Method and equipment for detecting coolant leakage of power battery and storage medium
By installing multiple vibration sensors in the power battery and collecting resistance signals, extracting features using a classification network to identify the leakage of coolant of the power battery, the problem of high error detection rate in the prior art is solved, and detection accuracy and safety are improved.
Patent Information
- Application Number
- CN202411469464.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-01-10
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art has a high false detection rate when detecting leakage of power battery coolant, especially when the coolant is slowly leaking, which makes it difficult to accurately detect, resulting in a safety risk of power battery.
Using multiple vibration sensors and resistance signal acquisition, vibration and resistance characteristics are extracted through classification networks (including encoder, decoder and classification head structures), and fused into multimodal characteristics to identify whether the cooler leaks coolant and improve detection Accuracy.
When the coolant is slowly leaking, the detection sensitivity is high, which significantly improves the detection accuracy of the coolant leakage of the power battery and enhances the safety of the power battery.
Smart Images
Figure CN119290259A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application belong to the technical field of new energy, and in particular, relate to a method, device and storage medium for detecting coolant leakage in a power battery. Background Art
[0002] As new energy vehicles gradually become popular in people's daily lives, higher requirements are placed on the performance, life and safety of power batteries. Under actual application conditions, design defects, external collision and extrusion, vibration fatigue, extreme temperature and other reasons may cause the cooler inside the power battery to deform or break, resulting in coolant leakage, short circuit and arcing, which may lead to thermal runaway.
[0003] At present, the method of detecting whether a power battery is leaking coolant mainly relies on resistance. When the insulation resistance value between its conductive loop and the non-powered platform drops to a certain threshold, it can be considered that the power battery coolant is leaking.
[0004] However, in the case of a slow leak of coolant, the insulation resistance value will basically not decrease. Only when the coolant accumulates and soaks the conductive circuit (especially the high-voltage components) will the insulation resistance value drop significantly and reach a detectable level, resulting in a certain false detection rate in the detection of coolant leakage in the power battery, posing a risk to the power battery. Summary of the invention
[0005] In view of this, embodiments of the present application provide a method, device, and storage medium for detecting coolant leakage in a power battery, so as to improve the accuracy of detecting coolant leakage in a power battery.
[0006] A first aspect of an embodiment of the present application provides a method for detecting coolant leakage in a power battery, comprising:
[0007] During the driving of the vehicle, the motion state of the power battery in the vehicle is detected; the power battery is provided with a cooler for storing coolant; the housing of the cooler is provided with a plurality of vibration sensors;
[0008] If the motion state is a collision, collecting the original vibration signal generated by each of the vibration sensors, and collecting the resistance signal of the cooler;
[0009] Detecting the propagation states of the plurality of original vibration signals in the housing of the cooler according to the motion state;
[0010] If the propagation state is a sudden change, a preset classification network is loaded; the classification network includes a first encoder, a second encoder, a decoder and a classification head structure;
[0011] Inputting a target vibration signal formed by fusing the plurality of original vibration signals into the first encoder to sequentially extract target vibration features of multiple scales;
[0012] Inputting the resistance signal into the second encoder to sequentially extract target resistance features of multiple scales;
[0013] Inputting the target vibration features at multiple scales and the target resistance features at multiple scales into the decoder to fuse them into target multimodal features;
[0014] The target multimodal features are input into the classification head structure to classify whether the cooler leaks the coolant.
[0015] A second aspect of an embodiment of the present application provides a device for detecting coolant leakage in a power battery, comprising:
[0016] A motion state detection module is used to detect the motion state of a power battery in the vehicle during the driving process; the power battery is provided with a cooler for storing coolant; the housing of the cooler is provided with a plurality of vibration sensors;
[0017] A signal acquisition module, for acquiring original vibration signals generated by each of the vibration sensors and a resistance signal of the cooler if the motion state is a collision;
[0018] A propagation state detection module, used for detecting the propagation state of a plurality of the original vibration signals in the housing of the cooler according to the motion state;
[0019] A classification network loading module, used for loading a preset classification network if the propagation state is a sudden change; the classification network includes a first encoder, a second encoder, a decoder and a classification head structure;
[0020] A target vibration feature extraction module, used for inputting a target vibration signal formed by fusing a plurality of the original vibration signals into the first encoder to sequentially extract target vibration features of multiple scales;
[0021] A target resistance feature extraction module, used for inputting the resistance signal into the second encoder to sequentially extract target resistance features of multiple scales;
[0022] A target multimodal feature fusion module, used for inputting the target vibration features of multiple scales and the target resistance features of multiple scales into the decoder to fuse them into target multimodal features;
[0023] The coolant leakage classification module is used to input the target multimodal features into the classification head structure to classify whether the cooler leaks the coolant.
[0024] A third aspect of an embodiment of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for detecting coolant leakage of a power battery as described in the first aspect above is implemented.
[0025] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for detecting coolant leakage of a power battery as described in the first aspect above is implemented.
[0026] A fifth aspect of an embodiment of the present application provides a computer program product. When the computer program product is run on a computer, the computer is enabled to execute the method for detecting coolant leakage of a power battery as described in the first aspect.
[0027] In this embodiment, during the driving of the vehicle, the movement state of the power battery in the vehicle is detected; a cooler for storing coolant is configured in the power battery; the housing of the cooler is configured with multiple vibration sensors; if the movement state is a collision, the original vibration signals generated by each vibration sensor are collected, and the resistance signal of the cooler is collected; the propagation state of the multiple original vibration signals in the housing of the cooler is detected according to the movement state; if the propagation state is a mutation, a preset classification network is loaded; the classification network includes a first encoder, a second encoder, a decoder and a classification head structure; the target vibration signal formed by fusion of multiple original vibration signals is input into the first encoder to sequentially extract target vibration features of multiple scales; the resistance signal is input into the second encoder to sequentially extract target resistance features of multiple scales; the target vibration features of multiple scales and the target resistance features of multiple scales are input into the decoder to be fused into target multimodal features; the target multimodal features are input into the classification head structure to classify whether the cooler leaks coolant. This embodiment increases the amount of information of the internal resistance in terms of timing and increases the amount of information of the vibration signal externally. The vibration signal is strongly correlated with the deformation of the cooler and still has a high sensitivity in the case of a slow leak of the coolant, thereby improving the accuracy of detecting coolant leakage in the power battery and improving the safety of the power battery. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or prior art descriptions. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0029] Figure 1is a schematic diagram of a method for detecting coolant leakage in a power battery provided by an embodiment of the present application;
[0030] Figure 2 is a schematic diagram of a classification network provided in an embodiment of the present application;
[0031] Figure 3 is a schematic diagram of a device for detecting coolant leakage in a power battery provided in an embodiment of the present application;
[0032] Figure 4 It is a schematic diagram of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0033] In the following description, specific details such as specific system structures, technologies, etc. are proposed for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from hindering the description of the present application.
[0034] The technical solution of the present application is described below through specific embodiments.
[0035] Reference Figure 1 , which shows a schematic diagram of a method for detecting coolant leakage in a power battery provided by an embodiment of the present application, which may specifically include the following steps:
[0036] Step 101: Detect the motion state of a power battery in the vehicle while the vehicle is traveling.
[0037] In this embodiment, vehicles such as cars and trucks may be equipped with power batteries (such as lithium batteries, sodium batteries, etc.) as a power source.
[0038] The power battery is equipped with a cooler for storing coolant. For power batteries in square, cylindrical and other forms, the coolers are mostly at the module level and placed at the bottom of the power battery box. For power batteries in soft-pack forms, the coolers are mostly at the cell level.
[0039] Furthermore, the shell of the cooler is configured with multiple vibration sensors, which can utilize the characteristics of transmitting vibration waves in liquid, solid and other media to detect vibration signals. The types of vibration sensors may include piezoelectric sensors, elastic wave sensors, bone conduction sensors, etc.
[0040] During the driving of the vehicle, the BMS (Battery Management System) in the power battery is responsible for detecting whether there is a coolant leak in the cooler of the power battery.
[0041] At this time, the BMS can call on various sensors of the power battery in the vehicle to collect parameters when the vehicle is moving, and evaluate the movement state of the power battery in the vehicle as the vehicle moves based on these parameters.
[0042] In a specific implementation, sensors such as accelerometers can be used to collect the acceleration of the power battery in the vehicle on multiple axes (such as the X-axis, Y-axis, Z-axis, etc.), and multiple accelerations can be fused into a motion vector of the power battery in the vehicle using a linear or nonlinear method.
[0043] Taking linear as an example, the fusion process can be as follows:
[0044] W=α1·x+α2·y+α3·z
[0045]
[0046] Among them, W represents the motion vector, x represents the acceleration on the X-axis, y represents the acceleration on the Y-axis, z represents the acceleration on the Z-axis, and α1, α2, and α3 all represent weights.
[0047] The magnitude of the motion vector is compared with a preset offset threshold.
[0048] If the modulus of the motion vector is greater than or equal to the preset offset threshold, indicating that the power battery in the vehicle moves a large amount, it can be determined that the motion state of the power battery in the vehicle is a collision. At this time, the vehicle moves a large amount, and the power battery in the vehicle may collide with other structural parts in the vehicle.
[0049] Step 102: If the motion state is a collision, then collect the original vibration signals generated by each vibration sensor and collect the resistance signal of the cooler.
[0050] When the power battery in a vehicle collides with other structural parts in the vehicle, the cooler may be deformed during the collision, and there is a risk of coolant leakage. During this process, the original vibration signals generated by each vibration sensor and the resistance signal of the conductive circuit in the cooler can be collected.
[0051] In a specific implementation, when it is determined that the motion state of the power battery in the vehicle is a collision, a sampling time period may be determined on the time axis.
[0052] During this time period, the original vibration signals generated by each vibration sensor and the resistance signal of the conductive circuit in the cooler are synchronously collected.
[0053] The original vibration signal and the resistance signal are original electrical signals. At this time, the original vibration signal and the resistance signal may be preprocessed, such as filtering, smoothing, etc., to improve the quality of the original vibration signal and the resistance signal.
[0054] Among them, for the original vibration signal, the modulus of the motion vector can be mapped to a window using a piecewise function or the like; wherein the size of the window is positively correlated with the modulus of the motion vector, that is, the larger the modulus of the motion vector, the larger the size of the window, and conversely, the smaller the modulus of the motion vector, the smaller the size of the window.
[0055] The original vibration signal is filtered using a window (such as median filtering) so that the intensity of the filtering is adapted to the motion state of the power battery in the vehicle, thereby achieving adaptive filtering and improving the quality of the original vibration signal.
[0056] Step 103: Detect the propagation state of multiple original vibration signals in the shell of the cooler according to the motion state.
[0057] When the outer shell of the cooler is not deformed, the original vibration signal propagates in the outer shell of the cooler according to a certain propagation law. Therefore, the propagation law of the original vibration signal when propagating in the outer shell of the cooler can be predicted based on the motion state, and compared with the actual propagation law of the original vibration signal when propagating in the outer shell of the cooler, the propagation state of the original vibration signal in the cooler shell can be detected. Based on the propagation state, a preliminary screening of the task of detecting whether the cooler is leaking coolant is performed, which can effectively reduce the data volume of the task, improve the real-time performance of detecting whether the cooler is leaking coolant, and meet the needs of intelligent assisted driving of vehicles.
[0058] If the actual propagation law of the original vibration signal matches the predicted propagation law, it can be determined that the propagation state of the original vibration signal in the housing of the cooler is not abrupt.
[0059] If the actual propagation law of the original vibration signal does not match the predicted propagation law, it can be determined that the propagation state of the original vibration signal in the housing of the cooler is a mutation.
[0060] In a specific implementation, a side wall that is indicated by the direction of the motion vector (ie, opposite to the direction) and against which the collision occurs is determined in the housing of the cooler.
[0061] In an offline experimental environment, the relationship between the propagation of vibration signals between each side wall of the cooler shell and each vibration sensor can be fitted by means of adaptive filtering, spectrum division, neural network, etc., which is recorded as the transfer function between the side wall and the vibration sensor.
[0062] For each vibration sensor, the transfer function between the side wall and the vibration sensor is queried.
[0063] For each vibration sensor, the original vibration signal is restored to a candidate vibration signal generated when the side wall collides according to the transfer function.
[0064] For all candidate vibration signals, correlation coefficients between any two candidate vibration signals, such as the Pearson correlation coefficient, may be calculated.
[0065] Each correlation coefficient is compared with a preset correlation threshold. To improve the sensitivity of the detection, the correlation coefficient may be an empirical value with a lower value, such as 0.5.
[0066] If the cooler housing is not deformed, the candidate vibration signals restored from the vibration sensors should be similar.
[0067] If at least one correlation coefficient is less than or equal to a preset correlation threshold, it means that the similarity between the candidate vibration signals restored from at least two vibration sensors is low, and the outer shell of the cooler may be deformed on one of the propagation paths from the side wall to the vibration sensor. It can be determined that the propagation state of at least one original vibration signal in the outer shell of the cooler is a mutation.
[0068] Step 104: If the propagation state is a mutation, load the preset classification network.
[0069] In this embodiment, a classification network can be constructed and trained in advance based on deep learning, so that the classification network can perform binary classification based on multiple original vibration signals and resistance signals to identify whether the cooler is leaking coolant.
[0070] Among them, Figure 2 As shown, the classification network includes a first encoder Encoder_1, a second encoder Encoder_2, a decoder Decoder and a classification head structure Head, wherein the first encoder Encoder_1 is responsible for extracting features from multiple original vibration signals, the second encoder Encoder_2 is responsible for extracting features from resistance signals, the decoder Decoder is responsible for adjusting (such as dimensionality reduction) and fusing the features of the original vibration signals and the features of the resistance signals to meet the requirements of the classification task, and the classification head structure Head is responsible for performing specific classification tasks, namely, identifying whether the cooler is leaking coolant.
[0071] In this embodiment, the classification network (including the first encoder Encoder_1, the second encoder Encoder_2, the decoder Decoder and the classification head structure Head) is modularly designed to improve reusability and reduce development costs. The first encoder Encoder_1 and the second encoder Encoder_2 contain more parameters. The classification mode of the first encoder Encoder_1, the second encoder Encoder_2 and the decoder Decoder can reduce training time and reduce overfitting. In addition, by modifying the structures of the first encoder Encoder_1, the second encoder Encoder_2, the decoder Decoder and the classification head structure Head, the classification network can be applied to different types of coolers to improve the generalization ability and performance of the classifier.
[0072] If the propagation state of at least one original vibration signal in the housing of the cooler is a mutation, the classification network can be loaded into the memory for operation, waiting for further identification of whether the cooler is leaking coolant.
[0073] Step 105: Input the target vibration signal formed by fusing the multiple original vibration signals into a first encoder to sequentially extract target vibration features of multiple scales.
[0074] In this embodiment, if Figure 2 As shown, multiple original vibration signals can be fused into a target vibration signal, and the target vibration signal is input into the first encoder Encoder_1. The first encoder Encoder_1 extracts features of multiple scales from the target vibration signal in sequence, which are recorded as target vibration features.
[0075] In a structure of a first encoder Encoder_1, such as Figure 2 As shown, the first encoder Encoder_1 includes a first feature structure Feature_1, a first outer encoding block OutEncoderBlock_1, a second outer encoding block OutEncoderBlock_2 and a third outer encoding block OutEncoderBlock_3; accordingly, the target vibration features of multiple scales include a first target vibration feature, a second target vibration feature and a third target vibration feature.
[0076] Among them, the first feature structure Feature_1, the first outer encoding block OutEncoderBlock_1, the second outer encoding block OutEncoderBlock_2 and the third outer encoding block OutEncoderBlock_3 are all modules that encapsulate structures related to recurrent neural networks (RNN) (such as LSTM (Long Short Term Memory), GRU (Gated Recurrent Unit), self-attention mechanism, activation function, normalization layer), and can reuse the structure of extracting sequence features in a third-party network (such as Transformer, etc.), or can customize the structure, which is not limited in this embodiment.
[0077] In this structure, multiple original vibration signals are concatenated into a target vibration signal.
[0078] The target vibration signal is input into the first feature structure Feature_1 to extract features, which are recorded as basic vibration features.
[0079] The basic vibration feature is input into the first outer encoding block OutEncoderBlock_1 to extract the feature, which is recorded as the first target vibration feature.
[0080] The first target vibration feature is input into the second outer encoding block OutEncoderBlock_2 to extract features, which are recorded as the second target vibration feature.
[0081] The second target vibration feature is input into the third outer encoding block OutEncoderBlock_3 to extract features, which are recorded as the third target vibration feature.
[0082] Generally, the scale of the first target vibration feature is larger than the scale of the second target vibration feature, and the scale of the second target vibration feature is larger than the scale of the third target vibration feature.
[0083] Exemplarily, the basic vibration feature is the size of the target vibration signal downsampled 8 times, the first target vibration feature is the size of the target vibration signal downsampled 16 times, the second target vibration feature is the size of the target vibration signal downsampled 32 times, and the third target vibration feature is the size of the target vibration signal downsampled 64 times.
[0084] Step 106: Input the resistance signal into a second encoder to sequentially extract target resistance features of multiple scales.
[0085] In this embodiment, if Figure 2As shown, the resistance signal can be input into the second encoder Encoder_2, and the second encoder Encoder_2 sequentially extracts features of multiple scales from the resistance signal and records them as target resistance features.
[0086] In a structure of a second encoder Encoder_2, such as Figure 2 As shown, the second encoder Encoder_2 includes a second feature structure Feature2, a first inner encoding block InEncoderBlock_1, a second inner encoding block InEncoderBlock_2, and a third inner encoding block InEncoderBlock_3; accordingly, the target resistance features of multiple scales include a first target resistance feature, a second target resistance feature, and a third target resistance feature.
[0087] Among them, the second feature structure Feature2, the first inner encoding block InEncoderBlock_1, the second inner encoding block InEncoderBlock_2, and the third inner encoding block InEncoderBlock_3 are all modules that encapsulate structures related to recurrent neural networks (such as LSTM, GRU, self-attention mechanism, activation function, normalization layer), and can reuse the structure of extracting sequence features in a third-party network (such as Transformer, etc.), or can customize the structure, which is not limited in this embodiment.
[0088] In this structure, the resistance signal is input into the second feature structure Feature2 to extract features, which are recorded as basic resistance features.
[0089] The basic resistance feature is input into the first inner encoding block InEncoderBlock_1 to extract the feature, which is recorded as the first target resistance feature.
[0090] The first target resistance feature is input into the second inner encoding block InEncoderBlock_2 to extract the feature, which is recorded as the second target resistance feature.
[0091] The second target resistance feature is input into the third inner encoding block InEncoderBlock_3 to extract the feature, which is recorded as the third target resistance feature.
[0092] Generally, the dimension of the first target resistance characteristic is greater than the dimension of the second target resistance characteristic, and the dimension of the second target resistance characteristic is greater than the dimension of the third target resistance characteristic.
[0093] Exemplarily, the base resistance characteristic is the size of the target resistance signal downsampled 8 times, the first target resistance characteristic is the size of the target resistance signal downsampled 16 times, the second target resistance characteristic is the size of the target resistance signal downsampled 32 times, and the third target resistance characteristic is the size of the target resistance signal downsampled 64 times.
[0094] Step 107: Input the target vibration features at multiple scales and the target resistance features at multiple scales into a decoder and fuse them into target multi-modal features.
[0095] In this embodiment, if Figure 2 As shown in the figure, the target vibration features of multiple scales and the target resistance features of multiple scales are input into the decoder. The decoder captures the time dependency of the target vibration features and the target resistance features at different scales and increases the receptive field. It can focus on short-term and fine-grained patterns while capturing macro trends and periodicity. The target vibration features of multiple scales and the target resistance features of multiple scales are fused into a new feature, which is recorded as the target multimodal feature, effectively improving the quality of the target multimodal feature and thus improving the classification accuracy.
[0096] In a decoder structure, such as Figure 2 As shown, the decoder Decoder includes a first decoding block DecoderBlock_1, a second decoding block DecoderBlock_2 and a third decoding block DecoderBlock_3.
[0097] Among them, the first decoding block DecoderBlock_1, the second decoding block DecoderBlock_2 and the third decoding block DecoderBlock_3 are all modules that encapsulate structures related to sequence feature fusion (such as convolutional layer, pooling layer, fully connected layer, downsampling layer, etc.), and can reuse the structure of fusion sequence features in a third-party network (such as Transformer, etc.), or can customize the structure, which is not limited in this embodiment.
[0098] In this structure, the first target vibration feature and the first target resistance feature are concatenated into a new feature, which is recorded as the first original multimodal feature.
[0099] The first original multimodal feature is input into the first decoding block DecoderBlock_1 to extract features, which are recorded as the second original multimodal feature.
[0100] The second target vibration feature, the second target resistance feature and the second original multimodal feature are concatenated into a new feature, which is recorded as the third original multimodal feature.
[0101] The third original multimodal feature is input into the second decoding block DecoderBlock_2 to extract features, which are recorded as the fourth original multimodal feature.
[0102] The third target vibration feature, the third target resistance feature and the fourth original multimodal feature are concatenated into a new feature, which is recorded as the fifth original multimodal feature.
[0103] The fifth original multimodal feature is input into the third decoding block DecoderBlock_3 to extract features, which are recorded as target multimodal features.
[0104] Step 108: input the target multimodal features into the classification head structure to classify whether the cooler is leaking coolant.
[0105] In this embodiment, if Figure 2 As shown, the target multimodal features are input into the classification head structure Head, and the classification head structure Head performs the classification task and maps the target multimodal features to whether the cooler leaks coolant.
[0106] In a classification header structure Head structure, such as Figure 2 As shown in the figure, the classification head structure Head includes a long short-term memory network LSTM, a convolutional layer (Convolutional Layer, Conv), a pooling layer Pooling and a fully connected layer (Fully Connected Layer, FC).
[0107] In this structure, the target multimodal features are input into the long short-term memory network LSTM to extract the time series features, which are recorded as multimodal time series features.
[0108] The multimodal temporal features are input into the convolutional layer Conv to extract spatial features, which are recorded as multimodal spatial features.
[0109] The multimodal spatial features are input into the pooling layer to perform pooling operations (such as maximum pooling operations, average pooling operations, etc.) to obtain multimodal pooling features.
[0110] The multimodal time series features and the multimodal pooling features are concatenated into new features, which are recorded as multimodal cascade features.
[0111] The multimodal cascade features are input into the fully connected layer FC and mapped into the probability of the cooler casing being damaged.
[0112] If the probability is greater than or equal to a preset confidence threshold, it is determined that the cooler is leaking coolant.
[0113] If the probability is less than a preset confidence threshold, it is determined that the cooler does not leak coolant.
[0114] In this embodiment, during the driving of the vehicle, the movement state of the power battery in the vehicle is detected; a cooler for storing coolant is configured in the power battery; the housing of the cooler is configured with multiple vibration sensors; if the movement state is a collision, the original vibration signals generated by each vibration sensor are collected, and the resistance signal of the cooler is collected; the propagation state of the multiple original vibration signals in the housing of the cooler is detected according to the movement state; if the propagation state is a mutation, a preset classification network is loaded; the classification network includes a first encoder, a second encoder, a decoder and a classification head structure; the target vibration signal formed by fusion of multiple original vibration signals is input into the first encoder to sequentially extract target vibration features of multiple scales; the resistance signal is input into the second encoder to sequentially extract target resistance features of multiple scales; the target vibration features of multiple scales and the target resistance features of multiple scales are input into the decoder to be fused into target multimodal features; the target multimodal features are input into the classification head structure to classify whether the cooler leaks coolant. This embodiment increases the amount of information of the internal resistance in terms of timing and increases the amount of information of the vibration signal externally. The vibration signal is strongly correlated with the deformation of the cooler and still has a high sensitivity in the case of a slow leak of the coolant, thereby improving the accuracy of detecting coolant leakage in the power battery and improving the safety of the power battery.
[0115] It should be noted that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0116] Reference Figure 3 , shows a schematic diagram of a device for detecting coolant leakage of a power battery provided in an embodiment of the present application, which may specifically include the following modules:
[0117] The motion state detection module 301 is used to detect the motion state of the power battery in the vehicle during the driving process; the power battery is equipped with a cooler for storing coolant; the housing of the cooler is equipped with multiple vibration sensors;
[0118] A signal acquisition module 302, for acquiring original vibration signals generated by each of the vibration sensors and a resistance signal of the cooler if the motion state is a collision;
[0119] A propagation state detection module 303, used for detecting the propagation state of the plurality of original vibration signals in the housing of the cooler according to the motion state;
[0120] A classification network loading module 304 is used to load a preset classification network if the propagation state is a sudden change; the classification network includes a first encoder, a second encoder, a decoder and a classification head structure;
[0121] A target vibration feature extraction module 305 is used to input a target vibration signal formed by fusing a plurality of the original vibration signals into the first encoder to sequentially extract target vibration features of multiple scales;
[0122] A target resistance feature extraction module 306, configured to input the resistance signal into the second encoder to sequentially extract target resistance features of multiple scales;
[0123] A target multimodal feature fusion module 307 is used to input the target vibration features of multiple scales and the target resistance features of multiple scales into the decoder to fuse them into target multimodal features;
[0124] The coolant leakage classification module 308 is used to input the target multimodal features into the classification head structure to classify whether the cooler leaks the coolant.
[0125] Optionally, the motion state detection module 301 is further used for:
[0126] Collecting acceleration of a power battery in the vehicle on multiple axes;
[0127] fusing the plurality of accelerations into a motion vector of a power battery in the vehicle;
[0128] If the modulus of the motion vector is greater than or equal to a preset offset threshold, it is determined that the motion state of the power battery in the vehicle is a collision.
[0129] Optionally, the signal acquisition module 302 is further used for:
[0130] Identify time periods on the timeline;
[0131] In the time period, synchronously collecting original vibration signals generated by the vibration sensors and collecting resistance signals of the coolers;
[0132] Mapping the modulus of the motion vector into a window; the size of the window is positively correlated with the modulus of the motion vector;
[0133] A filtering process is performed on the original vibration signal using the window.
[0134] Optionally, the propagation state detection module 303 is further used to:
[0135] determining a side wall in a housing of the cooler indicated by the direction of the motion vector and against which the collision occurred;
[0136] For each of the vibration sensors, query a transfer function between the side wall and the vibration sensor;
[0137] For each of the vibration sensors, restoring the original vibration signal into a candidate vibration signal generated when the side wall collides according to the transfer function;
[0138] Calculating correlation coefficients between each of the candidate vibration signals;
[0139] If at least one of the correlation coefficients is less than or equal to a preset correlation threshold, it is determined that the propagation state of the original vibration signal in the housing of the cooler is a mutation.
[0140] Optionally, the first encoder includes a first feature structure, a first outer encoding block, a second outer encoding block and a third outer encoding block; the target vibration features of multiple scales include a first target vibration feature, a second target vibration feature and a third target vibration feature
[0141] The target vibration feature extraction module 305 is also used for:
[0142] splicing a plurality of the original vibration signals into a target vibration signal;
[0143] Inputting the target vibration signal into the first feature structure to extract basic vibration features;
[0144] Inputting the basic vibration feature into the first outer encoding block to extract a first target vibration feature;
[0145] Inputting the first target vibration feature into the second outer encoding block to extract a second target vibration feature;
[0146] The second target vibration feature is input into the third outer encoding block to extract the third target vibration feature.
[0147] Optionally, the second encoder includes a second characteristic structure, a first inner encoding block, a second inner encoding block, and a third inner encoding block; the target resistance features of the multiple scales include a first target resistance feature, a second target resistance feature, and a third target resistance feature;
[0148] The target resistance feature extraction module 306 is also used for:
[0149] Inputting the resistance signal into the second characteristic structure to extract basic resistance characteristics;
[0150] Inputting the basic resistance feature into the first inner coding block to extract a first target resistance feature;
[0151] Inputting the first target resistance feature into the second inner coding block to extract a second target resistance feature;
[0152] The second target resistance feature is input into the third inner encoding block to extract the third target resistance feature.
[0153] Optionally, the decoder includes a first decoding block, a second decoding block and a third decoding block;
[0154] The target multimodal feature fusion module 307 is also used for:
[0155] splicing the first target vibration feature and the first target resistance feature into a first original multimodal feature;
[0156] Inputting the first original multimodal feature into the first decoding block to extract a second original multimodal feature;
[0157] splicing the second target vibration feature, the second target resistance feature, and the second original multimodal feature into a third original multimodal feature;
[0158] Inputting the third original multimodal feature into the second decoding block to extract a fourth original multimodal feature;
[0159] splicing the third target vibration feature, the third target resistance feature and the fourth original multimodal feature into a fifth original multimodal feature;
[0160] The fifth original multimodal feature is input into the third decoding block to extract the target multimodal feature.
[0161] Optionally, the classification head structure includes a long short-term memory network, a convolutional layer, a pooling layer and a fully connected layer;
[0162] The coolant leakage classification module 308 is also used to:
[0163] Inputting the target multimodal features into the long short-term memory network to extract multimodal temporal features;
[0164] Inputting the multimodal temporal features into the convolutional layer to extract multimodal spatial features;
[0165] Inputting the multimodal spatial features into the pooling layer to perform a pooling operation to obtain a multimodal pooling feature;
[0166] Splicing the multimodal time series feature and the multimodal pooling feature into a multimodal cascade feature;
[0167] Inputting the multimodal cascade features into the fully connected layer and mapping them into the probability of the shell of the cooler being damaged;
[0168] If the probability is greater than or equal to a preset confidence threshold, it is determined that the cooler leaks the coolant.
[0169] An embodiment of the present application provides a device for detecting coolant leakage in a power battery. By using the device, each step in the aforementioned method embodiments can be implemented.
[0170] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment part.
[0171] Reference Figure 4 , shows a schematic diagram of a terminal device provided in an embodiment of the present application. Figure 4 As shown, the terminal device 400 in the embodiment of the present application includes: a processor 410, a memory 420, and a computer program 421 stored in the memory 420 and executable on the processor 410. When the processor 410 executes the computer program 421, the steps in each embodiment of the method for detecting coolant leakage in a power battery are implemented. Alternatively, when the processor 410 executes the computer program 421, the functions of each module / unit in each of the above-mentioned device embodiments are implemented.
[0172] Exemplarily, the computer program 421 may be divided into one or more modules / units, which are stored in the memory 420 and executed by the processor 410 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which may be used to describe the execution process of the computer program 421 in the terminal device 400.
[0173] The terminal device 400 may include, but is not limited to, a processor 410 and a memory 420. Those skilled in the art will appreciate that Figure 4 It is only an example of the terminal device 400 and does not constitute a limitation on the terminal device 400. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device 400 may also include input and output devices, network access devices, buses, etc.
[0174] The processor 410 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.
[0175] The memory 420 may be an internal storage unit of the terminal device 400, such as a hard disk or memory of the terminal device 400. The memory 420 may also be an external storage device of the terminal device 400, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 400. Further, the memory 420 may also include both an internal storage unit of the terminal device 400 and an external storage device. The memory 420 is used to store the computer program 421 and other programs and data required by the terminal device 400. The memory 420 may also be used to temporarily store data that has been output or is to be output.
[0176] An embodiment of the present application also discloses a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for detecting coolant leakage of a power battery as described in the aforementioned embodiments is implemented.
[0177] An embodiment of the present application further discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for detecting coolant leakage of a power battery as described in the aforementioned embodiments is implemented.
[0178] An embodiment of the present application further discloses a computer program product. When the computer program product is run on a computer, the computer is enabled to execute the method for detecting coolant leakage of a power battery as described in the aforementioned embodiments.
[0179] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application is described in detail with reference to the above-mentioned embodiments, a person skilled in the art should understand that the technical solutions described in the above-mentioned embodiments can still be modified, or some of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.
Claims
1. A method for detecting coolant leakage in a power battery, characterized in that: include: During the driving of the vehicle, the motion state of the power battery in the vehicle is detected; the power battery is provided with a cooler for storing coolant; The housing of the cooler is provided with a plurality of vibration sensors; If the motion state is a collision, collecting the original vibration signal generated by each of the vibration sensors, and collecting the resistance signal of the cooler; Detecting the propagation states of the plurality of original vibration signals in the housing of the cooler according to the motion state; If the propagation state is a sudden change, a preset classification network is loaded; the classification network includes a first encoder, a second encoder, a decoder and a classification head structure; Inputting a target vibration signal formed by fusing the plurality of original vibration signals into the first encoder to sequentially extract target vibration features of multiple scales; Inputting the resistance signal into the second encoder to sequentially extract target resistance features of multiple scales; Inputting the target vibration features at multiple scales and the target resistance features at multiple scales into the decoder to fuse them into target multimodal features; The target multimodal features are input into the classification head structure to classify whether the cooler leaks the coolant.
2. The method according to claim 1, characterized in that The detecting the motion state of the power battery in the vehicle includes: Collecting acceleration of a power battery in the vehicle on multiple axes; fusing the plurality of accelerations into a motion vector of a power battery in the vehicle; If the modulus of the motion vector is greater than or equal to a preset offset threshold, it is determined that the motion state of the power battery in the vehicle is a collision.
3. The method according to claim 2, characterized in that The collecting of the original vibration signals generated by the vibration sensors and the collecting of the resistance signals of the coolers include: Identify time periods on the timeline; In the time period, synchronously collecting original vibration signals generated by the vibration sensors and collecting resistance signals of the coolers; Mapping the modulus of the motion vector into a window; the size of the window is positively correlated with the modulus of the motion vector; A filtering process is performed on the original vibration signal using the window.
4. The method according to claim 2, characterized in that: The step of detecting the propagation state of the plurality of original vibration signals in the housing of the cooler according to the motion state comprises: determining a side wall in a housing of the cooler indicated by the direction of the motion vector and against which the collision occurred; For each of the vibration sensors, query a transfer function between the side wall and the vibration sensor; For each of the vibration sensors, restoring the original vibration signal into a candidate vibration signal generated when the side wall collides according to the transfer function; Calculating correlation coefficients between each of the candidate vibration signals; If at least one of the correlation coefficients is less than or equal to a preset correlation threshold, it is determined that the propagation state of the original vibration signal in the housing of the cooler is a mutation.
5. The method according to any one of claims 1 to 4, characterized in that The first encoder includes a first feature structure, a first outer encoding block, a second outer encoding block and a third outer encoding block; the target vibration features of the multiple scales include a first target vibration feature, a second target vibration feature and a third target vibration feature The step of inputting the target vibration signal formed by fusing the plurality of original vibration signals into the first encoder and sequentially extracting target vibration features of multiple scales comprises: splicing a plurality of the original vibration signals into a target vibration signal; Inputting the target vibration signal into the first feature structure to extract basic vibration features; Inputting the basic vibration feature into the first outer encoding block to extract a first target vibration feature; Inputting the first target vibration feature into the second outer encoding block to extract a second target vibration feature; The second target vibration feature is input into the third outer encoding block to extract the third target vibration feature.
6. The method according to claim 5, characterized in that The second encoder includes a second characteristic structure, a first inner encoding block, a second inner encoding block, and a third inner encoding block; the target resistance characteristics of the multiple scales include a first target resistance characteristic, a second target resistance characteristic, and a third target resistance characteristic; Inputting the resistance signal into the second encoder to sequentially extract target resistance features of multiple scales includes: Inputting the resistance signal into the second characteristic structure to extract basic resistance characteristics; Inputting the basic resistance feature into the first inner coding block to extract a first target resistance feature; Inputting the first target resistance feature into the second inner coding block to extract a second target resistance feature; The second target resistance feature is input into the third inner encoding block to extract the third target resistance feature.
7. The method according to claim 6, characterized in that The decoder comprises a first decoding block, a second decoding block and a third decoding block; The step of inputting the target vibration features at multiple scales and the target resistance features at multiple scales into the decoder to fuse them into target multimodal features includes: splicing the first target vibration feature and the first target resistance feature into a first original multimodal feature; Inputting the first original multimodal feature into the first decoding block to extract a second original multimodal feature; splicing the second target vibration feature, the second target resistance feature, and the second original multimodal feature into a third original multimodal feature; Inputting the third original multimodal feature into the second decoding block to extract a fourth original multimodal feature; splicing the third target vibration feature, the third target resistance feature and the fourth original multimodal feature into a fifth original multimodal feature; The fifth original multimodal feature is input into the third decoding block to extract the target multimodal feature.
8. The method according to claim 7, characterized in that The classification head structure includes a long short-term memory network, a convolutional layer, a pooling layer and a fully connected layer; The step of inputting the target multimodal feature into the classification head structure to classify whether the cooler leaks the coolant comprises: Inputting the target multimodal features into the long short-term memory network to extract multimodal temporal features; Inputting the multimodal temporal features into the convolutional layer to extract multimodal spatial features; Inputting the multimodal spatial features into the pooling layer to perform a pooling operation to obtain a multimodal pooling feature; Splicing the multimodal time series feature and the multimodal pooling feature into a multimodal cascade feature; Inputting the multimodal cascade features into the fully connected layer and mapping them into the probability of the shell of the cooler being damaged; If the probability is greater than or equal to a preset confidence threshold, it is determined that the cooler leaks the coolant.
9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for detecting coolant leakage of a power battery as described in any one of claims 1 to 8 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for detecting coolant leakage in a power battery as described in any one of claims 1 to 8 is implemented.