Power battery health degree real-time detection method based on pile end charging data

Abstract

The invention discloses a power battery health degree real-time detection method based on pile end charging data, and relates to the technical field of power battery state monitoring, and the method comprises the following steps: S1, collecting the voltage, current, temperature and time sequence data of a power battery output by a charging pile in a charging process in real time; s2, extracting characteristic parameters of a charging curve in a constant-current charging stage from the data in the step S1, wherein the characteristic parameters comprise a voltage inflection point, a charging acceptance rate and a polarization voltage; s3, inputting the characteristic parameters into a pre-trained SOH prediction model to obtain a preliminary SOH prediction value; and S4, correcting the preliminary SOH predicted value by adopting a temperature-multiplying power coupling correction algorithm. According to the method, data such as the voltage, the current and the temperature collected by the charging pile in the charging process are directly utilized for analysis, the operation that a battery needs to be disassembled in a traditional method is avoided, the structural integrity and the sealing performance of the battery are protected, and the safety and the convenience of detection are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power battery state monitoring, and in particular to a power battery health degree real-time detection method based on pile end charging data. BACKGROUND

[0002] In the field of new energy vehicles, the health degree (SOH) of power batteries is a key indicator for evaluating battery performance, remaining life and safety, and is of great significance to charging station operation and maintenance, second-hand vehicle evaluation and other scenarios. The existing technology has the following problems:

[0003] In the prior art, there are many deficiencies in the detection of power battery SOH. Traditional detection methods often require disassembly of the battery, are complex to operate and may affect the sealing and safety of the battery. The detection accuracy is easily affected by environmental temperature (such as low or high temperature) and charging rate, resulting in large errors, lack of real-time performance, difficulty in achieving dynamic health assessment throughout the life cycle, incomplete extraction of feature parameters, and insufficient accuracy of the mapping relationship with SOH, which cannot meet the actual application requirements. SUMMARY

[0004] The present application provides a power battery health degree real-time detection method based on pile end charging data to solve the problems raised in the background art.

[0005] To solve the above technical problems, the technical solution adopted by the present application is:

[0006] A power battery health degree real-time detection method based on pile end charging data, comprising the following steps:

[0007] S1: Real-time acquisition of voltage, current, temperature and time series data of power batteries output by charging piles during the charging process;

[0008] S2: Extracting feature parameters of the constant current charging phase charging curve from the data of step S1, the feature parameters including voltage inflection point, charging acceptance rate and polarization voltage;

[0009] S3: Inputting the feature parameters into a pre-trained SOH prediction model to obtain a preliminary SOH prediction value;

[0010] S4: Modifying the preliminary SOH prediction value using a temperature-rate coupling correction algorithm to obtain a high-precision final SOH value;

[0011] S5: Continuously updating the final SOH value based on real-time data from multiple charging processes to achieve dynamic monitoring of the health status of power batteries throughout their life cycle.

[0012] The further improvement of the technical scheme of the present application is that the SOH prediction model is a neural network model trained based on an error back propagation algorithm, and the model is trained through a sample library containing charging characteristic parameters and measured SOH values of batteries in different health states.

[0013] The further improvement of the technical scheme of the present application is that the temperature-rate coupling correction algorithm is realized through the following formula:

[0014] SOH_final=SOH_pre*(1+α*(T-T_ref)+β*(C-C_ref)

[0015] wherein SOH_final is a final SOH value after correction, SOH_pre is a preliminary SOH prediction value, T is a real-time collected temperature value, T_ref is a reference temperature, C is a real-time charging rate, C_ref is a reference charging rate, and α and β are a temperature correction coefficient and a rate correction coefficient obtained through experiment calibration.

[0016] The further improvement of the technical scheme of the present application is that the extraction of the characteristic parameters further includes at least one of a peak value and a corresponding voltage value in a charging capacity increment curve and a constant current charging stage duration.

[0017] The further improvement of the technical scheme of the present application is that a sampling interval of the real-time data collection is not greater than 1 second.

[0018] The further improvement of the technical scheme of the present application is that before the characteristic parameters are input into the SOH prediction model, the characteristic parameters are further subjected to data preprocessing, and the preprocessing includes at least one of a sliding average filtering and a median filtering to eliminate data noise.

[0019] The further improvement of the technical scheme of the present application is that the method further includes step S6: based on a historical SOH value sequence, a battery capacity attenuation model is established through an exponential smoothing or linear regression method, and a remaining service life of the power battery is predicted based on the model.

[0020] The further improvement of the technical scheme of the present application is that the SOH prediction model supports an online learning function, and the model can be incrementally updated by using the final SOH value after correction and the corresponding characteristic parameters as new samples, so as to adapt to the aging characteristics of the battery.

[0021] Due to the adoption of the above technical scheme, the present application has the following technical progress compared with the prior art:

[0022] 1. The application provides a kind of power battery health degree real-time detection method based on pile end charging data, by directly using the voltage, current, temperature and other data collected in charging process of charging pile to carry out analysis, avoid the operation of traditional method needing to disassemble battery, both protect the structural integrity and sealing of battery, also improve the security and convenience of detection.

[0023] 2. The application provides a kind of power battery health degree real-time detection method based on pile end charging data, by temperature-rate coupling correction algorithm to preliminary SOH prediction value real-time correction, effectively inhibit the influence of environmental temperature and charging rate on detection result, control SOH detection error within 3%, significantly improve the detection precision and reliability.

[0024] 3. The application provides a kind of power battery health degree real-time detection method based on pile end charging data, data sampling interval is only 1s, combined with online learning and incremental update mechanism, can realize the real-time estimation and continuous update of SOH, dynamically reflect the change trend of battery health state, and support remaining service life prediction, real-time is strong. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 It is the method flowchart of the application;

[0026] Figure 2 It is the battery capacity attenuation rate variation diagram with cycle number of the application. DETAILED DESCRIPTION

[0027] The application will be further described in detail in combination with examples:

[0028] Example 1

[0029] As Figure 1 , Figure 2 shown, the application provides a kind of power battery health degree real-time detection method based on pile end charging data, the implementation of method relies on the data acquisition system integrated in charging pile end, the system includes voltage sensor, current sensor, temperature sensor, timing unit and a controller with data processing and communication function, such as MCU or embedded system, temperature sensor is usually arranged at battery surface or near pole, in charging process, the system synchronously collects the voltage, current, temperature and corresponding time stamp of power battery at not less than 1 time / second frequency, forms time series data stream, and is uploaded to local server or edge computing device for processing by CAN bus or Ethernet etc.

[0030] Example 2

[0031] As Figure 1 , Figure 2As shown, based on Embodiment 1, this invention provides a technical solution: the voltage curve of the power battery during the constant current charging stage contains key features reflecting its internal health status. The data processing unit preprocesses the uploaded raw data, first using a moving average filter to smooth high-frequency noise, with a window size of 5-10 sampling points. Then, it identifies the stage of the charging process and precisely locks the constant current charging stage. Within the constant current charging stage data segment, the system extracts the following feature parameters:

[0032] Voltage inflection point: It is determined by calculating the first derivative dV / dt of the voltage curve and finding the local extremum of the rate of change of the derivative, i.e., the second derivative.

[0033] Charge acceptance rate: defined as the rate of voltage rise per unit time during the constant current phase, which can be obtained by linearly fitting the voltage-time curve of this phase and determining its slope.

[0034] Polarization voltage: Record voltage values ​​V1 and V2 at the start and end times of constant current charging, t1 and t2. The polarization voltage can be approximately calculated as V_pol = V2 - V1 - I * R0, where I is the current and R0 is the ohmic internal resistance of the battery. It can be obtained in advance through hybrid pulse power characteristic experiments and stored in the system.

[0035] Capacity increment curve characteristics: Differentiate the charging capacity Q with respect to the voltage V to obtain the dQ / dV curve, extract the peak height of the curve and its corresponding voltage value, and obtain the charging capacity Q by integrating with the current.

[0036] Example 3

[0037] like Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: the SOH prediction model preferably adopts a three-layer BP neural network, namely: input layer: the number of nodes is the same as the number of selected feature parameters. For example, if four features are selected, namely voltage inflection point, charge acceptance rate, polarization voltage and ICA peak voltage, then the number of nodes in the input layer is 4; hidden layer: the number of nodes can be adjusted according to the actual situation, usually set to 8 to 12; output layer: 1 node, outputting the SOH prediction value, ranging from 0% to 100%.

[0038] The training of the model is completed in an offline state. Charging data of power batteries of different brands, models and health states at various temperatures, such as 0℃, 10℃, 25℃, 40℃ and various charging rates, such as 0.2C, 0.5C and 1C, are obtained through experiments, and the feature parameters are extracted according to the above method. The real SOH values are accurately measured through the capacity calibration experiment to form a training sample library. The network is trained by using a Levenberg-Marquardt algorithm and other training algorithms until the mean square error reaches a predetermined target. The trained model parameters are deployed in the online detection system.

[0039] Embodiment 4

[0040] As shown in Figure 1 , Figure 2 On the basis of Embodiment 1, the application provides a technical solution: since the offline trained model may have a decreased accuracy when encountering a temperature or rate condition that has not been experienced, a correction algorithm is designed. The correction is based on the following empirical formula:

[0041] SOH_final = SOH_pre * [1 + a * (T_actual - T_ref) + β * (C_actual - C_ref)]

[0042] Wherein: SOH_final: the final SOH value after correction; SOH_pre: the preliminary prediction value of the neural network model; T_actual: the average temperature of the current charging process; T_ref: the reference temperature, usually set to 25℃; C_actual: the rate of the current charging; C_ref: the reference charging rate, usually set to 0.5C; a, β: correction coefficients, determined by experiments under different (T, C) combinations, comparison of the model predicted SOH and the real SOH, and fitting by using the least square method;

[0043] The system can perform a complete SOH estimation process once in each charging process. The SOH_final obtained each time and the corresponding feature parameters and environmental conditions can be stored as a new data point in the historical database. The system can regularly analyze the historical SOH data by using the exponential smoothing method or the linear regression method, fit the capacity decay curve, and predict the remaining service life of the battery. In addition, the system also supports online fine-tuning of the model, that is, new high-quality data, such as data obtained by charging under standard conditions, are added to the training set for incremental learning of the neural network model, so that the model can better adapt to the changes in the characteristics during the battery aging process.

[0044] The working principle of the power battery health degree real-time detection method based on pile end charging data will be described in detail below.

[0045] As shown in Figure 1 , Figure 2As shown, by collecting voltage, current, temperature and time series data in real time at the end of the charging pile during the charging process of the power battery, and then extracting the key characteristic parameters in the constant current charging stage charging curve which can represent the battery health state, such as voltage inflection point, charging acceptance rate, polarization voltage and capacity increment curve characteristics; input these characteristic parameters into the BP neural network model which is trained by a large number of offline experimental data in advance, and obtain the preliminary health degree prediction value; then, using the special temperature-rate coupling correction algorithm, the preliminary prediction value is compensated and corrected for environmental factors to obtain the final SOH value with high precision; the system can execute this process in each charging process, thereby realizing real-time, online and high-precision evaluation of the battery health state, and dynamically monitoring the performance degradation trend and predicting the remaining service life of the battery throughout its life cycle based on the continuously updated historical data.

[0046] The above has made a detailed description of the present application in general, but some modifications or improvements can be made on the basis of the present application, which is obvious to those skilled in the art. Therefore, the modifications or improvements without departing from the spirit of the present application are within the scope of the present application.

Claims

1. A method for real-time detection of the health of a power battery based on charging pile data, characterized in that: Includes the following steps: S1: Real-time acquisition of voltage, current, temperature and time series data of the power battery output by the charging pile during the charging process; S2: Extract the characteristic parameters of the charging curve during the constant current charging stage from the data in step S1. The characteristic parameters include the voltage inflection point, charging acceptance rate, and polarization voltage. S3: Input the feature parameters into the pre-trained SOH prediction model to obtain the preliminary SOH prediction value; S4: The initial SOH prediction value is corrected using a temperature-rate coupling correction algorithm to obtain a high-precision final SOH value; S5: The final SOH value is continuously updated based on real-time data from multiple charging processes to achieve dynamic monitoring of the health status of the power battery throughout its entire life cycle.

2. The method for real-time detection of power battery health based on charging pile data according to claim 1, characterized in that: The SOH prediction model is a neural network model trained based on the error backpropagation algorithm. This model is trained using a sample library containing charging characteristic parameters of batteries in different health states and measured SOH values.

3. The method for real-time detection of power battery health based on charging pile data according to claim 1, characterized in that: The temperature-rate coupling correction algorithm is implemented using the following formula: SOH_final=SOH_pre*(1+α*(T-T_ref)+β*(C-C_ref)) Wherein, SOH_final is the corrected final SOH value, SOH_pre is the preliminary SOH prediction value, T is the real-time acquired temperature value, T_ref is the reference temperature, C is the real-time charging rate, C_ref is the reference charging rate, and α and β are the temperature correction coefficient and rate correction coefficient obtained through experimental calibration.

4. The method for real-time detection of power battery health based on charging pile data according to claim 1, characterized in that: The extraction of the feature parameters also includes at least one of the peak value in the charging capacity increment curve and its corresponding voltage value, and the duration of the constant current charging stage.

5. The method for real-time detection of power battery health based on charging pile data according to claim 1, characterized in that: The sampling interval for real-time data acquisition is no more than 1 second.

6. The method for real-time detection of power battery health based on charging pile data according to claim 1, characterized in that: Before inputting the feature parameters into the SOH prediction model, the model further includes data preprocessing of the feature parameters, which includes at least one of moving average filtering and median filtering to eliminate data noise.

7. The method for real-time detection of power battery health based on charging pile data according to claim 1, characterized in that: It also includes step S6: Based on the historical SOH value sequence, establish a battery capacity decay model through exponential smoothing or linear regression methods, and predict the remaining service life of the power battery accordingly.

8. The method for real-time detection of power battery health based on charging pile data according to claim 2, characterized in that: The SOH prediction model supports online learning, which can use the corrected final SOH value and its corresponding feature parameters as new samples to incrementally update the model in order to adapt to the aging characteristics of the battery.

Images