Research on Feature Extraction and Decomposition Algorithm of Nonlinear Current Signal

摘要

Non-linear power load generates a large number of harmonics waves and sudden-changing and non-steady-state signals. The non-linear characteristics are obvious, which seriously affects the accuracy of energy measurement. Large-scale non-linear power behavior has uncertain characteristics, such as randomness of time and space, intermittentness, and fluctuation, which brings new problems to power grid planning, safe operation, and optimal scheduling. By studying the charging load characteristics, the collected current data is analyzed, the characteristics of the non-linear load signal are extracted, and non-linear power load models are constructed. The Fourier analysis algorithm is used to decompose multiple ripples, and the improved wavelet algorithm is used to complete the time domain and frequency domain transformations, which can accurately detect the characteristics of unsteady signals. Experiment results demonstrate that the signal feature recognition method by combining with Fourier and Wavelet transform is effective to extract the signal features of the entire DC charging process, identify and separate the fundamental and non-steady-state wave signals.