A Data-Driven Linear Robust Optimal Power Flow Model

The distribution network has numerous devices, as well as random input and output of loads, and power supplies. Its precise physical model is difficult to obtain. Exploring a data-driven robust power flow model that does not rely on precise parameters is necessary. This paper proposes a data-driven linear robust optimal power flow model. The branch power flow and square of node voltage amplitude are modeled as linear regression models, and model errors are introduced into the linear power flow model as uncertainty variables. Kernel density estimation is used to obtain the uncertainty set of model errors at confidence level α, This model does not require any network parameters and fully considers model uncertainty, effectively improving the security of power system operation under data-driven models.