A data-driven power battery multi-index health evaluation method
By employing a data-driven approach, combining mechanistic and neural network models, the problem of inaccurate state health assessment of power batteries has been solved, enabling rapid and accurate multi-indicator assessment, which is applicable to used car transactions and battery financing lease scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DALIAN INSTITUTE OF CHEMICAL PHYSICS CHINESE ACADEMY OF SCIENCES
- Filing Date
- 2026-05-18
- Publication Date
- 2026-06-12
AI Technical Summary
Current technologies for assessing the state health of power batteries are inaccurate and time-consuming, failing to meet the rapid and non-destructive requirements of used car transactions and battery financing leases. Traditional methods also ignore the impact of the battery's internal microscopic features on its lifespan.
A data-driven approach is adopted, combining mechanistic and neural network models. By collecting real-time and historical data, battery health characteristics are extracted, and a multi-indicator health assessment is performed using a convolutional neural network-bidirectional long short-term memory model to obtain an accurate battery health value.
It achieves accuracy and efficiency in battery health assessment, takes into account microscopic characteristics such as operating temperature environment, avoids data distortion, and has the advantages of being non-destructive and fast.
Smart Images
Figure CN122193948A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of battery health assessment technology, and in particular to a data-driven multi-index health assessment method for power batteries. Background Technology
[0002] With the rapid development of the new energy vehicle market, the demand for aftermarket businesses such as used car transactions, battery financing leases, and cascade utilization has surged. However, as the core and most expensive component of new energy vehicles, the accurate assessment of the state of health (SOH) of power batteries has always faced significant technical bottlenecks. First, the assessment of battery SOH suffers from a serious "black box" phenomenon. Traditional assessment methods mainly rely on simple charge-discharge tests (capacity grading method), which are extremely time-consuming and require disassembling the battery pack, failing to meet the needs of fast and non-destructive on-site operations in used car transactions and battery financing leases; or they rely solely on the SOH value reported by the BMS, but due to significant differences in BMS algorithms and sensor aging, this value is often distorted and cannot reflect the true condition of the vehicle. Existing residual value assessment systems lack multi-dimensional technical indicators. Current assessments are mostly based on macroscopic data such as vehicle age and mileage for rough depreciation, ignoring the profound impact of the battery's internal microscopic characteristics on battery life, resulting in large errors in the assessment results.
[0003] Therefore, a multi-index health assessment method for power batteries that combines accuracy and detection efficiency is needed. Summary of the Invention
[0004] In view of this, the present invention provides a data-driven multi-index health assessment method for power batteries, which integrates the battery state health calculated by the mechanism model and the battery state health estimated by the neural network model to obtain an accurate assessment value of the battery state health; thus solving the problem of inaccurate assessment of power battery state health in the prior art.
[0005] Therefore, the present invention provides the following technical solution:
[0006] A data-driven multi-index health assessment method for power batteries, comprising: Collect data on the vehicle's power battery; Battery health characteristics are extracted using the power battery data; The theoretical value of battery health is obtained through a mechanism model based on the power battery data. Based on the battery state health characteristics, an estimated value of battery state health is obtained through a neural network model; The theoretical value and the estimated value of battery health are weighted and summed to obtain the battery health status assessment value.
[0007] Furthermore, the power battery data includes: Real-time status data and historical statistics; The real-time status data includes: total battery pack voltage, total current, insulation resistance, a list of voltages for all individual cells, and a list of battery pack temperatures. The historical statistics include: cumulative mileage, cumulative charge / discharge capacity, cumulative fast charge counts, and cumulative slow charge counts.
[0008] Furthermore, the battery health characteristics include: individual cell voltage range, historical fast charging percentage, and temperature distribution entropy.
[0009] Furthermore, the neural network model includes: Convolutional Neural Network - Bidirectional Long Short-Term Memory Model.
[0010] Furthermore, based on the power battery data, a theoretical value for the battery's state of health is obtained through a mechanistic model, including: A second-order RC equivalent circuit model is constructed based on the measured voltage and current responses, and the current ohmic internal resistance and polarization internal resistance are obtained by the least squares method. The theoretical value of battery health is calculated based on the mapping relationship between internal resistance growth rate and battery health.
[0011] Furthermore, the temperature distribution entropy:
[0012] in, For the first The probability density of a temperature sampling point in the overall temperature distribution; This represents the total number of temperature sampling points. Temperature distribution entropy reflects the uniformity of the temperature field inside the battery pack.
[0013] Advantages and positive effects of the present invention: This method obtains the theoretical value of the state health of the power battery through a mechanistic model, and estimates the state health of the power battery through a neural network model. The final state health value of the power battery is obtained by combining the results of the two methods, thus avoiding inaccurate assessments caused by data distortion and other problems.
[0014] This method, when estimating the state health value of a power battery using a neural network model, integrates consistency indicators, behavioral indicators, and thermal management indicators, and considers the profound impact of micro-characteristics such as operating temperature environment on battery life, making the estimation results more accurate.
[0015] This method uses only directly collectable data and does not require disassembling the battery pack, offering the advantages of being non-destructive and fast. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 Flowchart of a data-driven multi-index health assessment method for power batteries. Detailed Implementation
[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0020] This invention provides a data-driven multi-index health assessment method for power batteries. Based on operational data from new energy vehicles at different stages of their life cycle, it achieves accurate calculation and value quantification of battery SOH and RUL (Remaining useful life) through portable non-destructive data acquisition and the fusion of a "mechanism and data" dual-driven model. First, deep data from the BMS (Battery Management System) is read via the vehicle's OBD (On-Board Diagnostics) interface to form a raw dataset containing the micro-state and historical behavior of the battery cells. Then, key feature parameters are extracted through feature engineering, including the single-cell voltage range, internal resistance change rate, temperature distribution entropy, and historical fast-charging percentage, to construct a battery health profile. Based on this, a multi-model fusion calculation is performed using ECM (Equivalent Circuit Model) and CNN-BiLSTM (Convolutional Neural Network-Bidirectional Long Short-Term Memory) to obtain a high-precision SOH estimate.
[0021] Combination Figure 1 As shown, a data-driven multi-index health assessment method for power batteries includes: S1. Collect and preprocess the vehicle's power battery data; S2. Extract battery health characteristics from the preprocessed data; S3. Obtain the theoretical value of battery health status through a mechanism model based on power battery data; S4. Based on the battery state health characteristics, obtain the estimated value of battery state health through a neural network model; S5. The theoretical value and the estimated value of battery health are weighted and summed to obtain the battery health status assessment value.
[0022] Example 1 A data-driven multi-index health assessment method for power batteries, comprising: S1. Collect and preprocess the vehicle's power battery data to obtain a complete dataset reflecting the battery's current physical state and historical aging process. 1) Collect real-time vehicle status data and historical statistical data through the vehicle's CAN (Controller Area Network) bus protocol; Real-time status data, including: List of total voltage, total current, insulation resistance, voltage of all individual cells, and temperature of the battery pack; Historical statistics include: cumulative mileage, cumulative charge / discharge capacity, cumulative fast charge counts, and cumulative slow charge counts.
[0023] 2) Clean and align the dataset data.
[0024] S2. Extract battery health characteristics from the preprocessed data; Battery health characteristics include: consistency indicators, behavioral indicators, and thermal management indicators.
[0025] 1) Using the individual unit voltage range as a consistency indicator, the formula is expressed as:
[0026] in, For the first i The voltage value of each battery cell; This refers to the voltage range of a single unit; This reflects the situation of the weakest link cell in the "barrel effect." The greater the range, the more severely the overall performance of the battery pack is limited, and the higher the risk of SOH degradation.
[0027] 2) The historical fast charging percentage is used as a behavioral indicator, expressed by the formula:
[0028] in, To accumulate fast charging times, To accumulate the number of slow charging cycles, This represents the historical percentage of fast charging.
[0029] 3) Using temperature distribution entropy as a thermal management indicator, the formula is expressed as:
[0030] in, For the first The probability density of a temperature sampling point in the overall temperature distribution; Temperature distribution entropy reflects the uniformity of the temperature field inside the battery pack. The higher the entropy value, the more disordered the temperature difference distribution and the more uneven the battery aging.
[0031] S3. Obtain the theoretical value of battery health status through a mechanism model based on battery data; 1) Construct a second-order RC equivalent circuit model based on the measured voltage and current response, and obtain the current ohmic internal resistance and polarization internal resistance through the least squares method.
[0032] 2) The theoretical value of battery state health is calculated based on the mapping relationship between internal resistance growth rate and battery state health.
[0033] S4. Based on the battery state health characteristics, obtain the battery life estimate and battery state health estimate through a neural network model; The battery state health characteristics are used as input and the battery lifetime estimate and battery state health estimate are output through a convolutional neural network-bidirectional long short-term memory model (CNN-BiLSTM).
[0034] S5. The theoretical and estimated battery state health values are weighted and summed to obtain the comprehensive battery state health value, expressed by the following formula:
[0035] in, This represents the theoretical value for battery state health. This is an estimate of the battery's state of health. This represents the overall battery health value. These are preset weighting coefficients.
[0036] Example 2 This embodiment obtains the comprehensive battery state health value and the estimated battery life value based on the method in Embodiment 1; The suggested value of the remaining battery value is calculated based on the comprehensive value of battery health and market supply and demand.
[0037] in, This is a suggested value for the remaining battery value. This is a market supply and demand adjustment coefficient. λ This refers to the fast charging loss factor. The value of recycled materials; For functional value, according to calculate; This represents the historical percentage of fast charging.
[0038] Example 3 A data-driven multi-index health assessment device for power batteries can be connected to the vehicle's OBD interface and automatically identify the vehicle's CAN bus protocol through a multi-protocol adaptation engine. Under static or short-term dynamic operating conditions, it actively queries and reads deep data from the BMS (Battery Management System). Based on the collected data, it outputs a comprehensive battery health value and a suggested remaining battery value.
[0039] Example 4 In the used car market, appraisers first use the portable intelligent diagnostic tool of this invention, inserted into the OBD port of the vehicle to be sold. Without disassembling the battery pack, the device completes a full read of the BMS data within 2-3 minutes using a proprietary communication protocol. The data includes the vehicle's current cumulative mileage, total number of charge / discharge cycles, the percentage of historical fast charging cycles, and a list of the individual voltages of the 96 battery cells at the current moment.
[0040] The collected data is uploaded to the cloud-based evaluation platform in real time. The feature extraction module in the cloud immediately analyzes the data and finds that although the vehicle's odometer reading is only 50,000 kilometers, the voltage of its No. 12 battery cell deviates from the average voltage by more than 50mV, and historical data indicates that the vehicle has been in a high-temperature fast-charging state for a long time. Based on these characteristics, a multi-model fusion algorithm calculates that the battery's actual SOH is only 82%, far lower than the theoretical value of 92% calculated based on mileage, and estimates that its RUL will rapidly degrade under current usage habits.
[0041] First, the "functional value" of the battery is calculated based on its 82% SOH (State of Harmony). Then, the "material recycling value" is calculated as a safety net, taking into account the current lithium carbonate recycling market. Finally, an objective residual value for the battery is derived. The system automatically generates a standardized report that includes "risk of cell inconsistency," "recommendation to reduce fast charging frequency," and "recommendation of a battery value of 35,000 yuan."
[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A data-driven multi-index health assessment method for power batteries, characterized in that, include: Collect data on the vehicle's power battery; Battery health characteristics are extracted using the power battery data; The theoretical value of battery health is obtained through a mechanism model based on the power battery data. Based on the battery state health characteristics, an estimated value of battery state health is obtained through a neural network model; The theoretical value and the estimated value of battery health are weighted and summed to obtain the battery health value.
2. The method according to claim 1, characterized in that, The power battery data includes: Real-time status data and historical statistics; The real-time status data includes: total battery pack voltage, total current, insulation resistance, a list of voltages for all individual cells, and a list of battery pack temperatures. The historical statistics include: cumulative mileage, cumulative charge / discharge capacity, cumulative fast charge counts, and cumulative slow charge counts.
3. The method according to claim 1, characterized in that, The battery health characteristics include: individual cell voltage range, historical fast charging percentage, and temperature distribution entropy.
4. The method according to claim 1, characterized in that, The neural network model includes: Convolutional Neural Network - Bidirectional Long Short-Term Memory Model.
5. The method according to claim 1, characterized in that, Based on the power battery data, a theoretical value for battery health is obtained through a mechanistic model, including: A second-order RC equivalent circuit model is constructed based on the measured voltage and current responses, and the current ohmic internal resistance and polarization internal resistance are obtained by the least squares method. The theoretical value of battery health is calculated based on the mapping relationship between internal resistance growth rate and battery health.
6. The method according to claim 3, characterized in that, The temperature distribution entropy: in, For the first The probability density of a temperature sampling point in the overall temperature distribution; This represents the total number of temperature sampling points. Temperature distribution entropy reflects the uniformity of the temperature field inside the battery pack.