论述了数据同化中的伴随方法与泛函分析中的伴随算子的关系 ,指出经典的拉格朗日乘子法可以作为伴随方法的理论基础。关于伴随方法应用中只能利用模式的伴随的观点 ,作者认为是有疑问的 ,并提出了新的观点。所作的数值模拟实验表明 ,方程的伴随有时明显优于模式的伴随 ,因而说明方程的伴随也应该引起足够的重视。 The relationship between the adjoint operator in data assimilation and the adjoint operator in functional analysis is discussed. It is pointed out that the classical Lagrange multiplier method can be used as the theoretical basis of adjoint method. With regard to the concomitant use of patterns in concomitant application of methods, the author considers that there is doubt and puts forward new ideas. Numerical simulations show that the companion of the equation is sometimes better than the companion of the model, so the companion of the equation should also pay enough attention.