US 2026/0149258 A1 · ARCING DETECTION METHOD AND SYSTEM

Abstract

The present invention provides an arcing detection method and system, in which an AI model derives an AI score from current data and preliminarily determines whether a current signal of interest is arcing based on a comparison between the AI score and an AI threshold. It preliminarily determines whether the determination is false. These two determination steps enable more accurate arcing detection. When a false determination count becomes greater than or equal to a false determination count threshold, the AI model is retrained and dynamically updated to adapt itself to the current situation. This dispenses with the need for deployment of multiple static AI models in the system, which requires larger storage memory. The AI model can adapt itself to the current situation without needing manual commissioning of parameters, effectively reducing manpower costs. Further, this dispenses with the need for regular interrogation of environmental changes, reducing the system's task overhead.

Claims19

Dependent claims

An arcing detection method comprising:

step S1, sampling a current signal of interest a plurality of times within a sampling period, obtaining current data;
step S2, deriving an AI score by an AI model from the current data, the AI score characterizes a probability of arcing of the current signal of interest;
step S3, determining whether the current signal of interest is arcing based on a comparison between the AI score and an AI threshold;
step S4, determining whether the determination made in step S3 is a false determination and, if so, incrementing a false determination count by one;
step S5, comparing the false determination count with a false determination count threshold and, if the false determination count is greater than or equal to the false determination count threshold, retraining the AI model and clearing the false determination count.

2 more elements

An arcing detection system comprising a sampling module and a processing module signal-connected to the sampling module,

the sampling module configured to sample a current signal of interest a plurality of times within a sampling period, thereby obtaining current data,
the processing module provided with an AI model, the AI model derives an AI score from the current data, the AI score characterizing a probability of arcing of the current signal of interest,
the processing module configured to determine whether the current signal of interest is arcing based on a comparison between the AI score and an AI threshold,
the processing module also configured to determine whether the determination is a false determination and, if so, increment a false determination count by one,
the processing module also configured to compare the false determination count with a false determination count threshold and, if the false determination count is greater than or equal to the false determination count threshold, retrain the AI model and clear the false determination count.

2 more elements

Dependencies are inferred from the claim text.

Arcing detection method and system

Life of this patent

Arcing detection method and system

Life of this patent

Abstract

Claims19