Voltage Situation Awareness of Photovoltaic Power Station Based on Combined Neural Network Model

Abstract

Photovoltaic power station output randomness can easily cause voltage large fluctuations of the grid-connection point, through situation prediction to advance regulation is an effective way to improve voltage stability. In order to improve voltage situation prediction accuracy, this paper proposes a voltage situation awareness method based on Convolutional Neural Networks (CNN) and Gated Recurrent Unit (GRU). Firstly, use acquisition unit extract voltage data and construct time series; Then analyze and understand the correlation between voltage data and related variables in time series by calculating voltage autocorrelation coefficient and maximal information coefficient (MIC); Then use CNN extract the high-dimensional features of the input data; Finally, accomplish voltage situation prediction by connected to GRU network. By the real data of a photovoltaic power station verifying, the results show that: compared with GRU and SVR models, our model prediction accuracy is higher.