论文部分内容阅读
以北方某供水水库为例,将人工免疫识别系统(Artificial immune recognition system,AIRS)作为一种新兴数据挖掘方法用于水库调度规则提取研究,所获调度规则对检验样本的分类精度为86.1%,缺水指数为2.14(10~(14)m~6),优于RBF提取调度规则的结果.进一步从调度规则与训练、检验样本之间的距离分布间接分析了在水库调度特征属性构成的高维非线性决策空间中亲和力测度方法、学习样本空间分布以及追加样本对AIRS提取供水调度规则的行为与性能的影响.结果表明,①HVDM(Heterogeneous value difference metric)距离测度方法能融入调度决策先验知识,使AIRS提取的规则在空间分布上更接近训练、检验样本;②所获规则分类性能不仅取决于训练样本空间分布,也取决于检验样本的空间分布;③追加近期调度资料可以不断更新或补充规则,使其空间分布更适应未来水文环境变化.
Taking a water supply reservoir in the north as an example, Artificial Immune Recognition System (AIRS) is used as an emerging data mining method for reservoir scheduling rules extraction. The classification accuracy of the test results obtained by the scheduling rules is 86.1% The water deficit index is 2.14 (10 ~ (14) m ~ 6), which is better than the results of RBF extraction scheduling rules.Furthermore, we analyze the distribution of the dispatching rules between training schedule and testing samples, The results show that: 1) The Heterogeneous value difference metric (HVDM) distance measurement method can incorporate the priori knowledge of the scheduling decision so that the spatial distribution of the learning sample and the additional sample influence the behavior and performance of the AIRS extraction water supply scheduling rule The rules extracted by AIRS are closer to training and testing samples in spatial distribution. ② The performance of the rules classification obtained depends not only on the spatial distribution of training samples but also on the spatial distribution of test samples. ③ Adding recent scheduling information can continuously update or supplement the rules, Make its spatial distribution more adapt to the changes of the future hydrological environment.