Impact of prosumers on the accuracy of load forecast with neural networks

This master thesis answers the research question how prosumers affect the accuracy of the day-ahead load forecast with neural networks. Prosumers are households consuming their self-produced electricity. The surplus electricity is fed into the grid and, if additional electricity is needed, it is consumed from the grid (Bundesnetzagentur, 2018). The answer to this research question is relevant for power utilities as with the promotion of renewable energies an increasing share of prosumers is expected in Switzerland. In contrast to related research on prosumers and load forecast, this thesis addresses the impact of different shares of prosumers on the load forecast for areas with several households. In order to answer this research question, the load forecast accuracies for a dataset without prosumers is compared to the ones of datasets with different shares of prosumers in an experimental setup. This master thesis is therefore following mainly an experimental research design. In the course of the experiments load data and further variables such as weather data or data related to the date are used to train the neural networks. Thereby a sliding window approach is applied. This implies, that the input variables are lagged up to seven days. To achieve the best possible load forecast accuracy for a dataset without prosumers and a dataset consisting only of prosumers, the neural networks are parametrized following the design research methodology....