A Modified Data-Driven Regression Model for Power Flow Analysis

As one of the most important procedure in system recognition, Power Flow (PF) analysis is the pivotal basis of achieving optimal operation and production. Due to the difficulty of obtaining system parameters, low calculation accuracy and high computation costs, model-driven approaches, such as DC Power Flow (DCPF), cannot provide satisfied results for real application. While with the massive accumulated data from industrial enterprise, data-driven linearization approaches of PF become an effective alternative. In this paper, we propose a modified data-driven regression model that reveals the mapping rules between the known and unknown variables in power system. Then, the update mode of this model is further derived, which facilitates PF calculations with different variations of bus types. A regression model of network loss is also built so as to reflect the efficiency of the system as well as enhance comprehensiveness of the proposed data-driven PF analysis. The experimental results of 8 standard test systems demonstrate that the proposed model has a significant improvement on both computational accuracy and computation speed comparing with model-driven approaches and the inverse regression model of PF.