The Optimal Control Problem Of The Error Term Of Numerical Weather Prediction Model

The two basic factors, restricting the accuracy of numerical weather prediction, are initial field error and model error. Simply relying on the pursuit of the refinement of numerical model can not effectively reduce the model error. Taking into account the con-tinuous evolution of the atmosphere, recent historical data must contain a large number of forecasts field information. These data are a series of special solutions of the approx-imate atmospheric accurate mode. If the model error is seen as the unknown term of the atmospheric accurate model, the problem is converted into the inverse problem of the partial differential equations-knowing the particular solution of the equations to determine the unknown term in the equation. If we can use the historical data to solve this kind of inverse problem effectively,that is to say,wc can obtain the unknown mode error term, substituting it into the numerical model must greatly improve the prediction effect of the numerical model. In this paper, using historical data, the partial differential equations initial boundary value problem corresponding to the original numerical weath-er prediction models is converted to the inverse problem, and finally we give the optimal control problem of mode error term. The second chapter gives the Newton-Conjugate gradient algorithm of optimal control problem under the constraints of partial differen-tial equations. In the third chapter, we illustrate the examples of linear and nonlinear evolution equations and carry out a series of related numerical experiments. The results of numerical experiments show that making full use of the historical data and converting the initial boundary value problem of the numerical prediction into the inverse problem can greatly improve forecasting effect. The idea of inverse problems is an effective way to improve the accuracy of numerical weather prediction.