Prediction of Direct Normal Irradiance Based on Ensemble Deep Learning Models

Abstract

Due to the intermittency and instability of solar power plant output, the key components are often in danger of being scrapped. Therefore, it is necessary to accurately predict the output of solar power plants. At present, the solar radiation intensity is considered to be one of the most important factors affecting the output of solar power plants. In essentially, the solar radiation intensity is a general concept, which is composed of several parameters. Among these parameters, direct normal Irradiance (DNI) prediction has become a hot and difficult research topic. In recent years, with the continuous improvement of hardware performance, a deep learning model has become one of the best methods to solve most time series prediction problems. In this paper, an ensemble deep learning model is proposed. This model integrates the convolutional neural network (CNN) and long short-term memory (LSTM) neural network. The former is responsible for extracting spatial features from data, while the latter is responsible for extracting temporal features. As a case study, the proposed method will predict the DNI of four typical days (spring equinox, summer solstice, autumn equinox and winter solstice) and its seven days before and after which is based on the meteorological data of Zhangbei. The results show that the deep learning model can be effectively applied to DNI prediction. However, the method proposed in this paper has a higher prediction accuracy compared with other models.