Typical Feature Extraction And Classification Of Complex Dynamic Power Signal Run Mode In Amplitude Domain

With the construction of the new power system in China and the continuous development of clean energy generation and the substitution of electricity in demand-side industries such as high-speed rail,load fluctuations in the power grid have exhibited rapid and random fluctuations.This has resulted in electricity metering errors,leading to significant economic losses.Therefore,how to reasonably extract the load characteristics that affect the accuracy of electric energy metering is an urgent problem to be solved.This paper studies the typical run waveform modes of complex dynamic current signals collected in practice,analyzes the fluctuation speed of their run waveform modes,and determines the important sensitive characteristics that affect the out-of-tolerance of electric energy meters.It has important theoretical significance and application value.(1)In this paper,a run waveform mode extraction algorithm with small time granularity in the dynamic current amplitude domain is proposed,and the run waveform mode in the current amplitude domain of the high-speed rail signal and the spot welding machine signal is extracted.On this basis,the LK-Shape run waveform mode clustering analysis method is proposed,which abstracts the typical run waveform mode and their rapid change characteristics in the current amplitude domain,and solves the dimensionality reduction problem of the dynamic current amplitude domain waveform mode.(2)In order to analyze the fluctuation speed characteristics of the dynamic current amplitude,the current signal is converted from the amplitude domain to the run domain,the run length characteristics of the high-speed rail signal and the spot welding machine signal and the correlation between the amplitude and the run sequence are analyzed,and the typical on/off run length modes of two types of dynamic current signals were extracted,and based on this,the parameters of OOK dynamic testing signals were improved.(3)The typical run waveform mode of dynamic current is extracted from the actual signal,a typical run waveform mode classification data set is constructed,and a one-dimensional LRes Net current typical run waveform mode classification model is proposed.Experimental results show that the method proposed in this paper improves the recognition accuracy of the typical run waveform mode of dynamic current.(4)analyze the amplitude-sensitive waveform features and fluctuation velocity features that affect the error of the energy meter,build an error-sensitive feature testing system for the energy meter,and verify through experiments that the typical features extracted in this paper are sensitive to the error of the energy meter,which is of great significance for the formulation of the energy meter error testing standards.