Study On Multivariate Self-Memory Model Of Hydrologic Nonlinear Time Series Analysis

Hydrological system is one of most complex system. The change of every hydrologic element might be affected by other factors. The self-memory model, originating from weather forecast, based on nonlinear and dynamics, uses system memory to simulate and forecast. However, self-memory model primarily researched the self-development law of hydrological system and seldom considered influencing factors, in past research. That is un-reasonable. In order to correct the shortage and open out the law of hydrologic phenomena better, building the model of considering the influence factors becomes necessary.Hotan, located in hinterland of Euro-Asia, the northwest of China, is an oasis based on irrigation agriculture as the main living style. Because of little precipitation and high evaporation, the research of evaporation is very important in this area for determination of agriculture irrigation system and quota. Furthermore, the minority take up the most of the population in Hotan. So, this research is also significant for improving people's life.Because of the above two reasons, this paper, based on previous research and plentiful literature reviews, took a trial and preliminary exploration about multivariate self-memory model and obtained some researches, as follows:(1) According basic inverse theory and self-memory theory, this paper did not only clear up modeling approach but also give the detailed process of the modeling. In all, it put forward a study way for the whole research.(2) A grey GM (1,n) self-memory model was built according to the modeling approach of multivariate self-memory model and the theory of grey GM (1,n). Then the model was used to forecast the annual evaporation capacity and monthly evaporating capacity in Hotan. Comparing the prediction effect with the model of grey GM (1,1), a conclusion was obtained that this model had a better forecast accuracy, when the observed data had obvious fluctuation.(3) Although the grey GM (1,n) self-memory model had a better forecast accuracy, it still could not meet the production needs. Therefore, a new self-memory hydrologic forecasting model was built, which was based on multi-dimensional phase-space reconstruction theory. Then, a better forecast accuracy was got comparing the prediction effect with grey GM(l,n) self-memory model. It met the practical needs better and would be applied in production practice.