CAJ | 학술논문

针对传统的投影寻踪技术中投影与寻踪两个核心问题进行改进，用实数编码的遗传算法代替高斯-牛顿进行投影方向优化、利用非线性参数Hermite多项式代替非参数方法估计岭函数，建立基于改进的投影寻踪预测技术水电站入库流量预报模型，并将模型用于杂谷脑流域水电站的入库流量预测，研究表明改进的投影寻踪技术能够以较快速度收敛于模型全局最优解，研究成果为杂谷脑梯级水电站中期调度奠定了基础，也为其他流域水电站中期入库流量预报提供了重要借鉴。
침대전통적투영심종기술중투영여심종량개핵심문제진행개진，용실수편마적유전산법대체고사-우돈진행투영방향우화、이용비선성삼수Hermite다항식대체비삼수방법고계령함수，건립기우개진적투영심종예측기술수전참입고류량예보모형，병장모형용우잡곡뇌류역수전참적입고류량예측，연구표명개진적투영심종기술능구이교쾌속도수렴우모형전국최우해，연구성과위잡곡뇌제급수전참중기조도전정료기출，야위기타류역수전참중기입고류량예보제공료중요차감。
In this paper, two core issues of the traditional forecasting technology, projection and pursuit, are improved. The real-coded Genetic Algorithms, which insteads of the Gauss-Newton method, is used to optimize the projection di-rection. And the nonlinear parameter Hermite polynomial, which takes the place of non-parametric method, is used to esti-mate the ridge function. The plant inflow forecasting model which is established on the basis of the improved projection and pursuit forecasting technology is applied to the inflow forecasting of Zagunao basin hydropower station. The study indicates that the improved Projection and Pursuit technology can converge to the global optimal solutions at a faster rate. The foun-dation of the interim scheduling of Zagunao cascade hydropower stations and the important reference of the mid-term inflow forecasting of other basins plant are provided by the research result.