Complex Disturbances Identification: A Novel PQDs Decomposition and Modeling Method

In the context of unprecedented attention to renewable energy, wind and photovoltaic power generation are widely used. However, this process introduces a large number of solid-state switching and non-linear loads, which makes power quality disturbances (PQDs) complex and brings unknown challenges to power-pollution control. As a prerequisite for power-pollution control, this paper proposes an automatic PQDs classification approach that is suitable for complicated phenomena. First, an ensemble intrinsic time-scale decomposition (EITD) method is proposed to decompose the PQDs, which overcomes the decomposition level's endpoint effect and frequency aliasing by adding Gaussian noise and integrating multiple sub-components. Then, utilizing the global depth-wise convolution layer and parameter rectified linear unit, a global depth-wise shuffle CNN (GSCNN) is proposed to improve the performance and reduce the number of parameters. Based on EITD and GSCNN, an automatic framework is proposed to identify and classify complex PQDs. Simulation experiments and hardware platform tests show that the proposed framework has superior performance for complex and even nonlinear disturbances under different noise.