论文详情
常规自适应估计滤波方法，采用残差变化梯度近似代替均方误差梯度，当估计值和原始值具有统计独立的特点时，这种近似估计方法效果较好，并且估计结果稳定。但是，当原始数据或估计数据出现非独立统计特点时，误差梯度变化剧烈，出现估计结果的剧烈振荡。针对自适应估计方法存在的问题，结合微地震资料的特点，研究了自适应估计滤波的改进方法。在算法设计中，考虑加入约束条件，使迭代方法向着期望收敛方向进行，即在迭代过程中，下一次预测矢量的数学期望，要小于上一次的数学期望；把权系数用其离差进行归一化，使递推估计结果不会剧烈震荡，估计过程稳定收敛且估计结果可靠；误差和权系数由原来的自身迭代结果取值，改进为曲线拟合后再取值；同时讨论了窗口长度的选取问题，将固定窗口改进为自动调节的变化窗口。模型和实际资料测试结果证明了改进的自适应滤波方法设计理论正确合理，处理效果改善明显。
Residual variation gradient estimation instead of mean square error gradient is used for the conventional adaptive filtering method.However,when the raw data or estimated data happened to have the characteristics of statistically non-independent,the error gradient dramatically changes,and the estimated results shake vigorously.Aiming at the problem of the adaptive method,improved adaptive estimation filtering method is proposed with the consideration of the characteristics of the micro-seismic data.Algorithm design takes constraints into consideration and makes iterative approach towards the expected convergence direction.In iterative process,the next forecast vector mathematical expectation is less than previous mathematical expectation.Weight coefficient is normalized by its deviation,so recursive estimation results are not turbulent.The estimation process is stable and convergent and the estimated results are reliable.The values of the error and weight coefficients are changed from iteration results by themselves to the ones after the curve fitting.Also we discuss the problems of the selection of the window length and improve the fixed window to the changed window of automatic adjustment.The results of model and the actual data test prove that the design concept of improved method is correct and reasonable and the application effect is improved significantly.