A method for coupled interaction of models with R

Abstract

A method for coupling and interacting the model with R, using OpenMP (Shared Memory Parallel Programming) to build a "work-sleep cycle framework" to control the model and the R platform to work and sleep alternately, that is, when the model is working, the R platform is sleeping and waiting. After the model is executed, it will go to sleep, and the R platform will start to work; after the R platform is finished, it will send an instruction to start the model to run, thus starting a new working cycle, so it works alternately. In the present invention, the concepts of PEST input templates and output instructions and model running functions are instantiated as model threads, and user R functions and scripts are instantiated as R threads. These two threads are managed by the "work-sleep cycle framework" described above for interaction coupling. The method of the present invention is characterized in that it realizes the collaborative work between the model and R without touching the source code of the model, and the model can be called freely on the R platform like an R function, thereby making full use of the powerful functions of the R platform to realize parameter optimization , sensitivity and uncertainty analysis, and visualization, greatly simplify model inversion and simulation analysis operations, and improve work efficiency and user experience.

Description

technical field [0001] The invention belongs to the field of earth environment science, in particular to a method for coupling and interacting a model with R. Background technique [0002] Mathematical models are commonly used in various fields of science and engineering. In the application of mathematical models, it is usually necessary to carry out model inversion and observe the operating status of the model, which involves parameter calibration, sensitivity analysis, uncertainty analysis, and visualization of operating status and results. Carrying out model inversion (such as model parameter calibration) usually requires some external algorithms. For example, in the field of environmental science simulation, the widely used Shuffled Complex Evolution (SCE) Algorithm is combined with the model. This process is very challenging, and usually requires understanding the model structure and algorithm interface, mastering the source code of the model, and requiring a lot of pr...

AI Technical Summary

Problems solved by technology

[0007] The first is the need to rewrite the model program in R. Since the model is often complex, rewriting the model requires the user to know every detail of the model and be proficient in the R language, which is very difficult to write, time-consuming and labor-intensive; and converted to the R language It is difficult to guarantee the post-operation efficiency and the risk is high

[0008] The second solution is that developers have to give up the easy-to-use R language. In order to adapt to the programming language of the model, they must re-learn the programming language of the required model, resulting in a large learning cost. If multiple models need to be applied, the model If it is written in multiple languages, the learning burden paid by the user will be heavier, and there is a risk that the execution efficiency of the model language is not high

[0010] (1) The first solution is difficult, time-consuming and labor-intensive, and has high risks

[0011] (2) The second solution requires users to learn one or more programming languages, which has high learning costs, and some models cannot be visualized, resulting in poor user experience

[0012] The knowledge and technical costs required by the two solutions are very high, which is very difficult for ordinary scientists and engineers, and is not conducive to quickly completing specific business needs

Images