Abstract
Buildings are responsible for a significant part of global energy consumption and represent an important source of flexibility for future energy systems. As electricity grids move toward higher levels of renewable generation, upgrading buildings to become active, responsive, and grid-interactive assets becomes essential for balancing loads, preventing congestion, and enhancing overall grid reliability. This dissertation develops methods and frameworks to evaluate and update buildings into Smart Grid-Interactive Efficient Buildings (SGEBs), enabling demand-side flexibility in emerging local flexibility markets (LFMs). The research focuses on five main themes: evaluating building smartness and smart readiness, modeling building energy systems, estimating flexibility potential, forecasting operational flexibility, and tackling deployment challenges.
The primary case study focuses on four large non-residential buildings in Ålesund, Norway, which were extensively monitored, calibrated, and analyzed. These pilots, representing different types and construction periods, form the core of the dissertation, allowing for the integration of assessment methods, modeling approaches, and flexibility evaluations. Additionally, supplementary case studies from Norway and Italy are included to expand the scope of the developed methods.
For the assessment of building smartness and smart readiness, a review of existing methods was conducted with an emphasis on their applicability to the Norwegian context. From this review, the Smart SRI framework was created by combining the European Smart Readiness Indicator (SRI) with the Norwegian Smart by Powerhouse (SbP) assessments. This adaptation considers local conditions, such as Norway’s high level of electrification. Additionally, the SGEB score was introduced as a new evaluation metric, similar to energy performance certificates (EPCs), to measure the extent of building–grid interaction and flexibility potential across systems.
For modeling and flexibility estimation, the four Ålesund pilots were calibrated using both data-driven indoor-space models and physics-based simulations with EnergyPlus and Python-based optimization. These models served as the basis for quantifying flexibility in LFMs. Two types of flexibility were evaluated: baseline-dependent flexibility, identified via thermal setpoint adjustments, and baseline-free flexibility, represented by a battery energy storage system (BESS) integrated with photovoltaics. Flexibility assessment involved Monte Carlo simulations that accounted for uncertainties in event timing and price signals, reflecting the nature of LFM conditions.
On the forecasting side, machine learning pipelines were developed and tested for both baseline and flexible-load prediction across multiple pilots, with evaluation focusing on accuracy, stability, and generalizability. The Interpretable Machine Learning for Energy Management System (IML-EMS) pipeline achieved the best performance. It is designed as an end-to-end package that can be seamlessly integrated into existing Building Management Systems (BMSs), while its differentiable structure makes it suitable for use in advanced control applications.
Finally, the dissertation examines the technical, methodological, and market challenges of implementing building flexibility in practice. The findings show that integrating targeted assessment methods with strong modeling and forecasting techniques can improve building flexibility, increase participation in flexibility markets, and aid the shift toward a more sustainable and resilient energy system.