基于变分方法的有限区域风场分解与重构I:理论框架和仿真实验

Wind partitioning and reconstruction with variational method in a limited domain I: theoretical frame and simulation experiments

众所周知,风场分解与重构最有效的方法就是引入速度势和流函数,其一般通过求解两个Poisson方程得到.由于速度势和流函数在边界上的耦合性质,有限区域风场分解是不唯一的,这对风场分解带来了很大困难.本文采用变分伴随结合正则化方法来克服风场分解的不唯一性,其核心是把速度势和流函数的边值作为控制变量来反演.目标泛函由两部分组成,一是衡量重构风场误差大小的观测项;二是保证风场分解问题适定的正则化项,其目的在于寻求具有气象意义的稳定正则化解.数值试验结果表明,在正确选取正则化参数后,利用变分伴随结合正则化方法进行有限区域风场分解与重构是有效可行的.

As is well known, the efficient method to wind partitioning and reconstruction is to introduce the velocity potential and stream function which are calculated from divergence and vorticity by solving two Poisson’s equations. Since velocity potential and stream function are coupled at the boundary of limited domain, the wind partitioning problem is nonunique. To vercome the nonuniqueness of the wind portioning, a new variational adjoint method combined with regularization is proposed in this paper, which is based on the control of velocity potential and stream function boundary values under Dirichlet conditions. The cost function is composed of two parts, one is the observation term to minimize the error of the reconstructed wind field, and the other is the regularization term to guarantee the uniqueness of the reconstruction problem by seeking a stable regularization solution within meteorological content. The results of numerical experiments demonstrate that after choosing an appropriate regularization parameter, the new variational adjoint method combined with regularization is efficient and suitable for wind portioning and reconstruction in a limited domain.