Abstract:

The rapid expansion of China's highway network has exacerbated the problems of transportation energy consumption and carbon emissions. It is therefore imperative to achieve efficient local consumption of distributed energy resources in highway domains through virtual power plants (VPPs). To address the "source-load mismatch" problem caused by the volatility of distributed energy output and the randomness of electric vehicle (EV) charging demand, it is of great significance to develop an optimal operational strategy for highway-domain VPPs that account for source-load uncertainty, with the goal of reducing the VPPs' operating costs. To tackle the uncertainty in distributed energy output and EV charging demand, this paper firstly employs the Markov Chain Monte Carlo (MCMC) method to generate a multi-timescale scenario set. It then proposes a hybrid framework combining MCMC and the probability distance-based reduction method for scenario generation and reduction. Subsequently, an optimization model for highway-domain VPPs is established with the objective of minimizing its operating costs, and a smart scheduling algorithm based on constrained proximal policy optimization (C-PPO) is proposed to solve this optimization model. Case study results show that the proposed optimization scheme reduces the operating cost of the highway-domain VPPs, validating the favorable economic performance of the proposed solution.

Key words: transportation-energy integration, source-load uncertainty, highway-domain VPP, operation optimization, policy optimization algorithm