Real-time Estimation of Distributed 380V/220V Photovoltaics Based on Deep Learning

The massive access of distributed photovoltaics (DPV) puts forward higher requirements for the current grid's safe and stable operation. Aiming at the problem that the automatic dispatch control system does not directly collect realtime operating data for most 380V/220V DPV, a real-time PV restoration algorithm combined with meteorological data and historical PV data, based on long short term memory neural networks (LSTM) is proposed. Taking DPV in the Tangshan area in Hebei Province, China, as the experimental object, the estimation accuracy of the LSTM model based on typical distributed site power and meteorological data can reach up to 92.6%, which verifies the accuracy of the model. With the increasing number of DPV, the proposed estimation method is of great significance for the dispatching department to monitor and manage the operation of DPV and to dispatch electric resources, loads, and storage.