Data-based Predictions of Load Profiles for Buildings for Flexible Optimization

Abstract

The flexible usage of modern buildings results in varying load profiles. This means that internal loads, which are often critical for both energy consumption and the thermodynamics of the building, can be of the type of residential or commercial buildings, or a combination of both. Nevertheless, typical usage patterns arise in residential and non-residential buildings. These electric load profiles can be measured, and based on this measurement data, dynamic models can be designed that serve as a basis for prediction. Such predictions, which are adapted to the specific use case, can subsequently be used for optimized operation management (heating/air conditioning, storage management, sector coupling, etc.). In the present work, dynamic mode decomposition is used for data-driven modeling and predicting the load profiles of buildings with mixed usage. This enables adaptive yet reliable predictions in buildings with time-varying mixed usage. Utilizing the structure of a Takagi-Sugeno fuzzy system for energy management a seamless weighting between residential and commercial usage becomes possible.