Abstract:

The large-scale grid connection of distributed generation introduces non-smooth local control constraints such as droop control, which easily leads to the convergence failure of the traditional power flow calculation method based on the forward and backward substitution method. Moreover, because the grid-connected distributed generation changes the power flow direction of the system, the traditional theoretical line loss calculation methods such as the equivalent resistance method and the pressure drop method are no longer applicable. In order to solve the above problems, this paper proposed a smoothing model considering on-load tap changer regulation and distributed generation droop control and constructed a fast calculation model of linearized theoretical line loss of a three-phase distribution network driven by data and physics fusion. On the basis of traditional linearization based on steady-state operation and first-order Taylor expansion linearization, the partial least squares method was used to compensate for the linearized error. Compared to pure physics-driven linearization, it still maintained high accuracy under overload conditions; compared to pure data-driven linearization, it retained branch topology information, making it suitable for scenarios with switch status changes. The proposed model compensated for linearized errors, greatly improving the convergence and calculation efficiency of the power flow model while ensuring linearized accuracy, and it could adapt to different load levels to achieve precise error compensation. Based on the actual 42-node three-phase distribution network system simulation, it was verified that the proposed model had high accuracy and could realize the robust and fast calculation of the theoretical line loss of the distribution network.

Key words: drive by data and physics fusion, linearized power flow, theoretical line loss calculation, droop control