Data-Based State of Health Model for Battery Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the state of health of electrical energy stores, such as vehicle batteries, are inaccurate due to reliance on physical ageing models, leading to unreliable predictions of remaining capacity and ageing, which is crucial for assessing residual value and operational efficiency.
Innovation Solution
A data-based state of health model is implemented in a central processing unit, trained using operating features from multiple devices, which selects key operating feature points for measurement to improve prediction accuracy, utilizing probabilistic regression models and active learning methods to reduce uncertainties and require minimal user intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a physical ageing model is used to determine state of health, then the method is simple to implement, but the prediction accuracy is low with model discrepancies of up to more than 5%
Solution Approach 1:
The patent introduces an intermediary data-based correction model that mediates between the simple physical ageing model and the actual state of health. This correction model, trained on measurement data from multiple devices, compensates for the inaccuracies of the physical model without requiring complete replacement of the simple model architecture.
Solution Approach 2:
The patent transforms the state of health determination from relying solely on physical parameters in a deterministic model to using statistical parameters and probability distributions in a data-based model. This allows the system to adapt to varying operating conditions and reduce prediction discrepancies.
2Measurement precision
If measurements are performed frequently to improve prediction accuracy, then the state of health data becomes more reliable, but the number of label generation operations and user intervention increases
Solution Approach 1:
The patent performs preliminary actions by pre-training the data-based correction model using measurement data from a plurality of devices during normal operation. This preparatory phase allows the model to learn patterns and relationships, so that future state of health predictions can be made with fewer additional measurements and less user intervention.
Solution Approach 2:
The patent implements a feedback mechanism where measurement results from selected devices are used to retrain and improve the data-based correction model. This continuous feedback loop allows the system to progressively improve prediction accuracy while minimizing the frequency of new measurements required.
3Measurement precision
If a data-based state of health model is trained on data from multiple devices, then the prediction accuracy improves, but the data processing and model training complexity increases
Solution Approach 1:
The patent creates a universal data-based correction model that can be applied across multiple devices with different energy stores. The model learns from aggregated data from a plurality of devices and can generalize to predict state of health for various device types, making the complex data processing effort worthwhile through broad applicability.
4Reliability
If the state of health model is continuously retrained with new measurement data, then the model remains accurate over time, but the computational resources and processing time increase
Solution Approach 1:
The patent implements periodic retraining of the data-based correction model at predetermined intervals or when triggered by specific conditions, rather than continuous retraining. This approach maintains model reliability over the service life of energy stores while significantly reducing computational energy consumption compared to continuous updates.
Data Source
AI summary
A method for operating a central processing unit which is communicatively connected to a plurality of devices having electrical energy stores, includes determining a data-based state of health model which is trained to assign a state of health to an operating feature point which characterizes operation of a corresponding electrical energy store of the electrical energy stores of the plurality of devices and results from a plurality of operating features, determining operating feature points for the electrical energy stores of the plurality of devices, and selecting at least one of the operating feature points based on state uncertainties of the states of health of all operating feature points of the corresponding electrical energy store, such that the at least one operating feature point has a highest relevance for improving the state uncertainties of the operating feature points determined.