The Virtual Data-sampling Method Based on Optimal DBN for High-Scale Distributed Photovoltaic Power Generation

Abstract

In a distribution power grid with a high proportion of distributed photovoltaic (DP) connections, the real-time acquisition of the distributed photovoltaic power generation data is a costly and heavy task. In this paper, a virtual data sampling method based on Deep Belief Network (DBN) combined with a Second-order oscillating-firefly algorithm (SOO-FA) is proposed. It enables virtual data sampling from all distributed PV plants in a region where only one distributed PV plant has a complete data-sampling device and the other distributed PV plants have less expensive current sampling terminals. The firefly algorithm is utilized to continuously adjust and optimize the weights of the deep belief network, and the optimal DBN model is used to carry out the virtual collection of the power output data of the distributed PV equipment within the region. Finally, the proposed method has high accuracy for the PV output power by 100 distributed PV devices in the regional range in the simulation section.