In-depth Characteristics Extraction Of Cell Voltage Fluctuation In Aluminum Reduction Cells

Stability research of aluminum reduction cells is one of the spotlights in aluminum electrolysis process. Diagnosing working conditions of an aluminum reduction cell from its online parameters in order to take proper measures is an important strategy to enhance its stability. Cell voltage is one of the parameters that can be detected online, whose fluctuation contains a lot of information concerning performance of an aluminum reduction cell. Extracting in-depth characteristics of cell voltage fluctuation is an important foundation of systems diagnosing working conditions for aluminum reduction cells.This dissertation researched on 320kA pre-baked anode aluminum reduction cells, collected many cell voltage signals of aluminum reduction cells under typical conditions, and used wavelet to denoise all signals collected. After that power spectral density (PSD) estimation method and wavelet packets were used to analyze cell voltage signals. Research results are as follows:(1) Using wavelet denoising technology can preserve important information in original signals and remove noise thus make signals smooth. Different cell voltage signals have different optimal wavelet functions.(2) PSD estimation based on nonparametric methods have poor resolution, and cannot extract characteristics of cell voltage fluctuation well, but PSD estimation based parametric methods have much better resolution and can extract characteristics of cell voltage fluctuation under some working conditions well. The PSD estimation results have clear physical meanings, but parametric methods cannot extract information from unstable signals containing many sudden changes well. Although subspace methods have better resolution than parametric methods do, but they are too complex to be used in engineering practices.(3) Using wavelet packets to decompose cell voltage signals from 1 to 3 levels, it can be found out that 3-level wavelet packet can extract energy eigenvectors of cell voltage fluctuation well, what's more, since the high dimension of the vectors, the algorithm is robust. After that wavelet packet decomposition was optimized, it can be found out that different signals have different optimal wavelet packet tree.(4) Comparing PSD estimation with wavelet algorithm, it is suggested that wavelet complement with PSD estimation, and they have different advantages and disadvantages. Fourier analysis is much simpler, less-time consuming, and has clear physical meanings. But wavelet transform has better qualities in processing unstable signals with a lot of sudden change information. A combination of these two methods can extract in-depth characteristics of cell voltage fluctuation very well.