Green Hydrogen Plant: Optimal control strategies for integrated hydrogen storage and power generation with wind energy (2108.00530v1)

Abstract: The intermittent nature of renewable energy resources such as wind and solar causes the energy supply to be less predictable leading to possible mismatches in the power network. To this end, hydrogen production and storage can provide a solution by increasing flexibility within the system. Stored hydrogen can either be converted back to electricity or it can be used as feed-stock for industry, heating for built environment, and as fuel for vehicles. This research examines the optimal strategies for operating integrated energy systems consisting of renewable energy production and hydrogen storage. Using Markov decision process theory, we construct optimal policies for day-to-day decisions on how much energy to store as hydrogen, or buy from or sell to the electricity market, and on how much hydrogen to sell for use as gas. We pay special emphasis to practical settings, such as contractually binding power purchase agreements, varying electricity prices, different distribution channels, green hydrogen offtake agreements, and hydrogen market price uncertainties. Extensive experiments and analysis are performed in the context of Northern Netherlands where Europe's first Hydrogen Valley is being formed. Results show that substantial gains in operational revenues of up to 51\% are possible by introducing hydrogen storage units and competitive hydrogen market-prices. This amounts to a \euro 126,000 increase in revenues per turbine per year for a 4.5 MW wind turbine. Moreover, our results indicate that hydrogen offtake agreements will be crucial in keeping the energy transition on track.

• The paper formulates a Markov Decision Process to optimize GHP operations, enhancing profit margins by up to 51% over systems without storage.
• The paper demonstrates that flexible hydrogen sales strategies outperform fixed contracts, leveraging dynamic market interactions for higher gains.
• The paper highlights that improved HES efficiency and optimized PPA fulfillment are vital for reducing energy losses and managing seasonal inventory effectively.

This paper, "Green Hydrogen Plant: Optimal control strategies for integrated hydrogen storage and power generation with wind energy" (Green Hydrogen Plant: Optimal control strategies for integrated hydrogen storage and power generation with wind energy, 2021), examines the optimal operational strategies for a Green Hydrogen Plant (GHP). A GHP is defined as an integrated energy system comprising a renewable energy source (like a wind farm) and a co-located Hydrogen Energy Storage (HES) facility. The HES includes an electrolyzer to convert electricity to hydrogen, storage tanks, and a fuel cell to convert hydrogen back to electricity. The core challenge addressed is the intermittent nature of renewable energy and how hydrogen storage can provide flexibility, enabling the operator to maximize profit by strategically interacting with both the electricity and hydrogen markets.

The authors formulate the problem as a Markov Decision Process (MDP) to derive dynamic optimal control policies for the GHP operator. The operator's goal is to maximize total expected profit over a one-year planning horizon by making daily decisions on:

1. How much electricity to buy from or sell to the spot market.
2. How much electricity to sell to fulfill obligations under a Power Purchase Agreement (PPA).
3. How much electricity to convert to hydrogen for storage.
4. How much stored hydrogen to convert back to electricity.
5. How much stored hydrogen to sell directly as a gas.

The state variables in the MDP capture the essential information needed for decision-making at each period (daily): the current time period, electricity market price, hydrogen market price, wind energy production, current hydrogen inventory level, and the amount of electricity yet to be delivered under the PPA for the current target interval. Decisions are constrained by capacities of the electrolyzer, fuel cell, storage facility, transmission line, and hydrogen distribution network. The transition between states is driven by the operator's decisions and exogenous factors like stochastic wind production and market prices, modeled as AR(1) processes for prices and Weibull distributions for wind speed.

A key aspect of the paper is the consideration of different operational settings for the hydrogen market:

• Setting A (Free Market): Hydrogen can be sold in any period with a variable amount, subject to a stochastic market price.
• Settings B($n^h$) (Semi-Free Market): Hydrogen can be sold in any period with a variable amount, but only once every $n^h$ periods (e.g., weekly or bi-weekly), subject to a stochastic market price.
• Settings C($n^h, \bar{p}^h, Q^h$) (Fixed Contracts): A fixed amount of hydrogen $Q^h$ must be sold every $n^h$ periods at a fixed price $\bar{p}^h$. Penalties are incurred if the contractual amount cannot be met.
• Settings D(s) (Electricity Storage Only): Hydrogen is produced and stored but not sold as gas. It's only used for electricity arbitrage (converting back to electricity). $D(H2)$ uses realistic HES efficiency (0.5 round-trip), while $D(B)$ uses a hypothetically high efficiency (0.9 round-trip) to compare against battery-like systems.
• Setting E (No Storage): The baseline case where there is no HES, and wind energy is either sold to the market/PPA or curtailed.

The MDP is solved using backward dynamic programming on a discretized state space. The numerical analysis is based on a case paper relevant to the Northern Netherlands (Europe's first Hydrogen Valley), using data for a 4.5 MW wind turbine, historical wind data fitted to Weibull distributions, and electricity prices fitted to an AR(1) process. Hydrogen prices in the free settings are initially modeled conservatively, similar to electricity prices.

Practical Implementation Details and Findings:

The numerical experiments provide significant practical insights:

1. Profitability of GHPs: Introducing HES and the opportunity to sell hydrogen (Setting A) substantially increases operational revenues compared to a system with no storage (Setting E) or storage used only for electricity arbitrage (Setting D(H2)). The analysis shows potential revenue increases of up to 51% compared to no storage.
2. Value of Hydrogen Sales: Selling hydrogen as a gas (Settings A, B, C) provides additional revenue streams beyond electricity arbitrage, even with current lower HES efficiencies compared to hypothetical high-efficiency storage. Setting A, offering the most flexibility, yields the highest profit among settings with hydrogen sales.
3. Hydrogen Offtake Agreements (HOAs): While fixed HOAs (Setting C) are less profitable than flexible market sales (Setting A), they are crucial for initiating the green hydrogen economy. HOAs can make GHPs viable, especially if the fixed hydrogen price is moderately higher than the expected electricity price (e.g., 11-13 €/MWh higher in the base case). This price difference can offset the energy losses from current electrolyzer efficiencies. HOAs also lead to less energy loss from excessive arbitrage compared to flexible market interaction strategies.
4. Impact of PPA Structure: The size of the PPA target significantly influences profitability. A small PPA can act as a price floor, improving revenue. However, excessively large PPA targets reduce the operator's flexibility to profit from market price fluctuations, potentially decreasing overall profit. The optimal PPA size depends on the system's specific characteristics. Dynamic decision-making helps optimize PPA fulfiLLMent by spreading sales over the interval rather than solely at the deadline.
5. Conversion Efficiency: Increased HES round-trip efficiency (e.g., from 0.5 to 0.9) significantly boosts profitability, particularly for settings constrained by hydrogen delivery obligations (Settings C), as less electricity needs to be purchased to meet hydrogen targets. In the flexible Setting A, higher efficiency leads to more hydrogen sales rather than necessarily more electricity arbitrage.
6. Inventory Management: Optimal strategies show seasonal patterns in hydrogen inventory levels, with higher levels maintained during periods of lower expected wind production variability (e.g., summer) to ensure supply and meet potential obligations. This suggests GHP operators could benefit from flexible or seasonal storage arrangements.

Scientific Contributions:

The paper claims the following scientific contributions:

• First in-depth paper analyzing optimal dynamic policies for renewable energy producers interacting with both electricity and hydrogen markets.
• First paper to thoroughly integrate PPA obligations and hydrogen offtake agreements into such a model.

Implementation Considerations:

• Data Requirements: Requires historical data for wind speed (or other renewable source), electricity prices, and potentially hydrogen prices (if available) to train the stochastic process models. Future projections of these factors are also needed for long-term planning.
• Discretization: The MDP approach requires discretizing continuous state variables (prices, inventory, PPA target). The level of discretization impacts computational complexity and solution accuracy.
• Computational Resources: Solving MDPs, especially with multiple state variables and large horizons, can be computationally intensive. Backward dynamic programming is feasible for the one-year horizon and discretized state space used, but scaling to longer horizons or finer granularity might require approximate dynamic programming techniques.
• Model Complexity: The model captures key operational decisions and market interactions but simplifies some aspects (e.g., assumes simultaneous decisions, ignores intra-day fluctuations). Real-world implementation might require more complex models or multi-stage optimization approaches.
• Market Uncertainty: The model relies on stochastic process models for prices and production. The accuracy of these models is crucial for the optimality of the derived policies. Real-world market volatility and unexpected events pose ongoing challenges.

In conclusion, this research provides a valuable framework for understanding the economic viability and optimal operation of integrated green hydrogen plants, highlighting the critical roles of hydrogen market opportunities, contractual agreements like PPAs and HOAs, and technological advancements in HES efficiency. The MDP approach offers a principled way to derive dynamic strategies that can adapt to changing market conditions and renewable energy availability.