A Non-Linear Auto-Regressive Exogenous Model for Feeder Power Loading Prediction in PV Rich Distribution Network

Photovoltaic (PV) power curtailment results in less utilization of available PV and it could be reduced significantly if the low voltage (LV) feeder loading could be predicted in advance with the desired accuracy. In this paper, a methodology is proposed based on the nonlinear auto-regressive model with eXogenous Inputs (NARX) to predict the feeder power loading one hour ahead considering high PV injections. The developed model does not require measured historic smart meter or PV generation data. The predicted output would be useful to maximize the PV energy utilization and will support DSO in the decision-making process for short-time congestion management. A case study has been performed to verify the performance of the developed NARX model, considering the measured feeder data from the Strijp-S region, in Eindhoven, the Netherlands. Obtained results from the proposed method are compared with the base case scenario and actual measured feeder power loading values which indicates the satisfactory performance of the proposed methodology.