Abstract
Virtual power plants (VPP) can increase reliability and efficiency of power systems with a high share of renewables. However, their adoption largely depends on their profitability, which is difficult to maximize due to the heterogeneity of their components, different sources of uncertainty and potential profit streams. This paper proposes two profit-maximizing operating strategies for a VPP that aggregates solar systems and electric vehicle (EV) chargers with vehicle-to-grid (V2G) support, and generates profit by trading energy in day-ahead and imbalance electricity markets. Both strategies solve a two-stage stochastic optimization problem. In the first stage, energy bids are placed by solving a sequence of linear programs, each formulated for a specific forecast scenario. In the second stage, given the day-ahead commitments and real-time measurements, the decisions with respect to charging or discharging
EVs are made sequentially for every hour and adjustments to the
day-ahead commitments are settled in the imbalance market. The
two strategies differ in how they solve the sequential decision making problem in the second stage. But, they both foresee the effect of
their current (dis)charge decisions on the feasibility of fulfilling the
EV charging demands using a one-step lookahead technique. The first strategy employs a heuristic algorithm to find a feasible charging schedule for every EV that is connected to a charger. The second one utilizes a soft actor-critic reinforcement learning method with a differentiable projection layer that enforces constraint satisfaction. We empirically evaluate the proposed operating strategies using real market prices, solar traces, and EV charging sessions obtained from a network of chargers in the Netherlands, and analyze how
the uptake of V2G could affect profitability of this VPP.
EVs are made sequentially for every hour and adjustments to the
day-ahead commitments are settled in the imbalance market. The
two strategies differ in how they solve the sequential decision making problem in the second stage. But, they both foresee the effect of
their current (dis)charge decisions on the feasibility of fulfilling the
EV charging demands using a one-step lookahead technique. The first strategy employs a heuristic algorithm to find a feasible charging schedule for every EV that is connected to a charger. The second one utilizes a soft actor-critic reinforcement learning method with a differentiable projection layer that enforces constraint satisfaction. We empirically evaluate the proposed operating strategies using real market prices, solar traces, and EV charging sessions obtained from a network of chargers in the Netherlands, and analyze how
the uptake of V2G could affect profitability of this VPP.
| Original language | English |
|---|---|
| Title of host publication | BuildSys 2022 - Proceedings of the 2022 9th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation |
| Pages | 119-128 |
| Number of pages | 10 |
| ISBN (Electronic) | 9781450398909 |
| DOIs | |
| Publication status | Published - 9 Nov 2022 |
Bibliographical note
Funding Information:This research was supported by funding from the Canada First Research Excellence Fund as part of the University of Alberta’s Future Energy Systems research initiative.
Publisher Copyright:
© 2022 ACM.
