A Data-and-Model-Driven Acceleration Approach for Large-Scale Network-Constrained Unit Commitment Problem With Uncertainty

Numerous binary variables and constraints greatly affect the computation of large-scale network-constrained unit commitment (NCUC) problems. With the integration of massive renewable energy and distributed loads, binary variables and network power flows face more frequent changes to cope with uncertainties. How to ensure the calculation efficiency and accuracy of NCUC problems against uncertainties and frequent changes becomes a common challenge for academia and industry. Inspired by data-driven and model-driven technologies, this paper presents a hybrid reduction approach for variables and constraints to accelerate the solving of large-scale NCUC problems, which achieves “one plus one much greater than two” effect. Specifically, an offline database of nodal net load clusters and the corresponding NCUC solutions is established, which also captures the impacts of changes in renewables, loads, topology, etc. on UC solutions. With this database, a data-and-model-driven reduction method is proposed to fix binary variables and reduce constraints, while considering the changes of prices and net loads in online NCUC instances. Moreover, the reduction method was further developed to accelerate various NCUC problems, including scenario-based methods, robust optimization methods, etc. Thus, the computation burden of online large-scale NCUC problems is greatly mitigated while ensuring good solution quality. Numerical tests implemented on IEEE 24-bus system, 118-bus system, and Polish 2383-bus system illustrate the efficiency of the proposed approach.