AI approach improves fire prevention
Eaton's center for intelligent power (CIP) was challenged by our product team to improve the accuracy of its legacy AFDDs. The requirement was to engineer an accurate arc fault detection system that can not only discern genuine arc fault events from non-threatening anomalies but also one that could improve over time based on accumulated data analytics.
First, the data science team collected extensive data from various AFDDs in operation, which included information on normal electrical patterns as well as anomalies that led to arc faults.
By investigating the collected data, the team has been able to identify common characteristics and signatures of arc faults. They are currently developing machine learning models that are capable of distinguishing between actual arc faults and other electrical events. These models are being 'trained' using the collected dataset to ensure high sensitivity and specificity.
The machine learning algorithms are being continuously optimized through iterative testing and validation. The research team is employing advanced analytics to improve the detection algorithms, in order to reduce the risk of incidence of false positives and negatives.
Furthermore, to validate the accuracy of the AFDDs, extensive real-world testing is being conducted. Feedback from these tests will be used to further enhance the algorithm, ensuring its reliability and trustworthiness in diverse electrical environments.
Recognizing the dynamic and complex nature of electrical installations and usage patterns, the research team is also developing a feedback system designed to ensure ongoing improvement. The AI models are designed to learn from new data, adapting and improving over time to maintain high accuracy levels.