Autoencoder Anomaly Detection for Electrical Energy Store Aging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the aging state of electrical energy stores, such as vehicle batteries, are inaccurate and costly, relying on physical aging models that require numerous sensors and do not account for usage-specific parameters, leading to inefficient anomaly detection and aging prediction.
Innovation Solution
A computer-implemented method using a hybrid data-based aging state model that combines physical and electrochemical models with data-driven corrections, employing autoencoders for anomaly detection based on operating variable profiles, allowing for continuous monitoring and prediction of aging states without direct sensor integration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical aging models with numerous sensors are used to determine aging state, then measurement precision is improved, but device complexity and manufacturing cost increase
Solution Approach 1:
The patent replaces the mechanical/sensor-based physical aging model with a data-driven machine learning approach. Instead of using numerous physical sensors to directly measure aging parameters, the system uses operating variable profiles (electrical and thermal data) processed through autoencoders and neural networks to determine aging state, thereby reducing sensor integration complexity while maintaining measurement precision
Solution Approach 2:
The patent introduces an intermediary data processing layer (autoencoders, feature extraction algorithms, and neural networks) that transforms readily available operating variable profiles into accurate aging state predictions. This intermediary approach avoids the need for direct sensor integration into the energy store while achieving precise aging determination through pattern recognition in operational data
2Measurement precision
If physical aging models are used, then aging state can be determined, but anomaly detection capability and usage-specific prediction accuracy deteriorate
Solution Approach 1:
The patent implements feedback mechanisms where the autoencoder continuously monitors reconstructed errors from operating variable profiles and provides feedback signals for anomaly detection. The system uses reconstruction errors as indicators of deviations from normal aging patterns, enabling real-time anomaly detection that complements the aging state determination and improves overall system reliability
Solution Approach 2:
The patent transitions from static physical aging models to dynamic machine learning models that adapt to usage-specific patterns. The neural networks are trained on operational data to capture dynamic aging behaviors under different usage conditions, enabling accurate predictions that account for varying operational profiles and improving anomaly detection capability
3Measurement precision
If direct sensor integration in energy store is implemented, then aging state measurement accuracy is improved, but manufacturing cost and production complexity increase
Solution Approach 1:
The patent replaces direct sensor integration into the energy store with a computational approach that uses existing operational data. Instead of manufacturing sensors into the battery cells or modules, the system processes electrical and thermal operating variables through machine learning algorithms to infer aging state, thereby maintaining measurement accuracy while simplifying manufacturing processes
Solution Approach 2:
The patent creates a virtual model (digital twin) of the aging process through machine learning that replicates the information obtained from direct sensing. The autoencoder and neural networks generate accurate aging state predictions by learning from operational patterns, effectively copying the functionality of direct sensors without the manufacturing complexity and cost
Data Source
AI summary
A computer-implemented method for determining an anomaly of a behavior of an electrical energy store in a technical device includes sensing an operating variable profile of at least one operating variable of the electrical energy store, and determining at least one feature from the sensed operating variable profile of the at least one operating variable of the electrical energy store. The method further includes evaluating an anomaly detection model using an autoencoder with a supplied input vector that includes or depends on the determined at least one feature, in order to determine a reconstructed input vector, and signaling an error based on a reconstruction error between the reconstructed input vector and the supplied input vector.