Battery charge state estimation method and device and storage medium

Abstract

The invention discloses a battery state-of-charge estimation method, a computer device and a storage medium, and the method comprises the steps: building a time sequence hybrid neural network model, obtaining at least one piece of to-be-estimated discharge data of a to-be-estimated battery, processing the to-be-estimated discharge data through the time sequence hybrid neural network model, and obtaining the to-be-estimated discharge data of the to-be-estimated battery. And carrying out physically constrained filtering processing on a processing result of the time sequence hybrid neural network model to obtain a charge state estimation value of the to-be-estimated battery, and the like. According to the method, the time sequence characteristics in the discharge data to be estimated can be captured by using the mixed time sequence neural network model and physically constrained filtering processing, so that the state of charge estimation value can be accurately estimated, and the SOC estimation precision can be improved. In addition, filtering with physical constraints is additionally applied to the output of the time sequence hybrid neural network, the output result of the time sequence hybrid neural network can be constrained and smoothed, and the precision of model estimation is further improved. The method is widely applied to the technical field of battery management.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a battery state of charge estimation method, a computer device and a storage medium. Background Art

[0002] Batteries are widely used to power consumer electronics, industrial electronics, electric equipment and other equipment. Especially with the demand for the development of new energy sources, batteries have been widely used in many fields such as power supply and energy storage, transportation and aerospace equipment. The battery management system is crucial to maintaining the safe and stable operation of the battery, among which the detection and management of the battery's state of charge (SOC) is the most critical. However, there is still a lack of technical means to directly measure the battery's state of charge. Generally, the battery's state of charge can only be estimated by detecting other parameters of the battery. However, since the battery's state of charge is affected by many factors, the current technology has the disadvantages of large estimation deviations and unstable results. Summary of the invention

[0003] In view of the current technical problems such as large deviation and unstable results in estimating the charge state of a battery, the present invention aims to provide a battery charge state estimation method, a computer device and a storage medium.

[0004] In one aspect, an embodiment of the present invention includes a method for estimating a battery state of charge, the method comprising the following steps:

[0005] Establishing a time series hybrid neural network model; the time series hybrid neural network model is used to capture time series features of input data, learn time dependencies of the captured time series features, and perform regression prediction on the learned time dependencies;

[0006] Acquiring at least one to-be-estimated discharge data of the to-be-estimated battery;

[0007] Using the time series hybrid neural network model to process each of the discharge data to be estimated;

[0008] The processing result of the time series hybrid neural network model is subjected to physical constraint filtering processing to obtain an estimated value of the state of charge of the battery to be estimated.

[0009] Furthermore, the establishing of the temporal hybrid neural network model comprises:

[0010] Establishing a time series convolutional neural network, a long short-term memory network and a fully connected layer; the time series convolutional neural network is used to capture the time series features of the input data, the long short-term memory network is used to learn the time dependency relationship of the time series features captured by the time series convolutional neural network, and the fully connected layer is used to perform regression prediction on the time dependency relationship learned by the long short-term memory network;

[0011] The temporal convolutional neural network, the long short-term memory network and the fully connected layer are connected in sequence to obtain the temporal hybrid neural network model.

[0012] Furthermore, the establishing of the temporal hybrid neural network model comprises:

[0013] Obtaining sample discharge data and sample state of charge data of the sample battery corresponding to the same time;

[0014] The time series hybrid neural network model is trained using the sample discharge data and the sample state of charge data.

[0015] Further, the using the sample discharge data and the sample state of charge data to train the time series hybrid neural network model includes:

[0016] Establishing a plurality of hyperparameter sets; each of the hyperparameter sets includes a plurality of hyperparameters;

[0017] Execute multiple rounds of iterations until the iteration end condition is met;

[0018] Return the hyperparameter set corresponding to the model to be evaluated with the best performance evaluation in the last round of iterations;

[0019] Using the returned hyperparameter set to set the temporal hybrid neural network model to obtain the trained temporal hybrid neural network model;

[0020] Wherein, any round of the iteration process includes the following steps:

[0021] Using each of the hyperparameter sets to set the temporal hybrid neural network model respectively, a plurality of models to be evaluated are obtained; wherein any of the models to be evaluated is obtained by setting the temporal hybrid neural network model by a corresponding hyperparameter set;

[0022] Using the sample discharge data and the sample state of charge data, respectively training each of the models to be evaluated;

[0023] Performing performance evaluation on each of the trained models to be evaluated;

[0024] For at least one of the models to be evaluated with the worst performance evaluation result, the hyperparameter sets corresponding to each of the models to be evaluated are updated.

[0025] Furthermore, the processing result of the time series hybrid neural network model is subjected to physical constraint filtering processing to obtain the estimated value of the state of charge of the battery to be estimated, including:

[0026] Determine the filter window size and first moment;

[0027] Determine a plurality of second moments within a time length equal to the size of the filtering window before the first moment; the first moment and each of the second moments are respectively a collection moment of a corresponding one of the discharge data to be estimated;

[0028] Obtaining the to-be-filtered data corresponding to each of the second moments;

[0029] Arranging the data to be filtered into a first time series according to time sequence;

[0030] For any of the data to be filtered, converting the data according to the physical constraints to which the battery to be estimated is subject between the first moment and the second moment corresponding to the data to be filtered, to obtain a corresponding conversion value;

[0031] Arranging the conversion values ​​into a second time series according to the order of the first time series;

[0032] Adding the processing result of the time series hybrid neural network model on the discharge data to be estimated collected at the first moment to the end of the second time series to obtain a third time series;

[0033] Perform weighted smoothing filtering on the third time series to obtain the state of charge estimation value corresponding to the first moment.

[0034] Further, the obtaining the to-be-filtered data corresponding to each of the second moments includes:

[0035] For any of the second moments:

[0036] When the state of charge estimation value corresponding to the second moment exists, taking the state of charge estimation value corresponding to the second moment as the to-be-filtered data corresponding to the second moment;

[0037] When there is no corresponding state of charge estimation value at the second moment, the processing result of the time series hybrid neural network model on the discharge data to be estimated collected at the second moment is used as the data to be filtered corresponding to the second moment.

[0038] Further, performing conversion according to the physical constraints of the battery to be estimated between the first moment and the second moment corresponding to the data to be filtered to obtain a corresponding conversion value includes:

[0039] Acquiring battery parameters of the battery to be estimated;

[0040] The data to be filtered is used as an initial value, the second moment is used as an integral lower limit, the first moment is used as an integral upper limit, and ampere-hour integration is performed according to the battery parameters to obtain the conversion value.

[0041] Further, performing weighted smoothing filtering on the third time series to obtain the state of charge estimation value corresponding to the first moment includes:

[0042] Perform Gaussian weight function calculation according to the third time series to obtain a weight set;

[0043] The third time series is weighted averaged according to the weight set to obtain the state of charge estimation value corresponding to the first moment.

[0044] On the other hand, an embodiment of the present invention also includes a computer device, including a memory and a processor, the memory is used to store at least one program, and the processor is used to load at least one program to execute the battery state of charge estimation method in the embodiment.

[0045] On the other hand, an embodiment of the present invention further includes a computer-readable storage medium storing a program executable by a processor. When the program executable by the processor is executed by the processor, it is used to execute the battery state of charge estimation method in the embodiment.

[0046] The beneficial effects of the present invention are as follows: the battery state of charge estimation method in the embodiment improves the estimation capability and accuracy of the state of charge estimation value of the battery to be estimated by using a hybrid time series neural network model and a physically constrained filtering process, and combining these two algorithms. The hybrid time series neural network model can capture the time series characteristics in the discharge data to be estimated, thereby accurately estimating the state of charge estimation value, and can improve the accuracy of the estimated SOC. In addition, additionally applying a physically constrained filter to the output of the time series hybrid neural network can constrain and smooth the output results of the time series hybrid neural network, further improving the accuracy and stability of the model estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of the steps of the battery state of charge estimation method in an embodiment;

[0048] Figure 2 Schematic diagram of the principle of training a temporal hybrid neural network model in an embodiment;

[0049] Figure 3 Schematic diagram of a process of using a trained time series hybrid neural network model to obtain a state of charge estimation value of a battery to be estimated in an embodiment. DETAILED DESCRIPTION

[0050] Terminology explanation:

[0051] Data-driven model: Data-driven means that the model adapts to the data, and the model changes according to different data. Typical data-driven models include deep data network models, etc.

[0052] Neural Network: A neural network is a computational model that simulates the way the human brain works. It is the basis of the fields of deep learning and machine learning. Neural networks consist of a large number of nodes (or "neurons") that are interconnected in the network and can process complex data inputs and perform various tasks such as classification, regression, pattern recognition, etc.

[0053] Battery Management System: A battery management system is an electronic device that can monitor and manage batteries. It collects and calculates parameters such as voltage, current, temperature, and SOC, and then controls the battery's charge and discharge process, thereby protecting the battery and improving its overall performance.

[0054] SOC: State of charge, is used to describe the current battery power, which can be described by the "ratio of remaining available capacity / maximum available capacity".

[0055] TCN: Temporal convolutional network, TCN model is a time series prediction model based on convolutional neural network, which has certain advantages over CNN in processing time series data.

[0056] LSTM: Long Short-Term Memory, is a time recurrent neural network that is suitable for processing and predicting important events with relatively long intervals and delays in time series.

[0057] Fully connected layer: The fully connected layer plays the role of a "classifier" in the entire convolutional neural network. If the operations such as the convolution layer, pooling layer, and activation function are to map the original data to the hidden layer feature space, the fully connected layer plays the role of mapping the learned "distributed feature representation" to the sample label space.

[0058] Neural networks can extract characteristic information from complex data for classification or feature recognition. Therefore, it is possible to consider using neural networks to identify some precisely measurable data of the battery to estimate the battery's state of charge. However, if only a single neural network model is used for SOC estimation, or the timing characteristics of the battery SOC are not taken into account in the design of the model framework, it is easy to lead to low efficiency and accuracy of the model. Moreover, the SOC value estimated by the data-driven neural network model often fluctuates, which does not match the actual SOC curve of the battery.

[0059] Based on the above principle, this embodiment provides a battery state of charge estimation method. Figure 1 , the battery state of charge estimation method comprises the following steps:

[0060] S1. Establish a temporal hybrid neural network model;

[0061] S2. Obtain at least one discharge data to be estimated of the battery to be estimated;

[0062] S3. Processing each discharge data to be estimated using a time series hybrid neural network model;

[0063] S4. Perform physical-constrained filtering on the processing results of the time series hybrid neural network model to obtain an estimated value of the state of charge of the battery to be estimated.

[0064] The battery state of charge estimation method in this embodiment can be executed by a computer.

[0065] In step S1, the time series hybrid neural network model has the performance of capturing the time series features of the input data input to the time series hybrid neural network model, learning the time dependency of the captured time series features, and performing regression prediction on the learned time dependency, and can effectively extract the time series features of the input data. Specifically, in the training stage of the time series hybrid neural network model, the input data of the time series hybrid neural network model can be the sample discharge data detected during the discharge process of the sample battery; in the stage of using the time series hybrid neural network model to estimate the state of charge of the battery to be estimated, the input data of the time series hybrid neural network model is the discharge data to be estimated detected during the discharge process of the battery to be estimated.

[0066] In this embodiment, the sample battery is a battery used to collect data to train the time series hybrid neural network model, and the battery to be estimated is a battery whose state of charge needs to be estimated using the trained time series hybrid neural network model. The sample battery and the battery to be estimated can both be lithium-ion batteries.

[0067] In this embodiment, the temporal hybrid neural network model established in step S1 includes a temporal convolutional neural network, a long short-term memory network and a fully connected layer, and these three parts are connected in sequence. Among them, the temporal convolutional neural network is used to capture the temporal features of the input data, the long short-term memory network is used to learn the time dependency of the temporal features captured by the temporal convolutional neural network, and the fully connected layer is used to perform regression prediction on the time dependency learned by the long short-term memory network.

[0068] In this embodiment, after executing step S1 to establish the temporal hybrid neural network model, the temporal hybrid neural network model can be trained. The principle of training the temporal hybrid neural network model is as follows: Figure 2 shown.

[0069] Reference Figure 2 , the discharge cycle of the sample battery can be measured by interpolation sampling in the charge and discharge cycle of the sample battery in the laboratory environment. For example, for any time t0, in a period of time before time t0 (including time t, t+1, ... t0-1, etc., where these time periods can be in seconds), the discharge voltage, discharge current and operating temperature values ​​of the sample battery are detected to obtain the sample discharge data of the sample battery at time t0 The sample state of charge data of the sample battery at time t0 is determined by using a coulomb meter or other methods. Thus, a pair of training data is formed. Since the time t0 is arbitrary, multiple pairs of training data can be obtained based on the same principle. After removing outliers and standardizing these training data, they are divided into the SOC estimation model training set and the SOC estimation model verification set in a ratio of 4:1.

[0070] When training the time series hybrid neural network model, take the sample discharge data collected at time t0 as an example, refer to Figure 2 , the sample discharge data collected at time t0 Input into the time series hybrid neural network, and the time series convolutional neural network TCN will process the input data. Capture time series features, use the long short-term memory network to learn the time dependency of the time series features captured by the time series convolutional neural network, and use the fully connected layer to perform regression prediction on the time dependency learned by the long short-term memory network. The output of the fully connected layer represents the time series hybrid neural network based on the input data. The estimated state of charge of the sample battery at time t0, i.e., the estimated value y″ i , and the sample state of charge data of the sample battery at time t0 Then the true value y iAfter inputting n sample discharge data, n estimated values ​​will be obtained. Using n sample state of charge data as the true value,

[0071] Root Mean Square Error (RMSE):

[0072] And the mean absolute error (MAE):

[0073] The time series hybrid neural network model is evaluated by using the error function. If the error function value is less than the threshold, it indicates that the training of the time series hybrid neural network model is completed.

[0074] In this embodiment, refer to Figure 2 , a hyperparameter optimization algorithm based on cluster training can be applied in the training process of the time series hybrid neural network model. Specifically, it includes the following steps:

[0075] S101. Establish multiple hyperparameter sets; each hyperparameter set includes multiple hyperparameters;

[0076] S102. Execute multiple rounds of iterations until the iteration end condition is met;

[0077] S103. Return the hyperparameter set corresponding to the model to be evaluated with the best performance evaluation in the last round of iteration;

[0078] S104. Use the returned hyperparameter set to set the temporal hybrid neural network model to obtain a trained temporal hybrid neural network model.

[0079] In step S101, potential hyperparameters such as learning rate, number of network layers and neurons can be selected to form a hyperparameter set in the format of (learning rate, number of network layers, neurons). Since the learning rate, number of network layers and neurons can take different values, each value combination can form a hyperparameter set, so multiple hyperparameter sets can be formed.

[0080] In step S102, the number of iterations L and the update rate rate are set. Taking the kth iteration process as an example, the time series hybrid neural network model is set using the hyperparameter sets updated in the k-1th iteration process to obtain multiple models to be evaluated. For example, after the k-1th iteration process is updated, multiple hyperparameter sets such as (learning rate 1, network layer 1, neuron 1), (learning rate 2, network layer 2, neuron 2), (learning rate 3, network layer 3, neuron 3) are obtained, and a time series hybrid neural network model with hyperparameters such as (learning rate 1, network layer 1, neuron 1) is established to obtain model 1 to be evaluated, and a time series hybrid neural network model with hyperparameters such as (learning rate 2, network layer 2, neuron 2) is established to obtain model 2 to be evaluated, and a time series hybrid neural network model with hyperparameters such as (learning rate 3, network layer 3, neuron 3) is established to obtain model 3 to be evaluated... thereby obtaining multiple models to be evaluated. Next, use the sample discharge data and the sample state of charge data to train these models to be evaluated, and obtain trained model 1 to be evaluated, trained model 2 to be evaluated, trained model 3 to be evaluated, etc. Multiple trained models to be evaluated are obtained. Next, the performance of each trained model to be evaluated is evaluated, for example, the root mean square error and mean absolute error of trained model 1 to be evaluated, the root mean square error and mean absolute error of trained model 2 to be evaluated, the root mean square error and mean absolute error of trained model 3 to be evaluated, etc. are obtained, and a batch of models to be evaluated with the worst performance evaluation results (largest error) and a batch of models to be evaluated with the best performance evaluation results (smallest error) are selected. Assuming that the model to be evaluated with the best performance evaluation result is model 3, and the model to be evaluated with the worst performance evaluation result is model 5, then the update rate rate can be used to update the hyperparameters (learning rate 5, number of network layers 5, neuron 5) of model 5 with the worst performance evaluation result. For example, (learning rate 5, number of network layers 5, neuron 5) are set to random numbers respectively to complete the update, or the hyperparameters (learning rate 3, number of network layers 3, neuron 3) of model 3 with the best performance evaluation result are used to fine-tune (learning rate 5, network layer 5, neuron 5) so that (learning rate 5, number of network layers 5, neuron 5) are updated to (learning rate 5, number of network layers 3, neuron 5), etc. to complete the update.

[0081] In the above process, if the kth iteration process is the 1st iteration process, then (learning rate 1, number of network layers 1, neuron 1) etc. can take random numbers.

[0082] After executing the kth round of iteration, let L=L-1, and determine whether L>0 is satisfied. If so, execute the k+1th round of iteration. If L≤0, do not execute the next round of iteration, and execute step S103.

[0083] In step S103, the hyperparameter set corresponding to the model to be evaluated with the best performance evaluation in the last round of iteration is returned. Assuming that the model to be evaluated with the hyperparameter set of (learning rate 4, number of network layers 3, neurons 5) has the best performance evaluation, then in step S104, the hyperparameter set returned by step S103 is used to set the temporal hybrid neural network model, that is, a temporal hybrid neural network model with parameters such as (learning rate 4, number of network layers 3, neurons 5) is established as a trained temporal hybrid neural network model.

[0084] The trained time series hybrid neural network model is used to obtain the estimated state of charge of the battery to be estimated. The process is as follows: Figure 3 shown.

[0085] In step S2, the estimated discharge data of the battery to be estimated can be obtained in the same format as the sample discharge data. For example, if it is desired to detect the estimated state of charge of the battery to be estimated at the current time t0, the discharge voltage, discharge current and operating temperature values ​​at the current time t0 and a period of time before the current time t0 (including t, t+1, ... t0-1, etc., where these moments can be in seconds) can be obtained, thereby obtaining the estimated discharge data of the battery to be estimated at the time t0.

[0086] In step S2, the estimated discharge data of the battery to be estimated at the current time t0 may be The data is input into the time series hybrid neural network model. Since the trained time series hybrid neural network model has the performance of extracting state of charge characteristics based on discharge data, the processing result of the time series hybrid neural network model on the estimated discharge data can represent the estimated state of charge value of the battery to be estimated at time t0.

[0087] In this embodiment, if the time series hybrid neural network model is directly used to estimate the discharge data to be estimated at the current time t0 The processing result itself is used as the estimated value of the charge state of the battery to be estimated at time t0. Since the current time t0 is arbitrary, the time t0 changes over time. The estimated discharge data measured each time can be processed by executing S3 using a time series hybrid neural network model, which results in a large jump in the estimated state of charge value at different times, which does not conform to the actual SOC change of the estimated battery. Therefore, in this embodiment, step S4 can be executed to achieve smoothing.

[0088] Specifically, in this embodiment, when executing step S4, that is, performing filtering processing subject to physical constraints on the processing result of the time series hybrid neural network model to obtain the estimated value of the state of charge of the battery to be estimated, the following steps can be specifically performed:

[0089] S401. Determine the filter window size and the first moment;

[0090] S402. Determine multiple second moments within a time length equal to the filter window size before the first moment;

[0091] S403. Obtain the data to be filtered corresponding to each second moment;

[0092] S404. Arrange the data to be filtered into a first time series according to the time sequence;

[0093] S405. For any data to be filtered, convert according to the physical constraints of the battery to be estimated between the first moment and the second moment corresponding to the data to be filtered, and obtain a corresponding conversion value;

[0094] S406. Arrange the converted values ​​into a second time series according to the order of the first time series;

[0095] S407. Add the processing result of the time series hybrid neural network model on the discharge data to be estimated collected at the first moment to the end of the second time series to obtain a third time series;

[0096] S408. Perform weighted smoothing filtering on the third time series to obtain a state of charge estimation value corresponding to the first moment.

[0097] In step S401, the first moment is set to be the current moment t0, and the filtering window size is n, that is, a time window that can accommodate n moments is established.

[0098] In step S402, starting from the first moment, i.e., the current moment t0, n-1 moments are pushed forward to determine the moments t0-1, t0-2, ..., t0-n+2, t0-n+1, ​​etc., and these moments are arranged in chronological order from the earliest to the latest, thus forming a time window including n moments t0-n+1, ​​t0-n+2, ..., t0-2, t0-1, t0, etc., among which t0-n+1, ​​t0-n+2, t0-2, t0-1, t0, etc. before the first moment t0 are the first moments t0-n+1, ​​t0-n+2, t0-2, t0-1, t0, etc.

[0099] -n+2...t0-2, t0-1 and other moments are respectively referred to as second moments. At the first moment and each second moment, the discharge data to be estimated at that time is collected from the battery to be estimated.

[0100] In step S403, the data to be filtered corresponding to each second moment is obtained. Specifically, taking the second moment t0-1 before the first moment t0 as an example, before the current moment moves to the first moment t0, the second moment t0-1 is actually the previous "current moment", so the second moment t0-1 is actually the "first moment" in history. If S401-S408 are executed with the second moment t0-1 as the "first moment" in history, then the weighted smoothing filter is actually executed with the second moment t0-1 as the "first moment" to obtain the charge state estimation value corresponding to the second moment t0-1, and this charge state estimation value is used as the data to be filtered corresponding to the second moment t0-1. If S401-S408 is not performed with the second moment t0-1 as the "first moment" in history, that is, the weighted smoothing filter is not performed with the second moment t0-1 as the "first moment", then the discharge data to be estimated collected at the second moment t0-1 (and the moment before the second moment t0-1) can be obtained. (where V represents the discharge voltage, I represents the discharge current, T represents the operating temperature, and the subscript represents the sampling time), the discharge data to be estimated collected at the second time t0-1 Input to the trained time series hybrid neural network for processing, and use the processing result of the time series hybrid neural network as the data to be filtered corresponding to the second moment t0-1

[0101] Based on the above principle, the data to be filtered corresponding to the second time t0-n+1 can be obtained The data to be filtered corresponding to the second time t0-n+2 ...the data to be filtered corresponding to the second moment t0-2 In step S404, the to-be-filtered data corresponding to each second moment are arranged in the time sequence of the second moment, thereby obtaining a first time series

[0102]

[0103] In step S405, for the first time series Each of the to-be-filtered data in the filter is converted according to the physical constraints imposed on the battery to be estimated between a first moment and a second moment corresponding to the to-be-filtered data to obtain a corresponding conversion value.

[0104] For example, for the data to be filtered corresponding to the second time t0-n+1 Its meaning is equivalent to the estimated result of the state of charge of the battery to be estimated at the second time t0-n+1. Therefore, according to the physical constraints of the battery to be estimated between the second time t0-n+1 and the first time t0, the data to be filtered corresponding to the second time t0-n+1 can be Convert to a corresponding conversion value.

[0105] In this embodiment, when executing step S405, the physical constraint of the battery to be estimated may be the discharge effect of the battery to be estimated, that is, assuming that the battery to be estimated is continuously discharged from the second time t0-n+1 to the first time t0, and the influence of the physical effect on the battery to be estimated during this process.

[0106] Specifically, in this embodiment, when executing step S405, the physical constraint considered is the ampere-hour integral. For example, for the data to be filtered corresponding to the second time t0-n+1 When using the ampere-hour integration method, As the initial value, t0-n+1 is used as the lower limit of integration, and the first moment t0 is used as the upper limit of integration. When the battery parameter of the battery to be estimated is the capacity C, it can be calculated that

[0107]

[0108] Where i(t) is a function of the discharge current of the battery to be estimated changing with time t. It is the data to be filtered corresponding to the second moment t0-n+1 The corresponding conversion value, The meaning of is the remaining power of the battery to be estimated at the first moment t0 under the physical constraint that the battery continues to discharge from the second moment t0-n+1 to the first moment t0.

[0109] Based on the same principle, we can use the formula

[0110]

[0111] Calculate the data to be filtered corresponding to the second time t0-n+2 The corresponding conversion value Continue to calculate to get the conversion value corresponding to the second moment t0-n+3 ...the conversion value corresponding to the second moment t0-2 The conversion value corresponding to the second time t0-1 In step S406, these conversion values ​​are arranged into a second time series in the same time sequence as the first time series, namely, t0-n+1, ​​t0-n+2...t0-2, t0-1.

[0112]

[0113] In step S407, the time series hybrid neural network model is used to obtain the discharge data to be estimated collected at the first moment. The processing result Will Add to the end of the second time series to get the third time series

[0114]

[0115] The length of the third time series is consistent with the size of the filtering window, which is n. The first n-1 values ​​in the third time series are equivalent to the estimated state of charge of the battery to be estimated at the first time t0 at the second time t0-n+1, ​​t0-n+2...t0-2, t0-1, etc. The last value in the third time series is It is the actual estimation result of the time series hybrid neural network model.

[0116] In step S408, the third time series Perform weighted smoothing filtering to obtain a state of charge estimation value corresponding to the first moment.

[0117] In this embodiment, Gaussian weighted smoothing filtering may be performed in step S408. Gaussian weighted smoothing filtering is to use a Gaussian weight function in weighted average moving filtering to assign a higher weight to data points closer to the current time point and a lower weight to data points farther from the current time point. The purpose of this is to emphasize recent data points more and reduce the influence of distant data points on the average value, thereby achieving a smoothing effect.

[0118] Before performing Gaussian weighted smoothing filtering, a weight set is calculated based on the Gaussian weight function. The formula of the Gaussian weight function is: Where w(i) is the weight of the i-th point, c is the current point, and σ is the standard deviation of the Gaussian function. In practice, the width of the weight function can be controlled by adjusting σ to balance the smoothing effect and the response speed to the data point. Specifically, in step S408, c = t0, and i = t0-n+1, ​​t0-n+2, ..., t0-2, t0-1, t0 are taken in sequence to calculate the weight set in Indicates the third time series The weight of... express The weight of express The weight of express The weight of .

[0119] In step S408, each weight in the weight set is multiplied by the corresponding value in the third time series, and the accumulated value is divided by the sum of the weights to obtain the SOC value constrained by the filtering at the first moment, i.e., the current moment t0. Right now

[0120] Calculated It is the estimated state of charge value of the battery to be estimated at the first time t0. Since the first time t0 can be selected arbitrarily or change with the passage of time, the estimated state of charge value of the battery to be estimated at any time or any current time can be calculated.

[0121] In this embodiment, by performing physically constrained filtering on the processing results of the time series hybrid neural network model, the state of charge estimation value of the battery to be estimated over a period of time can be smoothed and filtered according to the physical constraints on the battery to be estimated, thereby reducing the ups and downs of multiple state of charge estimation values ​​detected over time, and making the detected state of charge estimation value closer to the actual SOC curve of the battery to be estimated.

[0122] In this embodiment, by using a hybrid time series neural network model and physically constrained filtering processing, the two algorithms are combined to improve the estimation ability and accuracy of the estimated value of the state of charge of the battery to be estimated. The hybrid time series neural network model is composed of a time series convolutional neural network TCN and a long short-term memory network LSTM and a fully connected layer to form a time series hybrid neural network to estimate SOC, which can improve the accuracy of the estimated SOC. In addition, additional physical constraint filtering (specifically Gaussian weighted filtering) is applied to the output of the time series hybrid neural network, which can constrain and smooth the output results of the time series hybrid neural network, further improving the accuracy of the model estimation.

[0123] A computer program for executing the battery state of charge estimation method in this embodiment can be written and written into a computer device or storage medium. When the computer program is read out and run, the battery state of charge estimation method in this embodiment is executed, thereby achieving the same technical effect as the battery state of charge estimation method in the embodiment.

[0124] It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly fixed or connected to the other feature, or it may be indirectly fixed or connected to the other feature. In addition, the descriptions of up, down, left, right, etc. used in the present disclosure are only relative to the relative positional relationship of the components of the present disclosure in the accompanying drawings. The singular forms of "a", "" and "the" used in the present disclosure are also intended to include the plural forms, unless the context clearly indicates other meanings. In addition, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as those generally understood by those skilled in the art. The terms used in the specification of this embodiment are only for describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in this embodiment includes any combination of one or more related listed items.

[0125] It should be understood that, although the term first, second, third etc. may be adopted to describe various elements in the present disclosure, these elements should not be limited to these terms. These terms are only used to distinguish the same type of elements from each other. For example, without departing from the scope of the present disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as" etc.) provided by the present embodiment is only intended to better illustrate embodiments of the present invention, and unless otherwise required, the scope of the present invention will not be limited.

[0126] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed dedicated integrated circuit for this purpose.

[0127] In addition, the operations of the process described in this embodiment may be performed in any suitable order, unless otherwise indicated in this embodiment or otherwise clearly contradicted by the context. The process described in this embodiment (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as a code (e.g., executable instructions, one or more computer programs, or one or more applications) executed on one or more processors in common, by hardware or a combination thereof. A computer program includes a plurality of instructions that may be executed by one or more processors.

[0128] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the above steps in conjunction with a microprocessor or other data processor, the invention of this embodiment includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

[0129] The computer program can be applied to input data to perform the functions of the present embodiment, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on the display.

[0130] The above are only preferred embodiments of the present invention. The present invention is not limited to the above embodiments. As long as the technical effects of the present invention are achieved by the same means, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention. Within the scope of protection of the present invention, its technical solutions and / or implementation methods may have various modifications and changes.

Claims

1. A method for estimating a battery state of charge, characterized in that: The battery state of charge estimation method comprises: Establishing a time series hybrid neural network model; the time series hybrid neural network model is used to capture time series features of input data, learn time dependencies of the captured time series features, and perform regression prediction on the learned time dependencies; Acquiring at least one to-be-estimated discharge data of the to-be-estimated battery; Using the time series hybrid neural network model to process each of the discharge data to be estimated; The processing result of the time series hybrid neural network model is subjected to physical constraint filtering processing to obtain an estimated value of the state of charge of the battery to be estimated.

2. The battery state of charge estimation method according to claim 1, characterized in that: The method of establishing a temporal hybrid neural network model comprises: Establishing a time series convolutional neural network, a long short-term memory network and a fully connected layer; the time series convolutional neural network is used to capture the time series features of the input data, the long short-term memory network is used to learn the time dependency relationship of the time series features captured by the time series convolutional neural network, and the fully connected layer is used to perform regression prediction on the time dependency relationship learned by the long short-term memory network; The temporal convolutional neural network, the long short-term memory network and the fully connected layer are connected in sequence to obtain the temporal hybrid neural network model.

3. The battery state of charge estimation method according to claim 1, characterized in that: The method of establishing a temporal hybrid neural network model comprises: Obtaining sample discharge data and sample state of charge data of the sample battery corresponding to the same time; The time series hybrid neural network model is trained using the sample discharge data and the sample state of charge data.

4. The battery state of charge estimation method according to claim 3, characterized in that: The using the sample discharge data and the sample state of charge data to train the time series hybrid neural network model includes: Establishing a plurality of hyperparameter sets; each of the hyperparameter sets includes a plurality of hyperparameters; Execute multiple rounds of iterations until the iteration end condition is met; Return the hyperparameter set corresponding to the model to be evaluated with the best performance evaluation in the last round of iterations; Using the returned hyperparameter set to set the temporal hybrid neural network model to obtain the trained temporal hybrid neural network model; Wherein, any round of the iteration process includes the following steps: Using each of the hyperparameter sets to set the temporal hybrid neural network model respectively, a plurality of models to be evaluated are obtained; wherein any of the models to be evaluated is obtained by setting the temporal hybrid neural network model by a corresponding hyperparameter set; Using the sample discharge data and the sample state of charge data, respectively training each of the models to be evaluated; Performing performance evaluation on each of the trained models to be evaluated; For at least one of the models to be evaluated with the worst performance evaluation result, the hyperparameter sets corresponding to each of the models to be evaluated are updated.

5. The battery state of charge estimation method according to any one of claims 1 to 4, characterized in that: The step of performing filtering processing on the processing result of the time series hybrid neural network model subject to physical constraints to obtain an estimated value of the state of charge of the battery to be estimated includes: Determine the filter window size and first moment; Determine a plurality of second moments within a time length equal to the size of the filtering window before the first moment; the first moment and each of the second moments are respectively a collection moment of a corresponding one of the discharge data to be estimated; Obtaining the to-be-filtered data corresponding to each of the second moments; Arranging the data to be filtered into a first time series according to time sequence; For any of the data to be filtered, converting the data according to the physical constraints to which the battery to be estimated is subject between the first moment and the second moment corresponding to the data to be filtered, to obtain a corresponding conversion value; Arranging the conversion values ​​into a second time series according to the order of the first time series; Adding the processing result of the time series hybrid neural network model on the discharge data to be estimated collected at the first moment to the end of the second time series to obtain a third time series; Perform weighted smoothing filtering on the third time series to obtain the state of charge estimation value corresponding to the first moment.

6. The battery state of charge estimation method according to claim 5, characterized in that: The obtaining the to-be-filtered data corresponding to each of the second moments includes: For any of the second moments: When the state of charge estimation value corresponding to the second moment exists, taking the state of charge estimation value corresponding to the second moment as the to-be-filtered data corresponding to the second moment; When there is no corresponding state of charge estimation value at the second moment, the processing result of the time series hybrid neural network model on the discharge data to be estimated collected at the second moment is used as the data to be filtered corresponding to the second moment.

7. The battery state of charge estimation method according to claim 5, characterized in that: The converting according to the physical constraints of the battery to be estimated between the first moment and the second moment corresponding to the data to be filtered to obtain a corresponding conversion value includes: Acquiring battery parameters of the battery to be estimated; The data to be filtered is used as an initial value, the second moment is used as an integral lower limit, the first moment is used as an integral upper limit, and ampere-hour integration is performed according to the battery parameters to obtain the conversion value.

8. The battery state of charge estimation method according to claim 5, characterized in that: The performing weighted smoothing filtering on the third time series to obtain the estimated value of the state of charge corresponding to the first moment includes: Perform Gaussian weight function calculation according to the third time series to obtain a weight set; The third time series is weighted averaged according to the weight set to obtain the state of charge estimation value corresponding to the first moment.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store at least one program, and the processor is used to load at least one program to execute the battery state of charge estimation method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to execute the battery state of charge estimation method described in any one of claims 1 to 8 when executed by the processor.

Images