New energy automobile battery parameter detection method and system
By arranging a variety of sensors on the battery pack of new energy vehicles and combining deep learning models, the problem of uncatched impacts of battery dynamic characteristics and environmental factors in traditional detection methods is solved, and accurate evaluation and early warning of battery status is achieved, improving the service life of the battery and the safety and reliability of the car.
Patent Information
- Application Number
- CN202510929013.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-07
- Publication Date
- 2025-08-15
AI Technical Summary
Traditional new energy vehicle battery parameter detection methods are difficult to fully capture the dynamic characteristics of batteries under complex operating conditions, the impact of environmental factors has not been included in the evaluation, and the problem of inconsistency of single batteries lacks effective monitoring, resulting in large deviations in the detection results, affecting the safety and reliability of the automobile.
Multi-sensor fusion and deep learning model are adopted, and multiple sensors are arranged on the battery pack to collect data in real time, combined with adaptive Kalman filtering and linear normalization algorithm for pre-processing, and long-term memory networks are used to establish a battery parameter model, and feature fusion is performed through attention mechanisms, setting a threshold range for state evaluation.
It realizes accurate estimation of the remaining battery power and health status of the battery, improves the accuracy and real-time detection, and issues warnings in a timely manner to ensure the safe use of the battery, extend battery life and improve vehicle safety and reliability.
Smart Images
Figure CN120490842A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of new energy vehicles, and particularly relates to a method and system for detecting battery parameters of new energy vehicles. Background Art
[0002] With the rapid development of the new energy vehicle industry, power batteries, as core components of vehicles, accurately testing and evaluating their performance is crucial for vehicle safety, endurance, and service life. Traditional methods for testing battery parameters in new energy vehicles have numerous limitations. Firstly, some methods rely on single-sensor data, making it difficult to fully capture the dynamic characteristics of batteries under complex operating conditions. For example, estimating the remaining charge based solely on voltage data is susceptible to interference from factors such as battery aging and temperature fluctuations, resulting in significant deviations in the estimated results. Secondly, commonly used algorithms such as the Kalman filter lack adaptability to the nonlinear and time-varying nature of battery systems and are unable to accurately track the chemical reactions within the battery, making the remaining charge (SOC) and state of health (SOH) estimations insufficiently accurate. Furthermore, existing technologies often overlook the impact of environmental factors on battery performance. Battery performance can significantly change under extreme conditions, such as high and low temperatures. Traditional testing methods fail to incorporate key parameters such as ambient temperature into their evaluation, resulting in a disconnect between test results and actual conditions. At the same time, there is a lack of effective monitoring methods for the inconsistency of individual cells in the battery pack. The performance differences of individual cells will accelerate the aging of the entire battery pack, and existing detection methods make it difficult to identify and intervene in this problem early, seriously affecting the reliability and safety of new energy vehicles. Summary of the Invention
[0003] The present invention provides a high-precision, real-time new energy vehicle battery parameter detection method and system, which realizes accurate estimation and status assessment of the battery's remaining power and health status through multi-sensor fusion and deep learning model.
[0004] The technical solutions of the present invention are as follows: A method for detecting parameters of a new energy vehicle battery comprises the following steps: A voltage sensor, current sensor, and temperature sensor are placed on each cell of a new energy vehicle battery pack. A total voltage sensor and a total current sensor are set in the total circuit of the battery pack. An environmental sensor is placed near the battery pack to collect real-time data on the voltage, current, and temperature of the cells, the total voltage, total current, and ambient temperature of the battery pack. Preprocessing the collected data transmission; A battery parameter model is established using a long short-term memory network, with the preprocessed data used as input features and the remaining power and health status obtained from actual measurements used as labels to train the battery parameter model; The pre-processed real-time data is input into the established battery parameter model to calculate the remaining power and health status of the battery in real time, and the battery status is evaluated in combination with the preset threshold range.
[0005] Furthermore, the preprocessing is specifically as follows: The collected data is filtered using an adaptive Kalman filter algorithm, and the filtered data is normalized using a linear normalization algorithm.
[0006] Furthermore, the voltage sensor is a differential voltage sensor, the current sensor is a Hall effect current sensor, the temperature sensor is a digital temperature sensor, and the environmental sensor includes a temperature sensor.
[0007] Furthermore, the preprocessed data is used as input features, and the remaining power and health status obtained by actual measurement are used as labels. The battery parameter model is trained as follows: Extract the characteristics of the single cell voltage, current and temperature data according to the battery pack topology and construct the single cell feature vector; The total voltage and current data of the battery pack are integrated with the ambient temperature data to construct a system-level feature vector; The single cell feature vector and the system-level feature vector are used together as input features, and the remaining power and health status obtained by actual measurement are used as labels to train the model.
[0008] Furthermore, the state assessment step includes: Set the threshold range for remaining power and health status; Compare the remaining power calculated in real time with the corresponding threshold: if it is lower than the low power warning threshold, a charging reminder is triggered; if it is higher than the full power threshold, an overcharge warning is triggered; The health status calculated in real time is compared with the corresponding threshold: if it is lower than the performance degradation warning threshold, it indicates that the battery performance has deteriorated and maintenance is required; if it is lower than the failure threshold, a battery replacement alarm is triggered.
[0009] Furthermore, the attention weighted fusion is specifically as follows: Perform spatial attention calculation on the single cell feature vector output by the first input branch to generate the importance weight of each single cell feature; Perform channel attention calculation on the system-level feature vector output by the second input branch to generate the importance weights of each dimension of the system feature; The two attention weights are applied to the corresponding feature vectors respectively, and then the weighted feature vectors are merged through the tensor splicing operation to obtain the fused feature representation.
[0010] Furthermore, the battery parameter model adopts a multi-input branch structure, including: A first input branch receives the single cell feature vector; A second input branch receives the system-level feature vector; Feature fusion layer, which performs attention-weighted fusion on the outputs of the two input branches; The bidirectional LSTM layer and the fully connected layer output the remaining power and health status prediction values.
[0011] The present invention also provides a new energy vehicle battery parameter detection system, the system comprising: The sensor module includes voltage sensors, current sensors, and temperature sensors arranged on each single cell, a total voltage sensor and a total current sensor arranged in the main circuit, and an environmental sensor arranged near the battery pack, which is used to collect real-time data on the voltage, current, and temperature of the single cell, the total voltage and current data of the battery pack, and the ambient temperature data; The data processing module is used to pre-process the collected data, including filtering using an adaptive Kalman filter algorithm and normalizing using a linear normalization algorithm; A model training module is used to establish a battery parameter model using a long short-term memory network, using the preprocessed data as input features and the actual measured remaining charge and health status as labels to train the battery parameter model; The status assessment module is used to input the pre-processed real-time data into the established battery parameter model, calculate the remaining power and health status of the battery in real time, and evaluate the battery status in combination with the preset threshold range.
[0012] Compared with the prior art, the present invention has the following advantages: The present invention achieves real-time and accurate collection of key battery parameters and environmental data by rationally arranging multiple sensors on the individual cells and overall circuit of a new energy vehicle battery pack. It utilizes an adaptive Kalman filter and linear normalization algorithm for data preprocessing, effectively improving data quality. It employs a multi-input branch-structured long-short-term memory network combined with an attention mechanism to train a battery parameter model, significantly improving the accuracy of remaining power and health status predictions. It also performs status assessments and issues timely warnings by setting threshold ranges to ensure safe battery use. Compared to traditional detection methods, the present invention can more comprehensively and accurately grasp battery status, providing strong support for the efficient management and safe operation of new energy vehicle batteries, effectively extending the battery's service life and the safety and reliability of vehicle use. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] The accompanying drawings illustrate various embodiments generally by way of example and not limitation, and together with the description and claims, serve to explain embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive of the embodiments of the present apparatus or method.
[0014] Figure 1 A schematic flow chart of the method of the present invention is shown. DETAILED DESCRIPTION
[0015] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0016] like Figure 1 As shown, the present invention provides a method for detecting parameters of a new energy vehicle battery, comprising: First, a voltage sensor (for collecting cell voltage data), a current sensor (for collecting cell current data), and a temperature sensor (for collecting cell temperature) are placed on each cell in the new energy vehicle battery pack. A total voltage sensor and a total current sensor are installed in the battery pack's main circuit to monitor the operation of the entire battery pack. Furthermore, environmental sensors are placed around the battery pack to collect ambient temperature data, a key environmental parameter that affects battery performance.
[0017] Data preprocessing An adaptive Kalman filter algorithm is used to filter the collected data. The filter parameters are automatically adjusted based on the dynamic changes in the data, removing noise interference and improving data accuracy and reliability. The filtered data is normalized using a linear normalization algorithm, mapping the data uniformly to a specific numerical range.
[0018] Battery parameter model training Feature construction The pre-processed single-cell voltage, current, and temperature data are then subjected to feature extraction based on the battery pack topology. For example, for a series battery pack, statistical features such as the mean and standard deviation of each single-cell data can be extracted to construct a single-cell feature vector. The total voltage and total current data of the battery pack are integrated with the ambient temperature data, for example, they are directly spliced into a vector to construct a system-level feature vector. A long short-term memory (LSTM) network is used to build a battery parameter model. This model employs a multi-input branch structure: the first input branch receives individual battery feature vectors, and the second input branch receives system-level feature vectors. The outputs of these two input branches enter the feature fusion layer, where spatial attention is applied to the individual battery feature vectors from the first input branch to generate importance weights for each individual battery feature. Channel-wise attention is applied to the system-level feature vectors from the second input branch to generate importance weights for each dimension of the system feature. The two attention weights are then applied to the corresponding feature vectors. The weighted feature vectors are then merged using a tensor concatenation operation to produce a fused feature representation. The fused features are then passed through a bidirectional LSTM layer and a fully connected layer to output predicted remaining charge and state of health (SOH). The model uses measured remaining charge and SOH as labels and is trained using extensive historical data to optimize model parameters, enabling it to accurately predict the remaining charge and SOH based on the input data.
[0019] Battery status assessment Threshold setting Preset threshold ranges for remaining battery and health status. For example, the low battery warning threshold for remaining battery is set to 20%, and the full battery threshold is set to 95%. The performance degradation warning threshold for health status is set to 80%, and the failure threshold is set to 70%. Real-time evaluation The preprocessed real-time data is input into the established battery parameter model to calculate the battery's remaining charge and health status in real time. The real-time calculated remaining charge is compared with the corresponding threshold. If it is below the low-battery warning threshold, a charging reminder is triggered, prompting the user to charge the vehicle via the vehicle dashboard or mobile app. If it is above the full-charge threshold, an overcharge warning is triggered, reminding the user to stop charging immediately to avoid battery overcharge damage. The real-time calculated health status is compared with the corresponding threshold. If it is below the performance degradation warning threshold, the battery performance is degraded and maintenance is required, and the user is advised to visit a professional repair station for battery inspection and maintenance. If it is below the failure threshold, a battery replacement alarm is triggered, clearly informing the user that the battery is no longer functioning properly and needs to be replaced.
[0020] The above is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, can make equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, which should be covered by the scope of protection of the present invention.
Claims
1. A method for detecting parameters of a new energy vehicle battery, characterized in that: The steps include: A voltage sensor, a current sensor, and a temperature sensor are arranged on each single cell of the new energy vehicle battery pack. A total voltage sensor and a total current sensor are set in the total circuit of the battery pack. At the same time, an environmental sensor is arranged to collect the voltage, current, and temperature data of the single cell, the total voltage, total current data of the battery pack, and the ambient temperature data in real time; Preprocessing the collected data transmission; A battery parameter model is established using a long short-term memory network, with the preprocessed data used as input features and the remaining power and health status obtained from actual measurements used as labels to train the battery parameter model; The pre-processed real-time data is input into the established battery parameter model to calculate the remaining power and health status of the battery in real time, and the battery status is evaluated in combination with the preset threshold range.
2. The new energy vehicle battery parameter detection method according to claim 1, characterized in that: The pre-processing is specifically as follows: The collected data is filtered using an adaptive Kalman filter algorithm, and the filtered data is normalized using a linear normalization algorithm.
3. The new energy vehicle battery parameter detection method according to claim 1, characterized in that: The voltage sensor adopts a differential voltage sensor, the current sensor adopts a Hall effect current sensor, the temperature sensor adopts a digital temperature sensor, and the environmental sensor includes a temperature sensor.
4. The new energy vehicle battery parameter detection method according to claim 1, characterized in that: The preprocessed data is used as input features, and the remaining power and health status obtained from actual measurements are used as labels. The battery parameter model is trained as follows: Extract the characteristics of the single cell voltage, current and temperature data according to the battery pack topology and construct the single cell feature vector; The total voltage and current data of the battery pack are integrated with the ambient temperature data to construct a system-level feature vector; The single cell feature vector and the system-level feature vector are used together as input features, and the remaining power and health status obtained by actual measurement are used as labels to train the model.
5. The new energy vehicle battery parameter detection method according to claim 1, characterized in that: The state assessment step includes: Set the threshold range for remaining power and health status; Compare the remaining power calculated in real time with the corresponding threshold: if it is lower than the low power warning threshold, a charging reminder is triggered; if it is higher than the full power threshold, an overcharge warning is triggered; The health status calculated in real time is compared with the corresponding threshold: if it is lower than the performance degradation warning threshold, it indicates that the battery performance has deteriorated and maintenance is required; if it is lower than the failure threshold, a battery replacement alarm is triggered.
6. The new energy vehicle battery parameter detection method according to claim 1, characterized in that: The battery parameter model adopts a multi-input branch structure, including: A first input branch receives the single cell feature vector; A second input branch receives the system-level feature vector; Feature fusion layer, which performs attention-weighted fusion on the outputs of the two input branches; The bidirectional LSTM layer and the fully connected layer output the remaining power and health status prediction values.
7. The new energy vehicle battery parameter detection method according to claim 6, characterized in that: The attention weighted fusion is specifically as follows: Perform spatial attention calculation on the single cell feature vector output by the first input branch to generate the importance weight of each single cell feature; Perform channel attention calculation on the system-level feature vector output by the second input branch to generate the importance weights of each dimension of the system feature; The two attention weights are applied to the corresponding feature vectors respectively, and then the weighted feature vectors are merged through the tensor splicing operation to obtain the fused feature representation.
8. A new energy vehicle battery parameter detection system, characterized in that: The system comprises: The sensor module includes voltage sensors, current sensors, and temperature sensors arranged on each single cell, a total voltage sensor and a total current sensor arranged in the main circuit, and an environmental sensor arranged near the battery pack, which is used to collect real-time data on the voltage, current, and temperature of the single cell, the total voltage and current data of the battery pack, and the ambient temperature data; A data transmission module, used to transmit the data collected by the sensor module to the data processing module; A data processing module is used to pre-process the transmitted data, including filtering using an adaptive Kalman filter algorithm and normalizing using a linear normalization algorithm; A model training module is used to establish a battery parameter model using a long short-term memory network, using the preprocessed data as input features and the actual measured remaining charge and health status as labels to train the battery parameter model; The status assessment module is used to input the pre-processed real-time data into the established battery parameter model, calculate the remaining power and health status of the battery in real time, and evaluate the battery status in combination with the preset threshold range.