Power battery risk diagnosis method, device and equipment and storage medium

Abstract

The invention discloses a power battery risk diagnosis method, device and equipment and a storage medium, and the method comprises the steps: obtaining the early-stage training data of a power battery of a current new energy vehicle, training an initial diagnosis model according to the early-stage training data, and obtaining a target diagnosis model; acquiring charging state data of the current new energy vehicle during each charging; and inputting the charging state data into the target diagnosis model to obtain a power battery risk diagnosis result, so that the current risk state of the battery can be diagnosed, the false alarm rate is effectively reduced, the safety monitoring accuracy of the electric vehicle is improved, and the power battery risk diagnosis speed and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of battery management technology, and in particular to a power battery risk diagnosis method, device, equipment and storage medium. Background Art

[0002] The power battery of new energy vehicles may be overcharged, over-discharged, overheated, etc. due to communication failure between the charger and the battery management system (BMS), battery aging, manufacturing defects or external environmental factors (such as high temperature); these problems will not only shorten the service life of the battery and increase maintenance costs, but may also cause safety accidents and cause damage to users and vehicles; currently, vehicles generally judge by monitoring whether the current, voltage, temperature and other signals during the operation of the vehicle exceed the preset threshold, and the threshold setting is difficult; and it usually examines the instantaneous situation of these states, and lacks the changing trend in the time dimension and the mutual influence between variables, which is prone to false alarms. Summary of the invention

[0003] The main purpose of the present invention is to provide a power battery risk diagnosis method, device, equipment and storage medium, aiming to solve the technical problems in the prior art that battery monitoring threshold setting is difficult, there is a lack of change trend in the time dimension and mutual influence between variables, and false alarms are prone to occur.

[0004] In a first aspect, the present invention provides a power battery risk diagnosis method, the power battery risk diagnosis method comprising the following steps:

[0005] Acquire the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model;

[0006] Obtaining charging status data of each charging of the current new energy vehicle;

[0007] The charging status data is input into the target diagnosis model to obtain a power battery risk diagnosis result.

[0008] Optionally, the obtaining of preliminary training data of the power battery of the current new energy vehicle, training the initial diagnosis model according to the preliminary training data, and obtaining the target diagnosis model includes:

[0009] Obtain the effective charging trip data of current new energy vehicles;

[0010] Performing data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data;

[0011] An initial diagnosis model is constructed, and the initial diagnosis model is trained according to the previous training data to obtain a target diagnosis model.

[0012] Optionally, performing data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data includes:

[0013] Cleaning the effective charging trip data, removing data that does not conform to physical characteristics and empty values, and filling the removed values ​​by linear interpolation to obtain a data cleaning result;

[0014] Obtaining a state of charge (SOC) division point, resampling the total current, load current, battery temperature, battery voltage, ambient temperature, and health state (SOH) in the data cleaning result by using an average aggregation method based on the SOC division point, and calculating a corresponding time interval between two SOC points;

[0015] Reconstructed data is generated according to the resampled data and the time interval, and the reconstructed data is normalized to obtain preliminary training data.

[0016] Optionally, the constructing of the initial diagnostic model, training the initial diagnostic model according to the previous training data to obtain a target diagnostic model, includes:

[0017] The initial diagnosis model consists of an LSTM_ENCODER layer and an LSTM_DECODER layer;

[0018] Inputting the preliminary training data into the initial diagnosis model to obtain the final output result of the model;

[0019] Calculating the loss function of the initial diagnosis model according to the final output result of the model;

[0020] The model with the smallest loss function is taken as the target diagnosis model.

[0021] Optionally, calculating the loss function of the initial diagnosis model according to the final output result of the model includes:

[0022] The loss function of the initial diagnosis model is calculated according to the final output result of the model by the following formula:

[0023]

[0024] Among them, loss is the loss function, y pn is the final output result of the model, y is the target value of the model, and n is the number of predicted values ​​output by the model.

[0025] Optionally, the acquiring charging status data of each charging of the current new energy vehicle includes:

[0026] The total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle are obtained to generate charging status data.

[0027] Optionally, inputting the charging state data into the target diagnosis model to obtain a power battery risk diagnosis result includes:

[0028] Inputting the charging state data into the target diagnosis model to obtain a comprehensive loss value of all charging trips;

[0029] Obtain the hyperparameters of the target diagnostic model, and obtain the mean and standard deviation of the comprehensive loss value, and calculate the preset judgment threshold by the following formula:

[0030] N=μ+k×σ

[0031] Among them, N is the preset judgment threshold, μ is the average value of the comprehensive loss value, σ is the standard deviation of the comprehensive loss value, and k is the hyperparameter of the target diagnosis model;

[0032] comparing the comprehensive loss value with the preset determination threshold;

[0033] When the comprehensive loss value is less than the preset determination threshold, determining that the power battery risk diagnosis result is a risk-free state;

[0034] When the comprehensive loss value is not less than the preset determination threshold, it is determined that the power battery risk diagnosis result is a risky state.

[0035] In a second aspect, in order to achieve the above-mentioned object, the present invention further proposes a power battery risk diagnosis device, the power battery risk diagnosis device comprising:

[0036] A training module is used to obtain the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model;

[0037] A data acquisition module, used to acquire charging status data of each charging of the current new energy vehicle;

[0038] The diagnosis module is used to input the charging status data into the target diagnosis model to obtain a power battery risk diagnosis result.

[0039] In the third aspect, in order to achieve the above-mentioned purpose, the present invention also proposes a power battery risk diagnosis device, which includes: a memory, a processor, and a power battery risk diagnosis program stored in the memory and executable on the processor, and the power battery risk diagnosis program is configured to implement the steps of the power battery risk diagnosis method described above.

[0040] In a fourth aspect, in order to achieve the above-mentioned purpose, the present invention further proposes a storage medium, on which a power battery risk diagnosis program is stored, and when the power battery risk diagnosis program is executed by a processor, the steps of the power battery risk diagnosis method described above are implemented.

[0041] The power battery risk diagnosis method proposed in the present invention obtains the preliminary training data of the power battery of the current new energy vehicle, trains the initial diagnosis model according to the preliminary training data, and obtains the target diagnosis model; obtains the charging status data of each charging of the current new energy vehicle; inputs the charging status data into the target diagnosis model to obtain the power battery risk diagnosis result, which can diagnose the current risk status of the battery, effectively reduce the false alarm rate, improve the safety monitoring accuracy of the electric vehicle, and improve the speed and efficiency of the power battery risk diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present invention;

[0043] Figure 2 This is a flow chart of a first embodiment of a power battery risk diagnosis method according to the present invention;

[0044] Figure 3 A schematic diagram of cloud computing in the power battery risk diagnosis method of the present invention;

[0045] Figure 4 This is a flow chart of a second embodiment of a power battery risk diagnosis method according to the present invention;

[0046] Figure 5 This is a flow chart of a third embodiment of a power battery risk diagnosis method according to the present invention;

[0047] Figure 6 This is a flow chart of a fourth embodiment of a power battery risk diagnosis method according to the present invention;

[0048] Figure 7 Schematic diagram of the model architecture in the power battery risk diagnosis method of the present invention;

[0049] Figure 8 This is a schematic diagram of a flow chart of a fifth embodiment of a power battery risk diagnosis method according to the present invention;

[0050] Fig. 9 This is a functional module diagram of the first embodiment of the power battery risk diagnosis device of the present invention.

[0051] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0053] The solution of the embodiment of the present invention is mainly: by obtaining the preliminary training data of the power battery of the current new energy vehicle, the initial diagnosis model is trained according to the preliminary training data to obtain the target diagnosis model; the charging status data of each charging of the current new energy vehicle is obtained; the charging status data is input into the target diagnosis model to obtain the power battery risk diagnosis result, which can diagnose the current risk status of the battery, effectively reduce the false alarm rate, improve the safety monitoring accuracy of electric vehicles, and improve the speed and efficiency of power battery risk diagnosis, and solve the technical problems in the prior art that the battery monitoring threshold setting is difficult, lacks the change trend in the time dimension and the mutual influence between variables, and is prone to false alarms.

[0054] Reference Figure 1 , Figure 1 The figure is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present invention.

[0055] like Figure 1 As shown, the device may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed RAM memory, or it may be a stable memory (Non-Volatile Memory), such as a disk memory. The memory 1005 may optionally be a storage device independent of the aforementioned processor 1001.

[0056] Those skilled in the art will understand that Figure 1 The device structure shown in the figure does not constitute a limitation of the device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0057] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating device, a network communication module, a user interface module, and a power battery risk diagnosis program.

[0058] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and performs the following operations:

[0059] Acquire the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model;

[0060] Obtaining charging status data of each charging of the current new energy vehicle;

[0061] The charging status data is input into the target diagnosis model to obtain a power battery risk diagnosis result.

[0062] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and also performs the following operations:

[0063] Obtain the effective charging trip data of current new energy vehicles;

[0064] Performing data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data;

[0065] An initial diagnosis model is constructed, and the initial diagnosis model is trained according to the previous training data to obtain a target diagnosis model.

[0066] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and also performs the following operations:

[0067] Cleaning the effective charging trip data, removing data that does not conform to physical characteristics and empty values, and filling the removed values ​​by linear interpolation to obtain a data cleaning result;

[0068] Obtaining a state of charge (SOC) division point, resampling the total current, load current, battery temperature, battery voltage, ambient temperature, and health state (SOH) in the data cleaning result by using an average aggregation method based on the SOC division point, and calculating a corresponding time interval between two SOC points;

[0069] Reconstructed data is generated according to the resampled data and the time interval, and the reconstructed data is normalized to obtain preliminary training data.

[0070] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and also performs the following operations:

[0071] The initial diagnosis model consists of an LSTM_ENCODER layer and an LSTM_DECODER layer;

[0072] Inputting the preliminary training data into the initial diagnosis model to obtain the final output result of the model;

[0073] Calculating the loss function of the initial diagnosis model according to the final output result of the model;

[0074] The model with the smallest loss function is taken as the target diagnosis model.

[0075] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and also performs the following operations:

[0076] The loss function of the initial diagnosis model is calculated according to the final output result of the model by the following formula:

[0077]

[0078] Among them, loss is the loss function, y pn is the final output result of the model, y is the target value of the model, and n is the number of predicted values ​​output by the model.

[0079] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and also performs the following operations:

[0080] The total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle are obtained to generate charging status data.

[0081] The device of the present invention calls the power battery risk diagnosis program stored in the memory 1005 through the processor 1001, and also performs the following operations:

[0082] Inputting the charging state data into the target diagnosis model to obtain a comprehensive loss value of all charging trips;

[0083] Obtain the hyperparameters of the target diagnostic model, and obtain the mean and standard deviation of the comprehensive loss value, and calculate the preset judgment threshold by the following formula:

[0084] N=μ+k×σ

[0085] Among them, N is the preset judgment threshold, μ is the average value of the comprehensive loss value, σ is the standard deviation of the comprehensive loss value, and k is the hyperparameter of the target diagnosis model;

[0086] comparing the comprehensive loss value with the preset determination threshold;

[0087] When the comprehensive loss value is less than the preset determination threshold, determining that the power battery risk diagnosis result is a risk-free state;

[0088] When the comprehensive loss value is not less than the preset determination threshold, it is determined that the power battery risk diagnosis result is a risky state.

[0089] Through the above scheme, this embodiment obtains the preliminary training data of the power battery of the current new energy vehicle, trains the initial diagnosis model according to the preliminary training data, and obtains the target diagnosis model; obtains the charging status data of each charging of the current new energy vehicle; inputs the charging status data into the target diagnosis model to obtain the power battery risk diagnosis result, which can diagnose the current risk status of the battery, effectively reduce the false alarm rate, improve the safety monitoring accuracy of the electric vehicle, and improve the speed and efficiency of the power battery risk diagnosis.

[0090] Based on the above hardware structure, an embodiment of a power battery risk diagnosis method of the present invention is proposed.

[0091] Reference Figure 2 , Figure 2 Schematic diagram of the flow chart of the first embodiment of the power battery risk diagnosis method of the present invention.

[0092] In a first embodiment, the power battery risk diagnosis method comprises the following steps:

[0093] Step S10: Acquire the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model.

[0094] It should be noted that the preliminary training data is the battery training data collected and integrated in the early stage. After obtaining the preliminary training data of the power battery of the current new energy vehicle, the initial diagnosis model can be trained according to the preliminary training data to obtain the corresponding diagnosis model as the target diagnosis model.

[0095] Step S20: Obtain charging status data of each charging of the current new energy vehicle.

[0096] It should be understood that each time the vehicle is charged, the charging status data of the current charge is uploaded to the cloud, and the cloud server can obtain the charging status data of each charge of the current new energy vehicle in a timely manner.

[0097] Step S30: input the charging status data into the target diagnosis model to obtain a power battery risk diagnosis result.

[0098] It can be understood that after the charging status data is input into the target diagnosis model, the power battery risk diagnosis result output by the target diagnosis model can be obtained, thereby determining whether there is a risk in the current new energy vehicle.

[0099] Furthermore, the step S30 specifically includes the following steps:

[0100] Inputting the charging state data into the target diagnosis model to obtain a comprehensive loss value of all charging trips;

[0101] Obtain the hyperparameters of the target diagnostic model, and obtain the mean and standard deviation of the comprehensive loss value, and calculate the preset judgment threshold by the following formula:

[0102] N=μ+k×σ

[0103] Among them, N is the preset judgment threshold, μ is the average value of the comprehensive loss value, σ is the standard deviation of the comprehensive loss value, and k is the hyperparameter of the target diagnosis model;

[0104] comparing the comprehensive loss value with the preset determination threshold;

[0105] When the comprehensive loss value is less than the preset determination threshold, determining that the power battery risk diagnosis result is a risk-free state;

[0106] When the comprehensive loss value is not less than the preset determination threshold, it is determined that the power battery risk diagnosis result is a risky state.

[0107] It should be noted that each charging trip is divided into X1~X5 and then brought into the models M1~M5 to obtain 5 error values ​​loss. The average value of the 5 losses is taken to record the comprehensive loss value L of this charging trip; according to the above method, the comprehensive loss value LS of all charging trips marked as risk-free in the training set is calculated; the normal judgment threshold N can be calculated according to the above formula, where μ is the average value of LS, σ is the standard deviation of L1, and k is a hyperparameter that needs to be set according to the model effect. Usually k can be taken as 1.5.

[0108] In the specific implementation, Figure 3 As shown, Figure 3 This is a schematic diagram of cloud computing in the power battery risk diagnosis method of the present invention, see Figure 3, the cloud receives the data of a certain charging trip of the vehicle; select the effective charging trip, that is, the starting SOC <= 30% and the ending SOC >= 80% are satisfied; reconstruct, standardize and segment the data of this trip; substitute the segmented X into the model to calculate the comprehensive loss value LS of this trip, and compare LS with N. When LS < N, the vehicle is determined to be in a risk-free state. When LS >= N, the vehicle is determined to be in a risky state; compared with the traditional methods of monitoring thresholds such as voltage, temperature, and current, this embodiment pays more attention to the comprehensive influencing factors among the battery state parameters in the entire time domain of each charging, can effectively reduce the false alarm rate, and improve the safety monitoring accuracy of electric vehicles.

[0109] Through the above solution, this embodiment obtains the preliminary training data of the power battery of the current new energy vehicle, trains the initial diagnosis model according to the preliminary training data to obtain the target diagnosis model; obtains the charging state data of each charging of the current new energy vehicle; inputs the charging state data into the target diagnosis model to obtain the risk diagnosis result of the power battery, can diagnose the current risk state of the battery, effectively reduce the false alarm rate, improve the safety monitoring accuracy of electric vehicles, and improve the speed and efficiency of power battery risk diagnosis.

[0110] Furthermore, Figure 4 is a schematic flowchart of the second embodiment of the power battery risk diagnosis method of the present invention. As Figure 4 shown, based on the first embodiment, the second embodiment of the power battery risk diagnosis method of the present invention is proposed. In this embodiment, the step S10 specifically includes the following steps:

[0111] Step S11, obtain the effective charging trip data of the current new energy vehicle.

[0112] It should be noted that by extracting the effective charging trip, the effective charging trip data of the current energy vehicle can be obtained.

[0113] In specific implementation, the trip with the starting state of charge (SOC) SOC <= 30% and the ending SOC >= 80% can be taken as the effective charging trip for extracting the training set data, and the interval of SOC ∈ [30%, 80%] in each effective trip is taken.

[0114] Step S12, perform data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data.

[0115] It can be understood that after performing data cleaning and data reconstruction on the effective charging trip data, preliminary training data can be obtained.

[0116] Step S13: construct an initial diagnosis model, train the initial diagnosis model according to the previous training data, and obtain a target diagnosis model.

[0117] It should be understood that after the initial diagnostic model is constructed, the initial diagnostic model can be trained according to the previous training data to obtain a target diagnostic model.

[0118] Through the above scheme, this embodiment obtains the effective charging trip data of the current new energy vehicle; performs data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data; constructs an initial diagnosis model, and trains the initial diagnosis model according to the preliminary training data to obtain a target diagnosis model. The target diagnosis model can be obtained quickly, thereby improving the speed and efficiency of power battery risk diagnosis.

[0119] Further, Figure 5 FIG. 1 is a flow chart of a third embodiment of a power battery risk diagnosis method according to the present invention. Figure 5 As shown, a third embodiment of the power battery risk diagnosis method of the present invention is proposed based on the first embodiment. In this embodiment, the step S12 specifically includes the following steps:

[0120] Step S121, performing data cleaning on the effective charging trip data, eliminating data that does not conform to physical characteristics and null values, and filling the eliminated values ​​by linear interpolation to obtain a data cleaning result.

[0121] It should be noted that the data cleaning process is to eliminate data that does not conform to physical characteristics and empty values, and fill the eliminated values ​​through linear interpolation to generate data cleaning results of valid charging trip data.

[0122] Step S122, obtaining the state of charge SOC division point, and resampling the total current, load current, battery temperature, battery voltage, ambient temperature and health state SOH in the data cleaning result using the average aggregation method based on the SOC division point, and calculating the corresponding time interval between two SOC points.

[0123] It is understandable that after obtaining the state of charge SOC division point, the parameters in the data cleaning result can be resampled using the average aggregation method based on the SOC division point, and then the corresponding time interval between two SOC points can be calculated.

[0124] Step S123: Generate reconstructed data according to the resampled data and the time interval, and normalize the reconstructed data to obtain preliminary training data.

[0125] It should be understood that the early training data can be obtained by generating reconstructed data based on the resampled data and the time interval, and then normalizing the reconstructed data.

[0126] In the specific implementation, the data reconstruction process is to divide the interval SOC∈[0%,100%] into 501 points with △SOC=0.1%, that is, SOC=[0,30.1%,30.2%...79.9%,80%], and resample the total current, load current, battery temperature, battery voltage, ambient temperature, and SOH based on the divided SOC points using the average aggregation method, and calculate the corresponding time interval t between two SOC points (t=0 when SOC=0). At this time, the reconstructed data of a certain charging trip is X=[I1;I2;T;V;A;SOC;SOH;K;SV;EV;ST;ET;S_SOC;E_S,t], and X is normalized by the formula, where xmin is the minimum value of the column where x is located, and xmax is the maximum value of the column where x is located.

[0127]

[0128] Divide X into five intervals, namely, X1 = X[SOC∈[30,40)], X2 = X[SOC∈[40,50)], X3 = X[SOC∈[50,60)], X4 = X[SOC∈[60,70)], X5 = X[SOC∈[70,80)]; process all valid charging trips according to the above method; preferably, in the extraction of the effective charging trip, the starting SOC and the ending SOC of the effective trip can be divided to set the length of the charging interval for charging diagnosis according to the needs, that is, what SOC interval range needs to be met to diagnose the vehicle; the interval divided by X in the data reconstruction can also be defined according to the SOC range taken in the effective trip.

[0129] Through the above scheme, this embodiment cleans the effective charging trip data, eliminates data that does not conform to physical characteristics and null values, and fills the eliminated values ​​by linear interpolation to obtain data cleaning results; obtains the state of charge SOC division point, and resamples the total current, load current, battery temperature, battery voltage, ambient temperature and health state SOH in the data cleaning result based on the SOC division point by using the average aggregation method, and calculates the corresponding time interval between two SOC points; generates reconstructed data according to the resampled data and the time interval, normalizes the reconstructed data, and obtains preliminary training data, which can quickly obtain preliminary training data and improve the speed and efficiency of power battery risk diagnosis.

[0130] Further, Figure 6FIG. 4 is a flow chart of a fourth embodiment of a power battery risk diagnosis method according to the present invention. Figure 6 As shown, a fourth embodiment of the power battery risk diagnosis method of the present invention is proposed based on the first embodiment. In this embodiment, step S13 specifically includes the following steps:

[0131] Step S131: An initial diagnosis model is constructed by a LSTM_ENCODER layer and a LSTM_DECODER layer.

[0132] It should be noted that the initial diagnosis model consists of an LSTM_ENCODER layer and an LSTM DECODER layer.

[0133] Step S132: input the preliminary training data into the initial diagnosis model to obtain the final output result of the model.

[0134] It is understandable that after the preliminary training data is input into the initial diagnosis model, the final output result of the model can be obtained.

[0135] Step S133: Calculate the loss function of the initial diagnosis model according to the final output result of the model.

[0136] It should be understood that the loss function of the initial diagnosis model can be calculated based on the final output result of the model.

[0137] Furthermore, the step S133 specifically includes the following steps:

[0138] The loss function of the initial diagnosis model is calculated according to the final output result of the model by the following formula:

[0139]

[0140] Among them, loss is the loss function, y pn is the final output result of the model, y is the target value of the model, and n is the number of predicted values ​​output by the model.

[0141] Step S134: taking the model with the minimum loss function as the target diagnosis model.

[0142] It can be understood that, various initial diagnostic models are obtained, and the model with the minimum loss function is used as the target diagnostic model.

[0143] In the specific implementation, see Figure 7 , Figure 7 Schematic diagram of the model architecture in the power battery risk diagnosis method of the present invention, such as Figure 7As shown in the figure, the input of the model is X1~X5 divided in 3. Note that X1~X5 are trained separately, that is, the model obtained by training the X1 interval is only for the X1 interval. Thus, 5 models can be trained, which are respectively suitable for data in different SOC intervals. Taking X1 as an example, the sequence length of X1 is 100 and the number of features is 15, so its shape is (100,15). After the LSTM_ENCODER layer, the shape of X1 is (100,E), and after the LSTM_DECODER layer, the final output shape is (100,D), where E is the number of hidden units set in the encoding layer, which needs to be set according to the model performance. D is the feature dimension that the final model needs to output. Here, the model needs to reconstruct the temperature, battery voltage, and total current, so D=3; the final output result Yp of the model is in the shape of (100,3), and the target value of the model is in the shape of Y (100,3), which is the total current, total voltage, and battery temperature corresponding to X1; the loss function of the model is defined as, the model that reaches the minimum loss through iterative training is the optimal model corresponding to the X1 interval, and models M1~M5 are trained in this way.

[0144] Through the above scheme, this embodiment forms an initial diagnosis model by a LSTM_ENCODER layer and a LSTM_DECODER layer; inputs the previous training data into the initial diagnosis model to obtain the final output result of the model; calculates the loss function of the initial diagnosis model according to the final output result of the model; and uses the model with the smallest loss function as the target diagnosis model, which can diagnose the current risk status of the battery, effectively reduce the false alarm rate, improve the safety monitoring accuracy of electric vehicles, and improve the speed and efficiency of power battery risk diagnosis.

[0145] Further, Figure 8 FIG. 5 is a flow chart of a fifth embodiment of a power battery risk diagnosis method according to the present invention. Figure 8 As shown, a fifth embodiment of the power battery risk diagnosis method of the present invention is proposed based on the first embodiment. In this embodiment, the step S20 specifically includes the following steps:

[0146] Step S21, obtaining the total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle, and generating charging status data.

[0147] It should be noted that the preliminary training data is the status data of each charging process in the vehicle's history, including: total current I1, load current I2, battery temperature T, battery voltage V, ambient temperature A, SOC, SOH, cumulative mileage K, charging start voltage SV, charging end voltage EV, charging start temperature ST, charging end temperature ET, starting SOC S_SOC, and ending SOCE_SOC.

[0148] In the specific implementation, the vehicle risk status can be labeled according to the charging status data, that is, whether the battery under the corresponding charging trip is a risky vehicle. This can generally be labeled based on after-sales market feedback and professionals, or the cloud server can give the labeling information after logical analysis and judgment based on the current charging status data.

[0149] Through the above scheme, this embodiment generates charging status data by acquiring the total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle, which can improve the accuracy of safety monitoring of electric vehicles and improve the speed and efficiency of power battery risk diagnosis.

[0150] Accordingly, the present invention further provides a power battery risk diagnosis device.

[0151] Reference Fig. 9 , Fig. 9 This is a functional module diagram of the first embodiment of the power battery risk diagnosis device of the present invention.

[0152] In a first embodiment of the power battery risk diagnosis device of the present invention, the power battery risk diagnosis device comprises:

[0153] The training module 10 is used to obtain the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model.

[0154] The data acquisition module 20 is used to acquire the charging status data of each charging of the current new energy vehicle.

[0155] The diagnosis module 30 is used to input the charging state data into the target diagnosis model to obtain a power battery risk diagnosis result.

[0156] The training module 10 is also used to obtain the effective charging travel data of the current new energy vehicle; perform data cleaning and data reconstruction on the effective charging travel data to obtain preliminary training data; construct an initial diagnosis model, and train the initial diagnosis model according to the preliminary training data to obtain a target diagnosis model.

[0157] The training module 10 is also used to clean the effective charging trip data, eliminate data that does not conform to physical characteristics and empty values, fill the eliminated values ​​by linear interpolation, and obtain data cleaning results; obtain the state of charge SOC division point, and resample the total current, load current, battery temperature, battery voltage, ambient temperature and health state SOH in the data cleaning results based on the SOC division point using the average aggregation method, and calculate the corresponding time interval between two SOC points; generate reconstructed data based on the resampled data and the time interval, normalize the reconstructed data, and obtain preliminary training data.

[0158] The training module 10 is also used to form an initial diagnosis model consisting of an LSTM_ENCODER layer and an LSTM_DECODER layer; input the previous training data into the initial diagnosis model to obtain the final output result of the model; calculate the loss function of the initial diagnosis model according to the final output result of the model; and use the model with the minimum loss function as the target diagnosis model.

[0159] The training module 10 is also used to calculate the loss function of the initial diagnosis model according to the final output result of the model by the following formula:

[0160]

[0161] Among them, loss is the loss function, y pn is the final output result of the model, y is the target value of the model, and n is the number of predicted values ​​output by the model.

[0162] The data acquisition module 20 is also used to obtain the total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle to generate charging status data.

[0163] The diagnostic module 30 is further used to input the charging state data into the target diagnostic model to obtain the comprehensive loss value of all charging trips;

[0164] Obtain the hyperparameters of the target diagnostic model, and obtain the mean and standard deviation of the comprehensive loss value, and calculate the preset judgment threshold by the following formula:

[0165] N=μ+k×σ

[0166] Among them, N is the preset judgment threshold, μ is the average value of the comprehensive loss value, σ is the standard deviation of the comprehensive loss value, and k is the hyperparameter of the target diagnosis model;

[0167] comparing the comprehensive loss value with the preset determination threshold;

[0168] When the comprehensive loss value is less than the preset determination threshold, determining that the power battery risk diagnosis result is a risk-free state;

[0169] When the comprehensive loss value is not less than the preset determination threshold, it is determined that the power battery risk diagnosis result is a risky state.

[0170] Among them, the steps implemented by each functional module of the power battery risk diagnosis device can refer to the various embodiments of the power battery risk diagnosis method of the present invention, and will not be repeated here.

[0171] In addition, an embodiment of the present invention further provides a storage medium, on which a power battery risk diagnosis program is stored. When the power battery risk diagnosis program is executed by a processor, the following operations are implemented:

[0172] Acquire the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model;

[0173] Obtaining charging status data of each charging of the current new energy vehicle;

[0174] The charging status data is input into the target diagnosis model to obtain a power battery risk diagnosis result.

[0175] Furthermore, when the power battery risk diagnosis program is executed by the processor, the following operations are also implemented:

[0176] Obtain the effective charging trip data of current new energy vehicles;

[0177] Performing data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data;

[0178] An initial diagnosis model is constructed, and the initial diagnosis model is trained according to the previous training data to obtain a target diagnosis model.

[0179] Furthermore, when the power battery risk diagnosis program is executed by the processor, the following operations are also implemented:

[0180] Cleaning the effective charging trip data, removing data that does not conform to physical characteristics and empty values, and filling the removed values ​​by linear interpolation to obtain a data cleaning result;

[0181] Obtaining a state of charge (SOC) division point, resampling the total current, load current, battery temperature, battery voltage, ambient temperature, and health state (SOH) in the data cleaning result by using an average aggregation method based on the SOC division point, and calculating a corresponding time interval between two SOC points;

[0182] Reconstructed data is generated according to the resampled data and the time interval, and the reconstructed data is normalized to obtain preliminary training data.

[0183] Furthermore, when the power battery risk diagnosis program is executed by the processor, the following operations are also implemented:

[0184] The initial diagnosis model consists of an LSTM_ENCODER layer and an LSTM_DECODER layer;

[0185] Inputting the preliminary training data into the initial diagnosis model to obtain the final output result of the model;

[0186] Calculating the loss function of the initial diagnosis model according to the final output result of the model;

[0187] The model with the smallest loss function is taken as the target diagnosis model.

[0188] Furthermore, when the power battery risk diagnosis program is executed by the processor, the following operations are also implemented:

[0189] The loss function of the initial diagnosis model is calculated according to the final output result of the model by the following formula:

[0190]

[0191] Among them, loss is the loss function, y pn is the final output result of the model, y is the target value of the model, and n is the number of predicted values ​​output by the model.

[0192] Furthermore, when the power battery risk diagnosis program is executed by the processor, the following operations are also implemented:

[0193] The total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle are obtained to generate charging status data.

[0194] Furthermore, when the power battery risk diagnosis program is executed by the processor, the following operations are also implemented:

[0195] Inputting the charging state data into the target diagnosis model to obtain a comprehensive loss value of all charging trips;

[0196] Obtain the hyperparameters of the target diagnostic model, and obtain the mean and standard deviation of the comprehensive loss value, and calculate the preset judgment threshold by the following formula:

[0197] N=μ+k×σ

[0198] Among them, N is the preset judgment threshold, μ is the average value of the comprehensive loss value, σ is the standard deviation of the comprehensive loss value, and k is the hyperparameter of the target diagnosis model;

[0199] comparing the comprehensive loss value with the preset determination threshold;

[0200] When the comprehensive loss value is less than the preset determination threshold, determining that the power battery risk diagnosis result is a risk-free state;

[0201] When the comprehensive loss value is not less than the preset determination threshold, it is determined that the power battery risk diagnosis result is a risky state.

[0202] Those skilled in the art will appreciate that all or part of the steps in the above-mentioned implementation method can be accomplished by instructing related hardware through a program, and the program is stored in a storage medium, including a number of instructions for enabling a device (which may be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the method described in each implementation method of the present application; and the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0203] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0204] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0205] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A power battery risk diagnosis method, characterized in that: The power battery risk diagnosis method comprises: Acquire the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model; Obtaining charging status data of each charging of the current new energy vehicle; The charging status data is input into the target diagnosis model to obtain a power battery risk diagnosis result.

2. The power battery risk diagnosis method according to claim 1, characterized in that: The obtaining of the preliminary training data of the power battery of the current new energy vehicle, training the initial diagnosis model according to the preliminary training data, and obtaining the target diagnosis model includes: Obtain the effective charging trip data of current new energy vehicles; Performing data cleaning and data reconstruction on the effective charging trip data to obtain preliminary training data; An initial diagnosis model is constructed, and the initial diagnosis model is trained according to the previous training data to obtain a target diagnosis model.

3. The power battery risk diagnosis method according to claim 2, characterized in that: The data cleaning and data reconstruction of the effective charging trip data to obtain preliminary training data includes: Cleaning the effective charging trip data, removing data that does not conform to physical characteristics and empty values, and filling the removed values ​​by linear interpolation to obtain a data cleaning result; Obtaining a state of charge (SOC) division point, resampling the total current, load current, battery temperature, battery voltage, ambient temperature, and health state (SOH) in the data cleaning result by using an average aggregation method based on the SOC division point, and calculating a corresponding time interval between two SOC points; Reconstructed data is generated according to the resampled data and the time interval, and the reconstructed data is normalized to obtain preliminary training data.

4. The power battery risk diagnosis method according to claim 2, characterized in that: The constructing of the initial diagnostic model, training the initial diagnostic model according to the previous training data, and obtaining the target diagnostic model, includes: The initial diagnosis model consists of an LSTM_ENCODER layer and an LSTM_DECODER layer; Inputting the preliminary training data into the initial diagnosis model to obtain the final output result of the model; Calculating the loss function of the initial diagnosis model according to the final output result of the model; The model with the smallest loss function is taken as the target diagnosis model.

5. The power battery risk diagnosis method according to claim 4, characterized in that: The step of calculating the loss function of the initial diagnosis model according to the final output result of the model comprises: The loss function of the initial diagnosis model is calculated according to the final output result of the model by the following formula: Among them, loss is the loss function, y pn is the final output result of the model, y is the target value of the model, and n is the number of predicted values ​​output by the model.

6. The power battery risk diagnosis method according to claim 1, characterized in that: The acquiring of charging status data of each charging of the current new energy vehicle includes: The total current, load current, battery temperature, battery voltage, ambient temperature, SOC, SOH, cumulative mileage, charging start voltage, charging end voltage, charging start temperature, charging end temperature, starting SOC and ending SOC of each charging of the current new energy vehicle are obtained to generate charging status data.

7. The power battery risk diagnosis method according to claim 1, characterized in that: The step of inputting the charging state data into the target diagnosis model to obtain a power battery risk diagnosis result includes: Inputting the charging state data into the target diagnosis model to obtain a comprehensive loss value of all charging trips; Obtain the hyperparameters of the target diagnostic model, and obtain the mean and standard deviation of the comprehensive loss value, and calculate the preset judgment threshold by the following formula: N=μ+k×σ Among them, N is the preset judgment threshold, μ is the average value of the comprehensive loss value, σ is the standard deviation of the comprehensive loss value, and k is the hyperparameter of the target diagnosis model; comparing the comprehensive loss value with the preset determination threshold; When the comprehensive loss value is less than the preset determination threshold, determining that the power battery risk diagnosis result is a risk-free state; When the comprehensive loss value is not less than the preset determination threshold, it is determined that the power battery risk diagnosis result is a risky state.

8. A power battery risk diagnosis device, characterized in that: The power battery risk diagnosis device comprises: A training module is used to obtain the preliminary training data of the power battery of the current new energy vehicle, train the initial diagnosis model according to the preliminary training data, and obtain the target diagnosis model; A data acquisition module, used to acquire charging status data of each charging of the current new energy vehicle; The diagnosis module is used to input the charging status data into the target diagnosis model to obtain a power battery risk diagnosis result.

9. A power battery risk diagnosis device, characterized in that: The power battery risk diagnosis device includes: a memory, a processor, and a power battery risk diagnosis program stored in the memory and executable on the processor, wherein the power battery risk diagnosis program is configured to implement the steps of the power battery risk diagnosis method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a power battery risk diagnosis program, and when the power battery risk diagnosis program is executed by the processor, the steps of the power battery risk diagnosis method according to any one of claims 1 to 7 are implemented.

Images