum or the mean absolute error reaches the minimum, the value of the optimal particle at this time is the optimal value of the parameter to be optimized in the long short-term memory neural network;
[0078] Step 4: Training the improved long short-term memory neural network estimation model: using the standardized feature data as the input of the neural network model, the actual health status of the battery as the output of the network model, and the mean absolute error of the output value of the network model as the fitness value of the particles in the particle swarm optimization algorithm.
[0079] Step 5: Use the optimal parameter values obtained through training as the corresponding values in the long short-term memory neural network model to estimate the health status of the lithium-ion battery.
[0080] In order to verify the effectiveness and superiority of the present invention, the estimation effects of the algorithm proposed in the present invention are compared with those of the support vector regression (SVR) and BP neural network algorithms. Figure 4 and Figure 5 The three algorithms are used to estimate the battery status of B0005 and B0007 on the NASA cycle aging test data set, respectively, which indicates that the present invention can effectively estimate the health status of lithium-ion batteries and has higher estimation accuracy than the other two algorithms.
Claims
1. A lithium battery health status estimation method based on improved long short-term memory neural network, Features: The following steps are involved: Step 1: Perform charge and discharge experiments on each battery according to given parameter settings to obtain a lithium-ion battery experimental data set; Step 2: Calculate the actual health state of the battery according to the capacity, extract four aging features that can characterize the health state of the battery, standardize the feature data, and use the grey correlation analysis method to quantitatively demonstrate the correlation between the extracted aging features and the health state of the battery; the four aging features are constant current charging time, discharge duration, battery temperature peak, and battery incremental capacity curve peak horizontal and vertical coordinates; Step 3, initialize parameters and establish an improved long short-term memory neural network model, determine the parameters that need to be optimized in the long short-term memory neural network, and use the particle swarm algorithm to optimize; Step 4: Train the improved long short-term memory neural network estimation model: use the standardized feature data as the input of the neural network model, the actual health status of the battery as the output of the network model, and use the mean absolute error of the output value of the network model as the fitness value of the particles in the particle swarm optimization algorithm; Step 5: Use the optimal parameter values obtained through training as the corresponding values in the long short-term memory neural network model to estimate the health status of the lithium-ion battery.
2. The lithium battery health status estimation method using the improved long short-term memory neural network according to claim 1, Features: The calculation formula for the health status of the lithium-ion battery in step 2 is as follows: In the above formula, C current is the current capacity of the lithium battery, C 0 is the rated capacity of the lithium battery; The specific calculation process of the correlation between each aging feature and the battery health status is: First, for a given data set, determine the comparison sequence X = {x i (t)} and the reference sequence Y = {SOH t ,t=1,2,…n},where x i (t) represents the extracted features, and n represents the length of the sequence; Secondly, each sequence is standardized using the min-max method, and the calculation formula is as follows: Calculate the grey correlation coefficient: In the above formula, μ is the resolution coefficient, and its value range is (0,1), and here it is 0.5; The grey relational degree is calculated as:
3. The lithium battery health status estimation method of the improved long short-term memory neural network according to claim 1, Features: The parameters that need to be optimized in the long short-term memory neural network in step 3 are: the number of neurons in the first layer network, the number of neurons in the second layer network, and the number of neurons in the third layer network.
4. The lithium battery health status estimation method of the improved long short-term memory neural network according to claim 1, Features: The mathematical formula representing the information flow in the long short-term memory neural network model in step 3 is as follows: f t =σ(W f ·[h t-1 ,x t ]+b f ); i t =σ(W i ·[h t-1 ,x t ]+b i ); C t ′=tanh(W c ·[h t-1 ,x t ]+b c ); C t =f t *C t-1 +i t *C t ′; the t =σ(W o ·[h t-1 ,x t ]+b o ); h t =o t *tanh(C t ); In the above formula, f t ,i t , o t are the outputs of the forget gate, input gate, and output gate at time t, σ represents the sigmoid function, W f , W i , W o are the weights corresponding to the forget gate, input gate and output gate, respectively, f , b i , b o They are the deviations corresponding to the forget gate, input gate and output gate, respectively, x t is the input at time t, h t-1 is the output at time t-1.
5. The lithium battery health status estimation method of the improved long short-term memory neural network according to claim 1, Features: In step 3, the number of neurons in each layer is optimized by using a particle swarm optimization algorithm, specifically: Step 3.1, taking the mean absolute error output by the long short-term memory neural network model as the fitness value of the particle, calculating the initial fitness value according to the initial position of each particle and determining the initial individual optimal position and the global optimal position; Step 3.2: Update the position and speed of each particle, calculate the new fitness value based on the new position, and update the individual optimal position and global optimal position of the particle; Step 3.3: When the number of iterations reaches the maximum or the mean absolute error reaches the minimum, the value of the optimal particle at this time is the optimal value of the parameter that needs to be optimized in the long short-term memory neural network.
