Design and operational control methodology for large-scale photovoltaic system in green port application

Abstract

ABSTRACT Due to the complex-shading and ununiform-corrosion problems caused by the oceanic climate, the working conditions of photovoltaic (PV) system in port are poor. In this study, some effective PV topologies are developed and compared under typical oceanic conditions. In order to improve the output of port PV system, a novel maximum power point tracking (MPPT) method is developed, in which the convolutional neural network (CNN) and bidirectional long short time memory network (BiLSTM) are introduced to forecast the real-time optimal operation point. Finally, the developed topology and MPPT method are verified by a comprehensive set of case studies. Experimental results show that the total-cross-tied (TCT) and gradient (GT) topologies can achieve the best performance. However, by considering the complex structure and higher failure rate due to the large number of joints in TCT topology, the PV arrays installed on the roofs, in the free space of green belt, freight yard and near the seaside are suggested to be connected in GT topology to balance operation efficiency, failure rate and maintenance costs. Moreover, the developed CNN-BiLSTM-based MPPT method and PV topology can effectively work together, and the average forecasting error is just 3.482%, which outperforms the compared state-of-the-art methods significantly.