Data-driven distributionally robust stochastic optimal dispatching method of integrated energy system considering multiple uncertainties

The integrated energy system (IES) has attracted significant attention due to the advancements in multi-energy complementary technology. However, the uncertainties of renewable energy output and load variability, pose challenges to effectively implementing the dispatching plan. At present, the majority of research studies about uncertainties employ a unified modeling approach to address uncertain factors with diverse characteristics, resulting in compromised reliability and cost-effectiveness of the scheduling plan. In this paper, a two-stage distributionally robust stochastic optimization model is proposed to optimize the operation strategy wherein the uncertainties associated with renewable energy and user load are described by different model. For the load side uncertainties, to enhance the representativeness of the scenarios set, generative adversarial networks are employed for its construction. Compared with the source side, it is relatively stable, a scenario-based stochastic programming is adopted to obtain the optimal expectation of objective. For the source side, which has stronger uncertainties, to obtain the feasible dispatching plan under the worst case of renewable energy output, distributionally robust optimization is adopted. To evaluate the effectiveness of the proposed method, a typical-structure IES is discussed as a case study. The results shown that, compared with the traditional stochastic optimization method, the scheduling plan is more reliable; and more economic compared with robust stochastic optimization, the proposed method can save the cost by an average of 11 %, which is beneficial for decision-makers to achieve a balance between economy and reliability in practice.