Cubic Grey Relational Luong Attention Bidirectional Long Short-Term Memory based Dissolved Oxygen Prediction in River

Abstract

Accurate prediction of the level of DO is a significant issue to be addressed for ensuring rationality of river inhabitants. ln this work, to enhance DO detection accuracy with lesser error, Cubic Grey Coefficient and Luong Attention based Bidirectional Long Short Term Memory (CGC-LABiLSTM) is proposed. The Cubic Spline selects the parameter value obtained at similar data on an adjacent data from sample data therefore filling missing value. Also, first derivatives and second derivatives are applied to the resultant values to reduce data loss. Second, Taguchi Grey Relational Coefficient-based (TGRC) feature selection to choose the features of unique time sequence data of Link River below Keno Canal and Klamath River above Keno Dam nr Keno. Finally, the LABiLSTM is produced for DO prediction. The Luong attention based BiLSTM adjoins forward and backward propagation that can seize the attributes of DO and better the prediction accuracy and efficiency. The experimental results demonstrate that our CGCLABiLSTM method outperforms traditional DO Prediction methods based on RMSE, NSE. DO prediction accuracy and DO prediction time using link River below Keno Canal, Near Klamath falls with data gathered from two distinct locations for site 11509370 and site 11507501 respectively.