A New Method Study on River Flow Forecast Model with Impactsof Natural and Human Activities

摘要

Under the influence of natural conditions and human activities in common in modern society, forecast of many characteristics is constrained by certainty and uncertainty factors in common. The application of single physics and random models is being constrained increasingly. When forecasting river flow, if broadly looking river flow routing by Muskingum method as a special case of a first order single variable autoregressive model at two dimension time — space distribution conditions, the foreseeable period of autoregressive model can be extended by changing spatiotemporal location of the corresponding factor in the course of channel flow routing. It can avoid the defect that " residual values relates with their lagged values " in time series autoregressive model, meet the standard assumptions of regression theory that "interference items are unrelated each other" and provide new ideas to set up a mixed forecast model. According to integrating and superposing principles, the mixed model can be simply expressed as " Certain Items + Fuzzy Items + Random Items". When applicating, The model can be expressed and calculated in quantum by the length of foreseeable period and the relationship between predictors of primary and secondary, through the assumption and transform, etc. It still need to analyze error sources of practical problems, confirm the critical time at which factors characteristics transform, use time — varying parameters rationally to improve model precision.