Battery Health Prediction Using Machine Learning Correction
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, leading to unreliable predictions of remaining capacity and aging, which is crucial for financial assessment and operational efficiency.
Innovation Solution
A data-based state of health model is implemented, combining usage patterns and dynamic models to generate load variables, which are then used to predict the state of health, incorporating a hybrid model that corrects physical health model errors with probabilistic or artificial-intelligence-based regression models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a physical state of health model is used to determine the present state of health of the energy store, then the state of health can be ascertained, but the prediction is inaccurate with model errors of up to more than 5%
Solution Approach 1:
The patent introduces an intermediary correction model trained with machine learning methods that acts as a mediator between the physical state of health model and the actual state of health. This correction model learns from historical data to compensate for the inaccuracies of the physical model, thereby improving prediction accuracy while maintaining the benefits of the physical model's structure.
Solution Approach 2:
The patent transforms the state of health determination from relying solely on physical model parameters to incorporating data-driven parameters learned through machine learning. By changing from a purely physics-based parameter set to a hybrid set including empirically learned parameters, the system achieves higher accuracy in predicting state of health.
2Measurement precision
If sensors are installed in proximity to the energy store to directly measure state of health, then measurement accuracy improves, but manufacturing cost and complexity increase
Solution Approach 1:
The patent replaces the mechanical/sensor-based measurement system with a computational model. Instead of using physical sensors to directly measure state of health parameters, the system uses a data-based correction model that processes existing operational data to infer state of health, thereby eliminating the need for additional complex sensor infrastructure.
Solution Approach 2:
The patent creates a virtual copy of the state of health measurement function through machine learning models. Rather than physically measuring state of health with sensors, the system creates a computational representation that mirrors what direct measurement would provide, using trained models to replicate the measurement function software-based.
3Loss of information
If the physical health model is used to indicate present state of health, then current status is known, but prediction of future state of health based on usage behavior is inaccurate
Solution Approach 1:
The patent implements feedback mechanisms where the correction model continuously learns from historical state of health data and actual outcomes. By feeding back actual state of health measurements and usage patterns into the machine learning model, the system refines its predictions over time, enabling accurate forecasting of future state of health based on observed usage behaviors.
Solution Approach 2:
The patent performs preliminary training of the correction model using historical data before actual prediction tasks. By pre-training the model with extensive historical state of health and usage behavior data, the system prepares the prediction engine in advance, enabling it to accurately predict future state of health for different usage scenarios before they actually occur.
Data Source
AI summary
A computer-implemented method for predicting a modeled state of health of an electrical energy store having at least one electrochemical unit, in particular a battery cell, or by according to rule-based and/or data-based mapping even in an entire system. The method including providing a data-based state of health model trained to assign a modeled state of health to the electrical energy store based on characteristics of operating variables of the electrical energy store; generating a characteristic of at least one load variable based on a provided usage pattern using a usage model; generating the characteristics of operating variables based on the at least one load variable using a predefined dynamic model; and determining a predicted modeled state of health based on the generated characteristics of operating variables.