Denoising Research On Acoustic Wave Signal In Oil Well Dynamic Fluid Level Based On Sparse Decomposition

Oil well liquid level refers to the pumping well dynamic crude oil level in the normal production process, which is in the annular space between oil tube and casing below the surface of soil. As an important index to reflect the formation crude oil reserves and liquid supply capacity, the dynamic liquid level depth value is an important basis to realize the oil production and overall efficiency maximization. At present acoustic reflection method is widely adopted to measure dynamic liquid level depth value. However, due to the complicated underground environment, acoustic signals collected by acoustic sensor with a lot of noise, resulting in the original clear fluid acoustic curve is difficult to identify, and thus unable to calculate the fluid level depth value by the waveform identification. Therefore, the paper focuses on the acoustic wave reflected signal denoising problem of the dynamic liquid level. The main work and research results are as follows:Firstly, this paper takes acoustic wave reflected signal in oil well dynamic fluid level as research object and introduces an improved double-variable threshold function to modify the threshold selection method of wavelet transform, which overcomes the shortcomings that hard threshold function is discontinuous and soft threshold exists constant deviation, moreover, compresses when it is in the range of wavelet coefficients absolute value less than threshold value. The contrast of denoising results reveals that this method outperforms both soft threshold and hard threshold in the comparison of SNR and MSE, testing this threshold function shows notable effectiveness in denoising signal of reflection acoustic wave in oil well dynamic liquid level.Secondly, applying sparse decomposition to oil well dynamic liquid acoustic wave reflected signal denoising, studying sparse representation algorithm, introducing the principles and steps of the classic MP algorithm and analyzing its shortcomings in detail, and introducing OMP and StOMP algorithms to modify MP. Then through the denoising simulation experiment verified the OMP and StOMP algorithms, improved denoising effect, and compared the obtained results with classical wavelet threshold function denoising algorithm, indicating that the superiority of oil well dynamic liquid level signal denoising algorithm based on sparse decomposition.Thirdly, proposing two over-complete dictionaries of dynamic liquid level signal, one is to select generate function according to the prior knowledge of dynamic liquid level signal, then by discretizing the generation function to construct no training dynamic liquid level over-complete dictionary, the other one is to train dynamic liquid level sample data with KSVD algorithm, and combined with OMP algorithm to construct dynamic liquid level signal over-complete training dictionary. By comparing with the common Gabor dictionary in the field of sparse decomposition, verifying that the sparse decomposition algorithm based on the proposed two over-complete dictionaries greatly improved the signal denoising performance.