DC Arc Detection Using Mathematical Morphology and Pattern Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing DC arc detection methods in photovoltaic power generation systems face challenges in accurately detecting series arcs due to minimal changes in DC voltage and current amplitudes, are influenced by environmental factors, and suffer from false alarms or missed detections, especially in large outdoor systems.
Innovation Solution
A DC arc detection method based on mathematical morphology and pattern recognition, involving mathematical transformation and morphological calculations to extract energy proportions of basic and fluctuation components from electrical quantity spectra, using a trained pattern recognition model to determine series arc faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If radiation-based arc detection methods are used, then detection capability is improved, but environmental influence increases and applicability to outdoor systems decreases
Solution Approach 1:
The patent replaces radiation-based detection (optical/thermal fields) with electrical field-based detection. Instead of using sensors that detect light or heat from arcs, the invention uses current signal analysis through mathematical morphology and pattern recognition to detect arc faults electrically, thereby eliminating environmental sensitivity while maintaining detection capability
Solution Approach 2:
The patent introduces mathematical morphology operations and pattern recognition algorithms as intermediaries between the raw current signals and arc detection. These computational tools process the electrical signals to extract arc characteristics without being affected by environmental factors, serving as a bridge that enables reliable detection in outdoor conditions
2Ease of operation
If current threshold methods are used, then detection simplicity is improved, but detection accuracy for series arcs deteriorates
Solution Approach 1:
The patent transforms the detection approach by changing from simple amplitude threshold comparison to analyzing spectral energy distribution across multiple frequency bands. Instead of checking if current exceeds a threshold, the method calculates energy proportions in different frequency ranges and uses pattern recognition to identify arc characteristics, significantly improving series arc detection accuracy
Solution Approach 2:
The patent adds frequency domain analysis as an additional dimension to detection. By performing mathematical transformation to obtain spectral information and analyzing energy distribution across multiple frequency bands, the method moves from one-dimensional amplitude comparison to multi-dimensional spectral analysis, enabling accurate series arc detection
3Measurement precision
If spectral energy proportion extraction is performed, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent performs mathematical morphology operations and pattern recognition model training in advance during system setup or normal operation. The trained model stores characteristic patterns of arcs and normal operations, allowing real-time detection to use pre-computed reference data rather than performing complex calculations on every measurement, thus reducing online computational complexity while maintaining high accuracy
Data Source
Figure 1~2
Figure 3~5
Figure 6~7
AI summary
Disclosed are a DC arc detection method based on mathematical morphology and pattern recognition. The method includes: selecting characteristic bands; extracting energy proportions, in each characteristic band, of a basic value component and/or a fluctuation component of an electrical quantity spectrum in an arc fault state and a normal working state; establishing and training an arc detection pattern recognition model; acquiring an electrical quantity of a DC power system in an operating state; extracting the energy proportion, in each characteristic band, of the basic value component or the fluctuation component in the electrical quantity spectrum in the operating state; and inputting the energy proportion in each characteristic band and working state data of the DC power system in the operating state into the arc detection pattern recognition model for determination. According to the present invention, by means of performing mathematical transformation - mathematical morphological calculation on the electrical quantity, the energy proportions of the basic value components or fluctuation components of a plurality of characteristic bands of arc occurrence are extracted from the electrical quantity spectrum and inputted into the arc detection pattern recognition model, so as to determine whether a series arc fault has occurred, which is less affected by the fluctuation of a signal sampling absolute value, and achieves high precision.