基于支持向量机的水电站中长期径流组合预报

中长期径流预报是充分利用水资源、实现水电站优化运行的重要环节。以金沙江中游龙盘电站为研究对象，分别采用自回归滑动平均模型、最近邻抽样回归模型、BP神经网络建立了该水电站月径流预报模型，在分析三种模型预报结果具有一定互补性的基础上进一步建立了支持向量机分月组合预报模型。统计结果表明，与单一预报模型相比，该组合预报模型具有更高的精度和稳定性，为寻求水电站径流预报规律和制定中长期调度计划提供了技术支持。

Medium and Long-term Runoff Combinational Forecast Based on Support Vector Machine

Medium and long-term runoff forecast is a key step to make full use of water resources and realize the hydropower station optimal operation. Taking Longpan station for an example, located on the middle reaches of Jinshajiang River, autoregressive moving average model, nearest neighbor bootstrapping regression model and BP neural network are selected to establish monthly runoff forecast model. The analysis shows that there are certain complementarities between three kinds of model. So, monthly combinational forecast model based support vector machine (SVM) is established. Compared with single forecast model, the results show that combined forecasting model has higher accuracy and stability, which provides technical support for discovering runoff forecast rules and formulating medium and long-term dispatch plan of hydropower station.