Two-stage distributionally robust optimization model of integrated energy system group considering energy sharing and carbon transfer

Abstract

Due to the small scale and few functions of the single integrated energy system, the absorption capacity of wind turbine and photovoltaic is limited, the ability to cope with uncertainties is weak, and the space for optimal allocation of resources is limited. To solve the coordination problem of robustness, economy, environmental protection and efficiency, this paper forms an integrated energy system group (IESG) by means of energy sharing and carbon transfer, and innovatively proposes a two-stage distributionally robust optimization model (TSDRO) based on kernel density estimation (KDE) and Wasserstein metric. Firstly, the structure of IESG with carbon capture, utilization, and storage-power-to-gas (CCUS-P2G) system is introduced. Then, the nonparametric KDE method is applied to fit the probability density functions of the forecast power error of wind turbine and photovoltaic. Wasserstein metric is used to characterize the fuzzy uncertainty set of distributions. The cumulative distribution function of KDE is taken as the center, and the obtained distance is taken as the radius to form the Wasserstein ball of probability distribution. Based on affinely adjustable policy, a correlation model of real-time variables with respect to day-ahead variables is established. Finally, according to the dual theory and convex optimization theory, the TSDRO model is reformulated into a solvable model. The simulation results show that: (1) energy sharing and carbon transfer can improve the ability of IESG to cope with uncertainty and expand the boundary of resource optimal allocation, and the minimum expected operating cost under the worst distribution is $ 40,259.94. (2) CCUS-P2G system strengthens the synergistic relationship between electricity and carbon and reduces the carbon emission of the system by 128.2 t. (3) After testing, the results obtained by nonparametric KDE are closer to the true distribution and more objective. (4) The TSDRO model is data-driven and has the advantages of high solving efficiency and low decision-making conservatism. The solution time of the TSDRO model is 73.03 % less than that of the stochastic optimization model, and the operation cost is 1.69 % less than that of the robust optimization model, which achieves the balance of economy, robustness and environmental protection of IESG.