Knowledge-Data-driven Modeling Method of Multi-time Scale Sampling System

Multi-time scale sampling system refers to the system with multiple sampling frequencies. Due to the need to match between fast sampling variables and slow sampling variables, it is difficult to make full use of the information of fast sampling variables, resulting in insufficient modeling information. Therefore, to solve this problem, a knowledge-data-driven modeling method of multi-time scale sampling system was proposed to establish the multi-time scale sampling system model and improve the modeling accuracy. First, the target model was established using the slow time scale variable, and the filter interpolation method was designed to fill the vacancy value of the slow sampling variable. Second, the reference model was established by the fast time scale variable, and the model sharing mechanism was designed to transfer the sufficient model knowledge from the reference model to the target model for supplementing modeling information. Finally, the parameters of the target model were learned by using the knowledge of the reference model and the data of the target model to improve the accuracy of the target model. The proposed modeling method was applied to theoretical datasets. The experimental results show that the method can fully mine the modeling information and build a high-precision multi-time scale sampling system model.