Supercapacitor SoH Prediction Using Driver-Specific Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The long lifespan of supercapacitors makes it difficult to efficiently evaluate their State of Health (SoH), as the evaluation process is time-consuming due to their hundreds of thousands of cycles, and their deterioration is influenced by various factors.
Innovation Solution
A system and method using a neural network model to predict SoH for each charge/discharge cycle of an electric energy storage device, incorporating a data collection unit, vehicle control unit, and display device, which utilizes preconfigured SoH prediction neural networks based on driver type and road conditions to provide real-time SoH predictions and replacement alerts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the supercapacitor is evaluated through hundreds of thousands of charge/discharge cycles to determine its lifespan, then the evaluation accuracy is improved, but the evaluation time becomes excessively long
Solution Approach 1:
The patent applies preliminary action by collecting and analyzing charge/discharge data from early stages of supercapacitor operation to train a neural network model. This model then predicts the remaining lifespan without requiring the supercapacitor to undergo hundreds of thousands of actual cycles, thus obtaining accurate lifespan evaluation in advance while significantly reducing the time required for evaluation.
2Reliability
If the supercapacitor undergoes hundreds of thousands of charge/discharge cycles to evaluate its durability, then the durability assessment is improved, but the operational time required increases significantly
Solution Approach 1:
The patent uses copying by creating a virtual model of the supercapacitor's aging process through a neural network. Instead of physically subjecting the supercapacitor to hundreds of thousands of cycles, the system copies the essential characteristics of long-term operation into trained neural network models that can predict durability outcomes based on limited actual cycle data, thereby assessing reliability without requiring extensive operational time.
3Device complexity
If a single neural network model is used for SoH prediction across different driver types, then the system complexity is reduced, but the prediction accuracy for specific driver types deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the neural network model into multiple specialized models, each trained on charge/discharge data from specific driver types (e.g., city driving, highway driving, mixed driving). The system selects and applies the appropriate neural network model based on the detected driver type, thereby maintaining high prediction accuracy for each specific driving condition while managing complexity through modular model selection rather than a single monolithic model.
Data Source
AI summary
A system for predicting a State of Health (SoH) of an electric energy storage device mounted in a vehicle includes a data collection unit configured to acquire charge/discharge data of the electric energy storage device. The system further includes a vehicle control unit configured to acquire an SoH prediction neural network model for predicting an SoH for each charge/discharge cycle of the electric energy storage device, based on first charge/discharge data of the electric energy storage device acquired by the data collection unit. The vehicle control unit is also configured to predict and determine the SoH for each charge/discharge cycle of the electric energy storage device using the SoH prediction neural network model.