Case Study on the Application of Machine Learning in Hydrology

Sammanfattning

This thesis presents a case study on the application of artificial intelligence (AI) in hydrology using two approaches. This topic was chosen since AI is an important tool, and water is a vital resource. The first case is on leakage detection with artificial data, while the second is on precipitation prediction with real data. For leakage detection, a random forest (RF) model and an artificial neural network (ANN) model were developed in series. The RF detects if there has been a leak or not; the detected leaks are forwarded to the ANN. The ANN determines where in the water distribution network the leak is located. This design achieved an accuracy of 88.6% and a precision of 94.1%. For precipitation, six models were developed; one spatial model and five area-specific prediction models. The models developed for precipitation outperformed the current satellite products with R2 values above 0.5 for five out of six models. The results from both parts of the case study show promise, and hydrology is definitely a field into which AI should be further integrated.