As renewable energy becomes more cost-competitive, it is emerging as a viable energy source for chemical processes.
However, renewables exhibit intrinsic variability, whereas most chemical processes operate under predominantly steady-state conditions....
As renewable energy becomes more cost-competitive, it is emerging as a viable energy source for chemical processes.
However, renewables exhibit intrinsic variability, whereas most chemical processes operate under predominantly steady-state conditions.
This mismatch necessitates careful consideration in traditional decision-making frameworks—spanning design, scheduling, and control—to ensure stable and efficient operations.
This thesis proposes strategies for multi-timescale decision-making in renewable-integrated energy systems.
First, a design optimization framework is proposed for an industrial-scale, renewable-based green hydrogen production process.
Existing studies often overestimate the flexibility of electrolyzer operation, neglecting the degradation and reduced lifespan caused by frequent shutdowns.
Moreover, many analyses focus on specific hydrogen demand scenarios, thereby failing to capture the broader economic impact of cooperation across the value chain.
The proposed framework accounts for realistic electrolyzer constraints and incorporates various cost-reduction strategies—such as renewable-aligned operation of downstream processes
and demand decentralization—using a mixed-integer linear programming (MILP) model over one year of regional climate data.
Applied to regions in the United States where industrial hydrogen facilities are currently operational, the results show that optimal electrolyzer strategies can lower the Levelized Cost of Hydrogen (LCOH) by 5.3%.
Fully distributed end-user supply can further decrease LCOH by up to 39.8%, while supply tolerance contributes an additional 24.2% reduction.
Case studies also reveal that access to the power grid can reduce LCOH by 12.9% in Houston, highlighting the potential for cost improvements as the grid becomes cleaner.
These findings underscore the importance of optimizing operational strategies to enhance the economic performance and market viability of green hydrogen systems.
Second, a multi-timescale energy management system (EMS) is developed that leverages real-world numerical weather forecasts to mitigate renewable variability.
The proposed EMS integrates reinforcement learning (RL) with model predictive control (MPC), enabling safe and reliable operation under uncertain conditions.
This approach is validated on a comprehensive, experimentally verified dynamic model of an electrolysis system, using both forecasted and measured solar data from South Korea.
Compared to a rule-based baseline, the proposed framework boosts hydrogen production by 13.5% while minimizing number of annual shutdowns,
underscoring the value of weather forecasting for maintaining stability without sacrificing productivity or operational safety.
Additional case studies confirm its adaptability across different regional climates and its effectiveness in improving hydrogen production efficiency.
Finally, risk-aware bidding strategies are proposed for managing electricity price volatility in electricity markets.
While price signals offer economic incentives to balance supply and demand, they also expose renewable system operators to substantial financial risk.
To address this challenge, a robust optimization-based bidding framework is developed, using a temporal fusion transformer model to approximate price uncertainty.
This approach effectively reduces revenue variance 13.5% without sacrificing the average annual return.
To further improve profitability, a stochastic optimization technique is employed, generating price scenarios via stochastic differential equations.
Simulation results show higher returns compared to the average energy storage system (ESS) performance in the California Independent System Operator (CAISO) market,
indicating the framework’s potential to model market volatility more accurately.
Focusing on the distinct timescales of design, scheduling, and control, this thesis highlights the necessity of matching appropriate data and methodologies to each decision hierarchy.
By offering optimized strategies for renewable-integrated processes, it is expected to contribute to the broader transition of chemical processes toward cleaner,
more cost-effective renewable energy sources, ultimately fostering more sustainable energy systems.