基于自适应KS变换与JetLeaf Synth网络的复杂电能质量扰动先进诊断框架

The accurate diagnosis of power quality disturbances (PQDs) is crucial for improving energy efficiency and advancing the development of the smart grid. However, the widespread adoption of solar and wind power introduces numerous power electronic converters, complicating PQDs and heightening identification challenges. This article introduces a novel automatic detection framework based on the adaptive Kaiser S-transform (AKST) and JetLeaf Synth network (JSTN), enabling the automatic analysis and detection of intricate PQD signals. To begin, AKST is employed to analyze the time-frequency characteristics of PQD signals. By adaptively refining the parameters of the Kaiser window based on maximum energy concentration, the time-frequency resolution is effectively improved, providing more detailed information. Subsequently, JSTN is developed to automatically extract and recognize crucial distinctive features of PQDs from the time-frequency matrix generated by AKST. Within JSTN, it inherits the local detail capability of the twin leaf mixer (TLM) and the global context ability of the jet-stream transformer (JST), significantly enhancing diagnostic accuracy. The integration of AKST and JSTN results in a detection framework known as hybrid adaptive time-frequency JetLeaf SynthNet (HAJSTN), which is proposed to achieve accurate diagnosis of various PQDs. Multiple simulations and an extensive experimental activity validate that HAJSTN outperforms some advanced PQD identification methods, demonstrating its commendable performance.