2019/03/03 by Mana Jalali, Vassilis Kekatos, Jalali, Mana +5
Engineering · #Microgrid Control and Optimization #Smart Grid Energy Management #Optimal Power Flow Distribution
paper · pdf · doi:10.48550/arxiv.1903.01016
Smart inverters have been advocated as a fast-responding mechanism for\nvoltage regulation in distribution grids. Nevertheless, optimal inverter\ncoordination can be computationally demanding, and preset local control rules\nare known to be subpar. Leveraging tools from machine learning, the design of\ncustomized inverter control rules is posed here as a multi-task learning\nproblem. Each inverter control rule is modeled as a possibly nonlinear function\nof local and/or remote control inputs. Given the electric coupling, the\nfunction outputs interact to yield the feeder voltage profile. Using an\napproximate grid model, inverter rules are designed jointly to minimize a\nvoltage deviation objective based on anticipated load and solar generation\nscenarios. Each control rule is described by a set of coefficients, one for\neach training scenario. To reduce the communication overhead between the grid\noperator and the inverters, we devise a voltage regulation objective that is\nshown to promote parsimonious descriptions for inverter control rules.\nNumerical tests using real-world data on a benchmark feeder demonstrate the\nadvantages of the novel nonlinear rules and explore the trade-off between\nvoltage regulation and sparsity in rule descriptions.\n