System for analyzing and predicting the condition of electric vehicle traction batteries during a direct current (DC) fast charging process
PIS measurement via a charging column adapter predicts battery state using machine learning, addressing the limitations of existing systems by providing fast and cost-effective analysis without disassembly or specialized equipment.
Patent Information
- Application Number
- DE202025000976
- Authority / Receiving Office
- DE · DE
- Patent Type
- Utility models
- Current Assignee / Owner
- Filing Date
- 2025-04-23
- Publication Date
- 2025-07-10
- Estimated Expiration
- 2035-04-30
AI Technical Summary
Existing vehicle battery analysis systems provide only feedback on current capacity and energy availability without predicting remaining life, requiring complex and costly EIS measurements that necessitate battery disassembly and specialized equipment.
A pulse impedance spectroscopy (PIS) measurement using a measurement adapter between a charging column and electric vehicle generates battery data, processed by machine learning to predict battery state without disassembly or specialized equipment.
Enables fast, cost-effective battery state prediction directly from a charging station, integrating battery analysis into workshops without specialized knowledge or equipment, and allowing the battery to remain in the vehicle.
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
[0001] Currently, various solutions for analyzing traction batteries are available from vehicle manufacturers and diagnostic companies. Data is acquired by contacting the vehicle's own battery management system during a (partial) charging process or a test drive. However, unlike this invention, raw data is not captured and further processed; instead, data from the vehicle's own control unit is used. The results of these existing systems only provide feedback on the currently available remaining capacity when the vehicle is fully charged, or the currently available amount of energy when the vehicle is fully charged. No feedback is provided on the remaining service life (condition prediction).
[0002] Currently, an independent battery analysis can be performed without vehicle-specific data using a complex and time-consuming electrochemical impedance spectroscopy (EIS) measurement. This generated data can then be used, for example, to generate a condition prediction using machine learning methods. EIS measurement currently only works with the traction battery removed and requires expensive measurement technology. Performing the measurement also requires a special qualification level for working with high voltage, test bench expertise, and additional information to adjust certain test parameters, such as the frequency of alternating current.
[0003] The invention enables a public charging station, and the associated characteristic start-up process of the rapid charging process, to be used to generate and record data. Direct current instead of alternating current is used for the measurement method. This results in a pulse impedance spectroscopy (PIS) measurement rather than an EIS measurement. A developed measuring adapter, which is connected between the charging station (10) and the electric vehicle (4), generates the battery data. The battery data from the PIS measurement is then processed in a machine learning model, as described in claim 1, to obtain a condition prediction of the traction battery.
[0004] An example of the measuring adapter is shown in Fig. to see.
[0005] This invention requires no special measuring equipment or knowledge for measurement, but rather a standard DC charging station. Because the charging station is used as part of the measurement technology, battery analysis can also be easily integrated into inspection locations, such as workshops or testing organizations. The traction battery can also remain in the vehicle, and the invention represents a test that requires no disassembly or dismantling. In contrast to EIS measurements, PIS measurements are also faster and do not require any additional qualifications from the person performing them. List of reference symbols 1 LAN socket 2 Measurement technology 3 DIN rails 4 cable entry 5 current transformers 6 Current measuring module 7 Voltage measuring module 8 laboratory sockets 9 rocker switches 10 charging socket 11 Locking actuator
Claims
[1] A trained and validated machine learning model, into which the generated battery data from the pulse impedance spectroscopy measurement are fed, characterized by the fact that the model is cloud-based and can be connected via LAN (1) or WLAN to the measuring adapter, in Fig. , is connected. Between them is the measurement technology (2), which processes the data in a first step and serves as an interface to the cloud. The model includes a specially developed data processing procedure to produce a value for lifetime prediction.