基于集对分析的水资源相似预测模型

Set pair analysis based on similarity forecast model of water resources

准确预测水资源的丰枯变化可为水安全管理提供重要的科学依据。针对水资源过程的自相依性不强、水资源与其影响因子之间呈复杂的非线性关系这一特点,为避免用参数统计预测模型在确定和求解显式数学函数模型过程中所带来的不确定性问题,提出了用集对分析检验和互相关系数统计假设检验相结合识别影响水资源变化的主要物理因子,用集对分析方法从同、异、反三方面定量刻画水资源历史样本之间的相似性,然后用多个最相似的历史样本的水资源加权平均值作为当前水资源的预测值,进而建立了基于集对分析的水资源相似预测模型(SPA-SF)。研究结果说明:用SPA-SF预测水资源的丰枯变化,物理概念清晰,计算直观,预测精度较高;这套SPA-SF建模方案简便有效,预测技术完备,在历史样本丰富、代表性高的水文时间序列预测中具有应用价值。

Exact forecast of abundant and insufficient change of water resources can be used for the important scientific basis of water security management. In order to deal with weak dependence of water resources process and complex relation between water resources process and its influencing factors, the uncertainty problems are avoided during the process of selecting and resolving the obvious mathematics functions with parametric statistics prediction methods, both set pair analysis and statistic test of correlation coefficients can be used to determine the main physics vectors for affecting change of water resources, set pair analysis method can be used to quantificationally compute the similarity between the nonce and history main physics vectors from the views of identical, discrepant and contrary sides, the weighted average of the water resources values of the nearest neighbor history samples can be regarded as the predicted value of the nonce water resources, then the set pair analysis based on similarity forecast model of water resources change, named SPA-SF, is established. The application results of SPA-SF show that the statistic and physical concepts of SPA-SF are distinct, its computation method is visual, its precision is high, and the scheme of modeling SPA-SF is simple, effective and hydrological time series with plenty representative history perfect, so SPA-SF can be applied for forecasting different samples.