Within the framework of the DR-RISE project, partners at SUPSI have published a new scientific paper entitled “Nonlinear reconciliation: Error reduction theorems”, advancing the state of the art in forecast reconciliation methodologies.
Accurate and consistent forecasting is a cornerstone for the effective deployment of residential Demand Response (DR) solutions, as envisioned by DR-RISE. In particular, DR strategies rely on forecasts at multiple aggregation levels, from individual households to neighborhoods and grid-level indicators, which must remain coherent across spatial and temporal scales.
The paper addresses a key methodological challenge in this context by extending reconciliation techniques to nonlinear settings and providing theoretical guarantees on error reduction. These results strengthen the mathematical foundations of advanced forecasting pipelines that can be applied to energy demand, flexibility availability, and system-level indicators relevant for DR-RISE use cases.
In addition to the theoretical contribution, the authors released an open-source Python package implementing the proposed nonlinear reconciliation methods, supporting transparency, reproducibility, and future integration into applied energy analytics tools.
This research directly supports DR-RISE’s objective of enabling robust, data-driven decision-making for residential demand response and contributes to the development of more reliable and scalable forecasting frameworks for the European energy system.
Read the full paper: https://openreview.net/forum?id=dXRWuogm3J