Methods And System Development For Mid-long Term Hydrological Forecasting Of Qinghe Reservoir

With a long foreseen period,mid-long term hydrological forecasting can be combined with the short term hydrological forecasting and the super long term hydrological forecasting.People can make a decision when protecting the flood and fighting a drought, impounding water and abandoning water with the forecasting results.People can also take early measures to resolve the contradiction between the various departments of water co-ordination arrangements, in order to get maximum benefit. Qinghe Reservoir as an example, The original mid-long term hydrological forecasting system is out-of-date with a behind forecasting method and an inaccurate forecasting results. In this situation,the author of this paper collecte adequate hydrological data of Qinghe Reservoir,analysis of the evolution of Qinghe Reservoir runoff,study of existing forecasting methods,then make the appropriate study. This article main research content and the related conclusion are as follows:(1)Propose gray system-RBF neural network forecasting method consists of gray system and RBF neural network.Select the the runoff of each non-flood month as predictors.Through calculating gray correlation between predictors,Select the years in the past which had a similar early hydrological regime with the year need to be forecasted.Build a relationship between the predictors(the runoff of each non-flood month) and the predictand(the flood runoff) by using RBF neural network,then forecast the flood runoff.(2)Propose the layeres superposed forecasting method consists of a linear regression,periodic mean superposition and computing Euclidean distance. Divide the flood runoff into linear portion,cycle portion and random portion of the flood runoff. Forecast the three parts separately.Then overlay the various parts of the forecasting results. This method takes into account the certainty, periodicity and randomness of hydrological itself.(3)Apply probability density in mid-long term hydrological forecasting.Describe the range and probability of predictand by probability density function.Propose probability distribution error as the probability density between the actual value and the forecasting value, and take it as a standard for evaluating the forecasting results. Compared to relative error and absolute error,the probability distribution error can evaluating the forecasting results much objectively. By comparing the probability density error of this forecasting results between the two methods, gray system-RBF neural network forecasting method is more suitable for Qinghe Reservoir.(4)Based on the Gray system-RBF neural network forecasting method used in this paper,develop a workable mid-long term hydrological forecasting system with a friendly interface.