A Risk-Based Stochastic Two-Stage Optimization Framework for Virtual Power Plants: The Role of Energy Storage Systems and Demand Response
This paper proposes a risk-based stochastic two-stage MILP optimization framework is proposed for optimal scheduling of virtual power plants under renewable-generation and market-price uncertainties. The first stage determines day-ahead scheduling and bidding decisions (here-and-now decisions), while the second stage performs real-time operational adjustments (wait-and-see decisions) considering realized uncertainty scenarios. A two-stage stochastic mixed-integer linear programming (MILP) model is developed to maximize the VPP’s profit while managing uncertainties associated with renewable generation, load, and market prices. Conditional value at risk (CVaR) is employed to control operational risk under different risk aversion levels. A key contribution of this work is the explicit modeling of BESS charge/discharge cycles and their direct impact on battery degradation and overall profitability. The results demonstrate that ignoring the battery lifecycle model leads to excessive cycling (e.g., five cycles per day instead of two), which increases short-term revenue but significantly reduces long-term economic viability due to higher investment and operation costs. A comparative economic analysis over 2,500 days shows that incorporating degradation-aware strategies results in 15,000 USD lower net loss despite slightly lower daily profits. Furthermore, the synergistic operation of BESS with price-based and incentive-based DR programs effectively smooths the load curve, reduces peak-to-valley ratio from 2.0 to 1.63, and enhances grid integration of renewable sources. The findings highlight that for VPPs operating in energy storage-intensive environments, extending battery lifetime through controlled depth of discharge (DOD) and limited daily cycles is economically superior to maximizing short-term revenue via aggressive cycling.