Lead-acid battery state of charge estimation method and system

Abstract

The present invention discloses a method and system for estimating the state of charge (SOC) of a lead-acid battery, relating to the field of battery technology. The method obtains standard discharge data and discharge data of the lead-acid battery, uses a pre-trained discharge capacity detection model to perform capacity prediction on the standard discharge data and discharge data, respectively, to obtain a discharge capacity value and a standard discharge capacity value, and then performs SOC analysis on the discharge capacity value and the standard discharge capacity value to obtain a corresponding SOC estimate for the lead-acid battery. This method solves the technical problem that the existing technology mainly uses a pre-trained feedforward neural network model to predict the SOC of lead-acid batteries, but most discharge data in engineering projects is incomplete discharge, making it difficult for the model to obtain accurate estimation results.

Description

Technical Field

[0001] The present invention relates to the technical field of batteries, and in particular to a method and system for estimating the state of charge of a lead-acid battery. Background Art

[0002] The state of charge (SOC) of lead-acid batteries is a key parameter for measuring their remaining charge, directly influencing their use in energy storage systems, smart warehousing, and other application scenarios (e.g., battery shipment selection). Therefore, accurately determining the SOC of lead-acid batteries is crucial.

[0003] Currently, existing technologies mainly predict the state of charge of lead-acid batteries through pre-trained feedforward neural network models. However, most discharge data in engineering projects are incomplete discharges, which makes it difficult for the model to obtain accurate estimation results. Summary of the Invention

[0004] The present invention provides a method and system for estimating the state of charge of a lead-acid battery, which solves the technical problem that the existing technology mainly predicts the state of charge of a lead-acid battery through a pre-trained feedforward neural network model, but most discharge data in engineering are incomplete discharge, which makes it difficult for the model to obtain accurate estimation results.

[0005] A first aspect of the present invention provides a method for estimating the state of charge of a lead-acid battery, comprising:

[0006] Obtain standard discharge data and discharge data of lead-acid batteries;

[0007] Extracting features from the standard discharge data and the discharge data using a pre-trained multi-scale feature network of a discharge capacity detection model to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector;

[0008] Performing discharge capacity prediction on the multimodal fusion feature vector and the standard modal fusion feature vector respectively by using the residual predictor of the discharge capacity detection model to obtain a discharge capacity value and a standard discharge capacity value;

[0009] A state of charge analysis is performed on the discharge capacity value and the standard discharge capacity value to obtain an estimated state of charge value corresponding to the lead-acid battery.

[0010] Optionally, the multi-scale feature network includes a current integration module, a main network and a feature fusion layer. The step of extracting features from the standard discharge data and the discharge data respectively using the multi-scale feature network of the pre-trained discharge capacity detection model to obtain a multimodal fusion feature vector and a standard modal fusion feature vector includes:

[0011] Performing ampere-hour integration operations on the standard discharge data and the discharge data respectively by the current integration module to obtain an ampere-hour integration vector and a standard ampere-hour integration vector;

[0012] Performing multi-feature extraction on the standard discharge data and the discharge data through the main network to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector;

[0013] Performing feature fusion on the attention feature vector, the enhancement feature vector, and the ampere-hour integral vector through a feature fusion layer to obtain a multimodal fusion feature vector;

[0014] The standard attention feature vector, the standard enhanced feature vector and the standard ampere-hour integral vector are subjected to feature fusion through a feature fusion layer to obtain a standard modality fusion feature vector.

[0015] Optionally, the main network includes a statistical feature module, a temperature adaptive gating module, and a memory enhancement recovery module. The step of performing multi-feature extraction on the standard discharge data and the discharge data through the main network to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector includes:

[0016] Performing feature extraction on the standard discharge data and the discharge data respectively by the statistical feature module to obtain a statistical feature vector and a standard statistical feature vector;

[0017] Performing temperature gating modulation on the statistical feature vector and the standard statistical feature vector respectively by the temperature adaptive gating module to obtain a temperature modulation vector and a standard temperature modulation vector;

[0018] Performing memory enhancement and recovery processing on the statistical feature vector and the temperature modulation vector by the memory enhancement and recovery module to obtain an attention feature vector and an enhanced feature vector;

[0019] The memory enhancement recovery module performs memory enhancement recovery processing on the standard statistical feature vector and the standard temperature modulation vector to obtain a standard attention feature vector and a standard enhancement feature vector.

[0020] Optionally, the memory enhancement recovery module includes a long short-term memory network, an attention mechanism module, and a memory enhancement module. The step of performing memory enhancement recovery processing on the statistical feature vector and the temperature modulation vector by the memory enhancement recovery module to obtain an attention feature vector and an enhanced feature vector includes:

[0021] Performing time series feature extraction on the statistical feature vector and the temperature modulation vector through the long short-term memory network to obtain a time series feature vector;

[0022] Extract key memory features from the time series feature vector using the attention mechanism module to obtain an attention feature vector;

[0023] The memory enhancement module performs capacity recovery simulation processing on the time series feature vector to obtain an enhanced feature vector.

[0024] Optionally, the training process of the discharge capacity detection model is specifically as follows:

[0025] Acquire a plurality of historical discharge data, perform data preprocessing on each of the historical discharge data, and obtain a discharge feature set;

[0026] Using the discharge feature set to train a preset initial discharge capacity detection model to obtain training discharge capacity data;

[0027] Calculating a training loss function value of the discharge feature set based on the training discharge capacity data based on a preset mean absolute error function;

[0028] When the training loss function value is greater than or equal to a preset standard loss function value, adjusting the network parameters of the initial discharge capacity detection model, and jumping to the step of training the preset initial discharge capacity detection model using the discharge feature set to obtain training discharge capacity data, until the training loss function value is less than the standard loss function value;

[0029] When the training loss function value is less than the standard loss function value, a discharge capacity detection model is generated.

[0030] Optionally, the step of performing state of charge analysis on the discharge capacity value and the standard discharge capacity value to obtain an estimated state of charge value corresponding to the lead-acid battery includes:

[0031] performing difference processing on the standard discharge capacity value and the discharge capacity value to obtain a corresponding first difference;

[0032] The first difference is compared with a preset rated capacity to obtain an estimated state of charge value corresponding to the lead-acid battery.

[0033] A second aspect of the present invention provides a lead-acid battery state of charge estimation system, comprising:

[0034] an acquisition module, configured to obtain standard discharge data and discharge data of a lead-acid battery, and perform capacity prediction on the standard discharge data and the discharge data respectively using a pre-trained discharge capacity detection model, wherein the discharge capacity detection model includes a multi-scale feature network and a residual predictor;

[0035] a feature extraction module, configured to perform feature extraction on the standard discharge data and the discharge data respectively through the multi-scale feature network to obtain a multimodal fusion feature vector and a standard modal fusion feature vector;

[0036] a capacity prediction module, configured to perform discharge capacity prediction on the multimodal fusion feature vector and the standard modal fusion feature vector respectively through the residual predictor to obtain a discharge capacity value and a standard discharge capacity value;

[0037] The state of charge estimation module is used to perform a state of charge analysis on the discharge capacity value and the standard discharge capacity value to obtain a state of charge estimation value corresponding to the lead-acid battery.

[0038] A third aspect of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the lead-acid battery state of charge estimation method as described in any one of the above items.

[0039] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the lead-acid battery state of charge estimation method as described in any one of the above items.

[0040] A fifth aspect of the present invention provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the lead-acid battery state of charge estimation method as described in any one of the above items.

[0041] It can be seen from the above technical solutions that the present invention has the following advantages:

[0042] The present invention obtains standard discharge data and discharge data of a lead-acid battery and uses a pre-trained discharge capacity detection model to perform capacity prediction on the standard discharge data and discharge data, respectively. The discharge capacity detection model includes a multi-scale feature network and a residual predictor. The multi-scale feature network is used to extract features from the standard discharge data and discharge data to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector. The residual predictor is used to perform discharge capacity prediction on the multi-modal fusion feature vector and the standard modal fusion feature vector to obtain a discharge capacity value and a standard discharge capacity value. The discharge capacity value and the standard discharge capacity value are subjected to state-of-charge analysis to obtain a state-of-charge estimate corresponding to the lead-acid battery. This overcomes the technical problem that the prior art mainly uses a pre-trained feedforward neural network model to predict the state of charge of a lead-acid battery, but most discharge data in engineering is incomplete discharge, making it difficult for the model to obtain accurate estimation results. Compared with the traditional state of charge prediction method, the present invention predicts the capacity of incomplete discharge data and standard discharge data by constructing a discharge capacity detection model to obtain the discharge capacity value and the standard discharge capacity value, and then estimates the state of charge of the lead-acid battery based on the discharge capacity value and the standard discharge capacity value. The state of charge estimation value corresponding to the lead-acid battery improves the accuracy of the state of charge estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 A flowchart of a method for estimating the state of charge of a lead-acid battery provided in Example 1 of the present invention;

[0045] Figure 2 A flowchart of a method for estimating the state of charge of a lead-acid battery provided in accordance with the second embodiment of the present invention;

[0046] Figure 3 A schematic structural diagram of a discharge capacity detection model provided in the second embodiment of the present invention;

[0047] Figure 4 This is a structural block diagram of a lead-acid battery state of charge estimation system provided in Example 3 of the present invention;

[0048] Figure 5 This is a structural block diagram of an electronic device provided in Example 4 of the present invention. DETAILED DESCRIPTION

[0049] An embodiment of the present invention provides a method and system for estimating the state of charge of a lead-acid battery, which is used to solve the technical problem that the existing technology mainly predicts the state of charge of a lead-acid battery through a pre-trained feedforward neural network model, but most discharge data in engineering are incomplete discharge, which makes it difficult for the model to obtain accurate estimation results.

[0050] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0051] See also Figure 1 , Figure 1 This is a flowchart of the steps of a method for estimating the state of charge of a lead-acid battery provided in Example 1 of the present invention.

[0052] The present invention provides a method for estimating the state of charge of a lead-acid battery, comprising:

[0053] Step 101: Obtain standard discharge data and discharge data of a lead-acid battery.

[0054] Standard discharge data refers to the data set collected from a complete discharge test of a lead-acid battery under strictly controlled standardized conditions, including but not limited to standardized voltage, current, temperature, and timestamp.

[0055] Discharge data refers to the set of real-time parameters recorded by a lead-acid battery during the discharge process, including but not limited to voltage (i.e., the real-time terminal voltage of a single cell), current (i.e., the discharge current value), temperature (i.e., the battery surface or ambient temperature), and timestamp (i.e., the time when the data is collected).

[0056] In an embodiment of the present invention, standard discharge data and discharge data of a lead-acid battery are obtained, and the standard discharge data and the discharge data are respectively input into a pre-trained discharge capacity detection model, wherein the discharge capacity detection model includes a multi-scale feature network and a residual predictor.

[0057] Step 102 : extract features from the standard discharge data and the discharge data respectively through the multi-scale feature network of the pre-trained discharge capacity detection model to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector.

[0058] In an embodiment of the present invention, the current integration module performs ampere-hour integration operations on the standard discharge data and the discharge data, respectively, to obtain an ampere-hour integration vector and a standard ampere-hour integration vector. The main network performs multi-feature extraction on the standard discharge data and the discharge data, respectively, to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector. The feature fusion layer performs feature fusion on the attention feature vector, the enhanced feature vector, and the ampere-hour integration vector to obtain a multimodal fusion feature vector. The feature fusion layer performs feature fusion on the standard attention feature vector, the standard enhanced feature vector, and the standard ampere-hour integration vector to obtain a standard modal fusion feature vector.

[0059] Step 103 : Using the residual predictor of the discharge capacity detection model, discharge capacity is predicted for the multimodal fusion feature vector and the standard modal fusion feature vector respectively to obtain a discharge capacity value and a standard discharge capacity value.

[0060] The discharge capacity value refers to the discharged capacity of the lead-acid battery at the current moment.

[0061] The standard discharge capacitance value refers to the maximum discharge capacity of a lead-acid battery predicted by a discharge capacity detection model.

[0062] In an embodiment of the present invention, a residual predictor is used to predict the discharge capacity of the multimodal fusion feature vector and the standard modal fusion feature vector respectively, so as to obtain the discharge capacity value and the standard discharge capacity value corresponding to the lead-acid battery.

[0063] Step 104 : Perform state of charge analysis on the discharge capacity value and the standard discharge capacity value to obtain an estimated state of charge value corresponding to the lead-acid battery.

[0064] In the embodiment of the present invention, the standard discharge capacity value and the discharge capacity value are input into a preset state of charge estimation function to obtain a state of charge estimation value corresponding to the lead-acid battery.

[0065] It should be noted that the state of charge estimation function is specifically:

[0066] ;

[0067] in, is the estimated state of charge, is the standard discharge capacity value, is the discharge capacity value, is the rated capacity of the lead-acid battery.

[0068] In an embodiment of the present invention, by obtaining standard discharge data and discharge data of a lead-acid battery, a pre-trained discharge capacity detection model is used to perform capacity prediction on the standard discharge data and discharge data, respectively. The discharge capacity detection model includes a multi-scale feature network and a residual predictor. The multi-scale feature network is used to extract features from the standard discharge data and discharge data, respectively, to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector. The residual predictor is used to perform discharge capacity prediction on the multi-modal fusion feature vector and the standard modal fusion feature vector, respectively, to obtain a discharge capacity value and a standard discharge capacity value. The state of charge (SOC) of the discharge capacity value and the standard discharge capacity value is analyzed to obtain an estimated SOC value corresponding to the lead-acid battery. This overcomes the technical problem that the prior art mainly uses a pre-trained feedforward neural network model to predict the SOC of lead-acid batteries, but most discharge data in engineering is incomplete discharge, making it difficult for the model to obtain accurate estimation results. Compared with the traditional state of charge prediction method, the present invention predicts the capacity of incomplete discharge data and standard discharge data by constructing a discharge capacity detection model to obtain the discharge capacity value and the standard discharge capacity value, and then estimates the state of charge of the lead-acid battery based on the discharge capacity value and the standard discharge capacity value. The state of charge estimation value corresponding to the lead-acid battery improves the accuracy of the state of charge estimation.

[0069] See also Figure 2 , Figure 2 This is a flowchart of the steps of a method for estimating the state of charge of a lead-acid battery provided in Example 2 of the present invention.

[0070] The present invention provides a method for estimating the state of charge of a lead-acid battery, comprising:

[0071] Step 201: Obtain standard discharge data and discharge data of a lead-acid battery.

[0072] In an embodiment of the present invention, standard discharge data and discharge data of a lead-acid battery are obtained, and capacity prediction is performed on the standard discharge data and the discharge data respectively using a pre-trained discharge capacity detection model, wherein the discharge capacity detection model includes a multi-scale feature network and a residual predictor.

[0073] It should be noted that the training process of the discharge capacity detection model is as follows:

[0074] A1. Acquire multiple historical discharge data, perform data preprocessing on each historical discharge data, and obtain a discharge feature set.

[0075] Historical discharge data refers to a set of time series parameters recorded during the past discharge process of a lead-acid battery, including but not limited to historical voltage, historical current, historical temperature, historical discharged capacity, and historical timestamps.

[0076] Data preprocessing refers to the sequential processing of data cleaning, discrete data alignment, feature standardization, and label generation on historical discharge data.

[0077] In an embodiment of the present invention, a plurality of historical discharge data are acquired, and data cleaning, discrete data alignment, feature standardization, and label generation are performed on each historical discharge data in sequence to obtain a discharge feature set.

[0078] A2. Use the discharge feature set to train a preset initial discharge capacity detection model to obtain training discharge capacity data.

[0079] The training discharge capacity data refers to the discharged capacity predicted by the initial discharge capacity detection model based on the discharge feature set.

[0080] In an embodiment of the present invention, the discharge feature set is input into a preset initial discharge capacity detection model for training to obtain training discharge capacity data.

[0081] A3. Based on the preset mean absolute error function, calculate the training loss function value of the discharge feature set according to the training discharge capacity data.

[0082] In an embodiment of the present invention, the training discharge capacity data and the discharge feature set are input into a preset mean absolute error function to obtain a training loss function value.

[0083] The mean absolute error function is specifically:

[0084] ;

[0085] in, is the training loss function value, is the total number of samples, is the discharged capacity predicted by the model for the i-th sample, is the actual discharged capacity of the i-th sample, and i is the index of the sample.

[0086] A4. When the training loss function value is greater than or equal to the preset standard loss function value, adjust the network parameters of the initial discharge capacity detection model, and jump to the step of using the discharge feature set to train the preset initial discharge capacity detection model to obtain training discharge capacity data until the training loss function value is less than the standard loss function value.

[0087] In an embodiment of the present invention, when the training loss function value is greater than or equal to the preset standard loss function value, the random gradient descent method is used to adjust the network parameters of the initial discharge capacity detection model, and the execution jumps to A1-A4 until the training loss function value is less than the standard loss function value.

[0088] A5. When the training loss function value is less than the standard loss function value, a discharge capacity detection model is generated.

[0089] In an embodiment of the present invention, when the training loss function value is less than the standard loss function value, the training is stopped and a discharge capacity detection model is generated.

[0090] Step 202 : extract features from the standard discharge data and the discharge data respectively through the multi-scale feature network of the pre-trained discharge capacity detection model to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector.

[0091] Further, see Figure 3 As shown, the multi-scale feature network includes a current integration module, a main network and a feature fusion layer. Step 202 includes the following sub-steps:

[0092] S11. Performing ampere-hour integration operations on the standard discharge data and the discharge data respectively through a current integration module to obtain an ampere-hour integration vector and a standard ampere-hour integration vector.

[0093] In an embodiment of the present invention, a current integration module performs an ampere-hour integration operation on the standard discharge data and the discharge data, respectively, to obtain an ampere-hour integration vector and a standard ampere-hour integration vector. For example, the current integration module processes the discharge data by extracting current data from the discharge data and integrating the current data using the ampere-hour integration method to obtain an ampere-hour integration vector.

[0094] It's worth noting that the discharge capacity detection model incorporates a current integration module, explicitly embedding the ampere-hour integration formula and combining physical principles with data-driven analysis. This enables the model to more accurately estimate the battery state in the initial stages, laying the foundation for subsequent precise predictions.

[0095] S12. Perform multi-feature extraction on the standard discharge data and the discharge data through the main network to obtain the attention feature vector, the enhanced feature vector, the standard attention feature vector and the standard enhanced feature vector.

[0096] Further, see Figure 3 As shown, the main network includes a statistical feature module, a temperature adaptive gating module and a memory enhancement recovery module, and S12 includes the following sub-steps:

[0097] S121 . Perform feature extraction on the standard discharge data and the discharge data respectively through a statistical feature module to obtain a statistical feature vector and a standard statistical feature vector.

[0098] In the embodiment of the present invention, the statistical feature module extracts features from the standard discharge data and the discharge data to obtain statistical feature vectors and standard statistical feature vectors. Figure 3As shown in the figure, the statistical feature module includes two one-dimensional convolutional layers, a one-dimensional average pooling layer, a one-dimensional maximum pooling layer, and a feature fusion layer. The statistical feature module processes discharge data as follows: It extracts time-domain convolution features through two one-dimensional convolutional layers to obtain a discharge time-domain feature vector. It then smoothes the discharge time-domain feature vector through a one-dimensional average pooling layer to obtain a discharge time-domain smoothed vector. It then extracts extreme value features from the discharge time-domain feature vector through a one-dimensional maximum pooling layer to obtain a discharge extreme value feature vector. It then concatenates the discharge time-domain smoothed vector and the discharge extreme value feature vector through a feature fusion layer to obtain a statistical feature vector.

[0099] It should be noted that the convolution kernel of the first one-dimensional convolution layer is a 5-point wide kernel, and the convolution kernel of the second one-dimensional convolution layer is a 3-point narrow kernel. The cross-scale feature perception of the discharge data is achieved through the difference in convolution kernel size.

[0100] S122 . Perform temperature gating modulation on the statistical characteristic vector and the standard statistical characteristic vector respectively through the temperature adaptive gating module to obtain a temperature modulation vector and a standard temperature modulation vector.

[0101] In an embodiment of the present invention, a temperature-adaptive gating module performs temperature-gated modulation on a statistical feature vector and a standard statistical feature vector, respectively, to obtain a temperature modulation vector and a standard temperature modulation vector. For example, the temperature-adaptive gating module processes the statistical feature vector by: obtaining a temperature feature from the statistical feature vector; determining a temperature gating compensation weight for the temperature feature based on a sigmoid activation function (i.e., an S-shaped function) in the temperature-adaptive gating module; and point-by-point multiplication of the temperature gating compensation weight with the convolution feature of the statistical feature vector to obtain a temperature modulation vector.

[0102] Notably, the temperature-adaptive gating module's gating mechanism has only 32 parameters, improving the predictive efficiency of the discharge capacity detection model. This module also enables environmentally adaptive feature modulation. Battery performance is significantly affected by temperature, and this module dynamically adjusts features based on temperature changes. At low temperatures (e.g., -20°C), the error fluctuation range is significantly reduced, significantly improving the model's adaptability and accuracy across diverse temperature environments.

[0103] S123. Perform memory enhancement and recovery processing on the statistical feature vector and the temperature modulation vector through a memory enhancement and recovery module to obtain an attention feature vector and an enhanced feature vector.

[0104] Further, see Figure 3 As shown, the memory enhancement recovery module includes a long short-term memory network, an attention mechanism module and a memory enhancement module, and S123 includes the following sub-steps:

[0105] S1231. Perform time series feature extraction on the statistical feature vector and the temperature modulation vector through a long short-term memory network to obtain a time series feature vector.

[0106] In an embodiment of the present invention, an LSTM memory network (i.e., long short-term memory network) is used to sequentially perform feature concatenation and time series feature extraction on the statistical feature vector and the temperature modulation vector to obtain a time series feature vector.

[0107] S1232. Extract key memory features from the time series feature vector through the attention mechanism module to obtain an attention feature vector.

[0108] In an embodiment of the present invention, a dynamic weighted operation is performed on the temporal features of the temporal feature vector through an attention mechanism module to obtain an attention feature vector.

[0109] It's worth noting that the bidirectional long-short-term memory network incorporates an attention mechanism module. This bidirectional structure enables the model to consider both past and future information simultaneously, effectively addressing the problem of long-range dependency decay. The attention mechanism allows the model to focus on information at critical time steps, significantly reducing response latency in critical scenarios such as charge-discharge switching points and improving the model's ability to quickly respond to changes in battery status.

[0110] S1233. Perform capacity recovery simulation processing on the time series feature vector through the memory enhancement module to obtain an enhanced feature vector.

[0111] In the embodiment of the present invention, a memory enhancement module is used to perform capacity recovery simulation processing on the time series feature vector to obtain an enhanced feature vector.

[0112] It is worth mentioning that the memory enhancement module realizes refined modeling of battery capacity recovery by combining the gating mechanism with the cyclic network.

[0113] S124. Perform memory enhancement and recovery processing on the standard statistical feature vector and the standard temperature modulation vector through a memory enhancement and recovery module to obtain a standard attention feature vector and a standard enhancement feature vector.

[0114] In an embodiment of the present invention, a memory enhancement recovery module is used to perform memory enhancement recovery processing on the standard statistical feature vector and the standard temperature modulation vector to obtain a standard attention feature vector and a standard enhanced feature vector.

[0115] It's worth noting that the memory enhancement module simulates the battery's capacity recovery effect, implementing a compensation algorithm for battery hysteresis through specific gating and memory units (such as GRU). In shallow charge and discharge tests, the recovery characteristic prediction error was minimal, more accurately simulating the battery's capacity changes in actual use, and enhancing the model's ability to learn and simulate complex battery behavior.

[0116] S13. Perform feature fusion on the attention feature vector, the enhancement feature vector, and the ampere-hour integral vector through the feature fusion layer to obtain a multimodal fusion feature vector.

[0117] In an embodiment of the present invention, a multimodal fusion feature vector is obtained by performing a splicing operation on the attention feature vector, the enhancement feature vector and the ampere-hour integral vector through a feature fusion layer.

[0118] S14. Perform feature fusion on the standard attention feature vector, the standard enhancement feature vector and the standard ampere-hour integral vector through the feature fusion layer to obtain a standard modality fusion feature vector.

[0119] In an embodiment of the present invention, a standard attention feature vector, a standard enhancement feature vector and a standard ampere-hour integral vector are concatenated through a feature fusion layer to obtain a standard modality fusion feature vector.

[0120] Step 203: Use a residual predictor to perform discharge capacity prediction on the multimodal fusion feature vector and the standard modal fusion feature vector respectively to obtain a discharge capacity value and a standard discharge capacity value.

[0121] It should be noted that the residual predictor is a double hidden layer residual structure, and the ReLU activation function (i.e., linear rectification function) is used between layers.

[0122] In an embodiment of the present invention, a residual predictor is used to perform discharge capacity prediction on the multimodal fusion feature vector and the standard modal fusion feature vector, respectively, to obtain a discharge capacity value and a standard discharge capacity value. For example, the residual predictor processes the multimodal fusion feature vector as follows: a global feature transformation is performed on the multimodal fusion feature vector using the first 128-node fully connected layer of the residual predictor to obtain a global discharge feature vector. Feature distillation is then performed using the first 64-node fully connected layer of the residual predictor to obtain the discharge capacity value.

[0123] It's worth noting that the discharge capacity detection model utilizes a dual-channel one-dimensional convolutional layer (Conv1D) in conjunction with a one-dimensional average pooling layer and a one-dimensional maximum pooling layer. This combination captures both transient and steady-state features from discharge data, comprehensively extracting characteristic information about the battery under different operating conditions. In noise testing, the model's signal-to-noise ratio tolerance improved, demonstrating that its feature extraction approach helps improve the model's ability to resist interference in complex environments.

[0124] Step 204 : performing difference processing on the standard discharge capacity value and the discharge capacity value to obtain a corresponding first difference value.

[0125] In the embodiment of the present invention, the difference between the standard discharge capacity value and the discharge capacity value is calculated to obtain the corresponding first difference.

[0126] Step 205 : performing ratio processing on the first difference and the preset rated capacity to obtain an estimated state of charge value corresponding to the lead-acid battery.

[0127] Rated capacity refers to the rated capacity of the lead-acid battery.

[0128] In the embodiment of the present invention, a ratio between the first difference and a preset rated capacity is calculated to obtain an estimated state of charge value corresponding to the lead-acid battery.

[0129] It's worth noting that, as shown in Table 1, at each time point, the discharged capacity (in AH) predicted by the discharge capacity detection model differed only slightly from the actual value. The largest discrepancies occurred at the beginning and end of discharge (possibly due to data boundary effects). The goodness of fit between the actual and predicted capacities was 0.999020, a near-perfect fit within the discharge capacity range (approximately 250–430 AH), demonstrating that the model's prediction deviation was very low across all states. Therefore, the present application can accurately predict the state of charge of lead-acid batteries even when the discharge data indicates incomplete discharge.

[0130] Table 1

[0131]

[0132] In an embodiment of the present invention, by obtaining standard discharge data and discharge data of a lead-acid battery, a pre-trained discharge capacity detection model is used to perform capacity prediction on the standard discharge data and discharge data, respectively. The discharge capacity detection model includes a multi-scale feature network and a residual predictor. The multi-scale feature network is used to extract features from the standard discharge data and discharge data, respectively, to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector. The residual predictor is used to perform discharge capacity prediction on the multi-modal fusion feature vector and the standard modal fusion feature vector, respectively, to obtain a discharge capacity value and a standard discharge capacity value. The state of charge (SOC) of the discharge capacity value and the standard discharge capacity value is analyzed to obtain an estimated SOC value corresponding to the lead-acid battery. This overcomes the technical problem that the prior art mainly uses a pre-trained feedforward neural network model to predict the SOC of lead-acid batteries, but most discharge data in engineering is incomplete discharge, making it difficult for the model to obtain accurate estimation results. Compared with the traditional state of charge prediction method, the present invention predicts the capacity of incomplete discharge data and standard discharge data by constructing a discharge capacity detection model to obtain the discharge capacity value and the standard discharge capacity value, and then estimates the state of charge of the lead-acid battery based on the discharge capacity value and the standard discharge capacity value. The state of charge estimation value corresponding to the lead-acid battery improves the accuracy of the state of charge estimation.

[0133] See also Figure 4 , Figure 4This is a structural block diagram of a lead-acid battery state of charge estimation system provided in Example 3 of the present invention.

[0134] The present invention provides a lead-acid battery state of charge estimation system, comprising:

[0135] The acquisition module 301 is used to obtain standard discharge data and discharge data of the lead-acid battery;

[0136] A feature extraction module 302 is configured to extract features from the standard discharge data and the discharge data using a multi-scale feature network of a pre-trained discharge capacity detection model to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector;

[0137] The capacity prediction module 303 is configured to perform discharge capacity prediction on the multimodal fusion feature vector and the standard modal fusion feature vector respectively through the residual predictor of the discharge capacity detection model to obtain a discharge capacity value and a standard discharge capacity value;

[0138] The state of charge estimation module 304 is configured to perform state of charge analysis on the discharge capacity value and the standard discharge capacity value to obtain a state of charge estimation value corresponding to the lead-acid battery.

[0139] Furthermore, the multi-scale feature network includes a current integration module, a main network and a feature fusion layer, and the feature extraction module 302 includes:

[0140] The current integration submodule is used to perform ampere-hour integration operations on the standard discharge data and the discharge data through the current integration module to obtain an ampere-hour integration vector and a standard ampere-hour integration vector;

[0141] A feature extraction submodule is used to perform multi-feature extraction on the standard discharge data and the discharge data through the main network to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector;

[0142] The first fusion submodule is used to perform feature fusion on the attention feature vector, the enhancement feature vector and the ampere-hour integral vector through a feature fusion layer to obtain a multimodal fusion feature vector;

[0143] The second fusion submodule is used to perform feature fusion on the standard attention feature vector, the standard enhancement feature vector and the standard ampere-hour integral vector through the feature fusion layer to obtain a standard modality fusion feature vector.

[0144] Furthermore, the main network includes a statistical feature module, a temperature adaptive gating module and a memory enhancement recovery module, and the feature extraction submodule includes:

[0145] A statistical feature unit is used to extract features from the standard discharge data and the discharge data respectively through a statistical feature module to obtain a statistical feature vector and a standard statistical feature vector;

[0146] A temperature gating unit is used to perform temperature gating modulation on the statistical characteristic vector and the standard statistical characteristic vector respectively through the temperature adaptive gating module to obtain a temperature modulation vector and a standard temperature modulation vector;

[0147] a first memory enhancement and recovery unit, configured to perform memory enhancement and recovery processing on the statistical feature vector and the temperature modulation vector through a memory enhancement and recovery module to obtain an attention feature vector and an enhanced feature vector;

[0148] The second memory enhancement recovery unit is used to perform memory enhancement recovery processing on the standard statistical feature vector and the standard temperature modulation vector through the memory enhancement recovery module to obtain a standard attention feature vector and a standard enhancement feature vector.

[0149] Furthermore, the memory enhancement recovery module includes a long short-term memory network, an attention mechanism module and a memory enhancement module, and the first memory enhancement recovery unit includes:

[0150] A timing subunit is used to extract timing features from the statistical feature vector and the temperature modulation vector through a long short-term memory network to obtain a timing feature vector;

[0151] The attention subunit is used to extract key memory features from the time series feature vector through the attention mechanism module to obtain the attention feature vector;

[0152] The memory enhancement subunit is used to perform capacity recovery simulation processing on the time series feature vector through the memory enhancement module to obtain an enhanced feature vector.

[0153] Furthermore, the training process of the discharge capacity detection model is as follows:

[0154] Acquire multiple historical discharge data, perform data preprocessing on each historical discharge data, and obtain a discharge feature set;

[0155] The preset initial discharge capacity detection model is trained using the discharge feature set to obtain training discharge capacity data;

[0156] Based on the preset mean absolute error function, the training loss function value of the discharge feature set is calculated according to the training discharge capacity data;

[0157] When the training loss function value is greater than or equal to the preset standard loss function value, adjusting the network parameters of the initial discharge capacity detection model, and jumping to the step of training the preset initial discharge capacity detection model using the discharge feature set to obtain training discharge capacity data until the training loss function value is less than the standard loss function value;

[0158] When the training loss function value is less than the standard loss function value, a discharge capacity detection model is generated.

[0159] Furthermore, the state of charge estimation module 304 includes:

[0160] a difference submodule, configured to perform difference processing on the standard discharge capacity value and the discharge capacity value to obtain a corresponding first difference;

[0161] The ratio submodule is used to perform ratio processing on the first difference and the preset rated capacity to obtain an estimated state of charge value corresponding to the lead-acid battery.

[0162] See also Figure 5 , Figure 5 This is a structural block diagram of an electronic device provided in Example 4 of the present invention.

[0163] An electronic device according to an embodiment of the present invention includes: a memory 401 and a processor 402, wherein the memory 401 stores a computer program; when the computer program is executed by the processor 402, the processor 402 executes the lead-acid battery state of charge estimation method as described in any of the above embodiments.

[0164] Memory 401 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Memory 401 has storage space 403 for program code 413 for executing any of the method steps described above. For example, storage space 403 for program code may include individual program codes 413 for implementing various steps in the method described above. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards, or floppy disks. The program codes may be compressed, for example, in a suitable format. When executed by a processing device, these codes cause the processing device to execute the various steps in the method described above. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards, or floppy disks. The program codes may be compressed, for example, in a suitable format. When these codes are executed by a computing and processing device, the computing and processing device is caused to execute the steps of the lead-acid battery state of charge estimation method described above.

[0165] The fifth embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for estimating the state of charge of a lead-acid battery according to any of the above embodiments is implemented.

[0166] Embodiment 6 of the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the lead-acid battery state of charge estimation method as described in any of the above embodiments.

[0167] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0168] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0169] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0170] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0171] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0172] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for estimating the state of charge of a lead-acid battery, characterized in that: include: Obtain standard discharge data and discharge data of lead-acid batteries; Standard discharge data refers to the discharge data collected from a complete discharge test under strictly controlled standardized conditions; Discharge data refers to the real-time discharge data recorded during the discharge process; Extracting features from the standard discharge data and the discharge data using a pre-trained multi-scale feature network of a discharge capacity detection model to obtain a multi-modal fusion feature vector and a standard modal fusion feature vector; The residual predictor of the discharge capacity detection model is used to predict the discharge capacity of the multimodal fusion feature vector and the standard modal fusion feature vector respectively, to obtain a discharge capacity value and a standard discharge capacity value; the discharge capacity value refers to the discharged capacity of the lead-acid battery at the current moment; The standard discharge capacitance value refers to the maximum discharge capacity of the lead-acid battery predicted by the discharge capacity detection model; Performing a state of charge analysis on the discharge capacity value and the standard discharge capacity value to obtain an estimated state of charge value corresponding to the lead-acid battery; The multi-scale feature network includes a current integration module, a main network and a feature fusion layer; Performing ampere-hour integration operations on the standard discharge data and the discharge data respectively by the current integration module to obtain an ampere-hour integration vector and a standard ampere-hour integration vector; Performing multi-feature extraction on the standard discharge data and the discharge data through the main network to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector; Performing feature fusion on the attention feature vector, the enhancement feature vector, and the ampere-hour integral vector through a feature fusion layer to obtain a multimodal fusion feature vector; Performing feature fusion on the standard attention feature vector, the standard enhancement feature vector, and the standard ampere-hour integral vector through a feature fusion layer to obtain a standard modality fusion feature vector; performing difference processing on the standard discharge capacity value and the discharge capacity value to obtain a corresponding first difference; The first difference is compared with a preset rated capacity to obtain an estimated state of charge value corresponding to the lead-acid battery.

2. The method for estimating the state of charge of a lead-acid battery according to claim 1, wherein: The main network includes a statistical feature module, a temperature adaptive gating module, and a memory enhancement recovery module. The steps of performing multi-feature extraction on the standard discharge data and the discharge data through the main network to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector include: Performing feature extraction on the standard discharge data and the discharge data respectively by the statistical feature module to obtain a statistical feature vector and a standard statistical feature vector; Performing temperature gating modulation on the statistical feature vector and the standard statistical feature vector respectively by the temperature adaptive gating module to obtain a temperature modulation vector and a standard temperature modulation vector; Performing memory enhancement and recovery processing on the statistical feature vector and the temperature modulation vector by the memory enhancement and recovery module to obtain an attention feature vector and an enhanced feature vector; The memory enhancement recovery module performs memory enhancement recovery processing on the standard statistical feature vector and the standard temperature modulation vector to obtain a standard attention feature vector and a standard enhancement feature vector.

3. The method for estimating the state of charge of a lead-acid battery according to claim 2, wherein: The memory enhancement recovery module includes a long short-term memory network, an attention mechanism module, and a memory enhancement module. The step of performing memory enhancement recovery processing on the statistical feature vector and the temperature modulation vector by the memory enhancement recovery module to obtain an attention feature vector and an enhanced feature vector includes: Performing time series feature extraction on the statistical feature vector and the temperature modulation vector through the long short-term memory network to obtain a time series feature vector; Extract key memory features from the time series feature vector using the attention mechanism module to obtain an attention feature vector; The memory enhancement module performs capacity recovery simulation processing on the time series feature vector to obtain an enhanced feature vector.

4. The method for estimating the state of charge of a lead-acid battery according to claim 1, wherein: The training process of the discharge capacity detection model is specifically as follows: Acquire a plurality of historical discharge data, perform data preprocessing on each of the historical discharge data, and obtain a discharge feature set; Using the discharge feature set to train a preset initial discharge capacity detection model to obtain training discharge capacity data; Calculating a training loss function value of the discharge feature set based on the training discharge capacity data based on a preset mean absolute error function; When the training loss function value is greater than or equal to a preset standard loss function value, adjusting the network parameters of the initial discharge capacity detection model, and jumping to the step of training the preset initial discharge capacity detection model using the discharge feature set to obtain training discharge capacity data, until the training loss function value is less than the standard loss function value; When the training loss function value is less than the standard loss function value, a discharge capacity detection model is generated.

5. A lead-acid battery state of charge estimation system, characterized in that: include: Acquisition module, used to obtain standard discharge data and discharge data of lead-acid batteries; Standard discharge data refers to the discharge data collected from a complete discharge test under strictly controlled standardized conditions; discharge data refers to the real-time discharge data recorded during the discharge process; a feature extraction module, configured to extract features from the standard discharge data and the discharge data respectively using a multi-scale feature network of a pre-trained discharge capacity detection model to obtain a multimodal fusion feature vector and a standard modal fusion feature vector; a capacity prediction module, configured to perform discharge capacity prediction on the multimodal fusion feature vector and the standard modal fusion feature vector respectively using a residual predictor of the discharge capacity detection model to obtain a discharge capacity value and a standard discharge capacity value; the discharge capacity value refers to the discharged capacity of the lead-acid battery at the current moment; The standard discharge capacitance value refers to the maximum discharge capacity of the lead-acid battery predicted by the discharge capacity detection model; a state of charge estimation module, configured to perform a state of charge analysis on the discharge capacity value and the standard discharge capacity value to obtain a state of charge estimation value corresponding to the lead-acid battery; The multi-scale feature network includes a current integration module, a main network and a feature fusion layer. The feature extraction module includes: a current integration submodule, configured to perform an ampere-hour integration operation on the standard discharge data and the discharge data respectively through the current integration module to obtain an ampere-hour integration vector and a standard ampere-hour integration vector; A feature extraction submodule, configured to perform multi-feature extraction on the standard discharge data and the discharge data respectively through the main network to obtain an attention feature vector, an enhanced feature vector, a standard attention feature vector, and a standard enhanced feature vector; A first fusion submodule is configured to perform feature fusion on the attention feature vector, the enhancement feature vector, and the ampere-hour integral vector through a feature fusion layer to obtain a multimodal fusion feature vector; A second fusion submodule is configured to perform feature fusion on the standard attention feature vector, the standard enhancement feature vector, and the standard ampere-hour integral vector through a feature fusion layer to obtain a standard modality fusion feature vector; The state of charge estimation module includes: a difference submodule, configured to perform difference processing on the standard discharge capacity value and the discharge capacity value to obtain a corresponding first difference; The ratio submodule is used to perform ratio processing on the first difference and a preset rated capacity to obtain an estimated state of charge value corresponding to the lead-acid battery.

6. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the lead-acid battery state of charge estimation method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for estimating the state of charge of a lead-acid battery according to any one of claims 1 to 4 is implemented.

8. A computer program product, characterized in that The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer is caused to execute the lead-acid battery state of charge estimation method according to any one of claims 1 to 4.

Images