Seizing unconventional arbitrage opportunities in virtual power plants: A profitable and flexible recruitment approach

Abstract

A virtual power plant (VPP) is typically a collection of distributed energy resources (DERs) aggregated by an energy service provider (ESP). However, recruiting DER owners to participate in a VPP is challenging. Therefore, we propose a profitable and flexible VPP recruitment-participation approach that incorporates both long-term regular recruitment and short-term casual recruitment. Casual recruitment caters to ambitious DER participants, consisting of fair and bet-on modes. The latter establishes a set of pre-determined payoff conditions, the fulfillment or non-fulfillment of which confers the participants a contractual right to get compensation from the ESP. To ensure the success of the proposed recruitment approach, we address two key problems. First, we introduce a new index, unconventional arbitrage opportunity (UAO), for evaluating future profits and propose a conditional time series generative adversarial network to predict UAO with weather conditions. Second, we introduce a payoff allocation method that combines fairness and incentives to motivate casual DER participants. The incentive coefficients are optimized using an improved deep reinforcement learning algorithm. Case studies are conducted to verify the proposed recruitment-participation approach, the effectiveness of the UAO prediction model, and the optimized incentive coefficients.