https://www.mdu.se/

Integrating bioenergy with carbon capture and storage (BECCS) into biomass-fired combined heat and power (CHP) plants offers crucial potential for achieving negative emissions. To respond to fluctuating heat demand and volatile electricity markets, CHP plants must operate dynamically, and this results in largely fluctuating operation of CO₂ capture. This dissertation aims to improve the dynamic operation of CO₂ capture in biomass-fired CHP plants to boost negative emissions through dynamic modelling, advanced control and potential assessment under different operating modes.

To provide systematic guidance for selecting appropriate modelling approaches, both first principles and machine learning (ML) approaches are established and compared. Systematic comparison is first conducted across three first-principles models (ideal static models, dynamic models with control, and dynamic models without control) under three varying operating parameters (flue gas flow rate, CO₂ concentration, and available heat). Three ML models (Informer, long short-term memory, and back-propagation neural network) are further compared across four applications (system identification, monitoring, optimisation, and performance estimation). Results show that no single model consistently outperforms the others across all cases. While Informer achieves the highest accuracy in most applications and for most target variables, model selection should be tailored to the specific application. Model predictive control (MPC) is then developed and evaluated for managing operational variability of CHP plants. MPC demonstrates superior controller performance over conventional proportional integral (PI) control, achieving a 47–62% reduction in settling time and recovery time, and a 66–74% reduction in integrated absolute errors for CO₂ capture rate.

With modelling foundations, negative emission potential is evaluated at both plant and national scales under two operating modes (OMs), both of which prioritise heat supply. OM1 maximises CO₂ capture by sacrificing electricity output while maintaining heat supply, achieving 8.7 MtCO₂/yr nationwide negative emissions at a levelized cost of CO₂ avoided of 36.9 $/tCO₂. OM2 maximises CO₂ capture while maintaining both heat and electricity supply, yielding 4.3 MtCO₂/yr positive emissions at 52.0 $/tCO₂ (but still reducing emissions by 6.3 MtCO₂/yr compared with the reference plant without CO₂ capture). The biogenic fraction of fuel emerges as the critical parameter, requiring minimum fractions of 32.8% and 84.3% for the two OMs to meet Sweden’s 3 MtCO₂/yr target.

The contributions of this work include: (i) systematic guidance for dynamic model selection tailored to different CO₂ capture applications, (ii) quantitative evidence of MPC’s superiority over PI control under realistic CHP dynamic scenarios, and (iii) a national-scale BECCS potential and cost assessment for Sweden under maintained heat supply constraints. Results demonstrate that Sweden’s BECCS climate targets (3–10 MtCO₂/yr by 2045) are technically achievable, as OM1 alone can deliver 8.7 MtCO₂/yr negative emissions. The choice between operating modes represents a fundamental trade-off between maximising carbon removal and maintaining electricity supply. These results offer quantitative guidance for policymakers weighing carbon removal ambitions against energy system constraints.