Lithium battery life prediction control method, system and detection device

Abstract

The application provides a lithium battery life prediction control method, system and detection equipment, relates to the lithium battery detection field, and utilizes non-invasive ultrasonic imaging technology to acquire an ultrasonic dendrite image of dendrite growth in a battery as original data. The method can directly utilize time sequence change characteristics of the dendrite to acquire a failure risk type under dendrite induction, accurately calculates residual life of the lithium battery under the risk type, realizes real-time detection of the dendrite state, and greatly improves detection precision of the lithium battery detection equipment.

Claims

1. A method for predicting and controlling the lifespan of a lithium battery, characterized in that, The method is applied to the controller of a lithium battery testing equipment, which further includes: a dendrite image generation module, a dendrite feature extraction module, and a battery life prediction module. The method includes: The dendrite image generation module performs time-delay superposition and reconstruction processing on the ultrasonic transmission signal of the target lithium battery to generate an ultrasonic dendrite image of the target lithium battery. A feature extraction operator is constructed based on the prior image features of the target dendrite region in the ultrasonic dendrite image, and the dendrite texture morphology features of the target dendrite region are obtained using the feature extraction operator; a temporal feature extraction operator is constructed using the temporal evolution data of the dendrite texture morphology features, and the temporal feature extraction operator is used to control the dendrite feature extraction module to extract the temporal variation features of the dendrites in the ultrasonic dendrite image; After obtaining the failure risk type of the target lithium battery under dendrite-induced failure based on the dendrite texture morphology features, the battery life prediction module is controlled by the dendrite time-series change features to calculate the remaining life of the target lithium battery under the failure risk type.

2. The lithium battery life prediction and control method according to claim 1, characterized in that, The step of generating an ultrasonic dendrite image of the target lithium battery by controlling the dendrite image generation module to perform time-delay superposition and reconstruction processing on the ultrasonic transmission signal of the target lithium battery includes: According to the preset ultrasonic frequency parameters, the ultrasonic transmitting unit in the dendrite image generation module transmits ultrasonic signals to the target cross-section of the target lithium battery. The ultrasonic receiving unit in the dendrite image generation module is controlled to receive the spindle-shaped ultrasonic transmission signal after the ultrasonic signal passes through the target lithium battery. Based on the ultrasonic frequency parameters, the timing parameters of the spindle-shaped ultrasonic transmission signal are obtained. The timing parameters are used to control the dendrite image generation module to perform time-delay superposition on the ultrasonic transmission signal of the target lithium battery, thereby obtaining the ultrasonic image of the target cross-section in the target lithium battery. The dendrite morphology parameters in the target cross section are determined using the cell structure parameters of the target lithium battery, and the dendrite feature extraction strategy of the ultrasonic image under the dendrite morphology parameters is obtained. The dendrite feature extraction strategy controls the dendrite image generation module to perform feature extraction and reconstruction processing on the ultrasonic image, thereby obtaining the ultrasonic dendrite image of the target lithium battery.

3. The lithium battery life prediction and control method according to claim 1, characterized in that, A feature extraction operator is constructed based on the prior image features of the target dendrite region in the ultrasonic dendrite image, and the dendrite texture morphology features of the target dendrite region are obtained using the feature extraction operator, including: The edge feature morphology parameters, texture feature morphology parameters, and grayscale distribution morphology parameters of the target dendrite region in the ultrasonic dendrite image are determined by the dendrite morphology parameters of the target lithium battery. Based on the edge feature morphology parameters, the texture feature morphology parameters, and the grayscale distribution morphology parameters, the image prior features of the target dendritic region are determined, and the feature extraction operator of the target dendritic region is constructed using the image prior features; The first convolution kernel corresponding to the feature extraction operator is determined using the dendrite morphology parameters, and a feature enhancement image of the ultrasonic dendrite image is generated using the first convolution kernel. After the dendrite feature extraction module extracts features from the dendrite texture in the feature-enhanced image by controlling the feature extraction operator, an image feature vector containing dendrite texture morphology features is obtained.

4. The lithium battery life prediction and control method according to claim 3, characterized in that, A temporal feature extraction operator is constructed using the temporal evolution data of the dendrite texture morphology features, and the temporal feature extraction operator is used to control the dendrite feature extraction module to extract the temporal variation features of dendrites in the ultrasonic dendrite image, including: The shape, size, and distribution parameters of dendrites in the target dendrite region are determined by the dendrite morphology parameters, and the temporal correlation parameters of the dendrites under the dendrite texture morphology features are determined by the shape parameters, size parameters, and distribution parameters. The dendrite texture morphology features are used to obtain the morphological evolution data of the dendrites, and the temporal feature extraction operator of the dendrites in the target dendrite region under the morphological evolution data is constructed through the temporal correlation parameters. The second convolution kernel corresponding to the temporal feature extraction operator is determined using the temporal correlation parameters, and the temporal evolution image of the image feature vector is generated through the second convolution kernel; The temporal feature extraction operator controls the dendrite feature extraction module to extract the temporal variation features of dendrites contained in the temporal evolution image.

5. The lithium battery life prediction and control method according to claim 1, characterized in that, Based on the dendrite texture morphology characteristics, the failure risk type of the target lithium battery under dendrite-induced conditions is obtained, including: Based on the dendrite texture morphology characteristics, the dendrite-induced short-circuit characteristics, dendrite-induced lithium plating characteristics, and dendrite-induced failure characteristics of the target dendrite region in the ultrasonic dendrite image are determined. A failure risk classification strategy for the target lithium battery under dendrite-induced conditions is constructed by using the dendrite-induced short-circuit characteristics, the dendrite-induced lithium plating characteristics, and the dendrite-induced failure characteristics. The failure probability values ​​corresponding to the dendrite-induced short circuit feature, the dendrite-induced lithium plating feature, and the dendrite-induced failure feature under the failure risk classification strategy are calculated using the dendrite texture morphology features. The failure risk type of the target lithium battery under dendrite induction is determined by the dendrite-induced feature type corresponding to the maximum probability value among the failure probability values.

6. The lithium battery life prediction and control method according to claim 5, characterized in that, The failure probability values ​​corresponding to the dendrite-induced short-circuit feature, the dendrite-induced lithium plating feature, and the dendrite-induced failure feature under the failure risk classification strategy are calculated using the dendrite texture morphology features, including: The failure type prediction model for the target lithium battery is determined based on the aforementioned failure risk classification strategy; wherein, the failure type prediction model is a superposition structure of a fully connected layer and a Softmax activation function layer, and the failure type prediction model uses a failure mechanism classification loss function during training. for: ; The failure mechanism label values ​​corresponding to the dendrite-induced short-circuit feature, the dendrite-induced lithium plating feature, and the dendrite-induced failure feature; For the first The true label of the class failure mechanism; The first output of the failure type prediction model The probability of a failure mechanism; To focus on weight parameters; After inputting the image feature vector containing the dendrite texture morphology features into the failure type prediction model, the battery life prediction module is controlled to obtain the dendrite-induced short circuit feature, the dendrite-induced lithium plating feature, and the dendrite-induced failure feature corresponding to the first dendrite-induced failure feature output by the failure type prediction model. Failure probability value of the failure mechanism.

7. The lithium battery life prediction and control method according to claim 5, characterized in that, The battery life prediction module is controlled by the dendrite timing variation characteristics to calculate the remaining life of the target lithium battery under the failure risk type, including: Based on the dendrite temporal variation characteristics, a linear regression activation function corresponding to the target dendrite region under the failure risk type is constructed. The linear regression activation function is used to control the battery life prediction module to obtain the remaining life of the target lithium battery under the failure risk type.

8. The lithium battery life prediction and control method according to claim 7, characterized in that, The linear regression activation function is used to control the battery life prediction module to obtain the remaining life of the target lithium battery under the failure risk type, including: The failure duration prediction model for the target lithium battery is determined based on the linear regression activation function; wherein, the failure duration prediction model is a superposition structure of fully connected layers and linear activation function layers, and the failure duration prediction model uses a failure duration regression loss function during training. Satisfy the following relationship: when hour, ;when hour, ; The actual remaining time of failure for the target lithium battery under the failure risk type; The predicted value of the remaining failure duration predicted by the failure duration prediction model; The preset threshold; The regularization coefficient is used. After inputting the image feature vector containing the dendritic texture morphology features into the failure duration prediction model, the battery life prediction module is controlled to obtain the remaining life of the target lithium battery under the failure risk type output by the failure duration prediction model.

9. A lithium battery life prediction and control system, characterized in that, The system is applied in the controller of a lithium battery testing equipment, which also includes: a dendrite image generation module, a dendrite feature extraction module, and a battery life prediction module. The system includes: Ultrasonic dendrite image generation module: used to control the dendrite image generation module to perform time-delay superposition and reconstruction processing on the ultrasonic transmission signal of the target lithium battery, and generate an ultrasonic dendrite image of the target lithium battery; Dendrite temporal variation feature extraction module: used to construct a feature extraction operator based on the prior image features of the target dendrite region in the ultrasonic dendrite image, and use the feature extraction operator to obtain the dendrite texture morphology features of the target dendrite region; then, construct a temporal feature extraction operator through the temporal evolution data of the dendrite texture morphology features, and use the temporal feature extraction operator to control the dendrite feature extraction module to extract the dendrite temporal variation features in the ultrasonic dendrite image; Lithium battery remaining life prediction module: After obtaining the failure risk type of the target lithium battery under dendrite-induced failure based on the dendrite texture morphology characteristics, the module uses the dendrite time sequence change characteristics to control the battery life prediction module to calculate the remaining life of the target lithium battery under the failure risk type.

10. A lithium battery testing device, characterized in that, The lithium battery testing equipment includes: a dendrite image generation module, a dendrite feature extraction module, a battery life prediction module, and a controller; the controller is connected to the dendrite image generation module, the dendrite feature extraction module, and the battery life prediction module, respectively. The controller includes a processor and a memory, the memory storing computer-executable instructions that can be executed by the processor, and the processor executing the computer-executable instructions to implement the lithium battery life prediction control method mentioned in any one of claims 1 to 8.

Images