Method for rapidly detecting health state of retired battery

Abstract

The invention is suitable for the technical field of battery prediction, and provides a method for rapidly detecting the health state of an out-of-service battery, and the method comprises the following operation steps: S1, collecting original data; s2, data processing; s3, extracting features of the data segment of each cycle; s4, a sample set is generated, each sample corresponds to one cycle and comprises a feature vector (features extracted from the cycle) and a label (the discharge capacity of the cycle), and a plurality of continuous cycles form the sample set; s5, model training; according to the method, the capacity (SOH) decline condition in the battery aging process is calculated through ampere-hour integration, continuous test data are split into single cyclic data to form a'label-feature 'sample set, and the SOH prediction model is trained through the sample set, so that the accuracy of battery health state prediction is improved, and the accuracy of SOH prediction is improved. And the battery health state of the recycled retired battery can be quickly evaluated.

Description

Technical Field

[0001] The present invention relates to the field of battery prediction technology, and more particularly, to a method for quickly detecting the health status of retired batteries. Background Art

[0002] As a major energy producer and consumer, my country is actively addressing energy and environmental issues. The rapid development of new energy power generation, with non-fossil energy consumption reaching 15.9% in 2020, has increased demand for energy storage systems. Simultaneously, the surge in new energy vehicle sales has exacerbated the problem of disposing of retired batteries. The cascade utilization of retired power batteries is highly compatible with the urgent needs of these two key sectors: energy storage and new energy vehicles.

[0003] After lithium batteries reach the end of their service life in electric vehicles, they can be further downgraded and reused through other methods to achieve the goals of saving resources, protecting the environment, and reducing costs. Battery reuse is usually measured by capacity. For example, batteries with a capacity between 100% and 80% are used in electric vehicles; batteries between 80% and 60% are used in energy storage; batteries between 60% and 20% are used for backup power; and batteries between 20% and 0% are directly disassembled and raw materials are recycled. Through health status assessment methods for retired lithium batteries, it is possible to effectively reorganize and classify retired batteries for reuse.

[0004] Traditional lithium battery testing usually uses a disassembly testing solution, which usually takes 1 day, greatly increasing the testing time. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a method for quickly detecting the health status of retired batteries, which improves the accuracy of battery health status prediction and can quickly evaluate the battery health status of recycled retired batteries.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for quickly detecting the health status of retired batteries includes the following steps: S1, raw data collection; S2, data processing; S2-1, use the sliding average filter algorithm to filter the noise in the data, where the window width = 30-40ms; S2-2, clean the collected data; S2-3, calculate the cycle capacity by integrating ampere-hours; S2-4, splitting the continuous test data into data segments of a single cycle according to the cycle; S3, extracting features from each cycle of data segments; S4. Sample set generation. Each sample corresponds to a cycle, including a feature vector (features extracted from the cycle) and a label (discharge capacity of the cycle). Multiple consecutive cycles form a sample set. S5, model training; S6, SOH quick prediction.

[0007] The present invention is further configured as follows: the data collection in S1 includes time series, current, voltage, and temperature, and there is no need to obtain the complete charge and discharge process, only the charge and discharge fragments in the 5-10% SOC range are required.

[0008] The present invention is further configured as follows: the noise filtering step in S2-1 includes: establishing a data buffer with a length of N to store the sampled data in sequence; Each time a new data is collected, the oldest data in the queue is discarded; Calculate the arithmetic mean of N data in the current queue as the filter output.

[0009] The present invention is further configured as follows: said S2-2 also includes adaptive correction of the battery data group, and the adaptive algorithm used in the adaptive correction is generally a gradient-based algorithm.

[0010] The present invention is further configured as follows: the ampere-hour integration formula in S2-3 is: capacity = ∫I(t) dt, where I(t) is the instantaneous current, t is the total discharge time, and the integration interval is the entire discharge process; Note that the current is negative during charging and positive during discharging (or take the absolute value according to the definition to ensure that the capacity is positive). For each cycle, calculate the ampere-hour integral of the discharge phase as the capacity of that cycle.

[0011] The present invention is further configured as follows: in S2-4, the cycle is split by finding the moment when the discharge ends, specifically: Discharge phase: The current is positive (or negative, depending on the definition, but generally the discharge current has the opposite sign to the charging current). The end point of the discharge phase (the current becomes 0 or enters a state of rest) is defined as the end point of the cycle. The discharge phase is used as the landmark phase of each cycle, because the capacity is calculated by discharge, from the beginning of the first discharge phase to the beginning of the second discharge phase (i.e., including charging, resting, discharging, and resting); Each discharge process is marked as a cycle, and the capacity of the discharge is recorded. At the same time, other data within the cycle (charging, resting, etc.) are used as the data of the cycle.

[0012] The present invention is further configured as follows: the features extracted in S3 include the charging stage: constant current charging time, voltage curve of constant current charging (including mean, variance, slope), constant voltage charging time, current curve characteristics of constant voltage charging and total charging time; Discharge stage: discharge time, discharge voltage curve (such as average value, minimum value, downslope, etc.) and temperature changes during discharge (such as maximum temperature, temperature rise); Resting stage: resting voltage change (such as relaxation voltage drop); Over the entire cycle: internal resistance (calculated from a pulse test, or estimated from the voltage jump at the start of charge or discharge if no pulse is present) and energy efficiency (discharge energy / charge energy).

[0013] The advantages of the present invention are: The present invention calculates the capacity (SOH) decline during the battery aging process through ampere-hour integration, and splits continuous test data into single cycle data to form a "label-feature" sample set. The SOH prediction model is trained based on this sample set, thereby improving the accuracy of battery health status prediction and enabling rapid assessment of the battery health status of recycled and retired batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a method for rapidly detecting the health status of retired batteries according to the present invention; Figure 2 This is a distribution diagram of data processing of the present invention. DETAILED DESCRIPTION

[0015] The present application will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.

[0016] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0017] See also Figure 1-2 , the present invention provides the following technical solutions: Specifically, it refers to a method for quickly detecting the health status of retired batteries, including the following steps: S1, raw data collection; Obtain battery cycle test data, typically including time series, current, voltage, temperature, etc. It is not necessary to obtain the complete charge and discharge process, only the charge and discharge fragments in the 5-10% SOC range (such as 30%→40% SOC) are required. The equipment used requires high-precision ADC sampling (≥16bit) and a current sampling frequency ≥10Hz (to capture dynamic response).

[0018] S2, data processing; S2-1, use the sliding average filter algorithm to filter the noise in the data, where the window width = 30-40ms; The specific steps include: 1. Create a data buffer with a length of N and store the sampled data in sequence; 2. Every time a new data is collected, the oldest data in the queue is discarded; 3. Calculate the arithmetic mean of the N data in the current queue as the filter output.

[0019] S2-2. Data cleaning is performed on the collected data. Data cleaning can review and verify the test data set to delete duplicate information, correct existing errors, and provide data consistency. Data cleaning of the battery cycle test data set can further ensure the accuracy of the test data; It also includes adaptive correction of the battery data group. The adaptive correction process is the process of making the data continuously approach the target. By adaptively correcting the test data group, the accuracy of the test data can be further ensured, thereby improving the accuracy of the test results. The adaptive algorithm used in the adaptive correction usually adopts a gradient-based algorithm.

[0020] S2-3. Calculate the cycle capacity (i.e., the SOH of the cycle) by integrating ampere-hours. Ampere-hour integration formula: Capacity = ∫I(t) dt, where I(t) is the instantaneous current, t is the total discharge time, and the integration interval is the entire discharge process.

[0021] Note that the current is negative during charging and positive during discharging (or take the absolute value according to the definition to ensure that the capacity is positive). For each cycle, calculate the ampere-hour integral of the discharge phase as the capacity of that cycle.

[0022] S2-4, splitting the continuous test data into data segments of a single cycle according to the cycle; Each cycle includes: step 1: constant current charge (CC) to the cut-off voltage, then constant voltage charge (CV) to the cut-off current; step 2: rest; step 3: constant current discharge (CC) to the cut-off voltage; step 4: rest.

[0023] The cycle can be split by finding the moment when the discharge ends, specifically: Discharge phase: The current is positive (or negative, depending on the definition, but generally the discharge current has the opposite sign to the charging current). The end point of the discharge phase (the current becomes 0 or enters a state of rest) is defined as the end point of the cycle. The discharge phase is used as the landmark phase of each cycle, because the capacity is calculated by discharge, from the beginning of the first discharge phase to the beginning of the second discharge phase (i.e., including charging, resting, discharging, and resting); Each discharge process is marked as a cycle, and the capacity of the discharge is recorded. At the same time, other data within the cycle (charging, resting, etc.) are used as the data of the cycle.

[0024] S3, feature extraction; The following features are extracted for each cycle data segment: 1. Charging stage: constant current charging time, voltage curve of constant current charging (including mean, variance, slope), constant voltage charging time, current curve characteristics of constant voltage charging and total charging time.

[0025] 2. Discharge stage: discharge time, discharge voltage curve (such as average value, minimum value, downward slope, etc.) and temperature changes during the discharge process (such as maximum temperature, temperature rise).

[0026] 3. Static stage: static voltage change (such as relaxation voltage drop).

[0027] 4. Whole cycle: internal resistance (can be calculated by pulse testing, or estimated by the voltage jump at the beginning of charge or discharge if there is no pulse) and energy efficiency (discharge energy / charge energy).

[0028] S4, sample set generation; Each sample corresponds to a cycle, including a feature vector (features extracted from the cycle) and a label (discharge capacity of the cycle (or SOH = discharge capacity / rated capacity)). Multiple consecutive cycles form a sample set.

[0029] At the same time, in order to predict SOH, historical data is usually required. Therefore, the label of each sample is the capacity of the current cycle. The features can include the features of the current cycle and the features of the previous cycles (forming time series features).

[0030] S5. Model training: The sample set formed in S4 is used to train the SOH prediction model.

[0031] S6, SOH quick prediction.

[0032] The above steps calculate the capacity (SOH) decay during battery aging through ampere-hour integration, split the continuous test data into single cycle data to form a "label-feature" sample set, and use this sample set to train the SOH prediction model, thereby improving the accuracy of battery health status prediction and enabling rapid assessment of the battery health status of recycled and retired batteries.

[0033] The above description is merely a preferred embodiment of the present application and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention herein is not limited to technical solutions formed by a specific combination of the aforementioned technical features, but also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for rapidly detecting the health status of retired batteries, characterized by: The steps are as follows: S1, raw data collection; S2, data processing; S2-1, use the sliding average filter algorithm to filter the noise in the data, where the window width = 30-40ms; S2-2, clean the collected data; S2-3, calculate the cycle capacity by integrating ampere-hours; S2-4, splitting the continuous test data into data segments of a single cycle according to the cycle; S3, extracting features from each cycle of data segments; S4. Sample set generation. Each sample corresponds to a cycle, including a feature vector (features extracted from the cycle) and a label (discharge capacity of the cycle). Multiple consecutive cycles form a sample set. S5, model training; S6, SOH quick prediction.

2. The method for rapidly detecting the health status of retired batteries according to claim 1, characterized in that: The data collected in S1 includes time series, current, voltage, and temperature, and there is no need to obtain the complete charge and discharge process, only the charge and discharge fragments in the 5-10% SOC range.

3. The method for rapidly detecting the health status of retired batteries according to claim 1, characterized in that: The noise filtering step in S2-1 includes: establishing a data buffer with a length of N to store the sampled data in sequence; Each time a new data is collected, the oldest data in the queue is discarded; Calculate the arithmetic mean of N data in the current queue as the filter output.

4. The method for rapidly detecting the health status of retired batteries according to claim 1, characterized in that: The S2-2 also includes adaptive correction of the battery data set. The adaptive algorithm used in the adaptive correction is usually a gradient-based algorithm.

5. The method for rapidly detecting the health status of retired batteries according to claim 1, characterized in that: The ampere-hour integration formula in S2-3 is: Capacity = ∫I(t) dt, where I(t) is the instantaneous current, t is the total discharge time, and the integration interval is the entire discharge process; Note that the current is negative during charging and positive during discharging (or take the absolute value according to the definition to ensure that the capacity is positive). For each cycle, calculate the ampere-hour integral of the discharge phase as the capacity of that cycle.

6. The method for rapidly detecting the health status of retired batteries according to claim 1, characterized in that: In S2-4, the cycle is split by finding the moment when the discharge ends, specifically: Discharge phase: The current is positive (or negative, depending on the definition, but generally the discharge current has the opposite sign to the charging current). The end point of the discharge phase (the current becomes 0 or enters a state of rest) is defined as the end point of the cycle. The discharge phase is used as the landmark phase of each cycle, because the capacity is calculated by discharge, from the beginning of the first discharge phase to the beginning of the second discharge phase (i.e., including charging, resting, discharging, and resting); Each discharge process is marked as a cycle, and the capacity of the discharge is recorded. At the same time, other data within the cycle (charging, resting, etc.) are used as the data of the cycle.

7. The method for rapidly detecting the health status of retired batteries according to claim 1, characterized in that: The features extracted in S3 include the charging stage: constant current charging time, voltage curve of constant current charging (including mean, variance, slope), constant voltage charging time, current curve characteristics of constant voltage charging and total charging time; Discharge stage: discharge time, discharge voltage curve (such as average value, minimum value, downslope, etc.) and temperature changes during discharge (such as maximum temperature, temperature rise); Resting stage: resting voltage change (such as relaxation voltage drop); Over the entire cycle: internal resistance (calculated from a pulse test, or estimated from the voltage jump at the start of charge or discharge if no pulse is present) and energy efficiency (discharge energy / charge energy).

Images