Projection Pursuit Clustering model for reservoir Flood Season Division

Abstract

Reservoir flood season division is an important foundation for optimizing reservoir seasonal flood limit levels and safely utilizing flood resource. In order to achieve multi-dimension clustering, time sequence partition and subjective and objective information utilization of reservoir flood season division, a Projection Pursuit Clustering model for reservoir Flood Season Division (PPC-FSD) was established in this paper. The applied results of PPC-FSD show that the first flood season, the main flood season and the last flood season of Panjiakou Reservoir of China are composed of from the first ten days of June to the first ten days of July, from the middle ten days of July to the middle ten days of August, and from the last ten days of August to the last ten days of September, respectively. The first flood season, the main flood season and the last flood season of Miyun Reservoir of China are composed of from June 15 to June 30, from July 1 to August 15, and from August 16 to September 15, respectively; the physics mechanism analysis results of reservoir flood season can expediently be utilized by using projection index function of PPC-FSD, the optimal projection direction and partition point values of flood season can be obtained by using accelerating genetic algorithm. The optimal projection values express the integration relative membership function values of periods of time belonging to the fuzzy set of flood season, which can be used to design reservoir flood limit dynamic levels, and the optimal projection direction can be used to identify key indexes affecting reservoir flood season division based on matrix of index characteristic values. PPC-FSD is visual, convenient and general, so it can be widely applied to reservoir flood management.