Jing Lu

Research on short-term module temperature prediction model based on BP neural network for photovoltaic power forecasting

As one of the most related parameters of photovoltaic (PV) power generation, the temperature of PV modules and its prediction play very important role in PV power forecasting. A short-term step-wise temperature prediction model for PV module based on back propagation (BP) neural network is proposed in this paper. Firstly, the impact factors of PV module temperature are determined according to the PV module physical characteristics and the correlation coefficient. Secondly, two different prediction methods, direct and step-wise modeling methods based on BP neural network are applied to build the prediction models respectively. Thirdly, the mapping models between the module temperature and the impact factors for step-wise prediction are established under each weathers types. Finally, the deviations of two different kinds of prediction models are analyzed and discussed using actual operation data. The results indicate that, other things equal, the step-wise prediction model has better accuracy than the direct prediction model.

SVM based cloud classification model using total sky images for PV power forecasting

The accuracy of photovoltaic (PV) power forecasting decreases drastically under cloudy weather due to the rapid, violent and irregular fluctuation of solar irradiance. Therefore, to improve the accuracy of PV power forecasting, a detailed study on the influence of clouds in different movement and evolution patterns on solar irradiance is very necessary. The classification and recognition of different kinds of clouds are the basic of the study on the effect between the cloud and irradiance. A Support Vector Machine (SVM) based cloud classification model using the high temporal and spatial resolution sky images captured via the total sky imager installed in the PV plant is established in this paper. Firstly, the influence on irradiance under clouds of different shapes and distributions in a sky image is analyzed and four different classes of clouds are distinguished taking into account the meteorology standard as well as the preceding analysis. Secondly, the spectral and textural features are extracted by the statistical tonal analysis and gray level cooccurrence matrix (GLCM) of the sky image. Finally, a c-support vector classification (C-SVC) model with radial basis function (RBF) kernel function is built to classify the different clouds in the sky images. The experimental results show that the proposed SVM model can make reasonable classification and efficient identification for the various clouds in the sky images of PV plant.