Cooperative dispatch of demand response and stability constrained transfer capability for inter-connected power systems: A hybrid learning-aided method

Reverse peak conditions in generation and load profiles, as well as limiting total transfer capability (TTC) of power transmission inter-corridors, are critical inducements of renewable curtailment. Despite demand response (DR) incents loads to better complement renewable generations, it can in turn aggravate hindrance of renewable energies' exporting consumption. To solve this issue, spatiotemporal cooperation between DR and TTC must be involved. Further challenge relates to complicated interaction between price-incentive DR and stability constrained TTC. A hybrid learning-aided cooperative dispatch method is thus proposed. Firstly, to mitigate over-conservative power transfer and reduce the high-dimensional dynamic security bounds, a neural network is introduced to fast track dynamic TTC varying with the dispatch model. It is then reformulated as mixed integer linear programming (MILP) and seamlessly integrated into the optimization process along with price-based DRs. At last, to help fast approach optimum of the DR-TTC-cooperative model, a physics-regularized temporal graph convolutional network is utilized to warmly initialize units' on/off status, DRs and energy storages' active states, and auxiliary integer variables introduced from MILP-reformed TTC estimator as well. Case studies on the modified IEEE 39-bus system verify that our method outperforms traditional method regarding security adherence and efficiency, and reveal the importance of DR and TTC's cooperation in releasing renewable consumption latency.