Method for constructing a battery SOH evaluation model and method for evaluating battery SOH value
By constructing a battery SOH evaluation method based on a similarity model, using a small amount of labeled data and a large amount of unlabeled data, the problem of high cost of battery SOH evaluation data collection is solved, and efficient and accurate battery SOH evaluation is achieved.
Patent Information
- Application Number
- CN202110376223.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-04-07
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-04-07
AI Technical Summary
In the prior art, battery SOH evaluation requires a large amount of training data marked with SOH tags, resulting in high cost and long time for data collection, which increases the difficulty and cost of evaluation.
By obtaining a large number of first battery data samples without SOH tags and a small number of second battery data samples with SOH tags, the sample interval is calculated, the sample pair is determined, and the similarity model is used for training, the battery SOH evaluation model is constructed to reduce the data Get difficulty and assess cost.
It effectively reduces the difficulty and cost of battery SOH evaluation, improves the accuracy and stability of evaluation, and uses a small amount of labeled data to establish an efficient SOH evaluation model.
Smart Images

Figure CN113111580B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of battery evaluation, and in particular to a method for constructing a battery SOH evaluation model and a method for evaluating a battery SOH value. Background Art
[0002] With the advancement of science and technology, batteries are widely used in various industries and occupy an important position. The performance of batteries will change with the battery usage time, charge and discharge capacity, and mileage of electric vehicles. At present, in order to evaluate SOH (State of Health, battery health), it is usually necessary to obtain a large number of SOH labels calibrated under different conditions as labels for model training data. In practical applications, it takes a lot of cost and time to collect a large amount of training data with SOH labels, which increases the difficulty and cost of battery SOH evaluation. Summary of the invention
[0003] In view of this, an object of the present invention is to provide a method for constructing a battery SOH evaluation model and a method for evaluating a battery SOH value, which can effectively reduce the difficulty and cost of battery SOH evaluation.
[0004] In a first aspect, an embodiment of the present invention provides a method for constructing a battery SOH evaluation model, comprising: obtaining a first training data set and a second training data set; wherein the first training data set includes a plurality of first battery data samples without SOH labels, and the second training data set includes a plurality of second battery data samples with SOH labels, and the number of samples in the second training data set is much smaller than the number of samples in the first training data set; based on the sample intervals between each of the first battery data samples in the first training data set, determining a plurality of sample pairs from the first training data set, and using each of the sample pairs to train a pre-constructed similarity model; and constructing a battery SOH evaluation model based on the trained similarity model and / or the second training data set.
[0005] In one embodiment, the step of determining a plurality of sample pairs from the first training data set based on the sample interval between each of the first battery data samples in the first training data set includes: randomly extracting first battery data samples from the first training data set as anchor samples; for each anchor sample, determining the positive samples and negative samples corresponding to the anchor sample from the first training data set according to the sample interval between the anchor sample and the remaining first battery data samples in the first training data set; and combining the anchor sample with the positive sample and / or negative sample corresponding to the anchor sample into a sample pair.
[0006] In one embodiment, the step of determining the positive sample and the negative sample corresponding to the anchor point sample from the first training data set according to the sample interval between the anchor point sample and the remaining first battery data samples in the first training data set includes: for each first battery data sample other than the anchor point sample in the first training data set, according to the usage time of the anchor point sample and the usage time of the first battery data sample, calculating the time interval between the anchor point sample and the first battery data sample; if the time interval is less than a first interval threshold, determining that the first battery data sample is a positive sample corresponding to the anchor point sample; if the time interval is greater than a second interval threshold, determining that the first battery data sample is is a negative sample corresponding to the anchor point sample; wherein the second interval threshold is greater than or equal to the first interval threshold; and / or; for each first battery data sample other than the anchor point sample in the first training data set, the mileage interval between the anchor point sample and the first battery data sample is calculated according to the mileage corresponding to the anchor point sample and the mileage corresponding to the first battery data sample; if the mileage interval is less than the third interval threshold, it is determined that the first battery data sample is a positive sample corresponding to the anchor point sample; if the mileage interval is greater than the fourth interval threshold, it is determined that the first battery data sample is a negative sample corresponding to the anchor point sample; wherein the fourth interval threshold is greater than or equal to the third interval threshold.
[0007] In one embodiment, the similarity model includes a feature mapping layer and a first regression layer; the step of using each of the sample pairs to train the pre-constructed similarity model includes: performing feature mapping on the sample pairs through the feature mapping layer to obtain the sample distance of the sample pairs; determining the degree of similarity between each sample in the sample pair according to the sample distance through the first regression layer; and updating the parameters of the similarity model according to the similarity and a preset loss function.
[0008] In one embodiment, the step of constructing a battery SOH evaluation model based on the trained similarity model and / or the second training data set includes: linking the feature mapping layer in the similarity model with the second regression layer to obtain a first SOH evaluation model, and using the second training data set to train the first SOH evaluation model; or, using the similarity model as a second SOH evaluation model; wherein the battery SOH evaluation model includes the first SOH evaluation model or the second SOH evaluation model.
[0009] In a second aspect, an embodiment of the present invention further provides a method for evaluating a battery SOH value, comprising: obtaining battery data to be evaluated; performing a health status evaluation on the battery data to be evaluated through a battery SOH evaluation model to obtain an SOH value corresponding to the battery data to be evaluated; wherein the battery SOH evaluation model is constructed according to the method for constructing a battery SOH evaluation model described in any one of the first aspects.
[0010] In one embodiment, the step of performing a health status assessment on the battery data to be evaluated through a battery SOH assessment model to obtain an SOH value corresponding to the battery data to be evaluated includes: if the battery SOH assessment model is a first SOH assessment model, inputting the battery data to be evaluated into the first SOH assessment model to obtain the SOH value output by the first SOH assessment model for the battery data to be evaluated.
[0011] In one embodiment, the step of performing health status assessment on the battery data to be evaluated through a battery SOH assessment model to obtain an SOH value corresponding to the battery data to be evaluated includes: if the battery SOH assessment model is a second SOH assessment model, inputting the battery data to be evaluated and a second training data set into the second SOH assessment model; calculating the degree of similarity between the battery data to be evaluated and each of the second battery data in the second training data set through the second SOH assessment model; determining target battery data from each of the second battery data based on the degree of similarity; and using the SOH label annotated with the target battery data as the SOH value corresponding to the battery data to be evaluated.
[0012] In a third aspect, an embodiment of the present invention further provides a device for constructing a battery SOH evaluation model, the device comprising: a data set acquisition module, used to acquire a first training data set and a second training data set; wherein the first training data set includes a plurality of first battery data samples without SOH labels, and the second training data set includes a plurality of second battery data samples with SOH labels, and the number of samples in the second training data set is much smaller than the number of samples in the first training data set; a model training module, used to determine a plurality of sample pairs from the first training data set based on the sample intervals between each of the first battery data samples in the first training data set, and train a pre-constructed similarity model using each of the sample pairs; a model construction module, used to construct a battery SOH evaluation model based on the trained similarity model and / or the second training data set.
[0013] In a fourth aspect, an embodiment of the present invention further provides a battery SOH value evaluation device, the device comprising: a data acquisition module, used to acquire battery data to be evaluated; an evaluation module, used to perform a health status evaluation on the battery data to be evaluated through a battery SOH evaluation model, and obtain the SOH value corresponding to the battery data to be evaluated; wherein the battery SOH evaluation model is constructed according to the battery SOH evaluation model construction method described in any one of the first aspects.
[0014] In a fifth aspect, an embodiment of the present invention further provides an electronic device, comprising a memory and a processor; the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method described in any one of the first and fourth aspects above are implemented.
[0015] In a sixth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to execute the method described in any one of the first and fourth aspects.
[0016] The embodiment of the present invention provides a method and device for constructing a battery SOH evaluation model. First, a plurality of first battery data samples without SOH labels are used as a first training data set, and a plurality of second battery data samples with SOH labels are used as a second training data set, wherein the number of samples in the second training data set is much smaller than the number of samples in the first training data set. Then, the sample interval between the first battery data samples is calculated, and a plurality of sample pairs are determined based on the sample interval, and the pre-constructed similarity model is trained using the sample pairs. Finally, a battery SOH evaluation model is constructed based on the trained similarity model and / or the second training sample data set. The above method can establish a battery SOH evaluation correction model through a large number of first battery data samples without SOH labels and a small number of second battery data samples with SOH labels, so as to evaluate the battery SOH value through the SOH evaluation model. Compared with the prior art that requires experimental calibration of the SOH labels of battery usage data under different conditions, the embodiment of the present invention can effectively reduce the difficulty of obtaining training data, thereby reducing the difficulty of evaluating the battery SOH, and can also significantly save the evaluation cost of the battery SOH.
[0017] The present invention provides a battery SOH evaluation method and device, which constructs a battery SOH evaluation model by the above-mentioned battery SOH evaluation model construction method, and then uses the battery SOH evaluation model to evaluate the battery data to be evaluated to obtain the SOH value of the battery data to be evaluated. The embodiment of the present invention can use the above-mentioned battery SOH evaluation model to efficiently and accurately evaluate the battery SOH value, and effectively reduce the difficulty and cost of battery SOH evaluation.
[0018] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.
[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 A schematic diagram of a method for constructing a battery SOH evaluation model provided by an embodiment of the present invention;
[0022] Figure 2 A schematic diagram of the structure of a similarity model provided by an embodiment of the present invention;
[0023] Figure 3 A schematic diagram of the structure of another similarity model provided by an embodiment of the present invention;
[0024] Figure 4 A schematic diagram of a flow chart of a battery SOH value evaluation method provided in an embodiment of the present invention;
[0025] Figure 5 A schematic diagram of a device structure for constructing a battery SOH evaluation model provided by an embodiment of the present invention;
[0026] Figure 6 A schematic diagram of the structure of a battery SOH value evaluation device provided in an embodiment of the present invention;
[0027] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0029] Batteries play an important role in people's production and life, and the performance of batteries will change with the battery usage time and charging and discharging. At present, in the process of battery SOH evaluation, since it takes a lot of cost and time to collect a large amount of training data with SOH labels, the difficulty and cost of battery SOH evaluation are increased. Based on this, the present invention provides a method for constructing a battery SOH evaluation model and a method for evaluating the battery SOH value, which can effectively reduce the difficulty and cost of battery SOH evaluation.
[0030] To facilitate understanding of this embodiment, a method for constructing a battery SOH evaluation model disclosed in an embodiment of the present invention is first introduced in detail, see Figure 1 A flow chart of a method for constructing a battery SOH evaluation model is shown, and the method may include the following steps:
[0031] Step S102, obtaining a first training data set and a second training data set. The first training data set includes a plurality of first battery data samples that are not labeled with SOH labels, and the second training data set includes a plurality of second battery data samples that are labeled with SOH labels, and the number of samples in the second training data set is much smaller than the number of samples in the first training data set. In practical applications, it is necessary to collect a large amount of battery usage data without SOH labels, and the battery usage data without SOH labels is the above-mentioned first battery data samples. In addition, a small amount of battery usage data with SOH labels can also be collected, and the battery usage data with SOH labels is the above-mentioned second battery data. In an optional embodiment, the ratio of the number of samples in the first training data set to the number of samples in the second training data set can be greater than a preset ratio. For example, the preset ratio is set to 10, the number of samples in the first training data set is 1000, and the number of samples in the second training data set is 100. At this time, it can be said that the number of samples in the first training data set is much larger than the number of samples in the second training data set.
[0032] Step S104, based on the sample intervals between each first battery data sample in the first training data set, multiple sample pairs are determined from the first training data set, and the pre-constructed similarity model is trained using each sample pair. The sample interval is used to characterize whether each first battery data sample is adjacent to each other. For example, the sample interval may include an interval in the time dimension or an interval in the mileage dimension. The sample pair may include a combination of any first battery data sample in the first training data set and its positive sample and / or negative sample. In one embodiment, a positive sample Y with a sample interval closer to the first battery data sample X and a negative sample Z with a sample interval farther from the first training data set may be determined based on the sample intervals between each first battery data sample. Then, the first battery data sample X and the positive sample Y may be used to obtain a group of sample pairs, the first battery data sample X and the negative sample Z may be used to obtain a group of sample pairs, and the first battery data sample X, the positive sample Y and the negative sample Z may be used to obtain a group of sample pairs. In addition, the similarity model may be trained using the above sample pairs so that the trained similarity model can better predict the degree of similarity between the input data samples.
[0033] Step S106, constructing a battery SOH evaluation model based on the trained similarity model and / or the second training data set. In some optional embodiments, a feature mapping layer can be extracted from the similarity model, and the feature mapping layer can be linked to the new regression layer to obtain a battery SOH evaluation model, and then the battery SOH evaluation model can be trained using the second training data set so that the battery SOH evaluation model can output the corresponding SOH value for the input battery data to be evaluated; the similarity model can also be directly used as the battery SOH model, and the battery SOH evaluation model can be used to predict the SOH of the battery data to be evaluated.
[0034] An embodiment of the present invention provides a method for constructing a battery SOH evaluation model. A battery SOH evaluation correction model can be established through a large number of first battery data samples without SOH labels and a small number of second battery data samples with SOH labels, so as to evaluate the battery SOH value through the SOH evaluation model. Compared with the prior art that requires experimental calibration of the SOH labels of battery usage data under different conditions, the embodiment of the present invention can effectively reduce the difficulty of obtaining training data, thereby reducing the difficulty of battery SOH evaluation, and can also significantly save the battery SOH evaluation cost.
[0035] For the aforementioned step S102, in order to obtain the first training data set, a large amount of battery usage data without SOH labels can be collected, and the battery usage data can be divided into multiple samples according to a certain distance standard as the first training data set P, wherein the above distance standard can include the number of days the battery is used, the cumulative charge and discharge power, the mileage of the electric vehicle, etc. In addition, in order to obtain the second training data set, a small amount of battery usage data with SOH labels can be collected, and the battery usage data can be divided into multiple samples according to the same standard as the second training data set Q.
[0036] In order to better evaluate the battery SOH, the embodiment of the present invention trains the similarity model according to the property that the battery SOH generally decreases monotonically and changes slowly with use. For ease of understanding, the embodiment of the present invention provides an implementation of the aforementioned step S104. Specifically, the following steps 1 to 3 may be referred to:
[0037] Step 1: randomly extract a first battery data sample from a first training data set as an anchor sample. For example, randomly extract an anchor sample X from a first training data set P.
[0038] Step 2: For each anchor point sample, determine the positive sample and negative sample corresponding to the anchor point sample from the first training data set according to the sample interval between the anchor point sample and the remaining first battery data samples in the first training data set. For example, for the remaining first battery data samples M except the anchor point sample X in the first training data set P, determine the positive sample Y and negative sample Z corresponding to the anchor point sample X according to the sample interval between the anchor point sample X and the first battery data sample M.
[0039] Step 3, the anchor sample is combined with the positive sample and / or negative sample corresponding to the anchor sample into a sample pair. For example, the anchor sample X and the positive sample Y form a positive sample pair, and the positive sample Y in the positive sample pair is represented by label 1, or the anchor sample X and the negative sample Z form a negative sample pair, and the negative sample Z in the negative sample pair is represented by label 0. In another embodiment, the anchor sample X and the corresponding positive sample Y and negative sample Z can also be combined into a sample pair, and the sample pair does not need to be labeled.
[0040] Based on this, the embodiment of the present invention also provides an implementation method of the aforementioned step 2, which can be specifically referred to in the following method 1 and / or method 2:
[0041] Method 1: (1) For each first battery data sample except the anchor sample in the first training data set, the time interval between the anchor sample and the first battery data sample is calculated based on the usage time of the anchor sample and the usage time of the first battery data sample; (2) If the time interval is less than the first interval threshold, the first battery data sample is determined to be a positive sample corresponding to the anchor sample; (3) If the time interval is greater than the second interval threshold, the first battery data sample is determined to be a negative sample corresponding to the anchor sample; wherein the second interval threshold is greater than or equal to the first interval threshold. Since the SOH of the battery decreases with the usage time of the battery, the positive and negative samples corresponding to the anchor sample can be determined from the first training data set based on the usage time. For example, if the usage time of the anchor sample is T1 and the usage time of the first battery data sample is T2, the time interval is ΔT. Assuming that the first interval threshold is 2 and the second interval threshold is 5, if ΔT<2, the first battery data sample can be determined to be a positive sample of the anchor point sample, and if ΔT>5, the first battery data sample can be determined to be a negative sample of the anchor point sample; assuming that the first interval threshold and the second interval threshold are 5, if ΔT<5, the first battery data sample can be determined to be a positive sample of the anchor point sample, and if ΔT>5, the first battery data sample can be determined to be a negative sample of the anchor point sample. It should be noted that the above values are only application examples, and the specific values of the interval thresholds can be set based on actual needs.
[0042] Method 2: (1) For each first battery data sample other than the anchor point sample in the first training data set, the mileage interval between the anchor point sample and the first battery data sample is calculated based on the mileage corresponding to the anchor point sample and the mileage corresponding to the first battery data sample; (2) If the mileage interval is less than the third interval threshold, the first battery data sample is determined to be a positive sample corresponding to the anchor point sample; (3) If the mileage interval is greater than the fourth interval threshold, the first battery data sample is determined to be a negative sample corresponding to the anchor point sample; wherein the fourth interval threshold is greater than or equal to the third interval threshold. Since the SOH of the battery decreases as the mileage of the battery increases, the positive and negative samples corresponding to the anchor point sample can be determined from the first training data set based on the mileage. In addition, the third interval threshold and the fourth interval threshold can also be limited based on actual needs, and the embodiment of the present invention does not limit this.
[0043] In practical applications, any one of the above-mentioned methods 1 and 2 can be selected to determine the sample pair, or a combination of method 1 and method 2 can be selected to determine the positive sample and negative sample corresponding to the anchor sample. In another embodiment, the positive sample can be the first battery data sample that is closest to the anchor sample in terms of usage time and mileage, and the negative sample can be the first battery data sample that is farthest from the anchor sample in terms of usage time and mileage.
[0044] In one embodiment, the similarity model includes a feature mapping layer and a first regression layer. Based on this, an embodiment of the present invention also provides an implementation method for training a pre-constructed similarity model using each sample pair, including: performing feature mapping on the sample pair through the feature mapping layer to obtain the sample distance of the sample pair; determining the similarity between each sample in the sample pair according to the sample distance through the first regression layer; and updating the parameters of the similarity model according to the similarity and a preset loss function. Among them, the feature mapping layer refers to the embedding layer. In order to effectively train the regression model with a small amount of labeled data and a large amount of unlabeled data, during model training, the sample pair is first input into the embedding layer to obtain the corresponding mapped feature vector pair, and then the mapped feature vector pair is input into the first regression layer, and the sample distance is used to determine whether the sample in the sample pair belongs to a similar sample, and the model error is calculated according to the sample similarity and a preset loss function (such as KL (Kullback-Leibler Divergence, relative entropy) divergence, ELBO (Evidence Lower Bound, evidence lower bound) etc.), and the parameters of the similarity model are updated using an optimization algorithm. For two batteries with similar SOH, the distance between their feature vectors is small, and for two batteries with large SOH differences, the distance between their feature vectors is large. The distance between feature vectors in the feature space reflects the size of the difference between the demand forecast targets, so that only less labeled data is needed to effectively build a regression model.
[0045] To facilitate understanding of the training process of the similarity model, the embodiments of the present invention provide some training examples of the similarity model, wherein the following network structure is only an application example and is not limited to the following network structure in actual application. (i) If the sample pair includes a combination of an anchor sample and a positive sample, or a combination of an anchor sample and a negative sample, the similarity model can use Siamese networks, see Figure 2The schematic diagram of the structure of a similarity model is shown, where network A and network B are networks with the same network structure and shared parameters. Generally, multi-layer CNN (Convolutional Neural Networks) or LSTM (Long Short-Term Memory) network structures can be used as needed. The input of network A and network B enters a distance metric, calculates the distance between the feature vectors output by network A and network B, and finally the output distance enters the first regression layer, such as a sigmoid gate to output the similarity probability or prediction result of the sample pair. (ii) If the sample pair includes anchor samples, positive samples and negative samples, the similarity model can use Triplet Network, see Figure 3 The structural diagram of another similarity model shown in FIG. 1 , the loss function generally uses Triplet loss.
[0046] Based on the above step S106, an embodiment of the present invention further provides an implementation method for constructing a battery SOH evaluation model based on the trained similarity model and / or the second training data set, wherein the battery SOH evaluation model includes a first SOH evaluation model or a second SOH evaluation model. In one implementation method, the feature mapping layer in the similarity model can be linked to the second regression layer to obtain a first SOH evaluation model, and the first SOH evaluation model is trained using the second training data set. In practical applications, the embedding layer in the similarity model is extracted and linked to the second regression layer to obtain the first SOH evaluation model, and the first SOH evaluation model is trained using data set B. During the training process, the parameters of the embedding layer are fixed or fine-tuned, and only the parameters of the second regression layer are updated. In another implementation method, the similarity model can be used as the second SOH evaluation model. At this time, the similarity model can be directly used to calculate the sample similarity between the sample to be evaluated and the second training data set, and the sample SOH label in the second training data set with the highest similarity is taken as the SOH value evaluated by the similarity model. The similarity can also be used as a weight to calculate the average value of the SOH labels of the second training data set as the final SOH value.
[0047] The present invention provides a method for evaluating the SOH value of a battery. Figure 4 A flow chart of a method for evaluating a battery SOH value is shown, and the method may include the following steps:
[0048] Step S402, obtaining battery data to be evaluated;
[0049] Step S404, the health status of the battery data to be evaluated is evaluated by the battery SOH evaluation model to obtain the SOH value corresponding to the battery data to be evaluated; wherein the battery SOH evaluation model is constructed according to the battery SOH evaluation model construction method provided in the above embodiment. In one embodiment, the battery data to be evaluated is evaluated according to the similarity model trained by the battery SOH evaluation model construction method, mainly including the health status of the battery, and the final evaluation result is output in the form of SOH value.
[0050] Based on the above discussion, the present invention also provides an implementation method of evaluating the health status of battery data to be evaluated through a battery SOH evaluation model to obtain the SOH value corresponding to the battery data to be evaluated, including: if the battery SOH evaluation model is a first SOH evaluation model, input the battery data to be evaluated into the first SOH evaluation model to obtain the SOH value output by the first SOH evaluation model for the battery data to be evaluated. The battery data set to be evaluated is input into the first SOH evaluation model, and the evaluation result is based on the SOH value output by the first SOH evaluation model.
[0051] The present invention also provides another implementation method for evaluating the health status of battery data to be evaluated by using a battery SOH evaluation model to obtain an SOH value corresponding to the battery data to be evaluated, including: if the battery SOH evaluation model is a second SOH evaluation model, inputting the battery data to be evaluated and the second training data set into the second SOH evaluation model; calculating the similarity between the battery data to be evaluated and each second battery data in the second training data set by the second SOH evaluation model; determining the target battery data from each second battery data based on the similarity; and using the SOH label annotated with the target battery data as the SOH value corresponding to the battery data to be evaluated. In practical applications, the similarity model is directly used to calculate the sample similarity between the sample to be evaluated and the second training data set, and the sample SOH label in the second training data set with the highest similarity is taken as the SOH value evaluated by the similarity model. The similarity can also be used as a weight to calculate the average value of the SOH labels of the second training data set as the final SOH value.
[0052] The present invention provides a battery SOH evaluation method, which can use the above-mentioned battery SOH evaluation model to efficiently and accurately evaluate the battery SOH value, and effectively reduce the difficulty and cost of battery SOH evaluation.
[0053] To sum up, the method for constructing a battery SOH evaluation model provided in an embodiment of the present invention reduces the dependence of the training model on the number of labeled SOH samples by training a similarity model with a large amount of battery usage data without SOH labels and a small amount of SOH labels, greatly reduces the experimental cost for labeling SOH, and improves the accuracy and stability of the model evaluation.
[0054] Based on the method for constructing a battery SOH evaluation model provided in the above embodiments, an embodiment of the present invention provides a device for constructing a battery SOH evaluation model, see Figure 5 A schematic diagram of a device structure for constructing a battery SOH evaluation model is shown, and the device includes at least the following parts:
[0055] The data set acquisition module 502 is used to acquire a first training data set and a second training data set; wherein the first training data set includes a plurality of first battery data samples that are not labeled with SOH labels, and the second training data set includes a plurality of second battery data samples that are labeled with SOH labels, and the number of samples in the second training data set is much smaller than the number of samples in the first training data set.
[0056] The model training module 504 is used to determine a plurality of sample pairs from the first training data set based on the sample intervals between each first battery data sample in the first training data set, and train a pre-built similarity model using each sample pair.
[0057] The model building module 506 is used to build a battery SOH evaluation model according to the trained similarity model and / or the second training data set.
[0058] The device for constructing the above-mentioned battery SOH evaluation model provided by the embodiment of the present invention can establish a battery SOH evaluation correction model through a large number of first battery data samples without SOH labels and a small number of second battery data samples with SOH labels, so as to evaluate the battery SOH value through the SOH evaluation model. Compared with the prior art that requires experimental calibration of the SOH labels of battery usage data under different conditions, the embodiment of the present invention can effectively reduce the difficulty of obtaining training data, thereby reducing the difficulty of battery SOH evaluation, and can also significantly save the battery SOH evaluation cost.
[0059] In one embodiment, the model training module 504 is also used to randomly extract first battery data samples from the first training data set as anchor samples; for each anchor sample, determine the positive samples and negative samples corresponding to the anchor sample from the first training data set based on the sample interval between the anchor sample and the remaining first battery data samples in the first training data set; and combine the anchor sample with the positive sample and / or negative sample corresponding to the anchor sample into a sample pair.
[0060] In one embodiment, the model training module 504 is also used to, for each first battery data sample other than the anchor point sample in the first training data set, calculate the time interval between the anchor point sample and the first battery data sample based on the usage time of the anchor point sample and the usage time of the first battery data sample; if the time interval is less than the first interval threshold, determine that the first battery data is a positive sample corresponding to the anchor point sample; if the time interval is greater than the second interval threshold, determine that the first battery data sample is a negative sample corresponding to the anchor point sample; wherein the second interval threshold is greater than or equal to the first interval threshold; and / or; for each first battery data sample other than the anchor point sample in the first training data set, calculate the mileage interval between the anchor point sample and the first battery data based on the mileage corresponding to the anchor point sample and the mileage corresponding to the first battery data sample; if the mileage interval is less than the third interval threshold, determine that the first battery data sample is a positive sample corresponding to the anchor point sample; if the mileage interval is greater than the fourth interval threshold, determine that the first battery data sample is a negative sample corresponding to the anchor point sample; wherein the fourth interval threshold is greater than or equal to the third interval threshold.
[0061] In one embodiment, the similarity model includes a feature mapping layer and a first regression layer; the model training module 504 is also used to perform feature mapping on the sample pairs through the feature mapping layer to obtain the sample distance of the sample pairs; determine the similarity between each sample in the sample pair according to the sample distance through the first regression layer; and update the parameters of the similarity model according to the similarity and a preset loss function.
[0062] In one embodiment, the model building module 506 is also used to link the feature mapping layer in the similarity model with the second regression layer to obtain a first SOH evaluation model, and train the first SOH evaluation model using the second training data set; or, use the similarity model as the second SOH evaluation model; wherein the battery SOH evaluation model includes the first SOH evaluation model or the second SOH evaluation model.
[0063] Based on the battery SOH value evaluation method provided in the above embodiment, the present invention also provides a battery SOH value evaluation device, see Figure 6 The schematic diagram of the structure of a battery SOH value evaluation device is shown, and the device comprises at least the following parts:
[0064] The data acquisition module 602 is used to acquire battery data to be evaluated.
[0065] The evaluation module 604 is used to evaluate the health status of the battery data to be evaluated through the battery SOH evaluation model to obtain the SOH value corresponding to the battery data to be evaluated; wherein the battery SOH evaluation model is constructed according to the battery SOH evaluation model construction method provided in the above embodiment.
[0066] The embodiment of the present invention further provides a battery SOH value evaluation device, which can use the above-mentioned battery SOH evaluation model to efficiently and accurately evaluate the battery SOH value, and effectively reduce the difficulty and cost of battery SOH evaluation.
[0067] In one implementation, the evaluation module 604 is further configured to, if the battery SOH evaluation model is a first SOH evaluation model, input the battery data to be evaluated into the first SOH evaluation model to obtain the SOH value output by the first SOH evaluation model for the battery data to be evaluated.
[0068] In one embodiment, the evaluation module 604 is also used to, if the battery SOH evaluation model is a second SOH evaluation model, input the battery data to be evaluated and the second training data set into the second SOH evaluation model; calculate the similarity between the battery data to be evaluated and each second battery data in the second training data set through the second SOH evaluation model; determine the target battery data from each second battery data based on the similarity; and use the SOH label annotated with the target battery data as the SOH value corresponding to the battery data to be evaluated.
[0069] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.
[0070] An embodiment of the present invention provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned implementation methods.
[0071] Figure 7 A structural diagram of an electronic device provided in an embodiment of the present invention, the electronic device 100 includes: a processor 70, a memory 71, a bus 72 and a communication interface 73, wherein the processor 70, the communication interface 73 and the memory 71 are connected via the bus 72; the processor 70 is used to execute an executable module stored in the memory 71, such as a computer program.
[0072] The memory 71 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 73 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.
[0073] The bus 72 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0074] Among them, the memory 71 is used to store programs, and the processor 70 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 70 or implemented by the processor 70.
[0075] The processor 70 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 70. The above processor 70 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to be executed, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 71, and the processor 70 reads the information in the memory 71 and completes the steps of the above method in combination with its hardware.
[0076] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.
[0077] If the functions are implemented in the form of software functional units 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 the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.
[0078] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims
1. A method for constructing a battery SOH evaluation model, characterized in that: include: Obtain a first training data set and a second training data set; wherein the first training data set includes a plurality of first battery data samples not labeled with SOH labels, and the second training data set includes a plurality of second battery data samples labeled with SOH labels, and the number of samples in the second training data set is much smaller than the number of samples in the first training data set; determining a plurality of sample pairs from the first training data set based on a sample interval between each of the first battery data samples in the first training data set, and training a pre-built similarity model using each of the sample pairs; Constructing a battery SOH evaluation model based on the trained similarity model and / or the second training data set; The step of determining a plurality of sample pairs from the first training data set based on the sample intervals between each of the first battery data samples in the first training data set includes: randomly extracting first battery data samples from the first training data set as anchor samples; for each anchor sample, determining, from the first training data set, positive samples and negative samples corresponding to the anchor sample based on the sample intervals between the anchor sample and the remaining first battery data samples in the first training data set; and combining the anchor sample with the positive sample and / or negative sample corresponding to the anchor sample into a sample pair; The similarity model includes a feature mapping layer and a first regression layer; the step of training the pre-constructed similarity model using each of the sample pairs includes: performing feature mapping on the sample pairs through the feature mapping layer to obtain sample distances of the sample pairs; determining the degree of similarity between each sample in the sample pair based on the sample distances through the first regression layer; and updating the parameters of the similarity model based on the degree of similarity and a preset loss function. The step of constructing a battery SOH evaluation model based on the trained similarity model and / or the second training data set includes: linking the feature mapping layer in the similarity model with the second regression layer to obtain a first SOH evaluation model, and using the second training data set to train the first SOH evaluation model; or, using the similarity model as a second SOH evaluation model; wherein the battery SOH evaluation model includes the first SOH evaluation model or the second SOH evaluation model.
2. The method according to claim 1, characterized in that The step of determining, from the first training data set, positive samples and negative samples corresponding to the anchor point sample, respectively, based on a sample interval between the anchor point sample and the remaining first battery data samples in the first training data set, comprises: For each first battery data sample other than the anchor point sample in the first training data set, calculating the time interval between the anchor point sample and the first battery data sample based on the usage time of the anchor point sample and the usage time of the first battery data sample; If the time interval is less than a first interval threshold, determining that the first battery data sample is a positive sample corresponding to the anchor point sample; if the time interval is greater than a second interval threshold, determining that the first battery data sample is a negative sample corresponding to the anchor point sample; wherein the second interval threshold is greater than or equal to the first interval threshold; and / or; For each first battery data sample other than the anchor point sample in the first training data set, calculating the mileage interval between the anchor point sample and the first battery data sample based on the mileage corresponding to the anchor point sample and the mileage corresponding to the first battery data sample; If the mileage interval is less than the third interval threshold, the first battery data sample is determined to be a positive sample corresponding to the anchor point sample; if the mileage interval is greater than the fourth interval threshold, the first battery data sample is determined to be a negative sample corresponding to the anchor point sample; wherein the fourth interval threshold is greater than or equal to the third interval threshold.
3. A method for evaluating the battery SOH value, characterized in that: include: Obtain battery data to be evaluated; The health status of the battery data to be evaluated is evaluated by using a battery SOH evaluation model to obtain an SOH value corresponding to the battery data to be evaluated; wherein the battery SOH evaluation model is constructed according to the battery SOH evaluation model construction method according to any one of claims 1-2.
4. The method according to claim 3, characterized in that The step of performing health status evaluation on the battery data to be evaluated using a battery SOH evaluation model to obtain an SOH value corresponding to the battery data to be evaluated includes: If the battery SOH evaluation model is a first SOH evaluation model, the battery data to be evaluated is input into the first SOH evaluation model to obtain an SOH value output by the first SOH evaluation model for the battery data to be evaluated.
5. The method according to claim 3, characterized in that The step of performing health status evaluation on the battery data to be evaluated using a battery SOH evaluation model to obtain an SOH value corresponding to the battery data to be evaluated includes: If the battery SOH evaluation model is a second SOH evaluation model, inputting the battery data to be evaluated and the second training data set into the second SOH evaluation model; Calculating the similarity between the battery data to be evaluated and each of the second battery data in the second training data set by the second SOH evaluation model; determining target battery data from each of the second battery data based on the similarity; The SOH tag annotated with the target battery data is used as the SOH value corresponding to the battery data to be evaluated.
6. A device for constructing a battery SOH evaluation model, characterized in that: include: a data set acquisition module, configured to acquire a first training data set and a second training data set; wherein the first training data set includes a plurality of first battery data samples not labeled with SOH labels, and the second training data set includes a plurality of second battery data samples labeled with SOH labels, and the number of samples in the second training data set is much smaller than the number of samples in the first training data set; a model training module, configured to determine a plurality of sample pairs from the first training data set based on a sample interval between each of the first battery data samples in the first training data set, and train a pre-built similarity model using each of the sample pairs; A model building module, configured to build a battery SOH evaluation model based on the trained similarity model and / or the second training data set; The model training module is specifically configured to: randomly extract first battery data samples from the first training data set as anchor samples; for each anchor sample, determine the positive sample and negative sample corresponding to the anchor sample from the first training data set based on the sample interval between the anchor sample and the remaining first battery data samples in the first training data set; and combine the anchor sample with the positive sample and / or negative sample corresponding to the anchor sample into a sample pair; The similarity model includes a feature mapping layer and a first regression layer; the model training module is specifically configured to: perform feature mapping on the sample pairs through the feature mapping layer to obtain sample distances of the sample pairs; determine the similarity between each sample in the sample pair based on the sample distances through the first regression layer; and update the parameters of the similarity model based on the similarity and a preset loss function; The model construction module is specifically used to: link the feature mapping layer in the similarity model with the second regression layer to obtain a first SOH evaluation model, and use the second training data set to train the first SOH evaluation model; or, use the similarity model as a second SOH evaluation model; wherein the battery SOH evaluation model includes the first SOH evaluation model or the second SOH evaluation model.
7. A battery SOH value evaluation device, characterized in that: include: A data acquisition module, used to acquire battery data to be evaluated; An evaluation module is used to perform a health status evaluation on the battery data to be evaluated through a battery SOH evaluation model to obtain an SOH value corresponding to the battery data to be evaluated; wherein the battery SOH evaluation model is constructed according to the battery SOH evaluation model construction method according to any one of claims 1-2.
8. An electronic device, characterized in that: including processor and memory; The memory stores a computer program, which, when executed by the processor, executes the method according to any one of claims 1 to 2, or executes the method according to any one of claims 3 to 5.
9. A computer storage medium, characterized in that Used to store computer software instructions used for the method according to any one of claims 1 to 2, or to store computer software instructions used for the method according to any one of claims 3 to 5.
