摘要
以郁江流域百色水库入库径流为研究对象,分别采用基于水文气象因子的逐步多元回归模型、人工神经网络模型(气象因子)和基于时间序列的混沌理论模型、最近邻抽样回归模型、小波分析法、神经网络—自回归模型共6个模型方法对百色水库年尺度、月尺度以及旬尺度入库径流进行中长期径流预报工作,随即采用平均相对误差、合格率、TS评分以及均方根误差4种评价指标对上述6个模型的预测结果进行精度评估,并依据熵权法的理论对上述4种指标进行客观赋权形成综合性指标分析体系,来确定最优模型以进行郁江流域百色水库不同尺度和预见期的入库径流中长期径流预报工作。结果表明:各模型预报结果中的平均相对误差相对较大,但其所对应的合格率以及TS评分指标均处于优良水平;月尺度预报过程中,各模型非汛期预报精度都要高于汛期预报精度。结合各模型自身特点以及综合性指标分析的基础上,可在年尺度中长期径流预报过程中采用神经网络—自回归模型、月尺度中长期径流预报过程中采用混沌理论模型,旬尺度中长期径流预报过程中依据不同的预见期分别采用人工神经网络模型(气象因子)以及小波分析法进行相关的中长期径流预报工作,从而为郁江流域百色水库制定未来中长期调度计划提供技术支持。
Taking the inflow from Baise Reservoir in Yujiang River Basin as the object, stepwise multiple regression(SMR) model, artificial neural network(ANN) model based on hydrometeorological factors and chaotic theory(CT) model, nearest neighbor bootstrapping regressive(NNBR) model, wavelet analysis(WA) model, artificial neural network-autoregressive model based on time series are used to forecast the annual, monthly and decadal scale inbound runoff from Baise Reservoir for the mid-term and long-term runoff forecasting. Then four evaluation indexes, namely, average relative error, passing rate, TS score and root mean square error(RMSE) are used to evaluate the accuracy of the prediction results of the above six models, and a comprehensive index analysis system is formed by objectively assigning weights to the above four indexes based on the theory of entropy weight method to determine the optimal model for the mid-term and long-term runoff forecasting of incoming runoff at different scales and forecasting periods in Baise Reservoir of the Yujiang River Basin. The results show that, the average relative error of each model is relatively large, and the corresponding pass rate and TS score indexes are all at excellent levels. In the monthly scale forecasting process, the forecast accuracy of each model in the non-flood period is higher than that in the flood period. Based on the characteristics of each model and comprehensive index analysis, the neural network-autoregressive model can be used in the annualscale Mid-term and long-term runoff forecasting process, the chaos theory model in the monthly scale mid-term and long-term runoff forecasting process, and the artificial neural network model(meteorological factor) and wavelet analysis in the decadal-scale mid-term and long-term runoff forecasting process according to different forecasting periods. The mid-term and long-term runoff forecasting is carried out by artificial neural network model(meteorological factor) and wavelet analysis for different forecasting periods, t
作者
唐振宇
梁国杰
张利平
陈森林
黄馗
TANG Zhen-yu;LIANG Guo-jie;ZHANG Li-ping;CHEN Sen-lin;HUANG Kui(State Key Laboratory of Water Resource and Hydropower Engineering Science,Wuhan University,Wuhan 430072,Hubei Province,China;Guangxi Power Grid Company,Nanning 530023,Jiangsu Province,China)
出处
《中国农村水利水电》
北大核心
2023年第1期82-88,94,共8页
China Rural Water and Hydropower
基金
广西电网有限责任公司委托项目“基于雨-水-电情下郁江流域低水头水库群优化调度技术研究”。
关键词
中长期径流预报
TS评分
综合性指标分析
郁江流域
mid-term and long-term runoff forecast
TS score
comprehensive index analysis
Yujiang River Basin
