Harmonic Aggregation Entropy: A Highly Discriminative Harmonic Feature Estimator for Time Series.

Area of Science:

• Signal Processing
• Electrical Engineering
• Data Analysis

Background:

• Harmonics are prevalent in power systems, causing increased energy consumption and risks to equipment safety and performance, especially in critical applications like large ships.
• Existing time-frequency analysis methods for harmonic detection suffer from high computational costs and limited feature extraction specificity.
• There is a critical need for advanced methods to accurately detect and characterize harmonic components in time-series data.

Purpose of the Study:

• To introduce a novel harmonic feature estimation method, Harmonic Aggregation Entropy (HaAgEn), designed for effective harmonic detection in time-series data.
• To address the limitations of traditional methods by developing a more computationally efficient and specific harmonic feature extraction technique.
• To validate the efficacy of HaAgEn in discriminating harmonic signals from background noise and improving detection accuracy.

Main Methods:

• The proposed method, HaAgEn, is based on bispectrum analysis, leveraging the unique distribution of harmonic signals within a bispectrum matrix.
• A novel Diagonal Bi-directional Integral Bispectrum (DBIB) technique is employed to extract harmonic features from the bispectrum matrix.
• Cross-entropy is utilized to calculate integration results (Ix and Iy) from DBIB on different frequency axes, forming the HaAgEn metric.

Main Results:

• HaAgEn demonstrates significantly higher sensitivity to harmonic components compared to other entropy-based methods, effectively reducing feature redundancy.
• The method successfully discriminates against background noise, providing more specific harmonic feature extraction.
• Sea trial data analysis showed HaAgEn achieving 96.8% accuracy in detecting harmonic components in shaft-rate electromagnetic field signals, outperforming existing methods.

Conclusions:

• HaAgEn offers a novel and effective technical approach for harmonic detection in industrial applications, particularly for time-series data analysis.
• The method provides a significant improvement in detection accuracy and efficiency over traditional techniques.
• The enhanced sensitivity and specificity of HaAgEn make it a valuable tool for ensuring the safety and performance of power systems and related equipment.