Cloud power battery state estimation method based on charging data

Abstract

The invention provides a cloud power battery state estimation method based on charging data, and the method comprises the steps: carrying out the cloud cleaning of charging segment data, discriminating the aging state similarity, the charging condition and the operation temperature of the charging segment data, classifying the charging segment data according to the aging state, the charging condition and the operation temperature, and carrying out the cloud cleaning of the charging segment data; obtaining classified charging fragment data; constructing a plurality of training samples for the charging fragment data belonging to the same kind; clustering the plurality of training samples based on a clustering algorithm, and judging the relative positions of starting points of different charging fragment data based on a clustering result, so as to determine the position of a splicing point; splicing the charging fragment data belonging to the same kind so as to obtain a complete charging curve under the current aging state and the running temperature; and based on the spliced complete charging curve, estimating the states of the power battery, including the state of charge, the state of health, the energy state, the remaining charging time and the charging power.

Description

Technical Field

[0001] The invention relates to a cloud-based power battery state estimation method based on charging data. Background Art

[0002] The power battery system is usually composed of hundreds or thousands of cells. Due to the limited computing power on the vehicle side, it is difficult to perform real-time synchronous state estimation of all cells. Based on the vehicle-cloud collaborative architecture, fine-grained state estimation of all cells in the power battery system is an important way to break through the hardware limitations of the vehicle side and achieve efficient power battery management. However, considering the data transmission and storage costs, the data uploaded from the vehicle side to the cloud has the characteristics of a large sampling period, which makes the traditional state estimation method based on battery external characterization parameters (voltage, temperature, etc.) or battery models unable to be applied in the cloud. Summary of the invention

[0003] In response to the above-mentioned technical problems that need to be solved urgently in the field, the present invention provides a cloud-based power battery status estimation method based on charging data, which aims to calculate the power battery status (including state of charge (SOC), state of health (SOH), state of energy (SOE), etc.) by only using fragmented and large sampling period battery voltage, current, and temperature data on the cloud, thereby laying the foundation for efficient management of power batteries in vehicle-cloud collaboration.

[0004] The present application provides a cloud-based power battery state estimation method based on charging data, comprising:

[0005] S1: After the charging segment data is cleaned in the cloud, the battery characteristic parameters are identified, and the charging segment data is classified according to the battery characteristic parameters to obtain the classified charging segment data;

[0006] S2: constructing multiple training samples for charging segment data classified into the same category; clustering the multiple training samples based on a clustering algorithm, and judging the relative positions of starting points of different charging segment data based on the clustering results, thereby determining the position of the splicing point; splicing the charging segment data classified into the same category to obtain a complete charging curve under the current aging state and operating temperature;

[0007] S3: Estimate the state of the power battery based on the spliced ​​complete charging curve.

[0008] According to at least one embodiment of the present application, the battery characteristic parameters include aging state similarity, charging conditions and operating temperature, and the charging conditions include fast charging and slow charging.

[0009] According to at least one embodiment of the present application, the segment data classified into the same category have similar aging states, similar operating temperature ranges, and the same charging conditions.

[0010] According to at least one embodiment of the present application, the charging segment data includes charging voltage, current, and operating temperature.

[0011] According to at least one embodiment of the present application, a machine learning method or a deep learning method is used to determine the similarity of the aging status of the charging segment data.

[0012] According to at least one embodiment of the present application, in step S2, each training sample is clustered using a Gaussian mixture model, wherein parameters of the Gaussian mixture model are solved using an expectation maxima algorithm.

[0013] According to at least one embodiment of the present application, step S3 includes:

[0014] S31: Calculate the absolute ampere-hour accumulation of each data point in the spliced ​​complete charging curve;

[0015] S32: Estimate the state of the power battery based on the assembled charging curve.

[0016] According to at least one embodiment of the present application, the state of the power battery is a healthy state, and step S32 includes:

[0017] In the spliced ​​complete charging curve, the maximum value of the absolute ampere-hour accumulation of each data point relative to the starting point of the charging curve is used as the estimated power battery charging capacity under the current aging state and temperature, which is used to characterize the health status of the power battery.

[0018] According to at least one embodiment of the present application, the state of the power battery is a charged state, and step S32 includes:

[0019] Using the estimated charge capacity, calculate the state-of-charge label for each data point in the training sample set;

[0020] Building a state-of-charge estimator based on the state-of-charge label;

[0021] For any charging fragment data of unknown status in the cloud, we first determine which fragment data set it belongs to, and then use the charge state estimator constructed by the fragment data set to which it belongs to obtain the charge state of each point in the fragment data.

[0022] According to at least one embodiment of the present application, the state of the power battery includes the state of charge, the state of health, the energy state, the remaining charging time, and the charging power. The present invention has the following beneficial effects:

[0023] (1) Traditional methods will fail for cloud battery data with large sampling periods and without initial state tags, but the methods described in the present invention are applicable.

[0024] (2) The method of the present invention can combine and use data of different battery cells with similar aging states, greatly improving the efficiency of using cloud battery data and reducing the battery offline test cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The following will further explain the above characteristics, technical features, advantages and implementation methods of the present application in a clear and understandable manner through the description of the preferred embodiments and in combination with the accompanying drawings. The following drawings are only intended to illustrate and explain the present application, and do not limit the scope of the present application. Among them:

[0026] Figure 1 The present invention shows a method for estimating the state of a power battery in the cloud based on charging data;

[0027] Figure 2 These are two fragment data used in one embodiment. DETAILED DESCRIPTION

[0028] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] As attached Figure 1 As shown, the specific steps of a cloud-based power battery state estimation method based on charging data described in the present invention are as follows:

[0030] S1: Data cleaning and classification: After the charging segment data is cleaned in the cloud, the battery characteristic parameters are judged, such as the similarity of the aging state of the charging segment data, and then the segment data with similar aging state (i.e., high similarity of aging state) are classified according to the charging condition (fast charging / slow charging, etc.) and operating temperature to obtain the classified charging segment data. The segment data classified into the same category have similar battery characteristic parameters, such as similar aging state, similar operating temperature range and the same charging condition. Specifically, the following steps are included:

[0031] S11: Data cleaning: Extract charging segment data of different units in the cloud. The charging segment data includes charging voltage, current, temperature and other data. Clean missing values, over-limit values, timing error values ​​and outliers.

[0032] S12: Aging state similarity determination: Use machine learning methods or deep learning methods to determine the aging state similarity of charging segment data. Machine learning methods or deep learning methods include but are not limited to support vector machines, K-nearest neighbors, random forests, fully connected neural networks, convolutional neural networks, recurrent neural networks, etc. For example, in this embodiment, a deep convolutional neural network is used to determine the aging state similarity of the cleaned charging segment data, and the specific method is described in patent application number: CN202410474514.4.

[0033] S13: Charging condition and operating temperature determination: determining the charging condition and operating temperature of the charging segment data;

[0034] S14: Data classification: Classify the charging segment data according to the aging state, charging condition (fast charging / slow charging, etc.) and operating temperature, so that the segment data classified into the same category have similar aging states, similar operating temperature ranges and the same charging conditions, and obtain the classified charging segment data.

[0035] S2: Charging segment data splicing: For charging segment data classified into the same category, continuous voltage and current data within a certain time range are intercepted to construct multiple training samples; each training sample is clustered based on a clustering algorithm, and the relative position of the starting point of different charging segment data is determined based on the clustering results, so as to determine the position of the splicing point; each charging segment data classified into the same category is spliced ​​to obtain a complete charging curve under the current aging state and operating temperature. Specifically, the following steps are included:

[0036] S21: constructing d-dimensional training samples: for charging segment data classified into the same category, extracting continuous voltage and current data within a certain time range, and constructing multiple training samples.

[0037] The two fragment data used in this embodiment are shown in the attached Figure 2 Segment data 1 starts constant current charging from 0% SOC, with a cumulative charge of about 40Ah; segment data 2 starts constant current charging from the middle section to 100% SOC, with a cumulative charge of about 35Ah.

[0038] In this embodiment, the following training samples are constructed for the data points of each charging segment data:

[0039]

[0040] where y j represents the training sample corresponding to data point j; U N represents each data point of the standardized voltage sequence, and the time interval between the data points is the cloud data sampling period; N represents the total number of samples generated by each charging segment data; Represents the standardized current series (IN,j I N,j+1 ...I N,j+d-2 ). In this embodiment, the sample dimension is 13.

[0041] S22: Classification of training samples and segment splicing: Cluster each training sample based on a clustering algorithm, and determine the relative position of the starting points of different charging segment data based on the clustering results, so as to determine the position of the splicing point. The clustering algorithm includes but is not limited to Gaussian mixture model, support vector machine, K-nearest neighbor, etc.

[0042] In this embodiment, the Gaussian mixture model is used to cluster the training samples, wherein the Gaussian mixture model parameters are solved using the expectation maximum algorithm. The sample corresponding to point p in segment 1 are classified into the same category. Therefore, point p is the splicing position of fragment 2 on fragment 1.

[0043] S23: splicing the charging segment data classified into the same category to obtain a complete charging curve under the current aging state and operating temperature.

[0044] By merging fragment 1 and fragment 2 at the splicing position (ie, point p), a complete charging curve is obtained.

[0045] S3: State estimation: Based on the spliced ​​charging curve, estimate the state of the power battery, including state of charge, health state, energy state, remaining charging time, charging power, etc. Specifically, it includes the following steps:

[0046] S31: Calculate the absolute ampere-hour accumulation of each data point in the spliced ​​complete charging curve.

[0047] In this embodiment, the starting point of segment 1 is the starting point of the complete charging curve, so the relative ampere-hour accumulation of each point in segment 1 (relative to the starting point of segment 1) is equal to the absolute ampere-hour accumulation. The absolute ampere-hour accumulation of each point in segment 2 is obtained by the following formula:

[0048]

[0049] in, represents the absolute ampere-hour accumulation at point p of segment 1; Indicates the absolute ampere-hour accumulation at each point in segment 2; Indicates the relative ampere-hour accumulation of each point in segment 2 (relative to the starting point of segment 2); N 2 Represents the total number of samples constructed in fragment 2.

[0050] S32: State estimation: Based on the spliced ​​charging curve, estimate the state of the power battery, including state of charge, health state, energy state, remaining charging time, charging power, etc.

[0051] For example, the maximum absolute ampere-hour accumulation of each data point in the spliced ​​complete charging curve relative to the starting point of the charging curve is the estimated charging capacity of the power battery under the current aging state and temperature, which can be used to characterize the state of health (SOH) of the power battery.

[0052] Refer to formula (2) to obtain the absolute ampere-hour accumulation Q of each sample a , then the estimated charge capacity for:

[0053]

[0054] In this embodiment, since segment 2 is a fully filled segment, This is the charge capacity estimated based on the splicing curve, based on which the battery's SOH can be calculated.

[0055] Using the estimated charging capacity, the state of charge (SOC) label of each data point in the training sample set is calculated, and then the SOC estimator is constructed based on the SOC label. For any unknown charging fragment data in the cloud, first determine which fragment data set it belongs to, and then use the SOC estimator constructed by the fragment data set to obtain the SOC of each point in the fragment data.

[0056] For state of charge (SOC) estimation, the state of charge estimator is first constructed. The state of charge calculation formula corresponding to each sample in the training sample set is as follows:

[0057]

[0058] in Represents the estimated state of charge of each training sample. Then the state of charge label of each category in step S32 can be obtained by the mean of the state of charge labels of all samples in the category, thereby constructing a GMM-based state of charge estimator. For any segment data to be estimated whose aging state and temperature are similar to those of the training sample, the state of charge estimation process is as follows: first, construct a sample based on formula (1), then use GMM to predict the category to which the sample to be estimated belongs, and the state of charge label of the category is the state of charge estimation result.

[0059] It should be pointed out that the present invention only takes the health status and charge status estimation as an embodiment, but after obtaining the complete charging curve through the method described in step S2, the energy state (SOE), remaining charging time, charging power, etc. can also be estimated, which all fall within the protection scope of the present invention.

[0060] The above description is only an illustrative embodiment of the present application and is not intended to limit the scope of the present application. Any equivalent changes, modifications and combinations made by any technician in the field without departing from the concept and principle of the present application shall fall within the scope of protection of the present application.

Claims

1. A cloud-based power battery state estimation method based on charging data, comprising: S1: After the charging segment data is cleaned in the cloud, the battery characteristic parameters are identified, and the charging segment data is classified according to the battery characteristic parameters to obtain the classified charging segment data; S2: construct multiple training samples for charging segment data classified into the same category; Clustering the multiple training samples based on a clustering algorithm, and judging the relative positions of the starting points of different charging segment data based on the clustering results, thereby determining the position of the splicing point; The charging segment data classified into the same category are spliced ​​to obtain a complete charging curve under the current aging state and operating temperature; S3: Estimate the state of the power battery based on the spliced ​​complete charging curve.

2. The method according to claim 1, wherein: Battery characteristic parameters include aging state similarity, charging conditions and operating temperature. Charging conditions include fast charging and slow charging.

3. The method according to claim 1, wherein: The fragment data grouped into the same category have similar aging states, similar operating temperature ranges, and the same charging conditions.

4. The method according to claim 1, wherein: The charging segment data includes charging voltage, current, and operating temperature.

5. The method according to claim 1, wherein: The similarity of aging status of charging segment data is determined by using machine learning methods or deep learning methods.

6. The method according to claim 1, wherein: In step S2, each training sample is clustered using a Gaussian mixture model, wherein the Gaussian mixture model parameters are solved using an expectation maxima algorithm.

7. The method according to claim 1, wherein: Step S3 includes: S31: Calculate the absolute ampere-hour accumulation of each data point in the spliced ​​complete charging curve; S32: Estimate the state of the power battery based on the assembled charging curve.

8. The method according to claim 7, wherein: The power battery is in a healthy state, and step S32 includes: In the spliced ​​complete charging curve, the maximum value of the absolute ampere-hour accumulation of each data point relative to the starting point of the charging curve is used as the estimated power battery charging capacity under the current aging state and temperature, which is used to characterize the health status of the power battery.

9. The method according to claim 8, wherein: The state of the power battery is the charged state, and step S32 includes: Using the estimated charge capacity, calculate the state-of-charge label for each data point in the training sample set; Building a state-of-charge estimator based on the state-of-charge label; For any charging fragment data of unknown status in the cloud, we first determine which fragment data set it belongs to, and then use the charge state estimator constructed by the fragment data set to which it belongs to obtain the charge state of each point in the fragment data.

10. The method according to claim 1, wherein: The status of the power battery includes charge state, health state, energy state, remaining charging time, and charging power.

Images