信号处理技术与残差分析联动的异常径流量判定方法

Anomalous runoff determination method by combining signal processing technique and residual analysis

对异常径流量的判定是保证流域安全的重要手段。针对单一判定法可能存在的误判,本文提出了信号处理技术与残差分析联动判定法。该方法从两种不同角度分析异常径流信息,首先以经验模态分解法(EMD)获得包含异常信息的高频分量,以此为基准得到异常阈值并识别相应异常值;然后以长短期记忆网络(LSTM)构建时序模型,得到包含异常信息的径流残差,进而确定异常阈值并判定相应异常值;最后融合两种方法的异常径流量判定结果。研究表明,提出的方法能够从不同角度挖掘异常径流信息,相互补充,可准确识别径流数据中不同类型的异常信息。

The determination of abnormal runoff is an important tool to ensure the safety of the watershed.To address the possible misjudgment problem of single determination method,this paper proposes a combined method based on signal processing technology and residual analysis.This method analyzes the anomalous runoff information from two different perspectives.First,the high-frequency component containing anomalous information is obtained by the empirical modal decomposition(EMD)method,which is used as a benchmark to obtain anomaly thresholds and identify the corresponding anomalous values;then,a time-series model is constructed by the long short-term memory network(LSTM)to obtain the runoff residuals containing anomalous information,which is used to determine the anomaly thresholds and determine the corresponding anomalous values;finally,the anomalous runoff determination results of the two methods are fused.The analysis shows that the proposed method can mine the anomalous runoff information from different perspectives and complement each other,and can accurately identify different types of anomalous information in runoff data.