Abstract: By studying the medium and long term runoff forecast model of Nanpan river basin, and making full use of the advantages of different types of runoff forecast models, an ensemble forecast model is established. Based on the forecast accuracy of each model, the combination weight of each model is assigned. The model can establish an adaptive selection technology based on the autonomous learning of machine, which can continuously and automatically optimize the combination weight coefficient of each model and improve the forecast accuracy. It provides scientific decision-making basis for the optimal reservoir operation and the formulation of power generation plan in Nanpan river basin.

Key words: medium and long term runoff forecast, ensemble forecast, adaptive selection technology, intelligent screening and sorting, optimal weighting method, forecast scheme, simulation accuracy

摘要： 对南盘江流域的中长期径流预报模型进行研究，充分利用不同类型径流预报模型的优点，建立一种集合预报模型，并根据各模型的预报精度排序，对各模型的组合权重进行分配。该模型可以基于机器的自主学习建立自适应挑选技术，不断自动优化各模型的组合权重系数，提高预测精度，为南盘江流域水库优化调度与发电计划制定提供科学的决策依据。

关键词: 中长期径流预报, 集合预报, 自适应挑选技术, 智能筛选排序, 最优加权法, 预报方案, 模拟精度