Battery RUL Forecasting With Feedback-Corrected Capacity Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing battery health estimation methods are device-specific, lack a self-correcting mechanism, and fail to detect anomalies, leading to inaccurate and computationally intensive predictions of Remaining Useful Life (RUL).

Innovation Solution

A method and electronic device that measures battery capacity values for each charging and discharging cycle, uses a battery capacity estimation model and data-driven model to forecast RUL, detects anomalies, and corrects forecasts through feedback, requiring minimal computation and device-agnostic implementation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple machine learning models are trained and compared to select the best RUL forecasting model, then forecasting accuracy is improved, but training time and computational resources increase

Engineering Contradiction:
ImproveRUL forecasting accuracyVSAvoidmodel training time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training multiple candidate machine learning models (LSTM, GRU, CNN-LSTM, etc.) before actual RUL forecasting tasks. These pre-trained models are stored in a model library and can be directly applied to new battery datasets without requiring full retraining, thus significantly reducing the time and computational resources needed for deployment while maintaining high forecasting accuracy

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent employs parameter changes by adjusting hyperparameters such as learning rate, batch size, sequence length, and model architecture parameters to optimize the performance of different machine learning models. By systematically varying these parameters during the model selection process, the system identifies the optimal configuration that achieves high RUL forecasting accuracy with reasonable training time

Inventive Principle:
Principle #35Parameter changes

2Loss of information

If feature importance analysis is performed to identify key degradation features, then model interpretability is improved, but computational overhead increases

Engineering Contradiction:
Improvemodel interpretabilityVSAvoidcomputational overhead
Core Design Contradiction:
Loss of informationVSUse of energy by stationary object

Solution Approach 1:

The patent replaces complex computational feature importance analysis methods with simpler, more efficient techniques. Instead of using computationally intensive methods like permutation importance or SHAP values, the system employs gradient-based importance scoring and attention mechanism analysis from the pre-trained models, which provide sufficient interpretability with much lower computational overhead

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent extracts only the most critical degradation features from the battery data by analyzing feature importance metrics from the trained models. Rather than processing all available features, the system identifies and focuses on the top few most influential features (such as capacity, resistance, temperature) that drive battery degradation, thereby reducing computational overhead while maintaining model interpretability

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4457530B1Method and electronic device for forecasting remaining useful life (RUL) of battery
Publication Date: 2026.04.29 SAMSUNG ELECTRONICS CO LTD
  • EP4457530B1 patent drawingFigure 1a~1c
  • EP4457530B1 patent drawingFigure 2
  • EP4457530B1 patent drawingFigure 3

AI summary

A method for forecasting remaining useful life (RUL) of a battery by an electronic device is provided. The method includes forecasting the RUL of the battery based on at least one capacity value of the battery estimated by at least one of a battery capacity estimation model and a data driven model, determining whether the at least one capacity value for the charging cycle and the discharging cycle is lower than the at least one capacity value estimated by at least one of the battery capacity estimation model and the data driven model and correcting the forecasting RUL of the battery by feeding back the at least one capacity value to at least one of the battery capacity estimation model and the data driven model.