改进的非参数时间序列模型在地下水位动态预报中的研究

Research on Groundwater Dynamic Level Prediction Based on Improved Nonparametric Time Series Model

针对地下水位动态预测常用方法中存在的不足,在借鉴相似预测理论思想的基础上,提出以最大相似度准则来刻画时间序列预测中各个样本的关联性,给出了以最大联系度来定量测度地下水位历史样本之间的相似度,提出了改进的非参数时间序列模型在地下水位动态预报的建模步骤和求解思路,并将五元联系度模型应用于中牟县18号观测井的地下水埋深年内预报中。预报结果表明,改进的非参数地下水位动态预报模型预测值与实测值具有较好的拟合优度,两者呈现较强的趋势性,同时模型的有效性和可靠性经后验差法检验表明预测精度较高,泛化能力较好。

Due to the shortage of current methods in groundwater level prediction,the rule of maximum similarity is proposed to describe the relevance between samples on the basis of theory of similarity forecast.Maximum connection degree is used to measure the similarities among historical samples of groundwater level quantitatively with time series consistency analysis.Steps of modeling and solving a five-element connection degree model for groundwater level prediction is applied in monthly and inter-annual depth forecast at 18#observation well in Zhongmu County.The results indicate that goodness of fit and trend between predictive value and measured value is optimal.Besides,the model is proved to be high prediction accuracy and better generalization with posterior-variance-test.