基于相似过程衍生的月径流概率预报模型研究

Monthly Runoff Probabilistic Forecast Model Based on Similar Process Derivation

针对传统月径流预报模型存在的缺陷,建立了相似过程衍生法与概率预报相结合的月径流概率预报模型。运用相似过程衍生法发布确定的预报结果,在定点预报的基础上利用概率预报提供一定置信水平下的预报区间作为模型预报结果。模型结构简单、易于构建且建模过程中无需考虑预报因子的选择问题。将该模型与BP神经网络模型进行对比仿真试验,结果表明该预报模型具有较好的预报精度,且合格率高于BP神经网络模型,可在水库月径流预报中推广应用。

Aiming at the deficiencies of traditional monthly runoff forecasting model,a monthly probability forecast model is established by combing similar process derivation method with probabilistic forecasting.Certain forecast result is given by using similar processes derivation.On the basis of point forecasting result,probabilistic forecasting model is used to calculate interval forecasting under a given confidence level.The structure of the model is simple,easy modeling and unnecessary to consider selection of predictor.Compared this model with BP neural network forecast model in simulated experiments,the results show that the proposed model has a high accuracy,and the qualified rate is higher than that of BP model,which can be popularized in monthly runoff forecasting.