Table 1 Limitations of existing PQD detection and classification methods.
From: Power system security and protection considering the integration of new energy power plants
Category | Limitation description |
|---|---|
Feature extraction | Conventional methods like Wavelet and Stockwell transforms often fail to capture the dynamic and non-stationary behavior of PQDs under varying noise conditions. |
Deep learning methods | CNNs and attention-based models require significant computational resources, hindering real-time implementation and increasing deployment complexity. |
Noise robustness | Performance degradation under high-noise environments is common, reducing classification accuracy in real-world grid conditions. |
Generalizability | Many methods are evaluated on synthetic datasets or limited test conditions, leading to poor adaptability in diverse and practical operating scenarios. |
Optimization techniques | Hybrid approaches using PSO or GA face slow convergence and sensitivity to parameter tuning, affecting reliability and adaptability. |
Model explainability | Black-box nature of advanced AI models limits interpretability, making it difficult for operators to trust and validate classification decisions. |
