Efficient constraint learning for data-driven active distribution network operation

Scheduling flexible sources to promote the integration of renewable generation is one fundamental problem for operating active distribution networks (ADNs). However, existing works are usually based on power flow models, which require network parameters (e.g., topology and line impedance) that may be unavailable in practice. To address this issue, we propose an efficient constraint learning method operate ADNs. This first trains multilayer perceptrons (MLPs) historical data learn mappings from decisions violations loss. Then, constraints can replicated by these MLPs without parameters. We further prove formulating a union disjoint polytopes approximate corresponding feasible region. Thus, proposed interpreted as piecewise linearization method, also explains its desirable ability replicate complex constraints. Finally, novel pruning developed remove useless binary variable solutions advance, enhance solution's reliability reduce computational complexity. Numerical experiments IEEE 33- 123-bus test systems validate achieve optimality, feasibility, efficiency simultaneously.