Groundwater Level Prediction Based on Hybrid GRU with Grey Wolf Optimizer Approach

Groundwater level (GWL) forecasting accuracy at many time scales is critical for water resource scheduling and agricultural management. However, forecasting GWL in some locations remains difficult due to a lack of hydrogeological data and a highly nonstationary, nonlinear, and complicated GW system. The creation of dependable GWL simulation models is critical for taking effective management measures. The present work proposes a novel and very accurate technique for forecasting GWL using a combination of the grey wolf optimizer (GWO) metaheuristic algorithm and gated recurrent unit (GRU) deep learning model to simulate monthly GW of Kendrapara district taking data of Kasoti observation well from 1998 to 2017. Based on coefficient of determination (R2 = 0.9921), Nash–Sutcliffe efficiency (NSE = 0.988), and root mean square error (RMSE = 0.965), the GRU-GWO hybrid model outperformed the standalone GRU model. In overall, the suggested model’s results are positive, and it gives a credible understanding for water resource planners performing future groundwater resource study.