Description
Collecting energetic performance data from buildings for energy community planning and modeling is a challenging and elaborate task, often left to the planner’s responsibility. Our tool aims to support and facilitate this undertaking.The modeling and composition of synthetic timeseries of heating and cooling demands of buildings, is typically approached in a deterministic way. Good examples of such approaches are the degree-day method or the regression of measured data via neural networks. As a consequence of their deterministic nature, these approaches only provide deterministic results that are solely based solely on the considered inputs without stochastic variations and are thus not capable of capturing the randomness of the data. Other methods aim to represent it as an ARMA process with noise, but no tool was found so far for such approach.
The method on which our tool is based is presented here. It represents the energy
consumption timeseries as a temporal succession of normal distributions with means and standard deviations that are context dependent (related to contextual timeseries such as temperature). These parameters are estimated with historical data of both the load and its environment. Such an approach allows us to represent not only central values such as mean or median but also variations with unchanged input.
Another advantage of the method is the possibility to generate various profiles for the same a context. Indeed, as the intermediary result of our computation is a probability distribution: the generated samples are always different, allowing the computation of different possible scenarios given the same conditions, which in turn, allows to assess the robustness of a given system.
| Period | 11 Sept 2024 |
|---|---|
| Event title | International Conference on Smart Energy Systems 2024 |
| Event type | Conference |
| Location | Aalborg, DenmarkShow on map |
| Degree of Recognition | International |