PSO-VMD-SVD的改进去噪方法研究  

Research on Improved Method for Denoising of PSO-VMD-SVD

作　　者：杜颖[1] 李宏[1] 刘庆强[1] 路敬祎[2,3] 李富[4] DU Ying;LI Hong;LIU Qingqiang;LU Jingyi;LI Fu(School of Electrical Engineering and Information,Northeast Petroleum University,Daqing 163318,China;Artificial Intelligence Energy Research Institute,Northeast Petroleum University,Daqing 163318,China;Heilongjiang Provincial Key Laboratory of Networking and Intelligent Control,Northeast Petroleum University,Daqing 163318,China;Drilling Company 1,Daqing Drilling Engineering Company,Daqing 163318,China)

机构地区：[1]东北石油大学电气信息工程学院,黑龙江大庆163318 [2]东北石油大学人工智能能源研究院,黑龙江大庆163318 [3]东北石油大学黑龙江省网络化重点实验室,黑龙江大庆163318 [4]大庆钻探工程公司钻井一公司,黑龙江大庆163318

出　　处：《吉林大学学报（信息科学版）》2021年第2期142-151,共10页Journal of Jilin University（Information Science Edition）

基　　金：国家自然科学基金资助项目(61873058;NSFC51575407);中国石油科技创新基金资助项目(2018D-5007-0302);黑龙江省自然科学基金资助项目(LH2020F005);东北石油大学青年科学基金资助项目(2018QNL-33);武汉科技大学冶金装备与控制教育部重点实验室开放基金资助项目(MECOF2019B01)。

摘　　要：针对VMD(Variational Mode Decomposition)在处理信号时,模态个数K和惩罚参数α的选择会影响信号分解的效果,提出一种改进PSO(Particle Swarm Optimization)优化VMD(PSO-VMD)的算法,通过改进阈值相关系数法选择有效模态分量,根据信号分析对高频噪声进行奇异值去噪重构。对仿真信号和管道泄露信号进行实验,通过选择多个K并结合相关系数法选择模态进行重构,比较不同K对应的重构信号与原信号的信噪比(SNR:Signal to Noise Ratio)、相关系数(CC:Coefficient)、平方绝对误差(SAE:Square Absolute Error)、均方误差(MSE:Mean Square Error)证明改进PSO优化VMD的可靠性。将该方法与改进的PSO-VMD结合豪斯特夫距离和奇异值(HD-SVD:Hustoff Distance-Singular Value Decompositio)、互信息和奇异值(MI-SVD:Mutual Information-Singular Value Decompositio)、相关系数和小波变换(CC-WT:Correlation Coefficient-Wavelet Transform)等方法对比,效果更佳。According to VMD(Variational Mode Decomposition) in processing the signal,the selection of mode number K and parameter α will affect the signal decomposition results,an optimized VMD(Variational Mode Decomposition) algorithm improving PSO(Particle Swarm Optimization)(PSO-VMD) was proposed. Based on the analysis of signal,high frequency noise is reconstructed by singular value denoising. Experiments on simulation signal and pipeline leakage signal,by selecting multiple K and combined with proposed the correlation method to select mode and reconstruct signal, compare SNR(Signal to Noise Ratio), correlation CC(Coefficient),SAE(Square Absolute Error),MSE( Mean Square Error) of reconstructed signal of different K and original signal,proved the reliability of optimizing VMD by improved PSO. The proposed method is compared with PSO-VMD combined with HD-SVD(Haussteff Distance and Singular Value),MI-SVD(Mutual Information and Singular Values),CC-WT(Correlation Coefficient and Wavelet Transform) and other methods.The proposed denoising method has the best effect.

关 键 词：VMD算法 粒子群优化 相关系数 奇异值 去噪 

分 类 号：TN911[电子电信—通信与信息系统]