Energy storage battery power estimation method and device based on multiple incremental features, electronic equipment and medium

Abstract

The application relates to the technical field of battery management, and specifically discloses a kind of energy storage battery power estimation method and device based on multiple incremental features, electronic equipment and medium, comprising the following steps: obtaining test data of battery under multiple ambient temperatures and operating conditions;Incremental feature extraction is carried out from the obtained test data, and a joint feature matrix is formed with the original test data;The joint feature matrix is pretreated to obtain pretreated data;The pretreated data is processed using a sliding window technique to obtain power evaluation data;The power evaluation data is input into a pre-constructed energy storage battery power estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model;The pre-trained SOC estimation model is migrated to a small data set battery to obtain the power of the small data set battery.The application not only improves the accuracy of SOC estimation, but also has strong adaptability and stability, providing strong support for the practical application of the battery management system.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and more specifically to a method, apparatus, electronic device, and medium for estimating the power of an energy storage battery based on multiple incremental features. Background Technology

[0002] In energy storage power stations, the State of Charge (SOC), or battery charge, is a key parameter representing the ratio of the battery's current remaining capacity to its maximum usable capacity. SOC provides users with battery capacity information, helps prevent overcharging and over-discharging of the battery system, and improves battery lifespan, playing a crucial role in the overall control of the energy storage system.

[0003] Currently, most mainstream data-driven methods for estimating battery SOC directly input collected parameters such as current and voltage into the model to obtain the battery SOC, without considering the process information contained in each parameter during the charge and discharge cycle, leading to a decrease in accuracy. At the same time, collecting sufficient battery data is expensive and time-consuming, and the time cost of model training is also not low.

[0004] Therefore, there is an urgent need for a method to estimate the energy capacity of energy storage batteries that can save the cost of model training and data collection while achieving high estimation accuracy. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention provides a method, device, electronic device and medium for estimating the power of energy storage batteries based on multiple incremental features, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for estimating the capacity of an energy storage battery based on multiple incremental features, comprising the following steps:

[0007] S1. Acquire test data of the battery under various ambient temperatures and operating conditions;

[0008] S2. Incremental feature extraction is performed from the acquired test data, and a joint feature matrix is ​​formed with the original test data;

[0009] S3. Preprocess the joint feature matrix to obtain preprocessed data;

[0010] S4. Use sliding window technology to process the preprocessed data to obtain power evaluation data;

[0011] S5. Input the power assessment data into the pre-built energy storage battery power estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model.

[0012] S6. Transfer the pre-trained SOC estimation model to a small dataset of batteries to obtain the battery power of the small dataset.

[0013] Preferably, in step S1, test data of the energy storage battery under various operating conditions at high temperature, normal temperature and low temperature are collected to obtain training samples; the main data collected include battery current, battery voltage and battery casing temperature data during battery operation.

[0014] Preferably, in step S2, the current, voltage and battery casing temperature collected at each time point are differentiated to extract incremental features, and the extracted incremental feature data and the original test data are integrated to form a joint feature matrix containing the original test data and incremental feature data.

[0015] The formula for extracting incremental features is expressed as follows:

[0016]

[0017] In the formula, v t c t and tem t dv represents the voltage, current, and temperature collected at time t, respectively. t ,dc t and dtem t These represent the extracted voltage increment, current increment, and temperature increment, respectively, with Δt representing the increment time interval.

[0018] The joint feature matrix, which includes the original test data and incremental feature data, is expressed by the following formula:

[0019]

[0020] In the formula, n is the length of the acquisition sequence.

[0021] Preferably, in step S3, the preprocessing includes outlier removal and normalization. Specifically, it involves removing obviously outlier values ​​from the joint feature matrix (data points exceeding ±3 standard deviations are considered outliers and deleted) to improve the normalization effect; then, the min-max normalization method is used to normalize all features in the joint feature matrix after outlier removal, using the following formula:

[0022]

[0023] In the formula, X is the feature dataset, min(X) is the minimum value in the dataset, max(X) is the maximum value in the dataset, and x i x' is the original data value. i These are the normalized data values.

[0024] Preferably, in step S4, the preprocessed data is processed using a sliding window technique to obtain power assessment data. Specifically, this includes setting the sliding window size to 100 and the sliding step size to 1 based on the preprocessed data to obtain the power assessment data, expressed by the following formula:

[0025]

[0026] In the formula, S i Let be the joint feature at time i, and n be the length of the acquisition sequence.

[0027] Preferably, in step S5, the energy storage battery capacity estimation model based on multi-incremental features includes:

[0028] Four stacked temporal convolutional networks (TCNs) with different dilation factors: capture long-term dependencies and local features in the power assessment data, and set up a residual connection at the input and output positions;

[0029] Bidirectional gated recurrent unit (BiGRU): By performing GRU operations in opposite directions, it traverses the time series data in both forward and backward directions to capture battery timing features and dependencies from both sides;

[0030] Sparse Attention layer: By introducing sparsification techniques, the computational cost is reduced while retaining the functionality of the original attention mechanism;

[0031] Fully connected layer (FC): Further processes the extracted features, combines them with a non-linear activation function, and maps them to generate the final SOC estimate.

[0032] Preferably, the stacked temporal convolutional networks with four different dilation factors specifically include:

[0033] The first layer is a causal dilated convolutional layer containing 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 1; an average pooling layer with a pooling window size of 3 and a stride of 1; and a batch normalization layer.

[0034] The second causal dilated convolutional layer contains 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 2; an average pooling layer with a pooling window size of 3 and a stride of 1; and a batch normalization layer.

[0035] The third causal dilated convolutional layer contains 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 4; an average pooling layer with a pooling window size of 3 and a stride of 1; and a batch normalization layer.

[0036] The fourth layer is a causal dilated convolutional layer containing 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 8; an average pooling layer with a pooling window size of 3 and a stride of 1; a batch normalization layer; and a Dropout layer with a coefficient of 0.001.

[0037] The fifth residual connection layer outputs the superposition of the input data and the output of the fourth causal dilated convolutional layer.

[0038] The bidirectional gated loop unit specifically includes: a first BIGRU layer with 128 hidden layers; L2 regularization is applied with an L2 coefficient of 0.05; and a Dropout layer with a coefficient of 0.005.

[0039] The sparse attention mechanism layer specifically includes: a Dropout layer with a coefficient of 0.002; and a flattening layer.

[0040] The fully connected layer specifically includes: a first fully connected layer with 32 hidden layers; a Dropout layer with a coefficient of 0.05; a second fully connected layer with 64 hidden layers; a Dropout layer with a coefficient of 0.05; and an output fully connected layer with 1 hidden layer.

[0041] Preferably, in step S6, the weight file of the pre-trained model is loaded and transferred to a small dataset of batteries to obtain the battery power. Specifically, the fine-tuning network structure strategy in transfer learning is used to freeze the TCN layer parameters of the pre-trained model, fine-tune the parameters of BiGRU, SparseAttention, and FC to obtain a fine-tuned power estimation model, and input the small battery evaluation data into the fine-tuned power estimation model to obtain the power of the small dataset of batteries.

[0042] On the other hand, to achieve the above objectives, the present invention also provides the following technical solution: a battery power estimation device based on multiple incremental features, the device comprising the following modules:

[0043] The test data acquisition module acquires test data of the battery under various ambient temperatures and operating conditions.

[0044] The incremental feature extraction module extracts incremental features from the acquired test data and combines them with the original test data to form a joint feature matrix;

[0045] The preprocessing module preprocesses the joint feature matrix to obtain preprocessed data.

[0046] The power assessment data acquisition module uses sliding window technology to process the preprocessed data to obtain power assessment data;

[0047] The pre-training module inputs the power assessment data into a pre-built energy storage battery power estimation model based on multiple incremental features for iterative training, resulting in a pre-trained SOC estimation model.

[0048] The power estimation module transfers the pre-trained SOC estimation model to a small dataset of batteries to obtain the battery power of the small dataset.

[0049] On the other hand, to achieve the above objectives, the present invention also provides the following technical solution: an electronic device, the electronic device comprising: a processor; and a memory for storing one or more programs;

[0050] When the one or more programs are executed by the processor, the processor performs the energy storage battery power estimation method based on multi-incremental features.

[0051] On the other hand, to achieve the above objectives, the present invention also provides the following technical solution: a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the aforementioned method for estimating the energy storage battery capacity based on multiple incremental features.

[0052] The beneficial effects of this invention are: it extracts incremental features from the collected current, voltage, and temperature parameters, and significantly improves the SOC estimation accuracy by fusing the incremental features; at the same time, through transfer learning, even when applied to datasets with small amounts of data, the method of this invention can quickly converge the model and achieve high estimation accuracy, saving the cost of model training and data collection; it can have good estimation accuracy, stability, and adaptability under complex operating conditions, providing strong support for the practical application of battery management systems. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the process steps of the energy storage battery power estimation method based on multiple incremental features in an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the framework of the energy storage battery power estimation method based on multiple incremental features in an embodiment of the present invention;

[0055] Figure 3 This is a schematic diagram of an energy storage battery power estimation model based on multiple incremental features in an embodiment of the present invention;

[0056] Figure 4 The following is a comparison chart of SOC estimation under different ambient temperatures in the embodiments of the present invention: (a) shows the comparison of estimation results at 25℃, (b) shows the comparison of estimation results at 10℃, (c) shows the comparison of estimation results at 0℃, and (d) shows the comparison of estimation results under varying temperatures from 10℃ to 25℃.

[0057] Figure 5 This is a schematic diagram of the energy storage battery power estimation device module based on multiple incremental features in an embodiment of the present invention;

[0058] Figure 6 This is a schematic diagram of the electronic device structure in an embodiment of the present invention;

[0059] In the diagram, 110 is the test data acquisition module; 120 is the incremental feature extraction module; 130 is the preprocessing module; 140 is the power assessment data acquisition module; 150 is the pre-training module; 160 is the power estimation module; 210 is the processor; and 220 is the storage device. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1

[0062] Please see Figures 1-4 This invention provides a technical solution: a method for estimating the capacity of energy storage batteries based on multiple incremental features, such as... Figure 1 and Figure 2 As shown, the specific steps include the following:

[0063] S1. Obtain test data of the battery under various ambient temperatures and operating conditions.

[0064] The test data includes current, voltage, and battery casing temperature collected during battery operation, all acquired using high-precision sensors. Meanwhile, the SOC label during model training is approximated using the ampere-hour integration method, calculated as follows:

[0065]

[0066] In the formula, Cap r Cap indicates the current remaining usable capacity of the battery. m This indicates the maximum usable capacity of the battery.

[0067] By analyzing data such as current, voltage, and temperature, a comprehensive assessment of the battery's capacity, lifespan, safety, and potential problems can be achieved.

[0068] It is understood that the SOC estimation device includes battery-powered electronic devices such as automotive control platforms, portable laptops, smartphones, or tablets.

[0069] S2. Incremental feature extraction is performed from the acquired test data, and a joint feature matrix is ​​formed with the original test data.

[0070] Incremental features, including current increment, voltage increment, and temperature increment, are extracted from battery test data. These incremental features reflect the state changes of the battery between adjacent time steps, providing detailed process information about the charging and discharging process. By fusing these dynamic features with the originally acquired features, the SOC estimation model can better capture the complex behavior of the battery under different operating conditions, thereby significantly improving the accuracy of SOC estimation.

[0071] The current, voltage, and battery casing temperature collected at each time point are differentiated to extract incremental features. The extracted incremental feature data and the original test data are integrated to form a joint feature matrix containing the original test data and incremental feature data.

[0072] Taking the acquisition of measurement data at a frequency of 1 Hz as an example, let the acquired voltage be V, current be C, and temperature be tem. Then, the incremental features are extracted using the following formula:

[0073]

[0074] In the formula, dc t ,dc t and dtem t These are the voltage increment features, current increment features, and temperature increment features extracted at time t, respectively. t c t and tem t These are the voltage, current, and temperature data at time t, respectively, where Δt is the characteristic change time interval, set to 1 second.

[0075] Subsequently, the extracted incremental feature data and the original measurement data are concatenated to obtain a joint feature matrix for further processing. The joint feature matrix is ​​as follows:

[0076]

[0077] In the formula, n is the length of the measurement data.

[0078] S3. Preprocess the joint feature matrix to obtain preprocessed data.

[0079] The preprocessing operations include removing outliers and normalization.

[0080] Specifically, removing outliers reduces noise and bias, improving the robustness and accuracy of the SOC estimation model based on multi-incremental features. Normalization makes the feature data comparable, accelerating the training speed of the subsequent SOC estimation model and improving its stability.

[0081] S31. Remove the data collected at other times that are abnormal from the collected data, and obtain the data after anomaly handling.

[0082] Specifically, data points exceeding ±3 standard deviations are considered outliers and deleted. The specific screening formula is as follows:

[0083]

[0084] In the formula, N is the number of data points collected, and x i Let be the i-th data point, μ be the mean of the collected data, and σ be the standard deviation of the collected data.

[0085] Based on the above, the range of outliers is determined, namely, the upper limit μ+3σ and the lower limit μ-3σ. If data point x... i Satisfy x i >μ+3σ or x i If the value is less than μ-3σ, then the data point x i If it is considered an outlier, remove it.

[0086] Outliers can affect the uniformity of measurement and cause the model to misunderstand the data, thus producing bias. The above-mentioned outlier removal operations can prevent this from happening.

[0087] S32. Normalize the data after anomaly handling using the max-min normalization method.

[0088] To avoid the adverse effects of dimensional differences on model training and to accelerate training speed, the max-min normalization method can be used to normalize the dataset. This method can adjust the numerical ranges of different features to a consistent level.

[0089] Specifically, min-max normalization eliminates dimensional differences between features by transforming the data to a fixed range, [0,1]. This not only allows feature data to be compared on the same scale, but also significantly improves the efficiency and effectiveness of subsequent SOC estimation model training.

[0090] First, taking the current characteristics in the collected data as an example, let the current collection sequence be X, the maximum value of the current characteristics be max(X), and the minimum value be min(X). Then, the normalized current value is obtained through the following formula (maximum-minimum normalized model):

[0091]

[0092] In the formula, x i x' represents the original current data value. iThis is the normalized current data value. Similarly, the other features in the joint feature matrix are normalized sequentially to obtain the normalized data.

[0093] S4. Use sliding window technology to process the preprocessed data to obtain power assessment data.

[0094] Based on the preprocessed data, a sliding window technique is used to process it to obtain battery capacity assessment data. Theoretically, this data contains historical information from multiple time steps, which is beneficial for fitting the SOC estimation model.

[0095] Specifically, the sliding window technique generates multiple samples by sliding a fixed-size window across time-series data. Each sample contains a segment of historical data and a corresponding prediction target. These samples can be used to train a supervised learning model, enabling it to predict future values ​​based on past observations.

[0096] Specifically, the window size is set to 100, and the sliding step size is 1. For the measurement data, assume its size is (num_samples, num_features), where num_samples is the data sequence length and num_features is the number of features (num_features = 6). The sliding window technique is used to convert it into three-dimensional data, i.e., power assessment data, with a shape of (num_new_samples, window_size, num_features), where num_new_samples is the number of samples after processing by the sliding window, and window_size is the size of the sliding window.

[0097] The power consumption assessment data is as follows:

[0098]

[0099] In the formula, S i The joint feature at time i, i.e., [v i ,c i ,tem i ,dv i ,dc i ,dtem i ], where n is the length of the acquisition sequence.

[0100] Preferably, the sliding window technique is also used, with the window size set to 100 and the step size to 1, to obtain the SOC value corresponding to the input data. The calculation formula is as follows:

[0101] SOC i =F(X) i )

[0102] In the formula, SOCi X represents the SOC value at time i. i This represents the values ​​of current, voltage, temperature, current increment, voltage increment, and temperature increment over the past 100 historical sampling times starting from time i.

[0103] Finally, the power assessment data used as input to the SOC estimation model is obtained.

[0104] S5. Input the power assessment data into the pre-built energy storage battery power estimation model based on multi-incremental features for iterative training to obtain a pre-trained SOC estimation model.

[0105] In this embodiment, when the SOC estimation model estimates the SOC at time i, the model input includes sliding window test data used to estimate the SOC at that time, i.e.:

[0106] SOC i =MulIFModel([S i ,S i-1 ,S i-2 ,...,S i-99 ])

[0107] Where MulIFModel represents the power estimation model based on multi-incremental features, and SOC i Let S be the SOC at time i, [S i ,S i-1 ,S i-2 ,...,S i-99 [] represents 100 historical data points and extracted incremental feature data starting from time i.

[0108] A pre-built energy storage battery capacity estimation model (SOC estimation model) based on multi-incremental features has the following structure: Figure 3 As shown, some Dropout operations between modules are omitted. The battery power estimation model based on multi-incremental features mainly includes a stacked TCN module, a BiGRU module, a sparse attention mechanism module, and a fully connected layer. The TCN module is used to capture long-term dependencies and local features in the evaluation data. The BiGRU module uses GRU operations in opposite directions to traverse the time-series data forward and backward, capturing temporal features and dependencies from both sides. The sparse attention mechanism module reduces computation by introducing sparsification techniques while retaining the functionality of the original attention mechanism. The fully connected layer further processes the extracted features and combines them with a non-linear activation function to generate the final SOC estimate.

[0109] The SOC estimation model is built based on the Tensorflow 2.16 deep learning framework, specifically including:

[0110] Four stacked temporal convolutional networks with different dilation factors:

[0111] The first layer is a causal dilated convolutional layer containing 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 1; an average pooling layer with a pooling window size of 3 and a stride of 1; and a batch normalization layer.

[0112] The second causal dilated convolutional layer contains 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 2; an average pooling layer with a pooling window size of 3 and a stride of 1; and a batch normalization layer.

[0113] The third causal dilated convolutional layer contains 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 4; an average pooling layer with a pooling window size of 3 and a stride of 1; and a batch normalization layer.

[0114] The fourth layer is a causal dilated convolutional layer containing 64 convolutional kernels of size 3, with a stride of 1 and an inflation factor of 8; an average pooling layer with a pooling window size of 3 and a stride of 1; a batch normalization layer; and a Dropout layer with a coefficient of 0.001.

[0115] The fifth residual connection layer outputs the sum of the input data and the output of the fourth causal dilated convolutional layer.

[0116] The bidirectional gated loop unit specifically includes: a first BIGRU layer with 128 hidden layers; L2 regularization is applied with an L2 coefficient of 0.05; and a Dropout layer with a coefficient of 0.005.

[0117] A sparse attention mechanism layer is connected after the two feature extraction modules mentioned above. The sparse attention mechanism layer specifically includes: a Dropout layer with a coefficient of 0.002; and a flattening layer.

[0118] The fully connected layer specifically includes: a first fully connected layer with 32 hidden layers; a Dropout layer with a coefficient of 0.05; a second fully connected layer with 64 hidden layers; a Dropout layer with a coefficient of 0.05; and an output fully connected layer with 1 hidden layer.

[0119] The specific parameter settings for the MulIFModel mentioned above are shown in Table 1:

[0120] Table 1 MulIFModel Parameter Settings Table

[0121]

[0122]

[0123] The computational process of the sparse attention mechanism is as follows:

[0124]

[0125] In the formula, Q, K, and V are the query matrix, key matrix, and value matrix, respectively; X is the input matrix; W Q W K W V These are the corresponding weight matrices.

[0126] Subsequently, the key matrix K is sampled to obtain the sampled matrix. Then query the i-th row q of matrix Q. i ∈Q, find information about The M value, that is:

[0127]

[0128] In the formula, To evaluate q i Is it an important indicator? for The j-th row; d is the scaling factor; d is the number of columns in K; L k Let K be the number of rows.

[0129] Next, obtain the q values ​​of u with the largest M value. i , this u q i Form a new query matrix Finally, we obtain sparse self-attention:

[0130]

[0131] Model training

[0132] First, battery samples were obtained: the publicly available Panasonic 18650PF dataset was used, which collected current, voltage, and temperature data from tests conducted at ambient temperatures of -20℃, -10℃, 0℃, 10℃, and 25℃. Steps S1 to S4 were performed on this dataset, and then the dataset was divided into training and test sets. Specifically, the UDDS, US06, HWFET, LA92, cycle1, and cycle2 operating condition data at each ambient temperature were used as the training set, and cycle3 and cycle4 were used as the test set to evaluate model performance. The cycle condition represents a random mixture of the four common automotive operating conditions mentioned above.

[0133] Subsequently, the hyperparameters for training the SOC estimation model were set, and their specific parameters are shown in Table 2. The loss function was set to mean squared error (MSE), and its calculation formula is as follows:

[0134]

[0135] In the formula, Let y be the estimated value of SOC. i This is the actual SOC value.

[0136] Table 2 Hyperparameter Settings for Model Training

[0137] hyperparameters set up Batchsize 1024 Optimizer Adam Lossfunction MSE Learningrate 0.001 Epochs 1000

[0138] The learning rate employs a decay strategy. Specifically, the initial learning rate is set to 0.001, and then decays every 50 cycles with a decay factor of 0.9. This setting allows the model to adjust more smoothly in the solution space, improving the stability and accuracy of model convergence.

[0139] During training, an Early Stopping strategy is used. This is a simple yet effective regularization technique that prevents overfitting by monitoring validation set performance and stopping training early when appropriate, thereby improving the model's generalization ability and performance. Its patience value is set to 50.

[0140] The model optimization algorithm is set to Adam, and the model parameters are optimized through backpropagation. After multiple training iterations, the model's predicted values ​​can continuously approach the true values.

[0141] After the model is trained, it can be used to estimate battery capacity. A comparison of the estimation results under different ambient temperatures is shown in the following figure. Figure 4 As shown in the figure, (a) represents the comparison of estimation results at 25℃, (b) represents the comparison of estimation results at 10℃, (c) represents the comparison of estimation results at 0℃, and (d) represents the comparison of estimation results under varying temperatures from 10℃ to 25℃. As can be seen from the figure, the SOC estimation performance incorporating incremental features is more stable than that incorporating only collected features, and the estimation accuracy is also higher.

[0142] S6. Transfer the pre-trained SOC estimation model to a small dataset of batteries to obtain the battery power of the small dataset.

[0143] Based on a pre-trained SOC power estimation model, the model is transferred by loading the weight file of the pre-trained model to obtain the pre-trained model. Specifically, in the network structure fine-tuning stage, the network parameters of the TCN layer in the pre-trained model are frozen, and the parameters of the remaining layers are trained. This is because the pre-trained model has already learned rich feature representations on large-scale datasets, and fine-tuning allows the model to better adapt to specific tasks and improve prediction accuracy.

[0144] The strategy for estimating battery capacity using transfer learning is implemented as follows:

[0145] First, battery samples were obtained. The publicly available LG 18650HG2 dataset was used for transfer learning. The data collected in this dataset is similar to that of the Panasonic 18650PF mentioned above, except that the materials of the two batteries are different. Steps S1 to S4 were performed on this dataset in the same way. Then, the pre-trained model was fine-tuned using the UDDS and US06 operating condition data in the dataset. The performance of the transfer learning model was evaluated using eight mixed operating conditions as a test set.

[0146] The next step is to set the hyperparameters for model training, which are the same as in step S5. After fine-tuning the pre-trained model, a new battery estimation model is obtained, which can then be used to evaluate its performance on a small dataset of batteries. The battery battery estimation method of this embodiment pre-trains a model on a large dataset and uses a transfer learning fine-tuning strategy to achieve battery estimation on a small dataset. The proposed method can increase the model's generalization ability and reduce the risk of overfitting by incorporating knowledge and experience from other datasets. Simultaneously, it fully utilizes existing data and model resources, avoiding the large amount of time and computational resources required to train a new model from scratch.

[0147] The energy storage battery capacity estimation method based on multi-incremental features of this invention extracts incremental features of the battery charging and discharging process, and can achieve good estimation accuracy under a wide range of operating conditions, including high and low temperature environments. It shows excellent performance in terms of MAE, RMSE and MAX error indicators in the estimation results, and effectively solves the problem of accuracy degradation caused by not considering battery charging and discharging process information in existing methods.

[0148] Based on the same inventive concept as the above-described method embodiments, this application also provides an apparatus for a multi-incremental feature-based energy storage battery capacity estimation method. This apparatus can achieve the functions provided by the above-described method embodiments, such as... Figure 5 As shown, the device includes the following modules:

[0149] Test data acquisition module 110 acquires test data of the battery under various ambient temperatures and operating conditions;

[0150] Incremental feature extraction module 120 extracts incremental features from the acquired test data and forms a joint feature matrix with the original test data;

[0151] Preprocessing module 130 preprocesses the joint feature matrix to obtain preprocessed data;

[0152] The power assessment data acquisition module 140 uses sliding window technology to process the preprocessed data to obtain power assessment data;

[0153] The pre-training module 150 inputs the power assessment data into a pre-built energy storage battery power estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model.

[0154] The power estimation module 160 transfers the pre-trained SOC estimation model to a small dataset of batteries to obtain the power of the small dataset of batteries.

[0155] Based on the same inventive concept as the above-described method embodiments, this application also provides an electronic device, such as... Figure 6 As shown, the device includes: a processor 210; and a memory 220 for storing one or more programs;

[0156] When the one or more programs are executed by the processor 210, the processor performs the energy storage battery power estimation method based on multi-incremental features.

[0157] The energy storage battery capacity estimation method based on multi-incremental features specifically includes the following:

[0158] Acquire test data for batteries under various ambient temperatures and operating conditions;

[0159] Incremental features are extracted from the acquired test data and combined with the original test data to form a joint feature matrix;

[0160] The joint feature matrix is ​​preprocessed to obtain preprocessed data;

[0161] The preprocessed data is processed using the sliding window technique to obtain power assessment data;

[0162] The power assessment data is input into a pre-built energy storage battery power estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model.

[0163] The pre-trained SOC estimation model is transferred to a small dataset of batteries to obtain the battery power of the small dataset.

[0164] Based on the same inventive concept as the above-described method embodiments, this application embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by processor 210, implements the energy storage battery power estimation method based on multi-incremental features.

[0165] The energy storage battery capacity estimation method based on multi-incremental features specifically includes the following:

[0166] Acquire test data for batteries under various ambient temperatures and operating conditions;

[0167] Incremental features are extracted from the acquired test data and combined with the original test data to form a joint feature matrix;

[0168] The joint feature matrix is ​​preprocessed to obtain preprocessed data;

[0169] The preprocessed data is processed using the sliding window technique to obtain power assessment data;

[0170] The power assessment data is input into a pre-built energy storage battery power estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model.

[0171] The pre-trained SOC estimation model is transferred to a small dataset of batteries to obtain the battery power of the small dataset.

[0172] In the several embodiments provided in this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0173] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0174] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a 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 to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0175] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0176] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0177] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0178] The terms "first" and "second" used in the embodiments are merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" can be interchanged in a specific order or sequence where permitted. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0179] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for state of charge estimation of an energy storage battery based on multiple incremental features, the method comprising: The method comprises the following steps: S1, obtaining test data of the battery under various environmental temperatures and operating conditions; S2, incremental feature extraction is performed from the obtained test data, and a joint feature matrix is formed with the original test data; S3, preprocessing the joint feature matrix to obtain preprocessed data; S4, using a sliding window technique to process the preprocessed data to obtain power evaluation data; S5, inputting the power evaluation data into a pre-constructed energy storage battery power estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model; S6, migrating the pre-trained SOC estimation model to a small data set battery to obtain the power of the small data set battery.

2. The multi-incremental-feature-based energy storage battery state of charge estimation method of claim 1, wherein: In step S1, test data of the energy storage battery under high temperature, normal temperature and low temperature, and various operating conditions is collected to obtain training samples; the main collected data includes battery current, battery voltage and battery shell temperature data during battery operation.

3. The multi-delta feature based energy storage battery state of charge estimation method of claim 1, wherein: In step S2, the current, voltage and battery shell temperature collected at each time point are differentiated to extract incremental features, and the extracted incremental feature data and the original test data are integrated to form a joint feature matrix containing the original test data and the incremental feature data; The extraction formula of the incremental feature is as follows: wherein v t , c t and tem t represent the voltage, current and temperature collected at time t, respectively, dv t , dc t and dtem t represent the extracted voltage increment, current increment and temperature increment, respectively, and Δt represents the increment time interval; The joint feature matrix containing the original test data and the incremental feature data is expressed by the following formula: In the formula, n is the length of the collected sequence.

4. The multi-incremental-feature-based energy storage battery state of charge estimation method of claim 1, wherein: In step S3, the preprocessing includes removing outliers and normalization, and the specific operation is as follows: the obviously abnormal values in the joint feature matrix are removed, and then all features in the joint feature matrix after the abnormal value processing are normalized by using the maximum and minimum normalization method, and the calculation formula is as follows: where X is the characteristic data set, min(X) is the smallest value in the data set, max(X) is the largest value in the data set, x i is the original data value, and x i is the normalized data value.

5. The multi-incremental-feature-based energy storage battery state of charge estimation method of claim 1, wherein: In step S4, the sliding window technique is used to process the preprocessed data to obtain power evaluation data, which specifically includes: according to the preprocessed data, the size of the sliding window is set to 100, the sliding step is set to 1, and the power evaluation data is obtained, and the formula is as follows: In the formula, S i is the joint feature at time i, and n is the length of the sequence collected.

6. The multi-incremental-feature-based energy storage battery state of charge estimation method of claim 1, wherein: In step S5, the energy storage battery power estimation model based on multiple incremental features comprises: Stacked four time convolution networks with different expansion factors: capture long-time dependencies and local features in the power evaluation data, and set a residual connection at the input and output positions; Bidirectional gated recurrent unit: traverse the time series data forward and backward through GRU operation in the opposite direction to capture battery time series features and dependencies from both sides; Sparse attention mechanism layer: by introducing a sparse technology, the calculation amount is reduced while the function of the original attention mechanism is retained; Fully connected layer: further processes the extracted features, combines a nonlinear activation function, and maps to generate the final SOC estimation.

7. The multi-incremental-feature-based energy storage battery state of charge estimation method of claim 6, wherein: The stacked four time convolution networks with different expansion factors specifically comprise: The first causal dilated convolution layer includes 64 convolution kernels with a kernel size of 3, a convolution step of 1, and an expansion factor of 1; the average pooling layer has a pooling window size of 3 and a step of 1; and the batch normalization layer; The second layer causal dilated convolution layer includes 64 convolution kernels with a kernel size of 3, a convolution step of 1, and a dilated factor of 2; an average pooling layer with a pooling window size of 3 and a step of 1; and a batch normalization layer; The third layer causal dilated convolution layer includes 64 convolution kernels with a kernel size of 3, a convolution step of 1, and a dilated factor of 4; an average pooling layer with a pooling window size of 3 and a step of 1; and a batch normalization layer; The fourth layer causal dilated convolution layer includes 64 convolution kernels with a kernel size of 3, a convolution step of 1, and a dilated factor of 8; an average pooling layer with a pooling window size of 3 and a step of 1; a dropout layer with a coefficient of 0.001; and a batch normalization layer; The fifth layer residual connection layer outputs a superimposed value of the input data and the output result of the fourth layer causal dilated convolution layer; The bidirectional gated recurrent unit specifically includes: a first layer BIGRU layer with a number of hidden layers of 128; an L2 regularization applied with an L2 coefficient of 0.05; and a dropout layer with a coefficient of 0.005; The sparse attention mechanism layer specifically includes: a dropout layer with a coefficient of 0.002; and a flattening layer. The full connection layer specifically includes: a first layer full connection layer with a number of hidden layers of 32; a dropout layer with a coefficient of 0.05; a second layer full connection layer with a number of hidden layers of 64; a dropout layer with a coefficient of 0.05; and an output full connection layer with a number of hidden layers of 1.

8. An apparatus for estimating the state of charge of an energy storage battery based on multiple incremental features according to any one of claims 1-7, wherein: The device includes the following modules: A test data acquisition module (110) acquires test data of a battery under various environmental temperatures and operating conditions; An incremental feature extraction module (120) extracts incremental features from the acquired test data and forms a joint feature matrix with original test data; A preprocessing module (130) pre-processes the joint feature matrix to obtain pre-processed data; An electric quantity evaluation data acquisition module (140) processes the pre-processed data using a sliding window technique to obtain electric quantity evaluation data; A pre-training module (150) inputs the electric quantity evaluation data into a pre-constructed energy storage battery electric quantity estimation model based on multiple incremental features for iterative training to obtain a pre-trained SOC estimation model; An electric quantity estimation module (160) migrates the pre-trained SOC estimation model to a small data set battery to obtain the electric quantity of the small data set battery.

9. An electronic device, comprising: The electronic device includes a processor (210) and a memory (220) for storing one or more programs; When the one or more programs are executed by the processor (210), the processor performs the energy storage battery electric quantity estimation method based on multiple incremental features according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is executed by the processor (210) to implement the energy storage battery electric quantity estimation method based on multiple incremental features according to any one of claims 1-7.

Images