Seasonal hydrogen storage decisions under constrained electricity distribution capacity (2007.08230v1)

Abstract: The transition to renewable energy systems causes increased decentralization of the energy supply. Solar parks are built to increase renewable energy penetration and to supply local communities that become increasingly self-sufficient. These parks are generally installed in rural areas where electricity grid distribution capacity is limited. This causes the produced energy to create grid congestion. Temporary storage can be a solution. In addition to batteries, which are most suitable for intraday storage, hydrogen provides a long-term storage option and can be used to overcome seasonal mismatches in supply and demand. In this paper, we examine the operational decisions related to storing energy using hydrogen, and buying from or selling to the grid considering grid capacity limitations. We model the problem as a Markov decision process taking into account seasonal production and demand patterns, uncertain solar energy generation, and local electricity prices. We show that ignoring seasonal demand and production patterns is suboptimal. In addition, we show that the introduction of a hydrogen storage facility for a solar farm in rural areas may lead to positive profits, whereas this is loss-making without storage facilities. In a sensitivity analysis, we show that only if distribution capacity is too small, hydrogen storage does not lead to profits and reduced congestion at the cable connection. When the distribution capacity is constrained, a higher storage capacity leads to more buying-related actions from the electricity grid to prevent future shortages and to exploit price differences. This leads to more congestion at the connected cable and is an important insight for policy-makers and net-operators.

• The paper analyzes optimal strategies for storing, buying, and selling energy in a solar park with local hydrogen storage operating under constrained grid capacity.
• It employs a Markov Decision Process to capture seasonal variations in solar production, demand, and prices, leading to distinct price thresholds for action.
• The paper demonstrates that integrating hydrogen storage can significantly boost profitability while also highlighting trade-offs in managing grid congestion.

This paper (2007.08230) examines the operational decisions for a solar park equipped with local hydrogen storage, connected to a local electricity grid with limited distribution capacity. The goal is to maximize expected future profits by deciding when to store energy as hydrogen, use hydrogen to satisfy local demand, sell excess electricity to the grid, or buy electricity from the grid. The problem is motivated by the increasing decentralization of renewable energy, particularly solar parks located in rural areas where grid infrastructure is often constrained, leading to congestion. Hydrogen storage is explored as a solution for long-term seasonal storage to bridge the mismatch between seasonal solar production peaks (summer) and higher demand or lower production periods (winter), while also potentially mitigating grid congestion.

The problem is modeled as a Markov Decision Process (MDP) over a one-year horizon, discretized into daily periods. The state of the system at the end of each day is defined by the current hydrogen inventory level, the prevailing electricity price, and the net solar energy production after local demand is met. Uncertainty in solar production, local demand, and electricity prices is incorporated. Solar production and demand exhibit seasonal patterns, while electricity prices are modeled using an AR(1) process.

The decision (action) at each step is the amount of energy to buy from or sell to the grid. This action is constrained by:

• The grid distribution capacity ($k^c$).
• The maximum hydrogen storage capacity ($m$).
• The maximum rate at which energy can be stored (electrolyzer capacity $k^+$).
• The maximum rate at which energy can be retrieved from storage (fuel-cell capacity $k^-$).
• The current net solar production $\bar{y}_t = Y_t - D_t$.
• The current hydrogen inventory $x_t$.

Energy stored in hydrogen incurs a conversion efficiency loss ($\alpha$). The reward function accounts for revenues from selling electricity to the grid, costs from buying electricity from the grid (which may include a price markup $c^+$), and a penalty ($s$) for unmet local demand, which is set sufficiently high to prevent this from occurring in an optimal policy.

The MDP is solved using backward dynamic programming over a discretized state and action space to find the optimal policy for each state and time period. The discretization involves dividing the inventory, net production, and price ranges into discrete intervals. The transition probabilities between states are derived from the stochastic processes governing net production and prices.

A base-case system based on a hypothetical solar park in the Netherlands is used for numerical analysis. It includes a 5 MWp solar park, a 1.25 MW (30 MWh/day) grid connection capacity, a 2.1 MW (50 MWh/day) electrolyzer and fuel cell, and a 1000 MWh hydrogen storage capacity, with a round-trip efficiency of 50%. Data for solar production, demand, and electricity prices from real-world sources are used to calibrate the stochastic models.

Key findings from the numerical analysis include:

1. Profitability: Adding hydrogen storage significantly increases profitability compared to a system without storage, which was loss-making in the base case (-4060.5 profit/year without storage vs. 6579.4 with storage). Ignoring seasonality in the policy optimization reduces profits by about 3% (6387.0 profit/year when ignoring seasonality).
2. Optimal Policy Structure: Optimal policies are characterized by price thresholds that dictate actions. For given inventory and net production levels, low prices typically lead to buying from the grid (up to maximum capacity), increasing prices lead to selling net overages or buying net shortages, moderate prices might lead to storing overages or using storage for shortages (avoiding grid interaction), and high prices lead to selling from inventory (up to maximum capacity). These thresholds vary significantly between summer and winter due to seasonal differences in net production and demand.
3. Congestion: The paper defines congestion as the full utilization of the grid connection capacity.
   • Under constrained distribution capacity ($k^c$), congestion is primarily caused by buying actions in winter, driven by the need to cover potential future shortages and exploit price differences.
   • Under higher distribution capacity, congestion is mainly caused by selling actions in summer, driven by high solar overproduction.
   • Counter-intuitively, increasing storage capacity ($m$) can increase congestion, especially when distribution capacity is already constrained. This is because more storage enables more buying actions in winter to hedge against future shortages and arbitrage prices, leading to more frequent full utilization of the import capacity.
4. Electrolyzer Utilization: Electrolyzer utilization is not a direct indicator of profitability. While higher utilization might seem desirable, larger electrolyzer capacities, even with lower utilization rates, tend to yield higher operational profits by allowing more flexible interaction with the grid to exploit price differences.
5. Sensitivity to Parameters:
   • Distribution Capacity ($k^c$): Increasing $k^c$ increases profit (up to a point where the system is no longer grid-constrained) and generally decreases overall congestion frequency, although the balance shifts from buying-induced (winter) to selling-induced (summer) congestion as $k^c$ increases. Very low $k^c$ can make the system infeasible or unprofitable.
   • Storage Capacity ($m$): Increasing $m$ generally increases profit, but its impact on congestion depends on $k^c$. With low $k^c$, increasing $m$ doesn't reduce congestion and can even increase it due to buying actions. With higher $k^c$, storage is used more for arbitrage, and its impact on congestion is less critical as $k^c$ is less binding.
   • Electrolyzer Capacity ($k^+$): Increasing $k^+$ increases profits by enabling more storage operations, even if utilization decreases. Lower $k^+$ makes storing less feasible, potentially increasing reliance on grid purchases during shortages.
   • Conversion Efficiency ($\alpha$): Higher efficiency increases profits and electrolyzer utilization, but also increases grid interaction and thus congestion frequency.
   • Price Markup ($c^+$): Higher price markups on buying electricity reduce profits and electrolyzer utilization (as arbitrage opportunities decrease) and also reduce grid interaction and congestion. This suggests price signals can be used to influence storage behavior and manage congestion.
   • Production Capacity ($w$): Increasing solar production capacity increases profits (linearly in the analyzed range) but also increases congestion, primarily selling-induced in summer due to larger overages.

In practice, implementing such a system involves setting up the physical infrastructure (solar panels, electrolyzer, fuel cell, hydrogen storage tank, grid connection) and deploying a control system that implements the optimal policy derived from the MDP. The MDP requires accurate models of seasonal solar production, local demand, and electricity prices, which can be derived from historical data. The discretized state space can become large, necessitating computational resources to solve the dynamic program. The optimal policy for the first year can be used as a stationary policy for subsequent years, assuming the underlying stochastic processes are recurring. The results highlight the complex interplay between different capacity constraints and market signals, suggesting that the design of decentralized renewable energy systems with storage and grid connections needs careful co-optimization.

Future research directions include expanding the model to incorporate multiple energy carriers (e.g., heat, transport hydrogen demand), exploring correlations between stochastic processes, considering physical hydrogen storage properties, and shifting focus to strategic planning (e.g., optimizing storage and grid connection sizes and locations) rather than purely operational decisions.