A lithium ion energy storage health condition diagnosis method based on linear regression method

Abstract

A lithium ion energy storage health condition diagnosis method based on a linear regression method belongs to the technical field of lithium ion batteries and comprises the following steps: obtaining state monitoring data of a battery pack at each stage, pre-processing and normalizing the state monitoring data, and obtaining a time series data set with stage labels. The time series data set is divided according to the stage labels to extract a standardized feature vector set corresponding to each stage. Based on the standardized feature vector set corresponding to each stage, a teacher model and student models corresponding to each stage are constructed in combination with the historical health state labels of the battery pack. The current state monitoring data of the battery pack is input into the teacher model to determine the stage corresponding to the current state monitoring data, and the current state monitoring data is input into the student model of the corresponding stage to output a health state diagnosis result. Through health diagnosis of the lithium ion battery, the initial signs of material damage can be found in advance.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of lithium ion batteries, and more particularly relates to a lithium ion energy storage health condition diagnosis method based on a linear regression method. BACKGROUND

[0002] At a polar scientific research station or a high mountain communication base station, the core component of a lithium ion energy storage system, a lithium battery, needs to be repeatedly used in freeze-thaw in the range of-40 DEG C to 20 DEG C. At this time, the temperature characteristics and voltage characteristics obtained in the prior art present a linear relationship that is completely different in the freezing stage and the liquid stage. For example: (1) in the first freeze-thaw cycle, the battery temperature lags behind the internal chemical reaction rate by more than 5 minutes; (2) after continuous freeze-thaw for many times, the internal resistance characteristics appear "corner" type mutation.

[0003] When the temperature drops to-40 DEG C, the organic solvent in the electrolyte will crystallize, stress will be generated between the active material and the separator, and the volume expansion or contraction of the crystallized electrolyte will cause micro-cracks on the electrode sheet. If not discovered in time, long-term cycling will cause electrode sheet peeling, separator penetration failure, and more easily cause capacity drop or internal short circuit. When the temperature rises to 0 DEG C, local melting and thermal expansion and contraction are superimposed, which will repeatedly act on the material interface, further aggravating the risk of interface fracture. Therefore, it is necessary to diagnose the health of the lithium ion battery energy storage to discover the internal resistance mutation or temperature lag anomaly in advance, and then determine whether there are early signs of material damage. SUMMARY

[0004] To solve the problems in the prior art, the purpose of the present application is to solve the above-mentioned defects, and further to provide a lithium ion energy storage health condition diagnosis method based on a linear regression method.

[0005] The application adopts the following technical solutions.

[0006] The first aspect of the application discloses a lithium ion energy storage health condition diagnosis method based on a linear regression method, applied to a lithium ion battery, and the method comprises:

[0007] Obtaining state monitoring data of a battery pack at each stage, and pre-processing and normalizing the state monitoring data to obtain a time series data set with a stage label;

[0008] According to the stage label, the time series data set is divided to extract a standardized feature vector set corresponding to each stage;

[0009] Based on the standardized feature vector set corresponding to each stage, a teacher model and a student model corresponding to each stage are constructed in combination with the historical health state label of the battery pack;

[0010] input the current state monitoring data of the battery pack to the teacher model to determine the corresponding stage of the current state monitoring data, and input the current state monitoring data to the student model of the corresponding stage to output the health state diagnosis result;

[0011] The state monitoring data includes temperature data, voltage data, current data of the battery pack, and environment temperature data of the environment where the battery pack is located, and the stages are divided into a freezing period, a thawing period, and a normal temperature period.

[0012] Further, the state monitoring data of the battery pack at each stage is obtained, and the state monitoring data is preprocessed and normalized to obtain a time series data set with a stage label, including:

[0013] Temperature sensors, voltage sensors, and current sensors are arranged in the battery pack, and the temperature data, voltage data, and current data of the battery pack in the freezing period, thawing period, and normal temperature period are collected in real time by the temperature sensors, voltage sensors, and current sensors;

[0014] The temperature data, voltage data, and current data are integrated into a time series data set corresponding to each stage in chronological order.

[0015] Further, the state monitoring data of the battery pack at each stage is obtained, and the state monitoring data is preprocessed and normalized to obtain a time series data set with a stage label, including:

[0016] An environment temperature sensor is arranged in the environment where the battery pack is located to collect environment temperature data, and an initialization stage label sequence is constructed based on the environment temperature data; the initialization stage label sequence is the range of environment temperature data of each stage.

[0017] The state monitoring data is summarized, and the state monitoring data is cleaned and normalized to remove abnormal data to obtain a time series data set with a stage label aligned with a timestamp.

[0018] Further, the time series data set is divided according to the stage label to extract a standardized feature vector set corresponding to each stage, including:

[0019] The time series data set is divided into a freezing period data set, a thawing period data set, and a normal temperature period data set according to the stage label;

[0020] Temperature gradient lag, internal resistance corner mutation features, and voltage and current regular fluctuation features are extracted from the freezing period data set, the thawing period data set, and the normal temperature period data set, respectively;

[0021] The temperature gradient hysteresis amount, the internal resistance corner mutation feature, and the voltage and current conventional fluctuation feature are normalized and outlier smoothed, and a teacher model is used for knowledge distillation to generate a freezing period feature matrix, a thawing period feature matrix, and a normal temperature period feature matrix;

[0022] The freezing period feature matrix is composed of a temperature gradient hysteresis amount and a freezing period internal resistance mutation factor, the thawing period feature matrix is composed of a current change rate and a thawing period internal resistance mutation factor, and the normal temperature period feature matrix is composed of a voltage change rate, a current change rate, and a battery temperature.

[0023] Further, based on the standardized feature vector set corresponding to each stage, the historical health state label of the battery pack is combined to construct a teacher model and a student model corresponding to each stage, including:

[0024] Temperature data, voltage data, and current data of the battery pack are extracted from all stage time series data sets to construct a common class feature vector set, and a machine learning framework is trained based on the common class feature vector set to obtain the teacher model;

[0025] Based on the freezing period feature matrix, the thawing period feature matrix, and the normal temperature period feature matrix and the historical health state label of the battery pack corresponding to each stage, a student model corresponding to each stage is constructed by imitating the feature representation of the intermediate layer of the teacher model.

[0026] Further, the state monitoring data further includes an infrared thermal image time series frame set, and the method further includes:

[0027] A medium wave infrared camera is arranged outside the battery pack, and an infrared image is captured by the medium wave infrared camera to construct the infrared thermal image time series frame set;

[0028] The infrared thermal image time series frame set is subjected to data cleaning and normalization processing, and missing frames in the infrared thermal image time series frame set are filled by a front and back frame pixel difference linear interpolation method.

[0029] Further, the method further includes:

[0030] Each infrared image in the infrared thermal image time series frame set is subjected to denoising and non-uniform correction, and the infrared thermal image time series frame set is subjected to binary segmentation, and a hotspot area is extracted from the infrared image according to an infrared temperature threshold;

[0031] Based on the hotspot area and the health state diagnosis result output by each student model, a linear regression model for evaluating the teacher model is constructed;

[0032] The linear regression model is called to output a predicted hot pixel ratio according to the current hot region and the health status diagnosis result, and a corresponding stage is selected according to a comparison result between the predicted hot pixel ratio and the current hot region.

[0033] The second aspect of the present application discloses a lithium ion energy storage health status diagnosis device based on a linear regression method, which is used to realize the lithium ion energy storage health status diagnosis method based on the linear regression method in any one of the first aspect.

[0034] The data processing module is used to obtain state monitoring data of the battery pack at each stage, and pre-process and normalize the state monitoring data to obtain a time series data set with a stage label.

[0035] The feature extraction module is used to divide the time series data set according to the stage label to extract a standardized feature vector set corresponding to each stage.

[0036] The model training module is used to construct a teacher model and a student model corresponding to each stage based on the standardized feature vector set corresponding to each stage and in combination with a historical health status label of the battery pack.

[0037] The health status diagnosis module is used to input current state monitoring data of the battery pack into the teacher model to determine a corresponding stage of the current state monitoring data, and input the current state monitoring data into a student model corresponding to the stage to output a health status diagnosis result.

[0038] The state monitoring data includes temperature data, voltage data, current data of the battery pack, and environmental temperature data of an environment in which the battery pack is located, and the stages are divided into a freezing period, a thawing period, and a normal temperature period.

[0039] The third aspect of the present application discloses a terminal including a processor and a storage medium, and characterized in that:

[0040] The storage medium is used to store instructions.

[0041] The processor is used to operate according to the instructions to perform the steps of the method of the first aspect.

[0042] The fourth aspect of the present application discloses a computer readable storage medium having a computer program stored thereon, and characterized in that the program is executed by a processor to realize the steps of the method of the first aspect.

[0043] The present application has the following advantages compared with the prior art:

[0044] (1) The present application collects the state monitoring data of lithium ion batteries in the freezing period, thawing period and normal temperature period, and constructs a data set with time sequence and stage label according to the stage of the state monitoring data, which provides a data basis for the construction of teacher model and student model.

[0045] (2) The present application extracts feature vectors based on the time sequence data set of each stage, constructs the corresponding feature matrix of each stage through the knowledge distillation of the teacher model, trains the machine learning framework, and obtains the corresponding student model of each stage for the health status evaluation of lithium ion energy storage in each stage. Due to the difference between the parameters of each stage, the teacher model based on the machine learning framework is used to determine the stage of the state monitoring data to be evaluated, and then the student model corresponding to the data rule of each stage is used to evaluate the health status of lithium ion energy storage in the corresponding stage. The evaluation result has high accuracy, can find the sudden change of internal resistance or temperature lag anomaly in advance, and then judge whether the lithium ion battery has the initial sign of material damage. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a flowchart of a lithium ion energy storage health status diagnosis method based on a linear regression method provided by the present application;

[0047] Figure 2 is a flowchart of a phased feature extraction and preprocessing provided by the present application;

[0048] Figure 3 is a flowchart of a phased data evaluation provided by the present application;

[0049] Figure 4 is a structural diagram of a lithium ion energy storage health status diagnosis device based on a linear regression method provided by the present application;

[0050] Figure 5 is a schematic diagram of environment perception, stage dynamic division and data flow provided by the present application;

[0051] Figure 6 is a schematic diagram of a teacher-student model cooperative diagnosis process provided by the present application. DETAILED DESCRIPTION

[0052] The present application will be further described below in conjunction with the drawings. The following examples are only used to more clearly illustrate the technical solutions of the present application, and cannot be used to limit the protection scope of the present application.

[0053] As shown in Figure 1 , in one embodiment, a lithium ion energy storage health status diagnosis method based on a linear regression method includes the following steps:

[0054] Step S110, obtain the state monitoring data of the battery pack at each stage, and preprocess and normalize the state monitoring data to obtain a time series data set with stage labels.

[0055] The state monitoring data includes temperature data, voltage data, current data of the battery pack, and environmental temperature data of the environment in which the battery pack is located, and each stage is divided into a freezing period, a thawing period, and a normal temperature period.

[0056] In some embodiments, the present application provides a lithium-ion energy storage health condition diagnosis method based on a linear regression method, and step S110 specifically includes the following steps:

[0057] Step S111, temperature sensors, voltage sensors, and current sensors are arranged in the battery pack, and temperature data, voltage data, and current data of the battery pack in the freezing period, the thawing period, and the normal temperature period are collected in real time through the temperature sensors, the voltage sensors, and the current sensors.

[0058] Step S112, the temperature data, the voltage data, and the current data are integrated into a time series data set corresponding to each stage in chronological order.

[0059] In some embodiments, the present application provides a lithium-ion energy storage health condition diagnosis method based on a linear regression method, and step S110 specifically further includes the following steps:

[0060] Step S113, an environmental temperature sensor is arranged in the environment in which the battery pack is located to collect environmental temperature data, and an initialization stage label sequence is constructed based on the environmental temperature data; the initialization stage label sequence is the range of environmental temperature data of each stage, as shown in Figure 5 .

[0061] Step S114, the state monitoring data is summarized, and the state monitoring data is cleaned and normalized to remove abnormal data, to obtain a time series data set with stage labels aligned with timestamps.

[0062] In a specific embodiment, the present application provides a lithium-ion energy storage health condition diagnosis method based on a linear regression method, including steps 1-3:

[0063] Step 1, lithium-ion battery state monitoring data collection at each stage.

[0064] Temperature sensors, voltage sensors, and current sensors are arranged in the battery pack to be connected with the external temperature field partition acquisition unit, and the state monitoring data of the lithium-ion battery at each stage in the freezing period, the thawing period, and the normal temperature period, i.e., the temperature data, the voltage data, and the current data, are collected in real time.

[0065] The method includes the following steps:

[0066] Step 1.1, set up an ambient temperature sensor to monitor the ambient temperature , and initialize the phase label sequence based on the detected ambient temperature , the expression of which is:

[0067]

[0068] wherein, are different temperature thresholds, which can be -10°C and 0°C, respectively.

[0069] It should be noted that the state data classified only by the ambient temperature is not accurate, so the above classification is only a preliminary classification. In fact, the selection of the temperature threshold is not only affected by the system parameters of the lithium battery itself (such as: surface thermal resistance, internal thermal conductivity, phase change temperature), but also related to the environmental parameters (such as: boiling point) of the region (such as: high mountain area).

[0070] Step 1.2, aggregate all input data sets and perform data cleaning and normalization on all input data sets.

[0071] Specifically, the input data sets include: the battery temperature returned by the high-precision temperature probe in the battery pack in real time , the voltage value collected by the battery module voltage sensor , the current value collected by the battery pack current sensor , and the infrared thermal image time series frame set obtained by shooting with a medium-wave infrared camera installed on the opposite side of the battery pack . Among them, the number of in the infrared thermal image time series frame set is multiple, and each corresponds to a different pixel point .

[0072] It should be noted that normalization is used to ensure timestamp alignment, for example, the time corresponding to the infrared thermal image time series frame set is consistent with the of the voltage value and other state monitoring data. Data cleaning needs to delete abnormal data that is obviously out of the physical range, for example, the current value may have an obvious current upper limit and current lower limit, and the current value greater than the current upper limit and less than the current lower limit needs to be removed. In addition, if the missing frames in the infrared thermal image time series frame set are less than 5% of the total number of frames, linear interpolation is used to generate the pixel difference between the two frames. When the ambient temperature is floating at the temperature threshold, it should be avoided to jump between different states.

[0073] Step S120, dividing the time series data set according to the phase label to extract the standardized feature vector set corresponding to each phase.

[0074] In some embodiments, the present application provides a lithium-ion energy storage health condition diagnosis method based on a linear regression method, and step S120 specifically comprises the following steps:

[0075] Step S121, dividing the time series dataset into a freezing period dataset, a thawing period dataset, and a normal temperature period dataset according to the phase label.

[0076] Step S122, extracting temperature gradient hysteresis, internal resistance corner mutation characteristics, and voltage and current regular fluctuation characteristics from the freezing period dataset, the thawing period dataset, and the normal temperature period dataset, respectively.

[0077] Step S123, normalizing and performing outlier smoothing processing on the temperature gradient hysteresis, the internal resistance corner mutation characteristics, and the voltage and current regular fluctuation characteristics, and generating a freezing period feature matrix, a thawing period feature matrix, and a normal temperature period feature matrix through knowledge distillation of a teacher model.

[0078] The freezing period feature matrix is composed of temperature gradient hysteresis and freezing period internal resistance mutation factors, the thawing period feature matrix is composed of current change rate and thawing period internal resistance mutation factors, and the normal temperature period feature matrix is composed of voltage change rate, current change rate, and battery temperature.

[0079] In specific embodiments, the present application provides a lithium-ion energy storage health condition diagnosis method based on a linear regression method, and step 2, phased feature extraction and preprocessing. For data labeled as the freezing period, temperature gradient hysteresis is extracted; for thawing period data, internal resistance corner mutation characteristics are extracted; for normal temperature period data, voltage and current regular fluctuation characteristics are extracted; and each time series is normalized and smoothed for outliers.

[0080] As shown in FIG. 2, step 2 specifically comprises the following steps: Figure 2

[0081] Step 2.1, dividing the input dataset into a freezing period dataset, a thawing period dataset, and a normal temperature period dataset based on the phase label sequence.

[0082] Specifically, the freezing period dataset , the thawing period dataset , and the normal temperature period dataset are respectively represented as:

[0083]

[0084]

[0085]

[0086] wherein, ​The states 1, 2, and 3 of the battery correspond to three different stages, i.e., the freezing period, the thawing period, and the normal temperature period. is a timestamp is the number of samples.

[0087] Step 2.2, based on the freezing period dataset, the thawing period dataset, and the normal temperature period dataset, generate the corresponding freezing period feature matrix, the thawing period feature matrix, and the normal temperature period feature matrix.

[0088] Specifically, the freezing period feature matrix , the thawing period feature vector , specifically includes the following steps:

[0089] Step 2.2.1, calculate the sliding window average temperature , the expression is:

[0090]

[0091] In the formula, represents a sliding window, for example, the value can be 5, represents the battery temperature at time t, is the sliding window average temperature at time t.

[0092] Step 2.2.2, calculate the temperature gradient lag , the expression is:

[0093]

[0094] In the formula, is the temperature gradient lag at time t.

[0095] Step 2.2.3, calculate the instantaneous resistance mutation factor , the expression is:

[0096]

[0097] In the formula, represents the resistance mutation factor at time t, has no special meaning and should be less than all non-zero , which prevents the denominator from being 0.

[0098] Step 2.2.4, calculate the freezing period resistance mutation factor , the expression is:

[0099]

[0100] ​​​​In the formula, represents the voltage value at time . represents the instantaneous internal resistance mutation factor at time .

[0101] For the normal temperature period characteristic matrix , the following steps are specifically included:

[0102] Step 2.2.5, calculating the voltage change rate , as shown in the following formula:

[0103]

[0104] In the formula, represents the voltage value at time .

[0105] Step 2.2.6, calculating the current change rate , as shown in the following formula:

[0106]

[0107] In the formula, represents the current value at time .

[0108] Similarly, for the thawing period characteristic matrix , the following steps are specifically included:

[0109] Step 2.2.7, calculating the current change rate , the expression is:

[0110]

[0111] Step 2.2.8, calculating the internal resistance mutation factor , the expression is:

[0112]

[0113] It can be understood that the freezing period characteristic matrix, the thawing period characteristic matrix, and the normal temperature period characteristic matrix are essentially the knowledge distillation of .

[0114] In this embodiment, the freezing period characteristic matrix is composed of two types of parameters, the temperature gradient hysteresis and the freezing period internal resistance mutation factor, the thawing period characteristic matrix is composed of two types of parameters, the current change rate and the thawing period internal resistance mutation factor, and the normal temperature period characteristic matrix is composed of three types of parameters, the voltage change rate, the current change rate, and the battery temperature.

[0115] It should be noted that the composition parameters in the feature matrix of each stage are different because the purposes of health condition diagnosis are different in different stages. In the freezing stage, the problem to be considered is that the partial crystallization of the electrolyte causes a sudden increase in internal resistance. The internal resistance mutation factor directly quantifies the sharp rise in electrochemical impedance caused by crystallization, and the temperature gradient lag quantity is a thermal precursor of the process. The combination of the two can capture the signs several minutes (usually 10 minutes, depending on the size of the battery pack) before the real crystallization causes a serious impedance imbalance. Similarly, the problem to be considered in the melting stage is that the uneven melting of the crystal causes local material peeling and dendrite corrosion. The internal resistance mutation factor can capture the fluctuation of the impedance that first decreases and then oscillates, and the current change rate can capture the current peak at the moment of dendrite contact. The problem to be considered in the normal temperature stage is cycle capacity degradation. Since the voltage change rate can quickly capture the abnormal sinking of the discharge platform, the current change rate reflects the flow increase at the moment of short circuit, and the battery temperature verifies whether the heat release occurs. The above three complement each other and can timely issue an early fault signal before the capacity significantly decreases.

[0116] In the present embodiment, the freezing stage feature matrix can be a matrix with a length of 2 and a width of The machine learning framework constructed in this way is a student model. However, the difference between the student model and the teacher model is that the teacher model is obtained by training the machine learning framework with the temperature data , voltage data and current data of the battery pack in the three stages as input parameters; while the student model is generated by the knowledge distillation of the state monitoring data through the parameters specific to each stage, and the freezing stage feature matrix, the melting stage feature matrix and the normal temperature stage feature matrix are constructed by imitating the feature representation of the intermediate layer of the teacher model.

[0117] In step S130, based on the standardized feature vector set corresponding to each stage, the teacher model and the student model corresponding to each stage are constructed in combination with the historical health state label of the battery pack.

[0118] In some embodiments, the present application provides a lithium ion energy storage health condition diagnosis method based on a linear regression method, and step S130 specifically includes the following steps:

[0119] In step S131, the temperature data, voltage data and current data of the battery pack are extracted from the time series data set of all stages to construct a common class feature vector set, and a machine learning framework is trained based on the common class feature vector set to obtain a teacher model.

[0120] Step S132, respectively based on the freezing period feature matrix, thawing period feature matrix and normal temperature period feature matrix and the historical health state label of the corresponding stage battery pack, modeling is carried out through the imitation of the feature representation of the intermediate layer of the teacher model to construct the corresponding student model of each stage.

[0121] Step S140, input the current state monitoring data of the battery pack into the teacher model to determine the corresponding stage of the current state monitoring data, and input the current state monitoring data into the student model of the corresponding stage to output the health state diagnosis result.

[0122] In some embodiments, the present application provides a lithium ion energy storage health condition diagnosis method based on linear regression method, and the state monitoring data further includes an infrared thermal image time series frame set, and the method further includes the following steps:

[0123] Step S210, a medium wave infrared camera is arranged outside the battery pack, and an infrared image is captured by the medium wave infrared camera to construct an infrared thermal image time series frame set.

[0124] Step S220, data cleaning and normalization processing is performed on the infrared thermal image time series frame set, and the missing frames in the infrared thermal image time series frame set are filled by the front and rear frame pixel difference linear interpolation method.

[0125] In some embodiments, the present application provides a lithium ion energy storage health condition diagnosis method based on linear regression method, and the method further includes the following steps:

[0126] Step S310, denoising and non-uniform correction are performed on each infrared image in the infrared thermal image time series frame set, and binary segmentation is performed on the infrared thermal image time series frame set, and the hotspot area is extracted from the infrared image according to the infrared temperature threshold.

[0127] Step S320, based on the hotspot area and the health state diagnosis result output by each student model, a linear regression model for evaluating the teacher model is constructed.

[0128] Step S330, calling the linear regression model according to the current hotspot area and the health state diagnosis result, outputting the predicted hotspot pixel ratio, and selecting the corresponding stage according to the comparison result between the predicted hotspot pixel ratio and the current hotspot area.

[0129] In a specific embodiment, the present application provides a lithium ion energy storage health condition diagnosis method based on linear regression method, and step 3 is the construction and training of the teacher model and the student model. First, based on the data common to the freezing period data set, the thawing period data set and the normal temperature period data set, that is, the battery temperature collected by the high-precision temperature probe in the battery pack and the current value of the battery pack current sensor The machine learning framework (such as scikit-learn, TensorFlow) is trained to obtain a teacher model. Due to the differences between the parameters of each stage, the teacher model is used to perform knowledge distillation on the condition monitoring data to generate stage features corresponding to the three stages. Based on the three stage features, student models corresponding to each stage are constructed by imitating the feature representation of the intermediate layers of the teacher model, i.e., for the freezing period (parameters include temperature gradient lag, freezing period internal resistance mutation factor), thawing period (parameters include current change rate, thawing period internal resistance mutation factor), and normal temperature period (parameters include voltage change rate, current change rate, battery temperature), the student models corresponding to each stage are obtained. Step 3 addresses the problem of significant differences in freeze-thaw characteristics. The teacher-student model modeling method can process all stage parameters based on the teacher model, and knowledge distillation can extract stage features between stages. This not only ensures that each student model has good fitting ability for the corresponding stage parameters, but also avoids distortion of the overall model.

[0130] In this embodiment, during the teacher-student model modeling process, the training required settings and optimized hyperparameters include learning rate, batch size, and training rounds. Using hyperparameters as model inputs can effectively converge student model parameters and ensure complete model training. The loss function of the teacher model is composed of a hard loss function and a soft loss function. The hard loss function is used to represent the loss between the teacher model's predicted output and the correct result of the output, and the soft loss function is the loss between the teacher model's predicted output and the student model's predicted output.

[0131] It should be noted that the essence of constructing a teacher-student model is to first train a machine learning framework based on all parameters of all stages to construct a teacher model, then perform knowledge distillation on all parameters according to the stage label using the teacher model to obtain stage features corresponding to each stage, and finally construct student models for each stage by imitating the feature representation of the intermediate layers of the teacher model. After the teacher-student model is established, the parameter differences can be analyzed by the teacher model based on the current condition monitoring data to determine the corresponding stage, and the condition monitoring data to be evaluated is divided into stages, and then the health state diagnosis is performed by the student model of the corresponding stage, as shown in Figure 6 .

[0132] In this embodiment, when the actual temperature approaches the temperature threshold, the division of data may not be accurate, for example, data belonging to the freezing period is divided into the thawing period, or data of the thawing period is divided into the normal temperature period. Based on this, as shown in Figure 3 , the following steps are included:

[0133] Step A1, denoising and non-uniform correction are performed on each infrared image, and a hot spot area is extracted by using global threshold segmentation.

[0134] Specifically, the following needs to be binarized: The proportion of hot spot pixels is counted , as shown in the following formula:

[0135]

[0136] In the formula, is the total number of pixel points, is the number of pixel points whose values are greater than the infrared temperature threshold value at time , The number of pixel points whose values are greater than the infrared temperature threshold value at time , and the infrared temperature threshold value can be randomly selected or can be the average value of the pixel points in the melting period.

[0137] Step A2, a linear regression model is constructed to evaluate whether the above teacher model (i.e., the teacher model mentioned in step 3) is accurate.

[0138] Specifically, when the constructed linear regression model takes as input, the corresponding , is expressed as:

[0139]

[0140] In the formula, satisfies , where is the predicted value output by the teacher model, The number of may be multiple, but a unique can be determined by combining the least square method, so that the value of is minimized, is a weight coefficient, and 3 refers to the 3 parameters common to all the above stages, i.e., the battery temperature real-time returned by the high-precision temperature probe in the battery pack, the voltage value collected by the battery module voltage sensor, and the current value collected by the battery pack current sensor.

[0141] In the constructed linear regression model, based on the input , a predicted hot spot pixel ratio is obtained, when it is not determined whether to select the freezing period or the melting period, if is less than , the freezing period is selected, otherwise the melting period is selected; when it is not determined whether to select the melting period or the normal temperature period, if is less than If yes, then select the thawing period, otherwise select the normal temperature period.

[0142] The lithium-ion energy storage health condition diagnosis device based on the linear regression method provided by the present application is described below, and the lithium-ion energy storage health condition diagnosis device based on the linear regression method described below can be correspondingly referred to the lithium-ion energy storage health condition diagnosis method based on the linear regression method described above.

[0143] As shown in FIG. 1, in one embodiment, a lithium-ion energy storage health condition diagnosis device based on a linear regression method includes a data processing module, a feature extraction module, a model training module, and a health condition diagnosis module. Figure 4

[0144] The data processing module is configured to obtain state monitoring data of the battery pack at each stage, and to pre-process and normalize the state monitoring data to obtain a time series data set with stage labels.

[0145] The feature extraction module is configured to divide the time series data set according to the stage labels to extract a set of standardized feature vectors corresponding to each stage.

[0146] The model training module is configured to construct a teacher model and a student model corresponding to each stage based on the set of standardized feature vectors corresponding to each stage and the historical health condition labels of the battery pack.

[0147] The health condition diagnosis module is configured to input the current state monitoring data of the battery pack into the teacher model to determine the stage corresponding to the current state monitoring data, and to input the current state monitoring data into the student model corresponding to the stage to output the health condition diagnosis result.

[0148] The state monitoring data includes temperature data, voltage data, current data of the battery pack, and environmental temperature data of the environment in which the battery pack is located, and each stage is divided into a freezing period, a thawing period, and a normal temperature period.

[0149] The present disclosure can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium having computer readable program instructions embodied therewith, which cause a processor to carry out various aspects of the present disclosure.

[0150] ​Computer readable storage media can be tangible storage media which can retain and store instructions for use by an instruction execution device. Computer readable storage media can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer readable storage media include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0151] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0152] Computer readable program instructions for carrying out operations of the present disclosure can be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0153] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0154] These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions can also be stored in a computer readable storage medium that can include random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other data storage device. When the computer readable program instructions are loaded into the computer and other programmable data processing apparatus, a series of operational steps are implemented that provide processes such that the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0155] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0156] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0157] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, rather than limiting the present application, and although the present application has been described in detail with reference to the above-mentioned embodiments, it should be understood by those skilled in the art that the specific embodiments of the present application can be modified or equivalent replacements without departing from the spirit and scope of the present application, and any modifications or equivalent replacements shall be covered within the protection scope of the claims of the present application.

Claims

1. A lithium-ion energy storage health condition diagnosis method based on a linear regression method, characterized in that, The method is applied to a lithium ion battery, and the method comprises the following steps: obtaining state monitoring data of the battery pack at each stage, and pre-processing and normalizing the state monitoring data to obtain a time series data set with a stage label; dividing the time series data set according to the stage label to extract a standardized feature vector set corresponding to each stage; based on the standardized feature vector set corresponding to each stage, combining the historical health state label of the battery pack, constructing a teacher model and a student model corresponding to each stage; inputting the current state monitoring data of the battery pack into the teacher model to determine the stage corresponding to the current state monitoring data, and inputting the current state monitoring data into the student model corresponding to the stage to output a health state diagnosis result; wherein the state monitoring data comprises temperature data, voltage data and current data of the battery pack, and environmental temperature data of the environment in which the battery pack is located, and the stages are divided into a freezing period, a thawing period and a normal temperature period. 2.The lithium-ion energy storage health condition diagnosis method based on linear regression method according to claim 1, wherein, The method comprises the following steps: setting temperature sensors, voltage sensors and current sensors in the battery pack, and collecting temperature data, voltage data and current data of the battery pack in the freezing period, the thawing period and the normal temperature period through the temperature sensors, the voltage sensors and the current sensors; integrating the temperature data, the voltage data and the current data into a time series data set corresponding to each stage in chronological order. 3.The lithium-ion energy storage health condition diagnosis method based on linear regression method according to claim 2, characterized in that, The method further comprises the following steps: setting an environmental temperature sensor in the environment in which the battery pack is located to collect environmental temperature data, and constructing an initialization stage label sequence based on the environmental temperature data; the initialization stage label sequence is the range of environmental temperature data of each stage; summarizing the state monitoring data, and performing data cleaning and normalization on the state monitoring data to remove abnormal data, to obtain a time-stamped time series data set with a stage label. 4.The lithium-ion energy storage health condition diagnosis method based on linear regression method according to claim 1, wherein, The method comprises the following steps: dividing the time series data set into a freezing period data set, a thawing period data set and a normal temperature period data set according to the stage label; extracting temperature gradient lag, internal resistance corner mutation features and voltage and current regular fluctuation features from the freezing period data set, the thawing period data set and the normal temperature period data set, respectively; normalizing and smoothing the temperature gradient lag, the internal resistance corner mutation features and the voltage and current regular fluctuation features, and generating a freezing period feature matrix, a thawing period feature matrix and a normal temperature period feature matrix through knowledge distillation of a teacher model. The freezing period feature matrix is composed of a temperature gradient hysteresis and a freezing period internal resistance mutation factor, the thawing period feature matrix is composed of a current change rate and a thawing period internal resistance mutation factor, and the normal temperature period feature matrix is composed of a voltage change rate, a current change rate and a battery temperature. 5.The lithium-ion energy storage health condition diagnosis method based on linear regression method according to claim 4, characterized in that, The teacher model and the student models corresponding to the stages are constructed based on the standardized feature vector set corresponding to each stage and in combination with the historical health state label of the battery pack, and the construction includes: Temperature data, voltage data and current data of the battery pack are extracted from the time series data set of all stages to construct a common class feature vector set, and a machine learning framework is trained based on the common class feature vector set to obtain the teacher model; The freezing period feature matrix, the thawing period feature matrix and the normal temperature period feature matrix are respectively modeled based on the historical health state label of the battery pack in the corresponding stage, and the student models corresponding to the stages are constructed by imitating the feature representation of the intermediate layer of the teacher model. 6.The lithium-ion energy storage health condition diagnosis method based on linear regression method according to claim 1, wherein, The state monitoring data further includes an infrared thermal image time series frame set, and the method further includes: A medium wave infrared camera is arranged outside the battery pack, and an infrared image is captured by the medium wave infrared camera to construct the infrared thermal image time series frame set; The infrared thermal image time series frame set is subjected to data cleaning and normalization processing, and missing frames in the infrared thermal image time series frame set are filled by a front and rear frame pixel difference linear interpolation method. 7.The lithium-ion energy storage health condition diagnosis method based on linear regression method according to claim 6, characterized in that, The method further includes: Each infrared image in the infrared thermal image time series frame set is subjected to denoising and non-uniform correction, and the infrared thermal image time series frame set is subjected to binary segmentation, and a hotspot area is extracted from the infrared image according to an infrared temperature threshold; A linear regression model for evaluating the teacher model is constructed based on the hotspot area and the health state diagnosis result output by each student model; The linear regression model is called to output a predicted hotspot pixel ratio according to the current hotspot area and the health state diagnosis result, and the corresponding stage is selected according to the comparison result between the predicted hotspot pixel ratio and the current hotspot area.

8. A lithium-ion energy storage health condition diagnosis device based on a linear regression method, characterized by, A lithium ion energy storage health condition diagnosis method based on a linear regression method for realizing any one of claims 1 to 7, the device comprising: A data processing module is configured to obtain state monitoring data of a battery pack in each stage, and to pre-process and normalize the state monitoring data to obtain a time series data set with a stage label; A feature extraction module is configured to divide the time series data set according to the stage label to extract a standardized feature vector set corresponding to each stage; A model training module is configured to construct a teacher model and student models corresponding to each stage based on the standardized feature vector set corresponding to each stage and in combination with a historical health state label of the battery pack; A health condition diagnosis module is configured to input current state monitoring data of the battery pack into the teacher model to determine a corresponding stage of the current state monitoring data, and to input the current state monitoring data into a student model of the corresponding stage to output a health state diagnosis result. The state monitoring data includes temperature data, voltage data, current data of the battery pack and environment temperature data of an environment where the battery pack is located, and each stage is divided into a freezing period, a thawing period and a normal temperature period. 9.A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is configured to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method according to any one of claims 1-7.

Images