Exploring Fast Modelling Approaches for Groundwater Well Drawdown

Groundwater modelling plays a crucial role in managing water resources, especially in areas dependent on wells for water supply. This thesis focuses on groundwater modelling for well drawdown using analytical approaches, specifically Theis solution with boundary conditions, and machine learning techniques, employing MLPregressor in Python. The main objective of the research is to determine the drawdown of an observation well to establish a time frame for optimization aimed at providing assistance to a water supply company, (named Landeswasserversorgung and will be abbreviated by LW) in effectively managing energy prices for the purpose of pumping water from the wells. Initially, assumptions are made for parameters such as pumping rate, time intervals, transmissivity, storativity, radial distance, and distance from the river. However, later obtained information from LW, including the type of well system, locations of wells, and type of aquifer soil, enables refinement of the modelling approach. The Theis solution model is used to calculate drawdown values for assumed pumping rate values and predict drawdown values for real pumping rate values using an approximation approach. In addition, a machine learning model based on MLPregressor in Python is developed to perform three different cases. The first case is to predict drawdown values for random pumping rate values using Theis solution. The second case is to predict drawdown values for real pumping rate values using Theis solution. The third case is to predict drawdown values for real pumping rate values using both the pumping rate values and drawdown values provided by LW as inputs. The research findings test the effectiveness of both analytical and machine learning approaches for predicting well drawdown, and contribute to the field of groundwater modelling. Further research can be conducted to refine and validate these models, and explore their applicability in other hydrogeological settings. Keywords: Groundwater modelling, Pumping rates, Drawdown, Theis solution, Machine Learning, MLPregressor, aquifer, Multi-well System, Landeswasserversorgung, Data Frame