Optimal Bidding Strategy of Virtual Power Plants Considering Internal Power Market for Distributed Generation Units

Abstract

Distributed Generation (DG), by itself, do not have the flexibility and sufficient capacity to take part in the power market. Aggregating such resources in a Virtual Power Plant (VPP) can solve this problem. The VPP, as a representative of DGs, participates in the power market and attempts to maximize its own profit. While the VPP participates in the wholesale power market, an internal market is formed within it, as well. DGs compete with each other in this internal market to achieve more profit. In this case, a bi-level problem will be formed. In the first level, VPPs and in the second level, DGs compete to gain more profit. In this regard, the bidding strategy plays a key role for maximizing the profit of VPPs and DGs. To this aim, the Q-learning algorithm is employed to find the optimal bidding strategy within this competitive environment. In this method, VPPs and DGs (as learning agents) provide the optimal bid in the competitive environment. This is performed to gain more profit in the power market and exhibit an appropriate performance, according to available limitations. It is observed that considering the uncertainty in the DG capacity will cause the simulation of VPPs' and DGs' behaviors for bidding in the power market environment, to be more realistic. Also, it will reduce possible penalties. On the other hand, considering the step-wise bidding strategy, they can have an intelligent profit making.