Surrogate Modeling for Green Hydrogen Site and Electrolyzer Sizing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Designing large-scale green hydrogen production plants that utilize intermittent renewable energy sources is challenging due to unpredictable power supply, leading to inefficient operation, high costs, and potential electrolyzer degradation, with existing methods failing to optimize electrolyzer numbers, sizes, and site locations effectively.
Innovation Solution
A computer-implemented method using physics-informed machine learning models and rigorous dynamic process models to create a surrogate process model, enabling multi-objective design optimization that considers site selection variables and constraints, optimizing electrolyzer and storage device numbers, and predicting optimal sites for renewable energy plants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional design methods are used for green hydrogen production plants, then the design process is simpler, but the levelized cost of hydrogen production is higher and operational efficiency is lower
Solution Approach 1:
The patent creates a digital twin (virtual replica) of the hydrogen production plant that mirrors the physical system's behavior, dynamics, and constraints. This digital copy enables comprehensive optimization simulations without affecting the actual plant, allowing complex multi-objective optimization to be performed on the virtual model and then applied to the physical system, thereby improving operational efficiency while keeping the actual design process manageable.
Solution Approach 2:
The patent performs design optimization before the actual plant construction and operation. By using the digital twin to simulate and optimize the plant design in advance, considering multiple objectives simultaneously (cost, efficiency, reliability), the optimal configuration is determined prior to implementation, ensuring high operational efficiency from the start without requiring complex adjustments during operation.
2Quantity of substance
If traditional design methods are used, then computational resources are saved, but the optimization of electrolyzer numbers, sizes, and site locations is insufficient
Solution Approach 1:
Instead of performing computationally intensive optimizations directly on the physical system or using excessive computational resources, the patent creates a digital twin that replicates the system's behavior. This virtual copy allows for high-precision multi-objective optimization to be performed on the digital model, achieving accurate optimization of electrolyzer numbers, sizes, and site locations without requiring disproportionate computational resources for actual implementation.
3Object-generated harmful factors
If intermittent renewable energy sources are used, then carbon emissions are reduced, but the power supply becomes unpredictable leading to electrolyzer degradation
Solution Approach 1:
The patent uses the digital twin to simulate various renewable energy scenarios and their impact on electrolyzer operation before actual implementation. By predicting the effects of intermittent power supply on electrolyzer degradation in the virtual model, the system can pre-determine optimal operating strategies, buffer configurations, and control parameters that protect the electrolyzer from degradation while maintaining high renewable energy utilization, thus preserving reliability when using clean energy sources.
Solution Approach 2:
The digital twin continuously monitors and analyzes the relationship between renewable energy input variability and electrolyzer performance. This feedback mechanism allows the system to learn from simulated operations and adjust control strategies to mitigate the harmful effects of intermittent power supply on electrolyzer stability, enabling the system to maintain reliability while maximizing the use of carbon-free renewable energy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present disclosure provide optimal site selection framework. A surrogate process model may be generated based on one or more physics-informed machine learning models and a rigorous dynamic process model. The surrogate process model may comprise a representation of a process including one or more target assets and one or more secondary assets. The process input for the one or more target assets may comprise power output from a renewable energy source. A design optimization algorithm representative of a multi-objective design optimization problem may be generated. The design optimization algorithm may comprise one or more site selection variables and one or more site selection constraints. One or more outputs may be generated by executing an optimization model based on the surrogate process model and the design optimization algorithm. The one or more outputs may comprise predicted optimal sites and an optimal conceptual design for the process.