摘要
鉴于其优越的预报性能,将相关向量机(RVM)应用到中长期径流预报中,并在相空间重构的基础上,建立了基于相关向量机的径流预报模型.该模型首先对径流时间序列进行相空间重构,并以重构后的径流序列作为模型输入;其次,采用粒子群优化(PSO)算法识别模型参数,利用优化所得重构参数验证时间序列具有混沌特性,在模型内循环过程中采用EM算法迭代估计超参数,并将RVM与应用较为广泛的最小二乘支持向量机(LSSVM)和自动回归滑动平均模型(ARMA)进行了比较分析,结果表明该模型具有较好的泛化能力;最后,基于水文过程变化的不确定性、RVM描述输出值的不确定度以及相应概率下的预报区间,使得调度人员在决策中能考虑预报的不确定性,定量估计各种决策的风险和效益.
Due to the superior forecasting performance,relevance vector machine(RVM) was applied to mid-and long-term runoff forecasting,and based on the phase space reconstruction,the runoff relevance vector machine forecasting model was established.Firstly,the runoff time series was reconstructed in the phase space,and the reconstructed series was as the proposed model input;Secondly,the particles swarm optimization(PSO) algorithm was applied to identifying the model parameters and chaotic properties of time series.The EM algorithm was used to estimate hyper-parameters in the inherent cycle,RVM was compared with widely used least squares support vector machine(LSSVM) and auto-regressive moving average model(ARMA).The test results show that the model has good generalization ability;Finally,in terms of the uncertainty of hydrological processes,the scheduling staffs consider the uncertainties in forecasting,and quantitatively estimate the risks and benefits in decision-making based on the uncertainty of RVM output values and the probability forecast interval.
出处
《大连理工大学学报》
EI
CAS
CSCD
北大核心
2012年第1期79-84,共6页
Journal of Dalian University of Technology
基金
水利部公益性行业专项资助项目(201001024)
国家自然科学基金资助项目(51109025)
关键词
相空间重构
相关向量机
长期径流预报
PSO算法
phase-space reconstruction
relevance vector machine
long-term runoff forecast
PSO algorithm