Data center battery capacity prediction method, device and equipment

Abstract

The application discloses a data center battery capacity detection method and a battery capacity attenuation coefficient prediction model construction method. In the method, the model construction method is based on state monitoring data in a historical charging and discharging process of a battery, obtains battery capacity attenuation coefficient data and corresponding battery state characteristics, and learns a battery capacity attenuation coefficient prediction model from the data. In this way, no additional experiments or additional sensors are needed, the workload is small and the method is easy to implement, and therefore, the model construction efficiency can be effectively improved. Meanwhile, the processing mode fully considers the influence of factors such as the battery internal resistance, temperature, voltage, use time, use frequency and battery model on the actual capacity of the battery, and therefore, the model prediction accuracy can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of battery management technology, specifically to a method and apparatus for constructing a battery capacity degradation coefficient prediction model, a method and apparatus for predicting data center battery capacity, and electronic equipment. Background Technology

[0002] The actual capacity of batteries is a crucial foundational data point in data center operations and maintenance. Based on the battery capacity, parameters such as discharge time and state of charge can be calculated. Batteries with excessively low actual capacity can also be identified and replaced, improving the stability of data center batteries. The calculation of actual battery capacity primarily involves estimating the battery capacity degradation factor, and then multiplying the rated capacity by the degradation factor to obtain the actual capacity. Therefore, accurately estimating the battery capacity degradation factor is the core issue.

[0003] A typical method for estimating the battery capacity degradation coefficient includes the following steps: 1) Conduct a full discharge experiment on the battery (discharge the battery to release all its capacity), fully discharge and then fully charge it each time, record the number of discharges, internal resistance and temperature before discharge, and current and voltage during discharge, and divide the discharge amount of each discharge by the rated capacity of the battery to obtain the battery capacity degradation coefficient; 2) Train a data-driven battery capacity degradation coefficient prediction model, with the model input being internal resistance, temperature, number of battery cycles, and battery current, and the model output being the battery capacity degradation coefficient; 3) Predict the battery capacity degradation coefficient using the model.

[0004] However, in the process of realizing this invention, the inventors discovered that the above-mentioned technical solutions all have at least the following problems: This method requires a full discharge experiment, adding sensors and other equipment on-site to collect experimental data to obtain the features of the input model and the corresponding battery capacity decay coefficient data. Furthermore, the battery cycle count can only be obtained under a fully discharged scenario, resulting in a large workload and low overall accuracy. Therefore, how to simplify the model construction method to improve model construction efficiency and model prediction accuracy has become an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] This application provides a method for constructing a prediction model for battery capacity degradation coefficient, thereby addressing the problems of low model construction efficiency and low prediction accuracy in existing technologies. This application also provides an apparatus for constructing a prediction model for battery capacity degradation coefficient, a battery capacity prediction method and apparatus, and electronic equipment.

[0006] This application provides a method for predicting the capacity of data center batteries, including:

[0007] The state data at the start time and the state data at the target time of the battery to be predicted are obtained as the state characteristics of the battery to be predicted.

[0008] Based on the state characteristics of the battery to be predicted, the battery capacity decay coefficient is determined by the battery capacity decay coefficient prediction model.

[0009] The battery capacity of the battery to be predicted is determined based on the battery capacity decay coefficient of the battery to be predicted.

[0010] The model is constructed as follows: based on the state monitoring data during the historical charging and discharging process of the battery, multiple target batteries with two discharge durations exceeding the discharge duration threshold are obtained; based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold, the state characteristics and battery capacity decay coefficient of the target batteries are determined; based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, the model is constructed.

[0011] This application also provides a method for constructing a prediction model for battery capacity degradation coefficient, including:

[0012] Acquire historical state monitoring data during the charging and discharging process of the battery;

[0013] Based on the historical status monitoring data, multiple target batteries with two discharge durations exceeding the discharge duration threshold are identified.

[0014] Based on two discharge data points where the discharge duration of the target battery exceeds the discharge duration threshold, determine the state characteristics and battery capacity decay coefficient of the target battery.

[0015] Based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, a battery capacity decay coefficient prediction model is constructed.

[0016] Optionally, the state characteristics include: battery internal resistance at the factory, battery internal resistance, temperature and / or voltage before two discharges, battery historical discharge count, total battery usage time, and battery rated capacity.

[0017] Optionally, the step of acquiring multiple target batteries that have two discharge durations exceeding a discharge duration threshold based on the historical state monitoring data includes:

[0018] Based on historical condition monitoring data, determine the battery discharge range data;

[0019] Based on the discharge interval data, a target battery with two discharge durations exceeding the discharge duration threshold is obtained.

[0020] Optionally, determining the battery discharge range data based on historical status monitoring data includes:

[0021] Based on historical status monitoring data, obtain voltage change point data;

[0022] Based on the voltage change data, determine the battery's discharge range data.

[0023] Optionally, the historical status monitoring data includes battery voltage data generated by charging and discharging the battery pack, wherein the battery pack includes multiple individual batteries;

[0024] The step of determining the battery discharge range data based on the historical status monitoring data includes:

[0025] Based on the historical status monitoring data, obtain the discharge range data of each battery cell;

[0026] The common discharge range data of multiple batteries is used as the discharge range data of each battery.

[0027] Optionally, determining the battery discharge range data based on the historical status monitoring data further includes:

[0028] Based on the historical status monitoring data, determine the proportion of missing data for a single battery cell relative to the total battery pack data.

[0029] If the ratio is less than or equal to the ratio threshold, then linear interpolation is performed using the data before and after the missing data;

[0030] If the ratio is greater than the ratio threshold, then the data for that battery cell is removed.

[0031] Optional, also includes:

[0032] Select batteries from multiple target batteries whose capacity decay coefficients are between the first threshold and the second threshold.

[0033] The model is constructed based on the correspondence between the selected multiple batteries.

[0034] Optional, also includes:

[0035] The historical status monitoring data is downsampled;

[0036] Based on the sampled historical status monitoring data, the multiple target batteries are obtained.

[0037] Optionally, the historical state monitoring data includes: state change data generated during float charging and state change data generated during discharge testing.

[0038] This application also provides a device for constructing a prediction model for battery capacity degradation coefficient, comprising:

[0039] The historical data acquisition unit is used to acquire historical state monitoring data during the charging and discharging process of the battery;

[0040] The battery selection unit is used to acquire multiple target batteries that have two discharge durations greater than the discharge duration threshold based on the historical status monitoring data.

[0041] The training data generation unit is used to determine the state characteristics and battery capacity decay coefficient of the target battery based on two discharge data where the discharge duration of the target battery is greater than the discharge duration threshold.

[0042] The model training unit is used to construct a battery capacity decay coefficient prediction model based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient.

[0043] This application also provides a battery capacity prediction device, comprising:

[0044] The feature acquisition unit is used to acquire the initial state data and the target state data of the battery to be predicted, as the state features of the battery to be predicted.

[0045] The attenuation coefficient prediction unit is used to determine the battery capacity attenuation coefficient of the battery to be predicted based on the state characteristics of the battery to be predicted, using a battery capacity attenuation coefficient prediction model.

[0046] The capacity calculation unit is used to determine the battery capacity of the battery to be predicted based on the battery capacity decay coefficient of the battery to be predicted.

[0047] The model is constructed as follows: based on the state monitoring data during the historical charging and discharging process of the battery, multiple target batteries with two discharge durations exceeding the discharge duration threshold are obtained; based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold, the state characteristics and battery capacity decay coefficient of the target batteries are determined; based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, the model is constructed.

[0048] This application also provides an electronic device, including:

[0049] Processor and memory;

[0050] The memory is used to store a program for implementing a battery capacity prediction method. After the device is powered on and the program for the method is run by the processor, the following steps are performed: acquiring the initial state data and target state data of the battery to be predicted as the state characteristics of the battery to be predicted; determining the battery capacity decay coefficient of the battery to be predicted based on the state characteristics of the battery to be predicted using a battery capacity decay coefficient prediction model; determining the battery capacity of the battery to be predicted based on the battery capacity decay coefficient; wherein, the model is constructed in the following manner: acquiring multiple target batteries with two discharge durations exceeding a discharge duration threshold based on state monitoring data during the historical charging and discharging process of the battery; determining the state characteristics and battery capacity decay coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold; constructing the model based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient.

[0051] This application also provides an electronic device, including:

[0052] Processor and memory;

[0053] The memory is used to store a program for constructing a battery capacity degradation coefficient prediction model. After the device is powered on and the program of the method is run by the processor, the following steps are performed: acquiring historical state monitoring data during the charging and discharging process of the battery; acquiring multiple target batteries with two discharge durations exceeding the discharge duration threshold based on the historical state monitoring data; determining the state characteristics and battery capacity degradation coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold; and constructing a battery capacity degradation coefficient prediction model based on the correspondence between the state characteristics of multiple target batteries and the battery capacity degradation coefficient.

[0054] This application also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the various methods described above.

[0055] This application also provides a computer program product including instructions that, when run on a computer, cause the computer to perform the various methods described above.

[0056] Compared with the prior art, this application has the following advantages:

[0057] The battery capacity degradation coefficient prediction model construction method provided in this application is based on the state monitoring data of the battery during historical charging and discharging processes. It acquires battery capacity degradation coefficient data and corresponding battery state characteristics, and learns a battery capacity degradation coefficient prediction model from this data. This eliminates the need for additional experiments or sensors, reducing workload and making implementation easier, thus effectively improving model construction efficiency. Furthermore, this approach fully considers the impact of factors such as battery internal resistance, temperature, voltage, usage time, number of uses, and battery model on the actual battery capacity, thereby effectively improving the model's prediction accuracy.

[0058] The data center battery capacity prediction method provided in this application acquires the initial and target state data of the battery to be predicted as its state characteristics. A battery capacity decay coefficient prediction model constructed in this manner determines the battery capacity decay coefficient based on the state characteristics of the battery. Finally, the battery capacity is determined based on the battery capacity decay coefficient. This eliminates the need for additional experiments or sensors, reducing workload and ease of implementation, thus effectively improving battery capacity prediction efficiency. Furthermore, this approach fully considers the impact of factors such as battery internal resistance, temperature, voltage, usage time, number of uses, and battery model on the actual battery capacity, thereby effectively improving the accuracy of battery capacity prediction. Attached Figure Description

[0059] Figure 1 A flowchart illustrating an embodiment of the battery capacity attenuation coefficient prediction model construction method provided in this application;

[0060] Figure 2a A schematic diagram of voltage mutation points in an embodiment of the model building method provided in this application;

[0061] Figure 2b A magnified schematic diagram of the voltage mutation point in an embodiment of the model building method provided in this application;

[0062] Figure 3 A schematic diagram of the attenuation coefficient of an embodiment of the model building method provided in this application;

[0063] Figure 4 A flowchart illustrating an embodiment of the data center battery capacity testing method provided in this application. Detailed Implementation

[0064] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0065] This application provides a method and apparatus for constructing a battery capacity degradation coefficient prediction model, a method and apparatus for predicting data center battery capacity, and electronic equipment. The various solutions are described in detail in the following embodiments.

[0066] First Embodiment

[0067] Please refer to Figure 1 This is a flowchart of the battery capacity degradation coefficient prediction model construction method of this application. In this embodiment, the method may include the following steps:

[0068] Step S101: Obtain historical state monitoring data during the charging and discharging process of the battery.

[0069] A storage battery is an electrochemical device that directly converts chemical energy into electrical energy. Specifically, a storage battery has the following characteristics: after discharging, it can be regenerated by charging, thus storing electrical energy as chemical energy; when discharging is needed, it converts chemical energy back into electrical energy. Examples include commonly used mobile phone batteries, car batteries (commonly known as accumulators), and data center batteries.

[0070] Taking data center batteries as an example, their status data during daily use can be monitored and recorded to form historical status monitoring data during the battery charging and discharging process. From the perspective of status attributes, the historical status monitoring data includes, but is not limited to: internal resistance, temperature, current, voltage, etc. before discharge, and may also include data such as battery location, battery number, and time. From the perspective of data generation method, the historical status monitoring data includes, but is not limited to: status change data generated during discharge testing, and status change data generated during float charging.

[0071] In practical applications, data centers typically conduct periodic discharge tests on their batteries, but not full discharge tests; it is unnecessary to discharge all of the battery capacity. For example, data centers may conduct battery discharge tests every six months and monitor and record data on changes in the battery's state during the discharge process.

[0072] A storage battery is a charging device. Besides supplying power to the conventional load, it also provides a float charge current; this operating mode is called float charge operation. Therefore, the battery can charge and discharge according to fluctuations in the power line voltage. When the load is light and the power line voltage is high, the battery charges; when the load is heavy or a power outage occurs, the battery discharges to share some or all of the load. In this way, the battery plays a voltage stabilizing role and is in standby mode.

[0073] In practical implementation, to obtain sufficient power supply voltage, data center batteries typically use a battery pack approach, such as a pack of 20 or 120 cells. Taking lead-acid batteries as an example, they can be repeatedly charged and discharged. Each individual cell has a voltage of 2V, and a battery pack consists of one or more cells. Common battery types include 2V, 4V, 6V, 8V, and 24V. For instance, the batteries used in data center equipment are six lead-acid batteries connected in series to form a 12V battery pack. In this embodiment, data center batteries are available in both 2V and 12V versions, with the 12V lead-acid battery consisting of six 2V batteries connected in series.

[0074] Step S103: Based on the historical status monitoring data, acquire multiple target batteries that have two discharge durations exceeding the discharge duration threshold.

[0075] The model construction method provided in this application embodiment obtains battery capacity decay coefficient data and corresponding battery state characteristics based on state monitoring data during the historical charging and discharging process of the battery. The batteries corresponding to these training samples need to meet certain conditions, namely: there are two discharge durations that are longer than the discharge duration threshold.

[0076] The discharge duration threshold can be set according to application requirements. In practical applications, a larger discharge duration threshold results in fewer eligible batteries and less training data, but a higher probability that the data is for discharge testing, thus leading to a more effective attenuation coefficient. Conversely, a smaller discharge duration threshold results in more eligible batteries and more training data, but a lower probability that the data is for discharge testing, leading to a less effective attenuation coefficient. For example, for data center batteries, a threshold of 15 minutes can be set to identify batteries with two discharge durations exceeding 15 minutes.

[0077] In one example, step S103 may include the following sub-steps:

[0078] Step S1031: Determine the discharge range data of the battery based on historical status monitoring data.

[0079] The discharge interval data refers to the battery's discharge time period data, such as the discharge time period being 17:21-17:53 on 2021 / 12 / 13.

[0080] In practice, battery discharge record data can be generated first based on historical status monitoring data; then, the battery discharge range data can be determined based on the battery discharge record data. The battery discharge record data includes, but is not limited to: battery identification, discharge time period, discharge voltage data, and battery data before discharge. The battery identification can be the battery device number, which is related to the battery's placement location. The discharge time period can include the start and end discharge times, the discharge voltage data can include voltage data at multiple points within the discharge time period, and the battery data before discharge can include the battery's internal resistance, temperature, and voltage before discharge. Table 1 shows the battery discharge record data in this embodiment.

[0081]

[0082]

[0083] In one example, step S1031 may include the following sub-steps:

[0084] Step S1031-1: Obtain voltage change point data based on historical status monitoring data.

[0085] Step S1031-3: Determine the discharge range data of the battery based on the voltage change point data.

[0086] In practice, a change-point detection algorithm can be used to obtain voltage change point data based on historical state monitoring data. During discharge, the battery voltage decreases. After discharge, the battery recharges, and the voltage increases. The change-point detection algorithm can identify the decrease and increase of individual cell voltages. If, within a certain period, the voltage of an individual cell first decreases and then increases, this is taken as the discharge range of that cell.

[0087] In practical applications, voltage data often contains abrupt changes, such as... Figure 2a The spikes shown can interfere with the determination of the discharge range. To solve this problem, step S1033 can be implemented as follows: First, by judging the changes in data values ​​and slopes, interfering abrupt changes in the data are identified and smoothed. After smoothing, voltage abrupt change points can be eliminated. Then, the discharge range is determined based on the smoothed data. Figure 2b This is what the burrs look like when magnified. The protruding parts are the original data, which are then smoothed into the straight lines below.

[0088] In one example, a data center typically discharges an entire battery pack, which comprises multiple individual cells. The historical state monitoring data includes battery voltage data generated during the charging and discharging of the battery pack. In this case, the discharge range is common to the entire battery pack. Step S1031 can be implemented as follows: 1) Obtain the discharge range data of each individual cell based on the historical state monitoring data; 2) Use the common discharge range data of multiple cells as the discharge range data of each individual cell. This method ensures that the common discharge range of the same battery pack is selected as the battery discharge range.

[0089] Step S1033: Based on the discharge interval data, obtain the target battery that has two discharge durations greater than the discharge duration threshold.

[0090] After determining the discharge range data of the battery, the data can be compared with the discharge duration threshold. Based on the comparison results, target batteries with two discharge durations exceeding the discharge duration threshold can be selected.

[0091] In practical applications, to obtain sufficient power supply voltage, data center batteries are typically grouped in sets of 20 or 120, with each group discharging together. The amount of raw voltage data generated during the discharge of a group of batteries is substantial; for example, each battery generates a voltage data point every 10 seconds. This necessitates acquiring data from multiple target batteries based on a large volume of historical state detection data.

[0092] To reduce computational load and improve model building efficiency, the method provided in this embodiment may further include the following steps: downsampling the historical state monitoring data; correspondingly, step S103 can be implemented in the following way: obtaining the multiple target batteries based on the sampled historical state monitoring data. For example, the original voltage data of the entire battery pack can be downsampled, such as recording one voltage data point every minute for original voltage data generated every 10 seconds, which can effectively reduce the amount of data and thus reduce computational complexity.

[0093] In practical applications, the original voltage data may be missing. To improve the quality of training data, the method provided in this embodiment may further include the following steps: Based on the historical state monitoring data, determine the proportion of missing data for a single battery cell relative to the total battery data; if the proportion is less than or equal to a proportion threshold, perform linear interpolation using data before and after the missing data; if the proportion is greater than the proportion threshold, remove the data for that battery cell. For example, if the missing data for a single battery cell accounts for less than 1% of the total data, perform linear interpolation using data before and after the missing data; if the missing data for a single battery cell exceeds 1% of the total data, remove the data for that battery cell.

[0094] Step S105: Based on two discharge data points where the discharge duration of the target battery exceeds the discharge duration threshold, determine the state characteristics and battery capacity decay coefficient of the target battery.

[0095] The correspondence between a set of state features of a target battery and its capacity degradation coefficient constitutes a training dataset. A target battery may have multiple sets of correspondences between state features and its capacity degradation coefficient; the correspondences between multiple state features and their capacity degradation coefficients for multiple target batteries constitute the entire training dataset.

[0096] The state characteristics include, but are not limited to: battery internal resistance at the factory, battery internal resistance, temperature and / or voltage before two discharges, historical discharge count, total battery usage time, and rated battery capacity. The battery internal resistance at the factory can be the internal resistance after the battery is manufactured or the internal resistance when the battery is first put into use; it can be a reference value provided by the battery manufacturer. In practical applications, if the manufacturer's internal resistance value is not available, this value can be obtained using an algorithm, such as by using historical internal resistance data, removing outliers, and taking the 10th percentile value, i.e., a value obtained from actual usage internal resistance data. The rated battery capacity can be used to distinguish different battery models.

[0097] The battery capacity degradation coefficient can be calculated based on two discharge records of the same selected battery. In this embodiment, the battery capacity degradation coefficient is equal to the ratio of the amount of electricity discharged when the battery voltage drops to the same value during the two discharges. Figure 3 As shown, the battery capacity degradation coefficient alpha = t2*I2 / t1*I1. Where t1 and t2 are the times required for the same battery to discharge to the same voltage twice (typically, this voltage value can be taken as the 15th percentile of the voltage during the first discharge), and I1 and I2 are the currents of the same battery during the two discharges. In practice, multiple voltage data points from the first discharge can be taken, and the degradation coefficient can be calculated multiple times using the above formula. The average of these multiple degradation coefficients can then be taken as the battery capacity degradation coefficient. For example, the 10th and 15th percentiles of the voltage during the first discharge can be taken separately, and the degradation coefficients can be calculated twice using the above formula. The average of the two degradation coefficients can then be taken as the battery capacity degradation coefficient.

[0098] For example, the features used for model training are x = [2.718, 25.3, 2.747, 25.4, 1.952, 1, 0.4, 0.4, 13.398, 13.387, 1073, 26, 360], and the capacity decay coefficient is y = [0.827]. From left to right, x represents: first discharge internal resistance, first discharge temperature, second discharge internal resistance, second discharge temperature, internal resistance reference value (factory value), historical discharge count, first discharge current, second discharge current, first discharge voltage, second discharge voltage, battery usage time (in days), time interval between two discharges (in days), and battery rated capacity (in ampere-hours, Ah). y is calculated using the two discharge voltage data; see [link to relevant documentation]. Figure 3 The graph shown.

[0099] In one example, due to anomalies in the data center battery data or battery replacements, data needs to be filtered to select those with battery capacity degradation coefficients between 0.7 and 1.0. To obtain representative degradation coefficients, the method may further include the following steps: selecting batteries from a plurality of target batteries whose capacity degradation coefficients are between a first threshold and a second threshold; and constructing the model based on the correspondence between the selected batteries. For example, if the first threshold is 0.7 and the second threshold is 1, then data with battery capacity degradation coefficients between 0.7 and 1.0 are selected.

[0100] Step S107: Based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, construct a battery capacity decay coefficient prediction model.

[0101] The model can be a tree model, such as LightGBM. The input data is the aforementioned state features, and the output data is the battery capacity degradation coefficient calculated using the above method. The model is trained on a large amount of training data, and saved after convergence. In specific implementations, open-source machine learning algorithms can be used during the training of the battery capacity degradation coefficient prediction model.

[0102] As can be seen from the above embodiments, the battery capacity degradation coefficient prediction model construction method provided in this application obtains battery capacity degradation coefficient data and corresponding battery state characteristics based on state monitoring data during the historical charging and discharging process of the battery, and learns the battery capacity degradation coefficient prediction model from this data. This eliminates the need for additional experiments or the installation of additional sensors, reducing workload and making implementation easier, thus effectively improving model construction efficiency. Furthermore, this approach fully considers the impact of factors such as battery internal resistance, temperature, voltage, usage time, number of uses, and battery model on the actual battery capacity, thus effectively improving the model prediction accuracy.

[0103] Second Embodiment

[0104] In the above embodiments, a method for constructing a battery capacity degradation coefficient prediction model is provided. Correspondingly, this application also provides a device for constructing a battery capacity degradation coefficient prediction model. This device corresponds to the embodiments of the above method. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0105] This application also provides a device for constructing a battery capacity decay coefficient prediction model, including: a historical data acquisition unit, a battery selection unit, a training data generation unit, and a model training unit.

[0106] The system includes a historical data acquisition unit for acquiring historical state monitoring data during the charging and discharging process of the battery; a battery selection unit for acquiring multiple target batteries with two discharge durations exceeding a discharge duration threshold based on the historical state monitoring data; a training data generation unit for determining the state characteristics and battery capacity decay coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold; and a model training unit for constructing a battery capacity decay coefficient prediction model based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient.

[0107] Optionally, the state characteristics include, but are not limited to: battery internal resistance at the factory, battery internal resistance, temperature and / or voltage before two discharges, battery historical discharge count, total battery usage time, and battery rated capacity.

[0108] Optionally, the battery selection unit includes a discharge range determination unit and a selection unit. The discharge range determination unit is used to determine the discharge range data of the battery based on historical status monitoring data; the selection unit is used to acquire target batteries that have two discharge durations exceeding a discharge duration threshold based on the discharge range data.

[0109] Optionally, the discharge range determination unit is specifically used to obtain voltage change point data based on historical state monitoring data; and to determine the discharge range data of the battery based on the voltage change point data.

[0110] Optionally, the historical state monitoring data includes battery voltage data generated by charging and discharging the battery pack, wherein the battery pack includes multiple individual batteries; the discharge range determination unit is specifically used to obtain the discharge range data of each battery cell based on the historical state monitoring data; and to use the common discharge range data of multiple batteries as the discharge range data of each battery cell.

[0111] Optionally, the discharge range determination unit is further configured to determine, based on the historical state monitoring data, the proportion of missing data of a single battery cell to the total battery data; if the proportion is less than or equal to a proportion threshold, then linear interpolation is performed using the data before and after the missing data; if the proportion is greater than the proportion threshold, then the data of that battery cell is removed.

[0112] Optionally, the device may further include: a training data selection unit. The training data selection unit is used to select batteries from a plurality of target batteries whose battery capacity decay coefficients are between a first threshold and a second threshold; and a model training unit, specifically used to construct the model based on the correspondence between the selected plurality of batteries.

[0113] Optionally, the device may further include: a data sampling unit for downsampling the historical state monitoring data; and a battery selection unit, specifically used to acquire the plurality of target batteries based on the sampled historical state monitoring data.

[0114] Optionally, the historical state monitoring data includes: state change data generated during float charging and state change data generated during discharge testing.

[0115] Third Embodiment

[0116] In the above embodiments, a method for constructing a prediction model for battery capacity degradation coefficient is provided. Correspondingly, this application also provides an electronic device. This device corresponds to the embodiments of the above method. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple, and relevant details can be found in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0117] An electronic device according to this embodiment includes a processor and a memory. The memory stores a program for implementing a method for constructing a battery capacity decay coefficient prediction model. After the device is powered on and the program of the method is run by the processor, the following steps are performed: acquiring historical state monitoring data during the charging and discharging process of the battery; acquiring multiple target batteries with two discharge durations exceeding a discharge duration threshold based on the historical state monitoring data; determining the state characteristics and battery capacity decay coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold; and constructing a battery capacity decay coefficient prediction model based on the correspondence between the state characteristics of the multiple target batteries and the battery capacity decay coefficient.

[0118] Fourth embodiment

[0119] Corresponding to the above-described method for constructing a battery capacity degradation coefficient prediction model, this application also provides a method for predicting data center battery capacity. The parts of this embodiment that are the same as those in the first embodiment will not be repeated here; please refer to the corresponding parts in Embodiment 1.

[0120] Please refer to Figure 4 This is a flowchart of the battery capacity prediction method of this application. In this embodiment, the method may include the following steps:

[0121] Step S401: Obtain the initial state data and target state data of the battery to be predicted, as the state characteristics of the battery to be predicted.

[0122] When using the model constructed in the above embodiments for actual prediction, the state characteristics x of the battery to be predicted can be from left to right as follows: the internal resistance of the battery during its first discharge in history, the temperature of the battery during its first discharge in history, the current internal resistance, the current temperature, the reference value of the internal resistance (factory value), the number of historical discharges, the current of the battery during its first discharge in history, the current discharge current (currently not discharged, estimated by dividing the current load power by the battery pack voltage), the voltage of the battery during its first discharge in history, the current voltage, the battery usage time (in days), and the time interval between the current and the first historical discharge (in days).

[0123] Step S403: Using the battery capacity decay coefficient prediction model, determine the battery capacity decay coefficient of the battery to be predicted based on the state characteristics of the battery to be predicted.

[0124] The model can be constructed as follows: Based on the state monitoring data during the historical charging and discharging process of the battery, obtain multiple second target batteries that have two discharge durations exceeding the discharge duration threshold; based on the two discharge data of the second target batteries with discharge durations exceeding the discharge duration threshold, determine the state characteristics and battery capacity decay coefficient of the second target batteries; construct the model based on the correspondence between the state characteristics of multiple second target batteries and the battery capacity decay coefficient.

[0125] Step S405: Determine the battery capacity to be predicted based on the battery capacity degradation coefficient of the battery to be predicted. The battery capacity to be predicted can be the product of the battery capacity degradation coefficient and the rated battery capacity.

[0126] The method provided in this application utilizes existing battery-related data from a data center and employs a reasonable and comprehensive model input feature design to fully consider the impact of various factors on the battery capacity degradation coefficient, thus ensuring the accuracy of the actual battery capacity prediction. The predicted actual battery capacity can be used to calculate the battery's discharge time and also to calculate the battery's state of charge (SOC). Experiments show that, on the validation dataset, the average error in predicting the actual battery capacity is 3.5% of the battery's rated capacity.

[0127] As can be seen from the above embodiments, the data center battery capacity prediction method provided in this application obtains the initial and target time state data of the battery to be predicted as its state characteristics. The battery capacity decay coefficient prediction model constructed in the above manner determines the battery capacity decay coefficient based on the state characteristics of the battery to be predicted. Based on the battery capacity decay coefficient, the battery capacity of the battery to be predicted is then determined. This eliminates the need for additional experiments or sensors, reducing workload and making implementation easier, thus effectively improving battery capacity prediction efficiency. Furthermore, this approach fully considers the impact of factors such as battery internal resistance, temperature, voltage, usage time, number of uses, and battery model on the actual battery capacity, thereby effectively improving the accuracy of battery capacity prediction.

[0128] Fifth embodiment

[0129] In the above embodiments, a method for predicting battery capacity is provided. Correspondingly, this application also provides a battery capacity prediction device. This device corresponds to the embodiments of the above method. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0130] This application also provides a battery capacity prediction device, including: a feature acquisition unit, an attenuation coefficient prediction unit, and a capacity calculation unit.

[0131] The feature acquisition unit is used to acquire the initial state data and target state data of the battery to be predicted as the state features of the battery to be predicted; the attenuation coefficient prediction unit is used to determine the battery capacity attenuation coefficient of the battery to be predicted based on the state features of the battery to be predicted through the battery capacity attenuation coefficient prediction model; the capacity calculation unit is used to determine the battery capacity of the battery to be predicted based on the battery capacity attenuation coefficient of the battery to be predicted.

[0132] The model is constructed as follows: based on the state monitoring data during the historical charging and discharging process of the battery, multiple target batteries with two discharge durations exceeding the discharge duration threshold are obtained; based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold, the state characteristics and battery capacity decay coefficient of the target batteries are determined; based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, the model is constructed.

[0133] Sixth Embodiment

[0134] In the above embodiments, a method for predicting battery capacity is provided. Correspondingly, this application also provides an electronic device. This device corresponds to the embodiments of the above method. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple, and relevant details can be found in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0135] This embodiment provides an electronic device, comprising: a processor and a memory; the memory for storing a program for implementing a battery capacity prediction method. After the device is powered on and the program for the method is run by the processor, the following steps are performed: acquiring the initial state data and target state data of the battery to be predicted as state characteristics of the battery; determining the battery capacity decay coefficient of the battery to be predicted based on the state characteristics of the battery using a battery capacity decay coefficient prediction model; determining the battery capacity of the battery to be predicted based on the battery capacity decay coefficient; wherein the model is constructed as follows: acquiring multiple target batteries with two discharge durations exceeding a discharge duration threshold based on state monitoring data during the historical charging and discharging process of the battery; determining the state characteristics and battery capacity decay coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold; constructing the model based on the correspondence between the state characteristics of the multiple target batteries and the battery capacity decay coefficient.

[0136] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

[0137] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0138] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0139] 1. Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0140] 2. Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A method for predicting the capacity of a data center battery, characterized in that, include: The state data at the start time and the state data at the target time of the battery to be predicted are obtained as the state characteristics of the battery to be predicted. Based on the state characteristics of the battery to be predicted, the battery capacity decay coefficient is determined by the battery capacity decay coefficient prediction model. The battery capacity of the battery to be predicted is determined based on the battery capacity decay coefficient of the battery to be predicted. The model is constructed as follows: Based on state monitoring data from the historical charging and discharging processes of the battery, the discharge interval data of the battery is determined; based on the discharge interval data, multiple target batteries with two discharge durations exceeding a discharge duration threshold are obtained; based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold, the state characteristics and battery capacity decay coefficient of the target batteries are determined; the state characteristics include: the battery's internal resistance, temperature, and / or voltage before the two discharges; the model is constructed based on the correspondence between the state characteristics of the multiple target batteries and the battery capacity decay coefficient.

2. A method for constructing a prediction model for battery capacity degradation coefficient, characterized in that, include: Acquire historical state monitoring data during the charging and discharging process of the battery; Based on the historical status monitoring data, the discharge range data of the battery is determined; Based on the discharge interval data, multiple target batteries with two discharge durations exceeding the discharge duration threshold are obtained; Based on two discharge data points where the discharge duration of the target battery exceeds the discharge duration threshold, the state characteristics and battery capacity decay coefficient of the target battery are determined; the state characteristics include: the battery's internal resistance, temperature, and / or voltage before the two discharges; Based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, a battery capacity decay coefficient prediction model is constructed.

3. The method according to claim 2, characterized in that, The status characteristics include: battery internal resistance at the time of manufacture, battery historical discharge count, total battery usage time, and battery rated capacity.

4. The method according to claim 2, characterized in that, Based on historical condition monitoring data, the discharge range data of the battery is determined, including: Based on historical status monitoring data, obtain voltage change point data; Based on the voltage change data, determine the battery's discharge range data.

5. The method according to claim 2, characterized in that, The historical status monitoring data includes battery voltage data generated by charging and discharging the battery pack, which includes multiple individual batteries; The step of determining the battery discharge range data based on the historical status monitoring data includes: Based on the historical status monitoring data, obtain the discharge range data of each battery cell; The common discharge range data of multiple batteries is used as the discharge range data of each battery.

6. The method according to claim 5, characterized in that, The step of determining the battery discharge range data based on the historical state monitoring data further includes: Based on the historical status monitoring data, determine the proportion of missing data for a single battery cell relative to the total battery pack data. If the ratio is less than or equal to the ratio threshold, then linear interpolation is performed using the data before and after the missing data; If the ratio is greater than the ratio threshold, then the data for that battery cell is removed.

7. The method according to claim 2, characterized in that, Also includes: Select batteries from multiple target batteries whose capacity decay coefficients are between the first threshold and the second threshold. The model is constructed based on the correspondence between the selected multiple batteries.

8. The method according to claim 2, characterized in that, Also includes: The historical status monitoring data is downsampled; Based on the sampled historical status monitoring data, the multiple target batteries are obtained.

9. The method according to claim 2, characterized in that, The historical status monitoring data includes: status change data generated during float charging and status change data generated during discharge testing.

10. A device for constructing a prediction model for battery capacity degradation coefficient, characterized in that, include: The historical data acquisition unit is used to acquire historical state monitoring data during the charging and discharging process of the battery; The battery selection unit is used to determine the discharge range data of the battery based on the historical status monitoring data; and to obtain multiple target batteries that have two discharge durations greater than the discharge duration threshold based on the discharge range data. The training data generation unit is used to determine the state characteristics and battery capacity decay coefficient of the target battery based on two discharge data where the discharge duration of the target battery is greater than the discharge duration threshold; the state characteristics include: the internal resistance, temperature and / or voltage of the battery before the two discharges; The model training unit is used to construct a battery capacity decay coefficient prediction model based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient.

11. A battery capacity prediction device, characterized in that, include: The feature acquisition unit is used to acquire the initial state data and the target state data of the battery to be predicted, as the state features of the battery to be predicted. The attenuation coefficient prediction unit is used to determine the battery capacity attenuation coefficient of the battery to be predicted based on the state characteristics of the battery to be predicted, using a battery capacity attenuation coefficient prediction model. The capacity calculation unit is used to determine the battery capacity of the battery to be predicted based on the battery capacity decay coefficient of the battery to be predicted. The model is constructed as follows: Based on state monitoring data from the historical charging and discharging processes of the battery, the discharge interval data of the battery is determined; based on the discharge interval data, multiple target batteries with two discharge durations exceeding a discharge duration threshold are obtained; based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold, the state characteristics and battery capacity decay coefficient of the target batteries are determined; the state characteristics include: the battery's internal resistance, temperature, and / or voltage before the two discharges; the model is constructed based on the correspondence between the state characteristics of the multiple target batteries and the battery capacity decay coefficient.

12. An electronic device, characterized in that, include: Processor and memory; The memory stores a program for implementing a battery capacity prediction method. After the device is powered on and the program is run by the processor, the following steps are performed: acquiring the initial state data and target state data of the battery to be predicted as state characteristics of the battery; determining the battery capacity decay coefficient of the battery to be predicted based on the state characteristics of the battery using a battery capacity decay coefficient prediction model; determining the battery capacity of the battery to be predicted based on the battery capacity decay coefficient; wherein the model is constructed in the following manner: determining the discharge interval data of the battery based on the state monitoring data during the historical charging and discharging process of the battery; acquiring multiple target batteries with two discharge durations exceeding the discharge duration threshold based on the discharge interval data; determining the state characteristics and battery capacity decay coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations exceeding the discharge duration threshold; the state characteristics include: the internal resistance, temperature, and / or voltage of the battery before the two discharges; and constructing the model based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient.

13. An electronic device, characterized in that, include: Processor and memory; The memory is used to store a program for constructing a battery capacity decay coefficient prediction model. After the device is powered on and the program of the method is run by the processor, the following steps are performed: acquiring historical state monitoring data during the charging and discharging process of the battery; determining the discharge interval data of the battery based on the historical state monitoring data; acquiring multiple target batteries with two discharge durations greater than the discharge duration threshold based on the discharge interval data; determining the state characteristics and battery capacity decay coefficient of the target batteries based on the two discharge data of the target batteries with discharge durations greater than the discharge duration threshold; the state characteristics include: the internal resistance, temperature and / or voltage of the battery before the two discharges; Based on the correspondence between the state characteristics of multiple target batteries and the battery capacity decay coefficient, a battery capacity decay coefficient prediction model is constructed.

Images