Entropy-Fused Enhanced Symplectic Geometric Mode Decomposition for Hybrid Power Quality Disturbance Recognition.

Area of Science:

• Electrical Engineering
• Signal Processing
• Machine Learning

Background:

• Power quality disturbances (PQDs) pose significant operational challenges in electrical networks.
• Accurate assessment of PQDs relies heavily on effective feature selection for classification models.
• High-dimensional feature vectors can increase model complexity and reduce recognition speed.

Purpose of the Study:

• To propose a robust feature extraction framework for accurate and efficient power quality disturbance identification.
• To develop a low-dimensional feature vector that enhances classifier performance.
• To improve the accuracy and robustness of power quality disturbance recognition models.

Main Methods:

• Improved Symplectic Geometric Mode Decomposition (ISGMD) for tri-band signal decomposition.
• Combination of refined generalized multiscale quantum entropy and reverse dispersion entropy for feature extraction.
• A deep extreme learning machine (ELM) algorithm for a double-layer composite classification model.

Main Results:

• The proposed method achieved an average recognition accuracy of 97.3% across various noise environments.
• Under complex mixed perturbations, accuracy remained above 96%.
• Demonstrated a 3.7% improvement in recognition accuracy compared to CNN + LSTM, with superior performance on small datasets.

Conclusions:

• The developed feature extraction and classification strategy accurately identifies power quality interferences.
• The method exhibits superior classification accuracy and robustness compared to traditional approaches.
• Validated effectiveness on simulation and measured data, showing high accuracy (99.10%) and noise resistance.