Optimal Dynamic Operation of Power to Ammonia Processes

The PowertoAmmonia (PtA) process is widely regarded as a key pathway to decarbonise global ammonia production and enable carbonfree energy storage of renewable power. Ammonia is synthesised from hydrogen and nitrogen via the Haber–Bosch (HB) process, which has traditionally operated under stable conditions to maximise production efficiency. However, the fluctuating nature of renewable energy introduces an entirely new operating regime for HB plants, requiring frequent and rapid load changes. This is particularly challenging because ammonia reactors exhibit highly nonlinear dynamics and openloop instabilities when subjected to such disturbances.

This thesis aims to develop modelling frameworks, optimisation strategies, and control architectures for flexible operation of the HB process in PtA plants, enabling integration with intermittent renewable power. A dynamic model of a fixedbed ammonia reactor is developed as the foundation of this work. The model uses a novel efficient formulation of the underlying partial differential algebraic equations, enabling fast simulation compared to traditional methods. The model captures the nonlinear dynamics and complex thermal behaviour of ammonia reactors subjected to variable conditions.

The model was extended to represent an adiabatic quenchcooled reactor (AQCR), incorporating heat exchangers and mixers. Simulations revealed the onset of oscillatory behaviour and extinction phenomena under load disturbances, highlighting the instability risks of flexible operation. Optimisation studies explored the tradeoff between conversion efficiency and stability, and a basic control strategy was proposed to maintain safe reactor operation under load ramps of up to 1% per minute.

The analysis was expanded to include two additional industrial reactor configurations: the adiabatic indirectcooled reactor (AICR) and the internal directcooled reactor (IDCR). The three reactor types were compared across an operating range of 30–130% of nominal load. The AICR and IDCR consistently achieved higher conversions than the AQCR and exhibited more favourable dynamic and stability characteristics.

To assess the impact of reactor integration with the synthesis loop, a rigorous dynamic model of the full HB loop is developed, including an AICR, compressors, steam turbines, and flash separators. Steadystate optimisation of the synthesis loop power consumption identified loop pressure as the most influential operating parameter, varying from 220 bar at high load to around 110 bar at low load. However, antisurge constraints in the main feed compressor were found to limit power efficiency at low load, creating a strong incentive for design improvements.

Building on these findings, the study investigates the synthesisloop compressor configuration, focusing on designs with multiple parallel compressor trains. The study demonstrates
that adding parallel compressors for both feed and recycle compression can significantly reduce power penalties from antisurge recycling at low load, achieving power reductions of 55%, 74% and 84% for two, three and four parallel compressor trains, respectively. An economic analysis further evaluates the tradeoff between added capital cost and operational energy savings, identifying the most costeffective configurations for flexible PtA operation. Designs with two, three and four compressor trains in parallel all reduced the total cost of the compressor units by about 20% compared to the singletrain configuration.

The study investigates the HB synthesis loop for selfoptimising variables. Using sensitivity analysis, the study identifies controlled variables that minimise economic loss when kept constant across the load range. Maintaining constant reactor bed inlet temperatures and H2/N2ratio results in nearoptimal power efficiency, whereas fixed loop pressure causes significant energy penalties at the upper and lower load extremes. These findings support a hybrid control strategy where some variables are controlled to constant setpoints across the load range, while the loop pressure is continuously updated through realtime optimisation.

The dynamic behaviour of the full HB synthesis loop is analysed through openloop simulations for step changes in key manipulated variables, including feed and recycle compressor power and reactor bed split fraction. The simulations reveal strong nonlinear dynamics and significant process interactions across multiple input–output pairs. These insights form the basis for designing a control architecture that provides stable and efficient operation across the full load range.

A control architecture is developed for the HB synthesis loop, combining realtime optimisation with selfoptimising control (SOC). Controlled and manipulated variable pairings were determined using relative gain analyses, supported by openloop dynamic simulations. To reduce interaction between control loops, decoupling strategies were incorporated. Dynamic simulation results show that the proposed control system achieves stable and efficient operation across the entire load range (10–120% of nominal) with accurate tracking of rapid load changes of up to 3% per minute. The selfoptimising control strategy maintained energy consumption close to the optimal benchmark, while reducing the need for frequent setpoint adjustments, extending equipment lifetime while maintaining high energy efficiency.

Overall, this thesis provides new insights into the dynamic behaviour of ammonia reactors and synthesis loops under renewabledriven operation. It introduces efficient modelling strategies, identifies key optimisation and design tradeoffs, and proposes practical control solutions to enable the transition from conventional steadystate plants to future flexible PtA systems.