1: \begin{abstract}

2: In light of the expanding relevance of Distributed Energy Resources (DERs) in advancing sustainable energy sectors, Virtual Power Plants (VPPs) have emerged as a key tool for aggregating heterogeneous DERs, enabling them to participate in wholesale energy markets. 

3: FERC (Federal Energy Regulatory Commission) order 2222 is designed to facilitate the participation of 

4: Supported by the Federal Energy Regulatory Commission’s (FERC) Order 2222, which equalizes the competitive landscape for distributed energy in various market services, the significance of DERs and VPPs has been intensified. Nonetheless, the inherent diversity and decentralization of DERs pose challenges to conventional centralized optimization methods in terms of efficiency and speed. Addressing this, our research introduces a novel machine-learning-based method for distributed optimization in VPPs. Central to our contribution is a multi-agent framework where each agent oversees multiple DERs, leveraging neural network approximators to expedite the solution search.  The proposed method uses gauge map to ensure hard feasibility to local constraints while effectively minimizing the necessity for auxiliary post-processing steps. Experimental results highlighted our method's advantages, revealing a solution time acceleration per iteration and a markedly reduced convergence time. The method consistently excels in optimization tasks that require repeated solving, demonstrating its adaptability and reliability across a range of problem situations.

5: \end{abstract}

6: