An online updated linear power flow model based on regression learning

The linear power flow (LPF) model is widely used in the optimization, operation, and control of distribution networks. These applications require the LPF model to be accurate, fast, and simple in order to simplify calculations as well as to efficiently perform operations and scheduling. In addition, it is difficult to realize the online update of parameters in the existing LPF models. The model retraining brings serious data burden and inefficiency. To serve these applications and comply with requirements, a brand new LPF model is proposed in this paper. A quadratic power flow model is trained by regression learning first, and then the proposed LPF model is derived by Taylor expansion. After only one initial regression learning, the proposed LPF model no longer needs retraining when updated. The refreshed parameter is simply updated online according to the real-time measurement data, which improves the generalization ability. In conclusion, the proposed LPF model is accurate, generalizable, and greatly minimizes the data consumption and running time. Performance analysis verifies these superiorities.