Data-driven Critical Factor Identification for Sending-end Planning with Multi-outfeed HVDC

Asynchronous multi-HVDC interconnections are increasingly widespread. This demands of strongly strengthen sending end to withstand huge reactive load by outfeed HVDC. A data-driven method is therewith proposed to discover critical influences in planning a strong sending-end. Instead of utilizing “rule-of-thumb” basis to assist planning, machine learning can disclose in-depth features that decide strength of a specified sending-end from historical operations and aided simulations. Towards this end, MIV method is applied to identify decisive factors, and a multilayer neural network-based version of the method is proposed. Group lasso is additionally advanced to learn a computational rules of sending-end strength. Benefited from data-driven method, interactions between HVDC and optimal power flow on AC system can be analyzed. Tests on typical multi outfeed HVDC sending-end grid reveal that the proposed method enables precise identification of critical influences and interpretation of strength inference rules, which can further guide accurate sending-end planning.