A Novel Similar Derivative Method To Forecast Monthly Runoffs

Runoff forecasting is of significant importance to reservoir operation, flood control,power generation and irrigation, is also the fundamental task that’s related to nationaleconomic development of our country. With long forecasting period, the mid-long termrunoff forecasting enable people to make a whole planning and balance all conflicts withina long time, when solving the problem of reservoir operation, flood control, powergeneration and irrigation, so that to gain more economically and socially. However,because of long forecasting period, the influence of uncertainty of runoff series iscomparatively significant to forecasting results and the accuracy of mid-long term forecastfailed to satisfy the need of practice.In this paper, according to the similarity of historical runoff, a forecasting model basedon similar derived method is established. This model is used to forecast the monthly runoffof the Three Gorge and the forecasting results are satisfied. The work and achievements inthis paper are as follows:（1） Created theories of functional similarity space, and defined similarity space, similarityaggregation, similarity neighborhoods and similarity mapping;（2） Established forecasting model based on the theory of functional similarity space. Therunoff process in the period of the research are in a certain functional space with thegeneral similar characteristics and the value the runoff can form a similar aggregation.What’s more, the similar characteristics in the similar neighborhoods of the functionare quite remarkable, which is called similar mapping. According to thesecharacteristics, some properties of a function can be derived from other functions in itsneighborhood. That’s similar derived method proposed by this paper;（3） The evaluated parameters in this model are β, m, n, R^min, R^maxand λ^{（i）}（i=1,2,…, n）.R^min, R^maxand λ^{（i）}（i=1,2,…, n）are evaluated by the square programming while m、n and β by enumeration;（4） Took year as the forecasting period and used similar derived method forecast themonthly runoff of the Three Gorge. Compared forecasting results by similar derivedmethod with the results by SAR model. And it shows that similar derived method iseffective and the results can fit the trend of observation. The determination coefficient,precision and percent of pass of results by similar derived method are better the theseby SAR.