Economic Analysis and Forecasting Potential Flexibility in District Heating

Abstract

Demand-side flexibility is a significant part of the global energy transformation and the rapid growth of renewable energy. The District Heating (DH) system, which represents a substantial amount of energy consumption in urban areas, due to its centralized infrastructure and thermal inertia, can offer potential flexibility. The accuracy in flexibility evaluation in a scalable way is still a challenge. This master’s thesis assesses the DH demand flexibility potential by breaking it down into temperature-dependent and temperature-independent components. The key focus of this study, the temperature-independent component, which is the key indicator for flexibility potential, can be adjusted without affecting the consumer’s comfort. Linear methods (Linear Regression, Ridge, Lasso) were used as a baseline, advanced machine learning algorithms (XGBoost, LightGBM), and recurrent neural networks (GRU, LSTM) were utilized to model the DH data to capture the breakdown.

The analysis was conducted based on real-life data of Helsinki’s DH network, and operational data sets were trained to validate the models and to estimate the flexibility potential of DH demand. A range of modeling techniques has been utilised on the data sets, and among all these, specifically gradient boosting, is effective in evaluating the flexibility potential of district heating (DH) systems. The findings highlighted the flexibility of the DH system towards market stability enhancement and facilitating the demand response strategies. However, several limitations have been identified in the scope of available datasets and fundamental speculations of the models. To establish the result as accurate and reliable, and impactful in real life, more research is required on enlarging the types of datasets and adopting hybrid modeling approaches.