A Virtual Data Collection Model of Distributed PVs Considering Spatio-Temporal Coupling and Affine Optimization Reference

The rapid development of distributed photovoltaic (DPV) made the shortage of data transmission channels and the difficulty of comprehensive coverage of measurement equipment becoming more significant. To enable a high-precision data collection of DPVs with fewer sensory devices and low computation costs, this paper proposes a virtual collection technology based on computational intelligence. To capture the spatial-temporal correlation of DPVs and find the optimal reference power station (RPS), a deep recurrent denoising autoencoder (D-RDAE)-based model is proposed in this paper. An affine artificial neural network (AANN) is constructed to tackle the uncertainty of solar radiation intensity and select RPSs. To address the high-dimensionality RPS selection, an improved honey badger algorithm (IHBA) with enhanced global search ability is proposed. The operation data of 33 DPVs in Nanjing, China, are used to train and verify the proposed method. The experimental results showed the effectiveness and superiority of the proposed method. Compared with DAE and RDAE, the deeply trained D-RDAE has the best capacity for finding the spatial-temporal correlation of DPVs. In addition, IHBA has the best global search ability compared with other 4 optimizers, and the AANN can better reduce the uncertainty of solar radiation intensity than robust and stochastic optimization techniques.

Automated summary

This paper proposes a virtual collection technology based on computational intelligence to enable high-precision data collection of distributed photovoltaics (DPVs). A deep recurrent denoising autoencoder (D-RDAE) is used to capture the spatial-temporal correlation of DPVs and find the optimal reference power station (RPS). An affine artificial neural network (AANN) is constructed to tackle the uncertainty of solar radiation intensity and select RPSs. Experimental results demonstrate the effectiveness and superiority of the proposed method.