BMS Model Optimization via Active Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current battery management system (BMS) algorithm models lack accuracy due to reliance on laboratory test data, which is resource-intensive and fails to account for real-world driving conditions and user habits, leading to inefficiencies in development and performance.
Innovation Solution
The method employs an active learning approach to select and weight sample vehicles from real driving data, optimizing the BMS model based on characteristic similarity and information entropy, reducing the need for laboratory testing and improving model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If laboratory battery test data is used to establish BMS algorithm model, then model parameters can be obtained, but the process consumes large amounts of manpower, material resources and time
Solution Approach 1:
The patent applies preliminary action by collecting and storing real driving data from multiple vehicles in advance through telematics systems. This pre-collected data repository enables subsequent model optimization without requiring time-consuming laboratory tests, allowing the system to leverage previously gathered real-world operating conditions directly for BMS model parameter calibration.
Solution Approach 2:
The patent uses copying by replacing physical laboratory battery tests with virtual data collection from real vehicles equipped with telematics systems. Instead of conducting actual capacity, internal resistance, and power tests in the lab, the system copies real-world battery operating data from multiple vehicles during normal driving, using this copied data to establish and optimize the BMS algorithm model.
2Ease of manufacture
If laboratory battery tests are designed based on experience, then tests can be conducted, but the experiment lacks theoretical guidance and has a certain gap with actual use
Solution Approach 1:
The patent implements feedback by continuously collecting real driving data from multiple vehicles through telematics systems and using this feedback to iteratively optimize the BMS algorithm model. The system compares model predictions with actual battery performance data from real vehicles, adjusting parameters based on this feedback loop to improve accuracy for actual use conditions rather than relying on experience-based laboratory test designs.
Solution Approach 2:
The patent applies parameter changes by transitioning from fixed laboratory test parameters to dynamic real-world operating parameters. Instead of using predetermined laboratory test conditions, the system utilizes actual battery operating parameters collected from vehicles during diverse real driving scenarios, allowing the model to adapt to actual use conditions through parameter variations encountered in real operation.
3Measurement precision
If a large number of battery test data are collected to improve model accuracy, then control accuracy can be improved, but the development progress of BMS system is reduced
Solution Approach 1:
The patent applies universality by creating a multi-functional system that simultaneously collects data for multiple purposes: model parameter calibration, model validation, and continuous optimization. The same telematics infrastructure and real driving data serve multiple functions in the development process, eliminating the need for separate dedicated laboratory testing campaigns and improving overall development efficiency while maintaining model accuracy.
Solution Approach 2:
The patent implements self-service by enabling the BMS model to optimize itself using real driving data collected from vehicles in normal operation. Instead of requiring external laboratory testing resources, the system uses its own operational data from the fleet to automatically calibrate and improve its algorithm parameters, reducing dependency on external testing infrastructure and accelerating development.
Data Source
AI summary
Provided are a method and a system for optimizing a BMS model. The method includes: creating a BMS model based on test data of a battery; selecting sample vehicles from real driving data of vehicles equipped with this type of battery through active learning approach, and giving a corresponding weight to each of the selected sample vehicles; optimizing the BMS model based on the data of the selected sample vehicles.