Grey-box modelling and forecasting of stormwater flow in sewer systems

The goal of this thesis is to wrap up my endeavours in the realm of modelling and probabilistic forecasting. The work is anchored in my two quite different papers and is therefore essentially two-legged. The first leg deals with the development and estimation of a grey-box model suited for forecasting, and this is the primary subject of the thesis. In this case, I study the modelling and forecasting of rainfall-response in a Danish stormwater tunnel, but the focus is really on the modelling process and its inherent challenges rather than the specific case. This exploration would therefore be relevant for practical grey-box modellers within many fields, e.g. wind power forecasting and financial forecasting. The second leg deals with forecast evaluation, and is exemplified by applying some evaluation metrics to the stormwater forecasting model developed in the first leg followed by a discussion of what kind of value is gained from such an evaluation effort. Again, the relevance of the demonstrated work on forecast evaluation is not limited to the specific case, but can be equivalently applied to probabilistic forecasts in other areas.