Research On The Ten-day Runoff Forecasting Method Of Huanren Basin In Flood Season And Its Application

The description of runoff input as one of the most important factors can impact the reservoir make optimal scheduling, it plays an important role in making water supply plans, power generation plans and real-time optimization scheduling decision-making process etc. Combined with reservoir runoff input information to predict the scheduling is an important non-engineering measures to fully exploit the potential of hydropower stations that have been built, to improve the comprehensive benefits of reservoir, to bring great economic benefits to the society without any increase in engineering investment. However, due to the complexity of the hydrological runoff forecasting and lack of breakthrough in understanding the mechanism and causes, there are still problems in forecasting scheduling theory and application, further research is needed. Therefore, under the guidance of the medium runoff forecasting and optimization scheduling, taking Huanren hydropower station as the research object, this article deeply discussed and researched the key technologies for the selection of predictors for medium runoff forecasting runoff and the formulation of Optimal scheduling rules combination of medium and long-term runoff forecasting on the basis of the incoming runoff characteristics and operating characteristics of Huanren reservoir. Main contents and conclusions of this study are summarized as follows:(1) The previous methods of forecasting ten-day runoff select definite predictors to forecast. This paper analyzes the different ten-day runoff characteristics of Huanren reservoir basin during flood season, improves the previous methods, proposes the approach of selecting various predictors for each ten-day in flood season to forecast ten-day runoff. The paper use a total of42years from1969-2010of Huanren reservoir to calculate and analyze, select1969to2000a total of32years of data as the rate period, from2001to2010a total of10years of data as the test period. The proposed method is applied to ten-day runoff forecasting of Huanren reservoir basin, during flood season. The most influential predictors for each ten-day in flood season are respectively selected as the variable predictors, combined with the multiple linear regression method and BP neural network method, the ten-day runoff forecasting models are constructed. The calculation results show that, the ten-day runoff forecasting model based on various predictors acquire higher forecasting precision and more ideal effects than the ten-day runoff forecasting model based on definite predictors both in rate period and in test period, which indicates that the variable predictor can better reflect the basin’s ten-day runoff variety and rules. Contrast of the two forecasting methods, the paper obtained the conclusion that the multiple linear regression method based on variable predictors is a more suitable method for ten-day runoff forecasting of this basin on the analysis of the forecasting results by the two methods. The ten-day runoff forecasting information forecasted by the proposed method is very important reference information for Huanren power station to make ten-day power dispatching scheduling program. The propose ten-day runoff forecasting method based on variable predictors is meaningful, significant and practical to other basins of China.(2) According to years of scheduling operation of Huanren hydropower station, the decision tree method is used to mine the dispatching scheduling rules of during flood season. Firstly, through analysis, the coming runoff of each month is impacted by different weather systems, so the flood season’s dispatching scheduling rules is divided into:May and June, July and August, September and October the three stages:to analysis. The beginning reservoir’s water level and ten-day average income flow are chosen to be impact variables and ten-day average power flow as a decision variable. In this paper the classic K-means clustering algorithm is used in the data pretreatment to grade the three variables into some phases, and then using the decision tree data mining technique which is widely used in practice to get different stages of the decision tree. Via the tree we get the reasoning mode or rules by the classification of three variables, and then in order to facilitate understanding and practical using,the rules are organized to the forms of if-then. At last, a few typical examples are selected to calculated by the rules got from the article, and compared to the results got from Huanren reservoir of conventional operation chart and Huanren reservoir actual scheduling process, through the contrast we find that the dispatching scheduling rules can reduce the abandoned water and improve the power benefit of Huanren hydropower plant.