Battery Cell Fault Classification for Early BESS Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery energy storage systems (BESS) face challenges in accurately detecting and classifying anomalous battery cells due to nonlinear time-varying dynamics and false alarms from rule-based methods, leading to potential catastrophic failures.
Innovation Solution
A system and method using a battery management system (BMS) to collect data, a prognostics and fault detection model, and an adjacency weighted, temporal and spectral distance informed neural network to classify outlier battery cells based on voltage, temperature, current, SOC, and cycle count data, enabling early detection of critical faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If rule-based methods with fixed thresholds are used to detect abnormal battery cells, then the system is simple to implement, but false alarms occur frequently and detection precision is poor
Solution Approach 1:
The patent replaces rule-based mechanical threshold checking with a neural network-based intelligent detection system. The neural network learns optimal detection thresholds and patterns from historical data, automatically adapting to different battery conditions and eliminating the need for manual threshold setting while significantly reducing false alarms and improving detection precision.
Solution Approach 2:
The patent transforms fixed detection thresholds into dynamic, adaptive parameters through machine learning. The neural network continuously adjusts detection parameters based on learned patterns from training data, enabling the system to adapt to varying battery conditions, states of charge, and operational scenarios, thereby improving detection accuracy without sacrificing implementation simplicity.
2Reliability
If conservative fixed thresholds are used to avoid false alarms, then false alarm rate decreases, but detection timing is delayed and abnormal situations are detected too late
Solution Approach 1:
The patent performs preliminary learning and pattern recognition during the training phase using historical battery data. The neural network pre-learns subtle indicators of abnormal conditions before actual operation, enabling it to detect early signs of battery issues during runtime without requiring conservative thresholds. This preliminary action allows the system to respond quickly to anomalies while maintaining low false alarm rates.
Solution Approach 2:
The patent implements a feedback mechanism where the neural network continuously learns from detected anomalies and adjusts its detection criteria. The system uses feedback from false alarms and missed detections to refine its detection patterns, progressively improving both the false alarm rate and detection timing through iterative optimization based on operational experience.
3Measurement precision
If multiple parameters (voltage, temperature, current, SOC, cycle count) are used for classification, then detection accuracy improves, but system complexity increases
Solution Approach 1:
The patent employs a universal neural network classifier that handles multiple parameters (voltage, temperature, current, SOC, cycle count) through a single integrated model. This multi-functional approach consolidates what would otherwise require multiple separate detection systems into one unified classifier, improving detection accuracy across all parameters while avoiding the complexity of implementing and coordinating multiple independent systems.
Solution Approach 2:
The patent merges multiple detection parameters and their analysis into a single neural network classification process. By combining voltage, temperature, current, SOC, and cycle count analysis within one unified model, the system achieves comprehensive multi-parameter detection accuracy while simplifying the overall system architecture compared to using separate analysis systems for each parameter.
4Measurement precision
If automated methods are used to identify abnormal cells with marginal deviations, then detection sensitivity improves, but false alarms increase due to noise
Solution Approach 1:
The patent uses feedback from training data to teach the neural network the difference between genuine anomalies and normal variations. During training, the system learns from labeled examples of both abnormal conditions and normal operational variations, enabling it to distinguish between meaningful deviations and noise during actual operation, thereby maintaining high sensitivity while minimizing false alarms.
Solution Approach 2:
The patent performs preliminary learning from historical data to establish baseline patterns of normal operation before actual detection begins. This preliminary training phase allows the neural network to recognize the range of normal variations and filter them out during operation, enabling the system to detect only genuine anomalies with high sensitivity while maintaining low false alarm rates.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method for detecting and classifying outlier battery cells operating abnormally in a storage battery of a battery energy storage system (BESS). A controller controls the operation of the BESS, and a battery management system (BMS) collects battery operational data from the storage battery and stores the battery data in a data repository. A prognostic agent coupled to the battery data repository uses the stored battery data to train a prognostics and fault detection model that is loaded in the controller and used to detect at least one outlier battery cell. Detected outlier battery cell and their operational data are classified using a data classification neural network to one of a plurality of fault types.