Intra-day scheduling framework for multi-energy microgrid incorporating flexible region of chemical industry load based on Bayesian nonparametric

Multi-energy microgrid has emerged as a crucial carrier for renewable energy utilization, supplying multi-energy to chemical industry load which can offer flexibility via adjustable production scheduling. However, it can be difficult for the intra-day scheduling of multi-energy microgrid owing to privacy concerns of chemical industry load and intra-day renewable energy uncertainties. To address this, an intra-day scheduling framework for multi-energy microgrid incorporating flexible region of chemical industry load based on Bayesian nonparametric is proposed in this paper. Firstly, a flexible region is constructed to characterize adjustable range of multi-energy inputs to the chemical industry load considering its production constraints. A calculation method based on vertex enumeration and Quickhull algorithm is proposed to formulate the flexible region. On this basis, essential flexibility-related information of chemical industry load can be directly utilized for scheduling of multi-energy microgrid, which can preserve its privacy. Secondly, an online-offline fitting method is proposed to construct a Gaussian mixture model to characterize the renewable energy uncertainties, with historical data captured via Dirichlet process mixture model (DPMM) and online data incorporated to update the model via incremental Gaussian learning. Finally, to solve the intra-day two-sided chance-constrained scheduling problem for the multi-energy microgrid, a second-order cone programming (SOCP) formulation is employed to ensure feasibility of the chance constraints. Case studies illustrate that the exploitation of chemical industry load flexibility and updating of Gaussian mixture model can effectively reduce operation costs. Besides, the proposed two-sided chance-constrained method has the advantage of low operational violation probability.