Abstract
The global horizontal irradiance (GHI), direct normal irradiance (DNI), temperature and other meteorological data are generally used for the photovoltaic (PV) power forecasting. Due to the multi-layered and complex factors between irradiance and PV power generation, a large amount of long-term operation data is required to train the model to achieve a high prediction accuracy. In order to reduce the data requirements, a coupled model based on solar irradiance and BP Neural Network is proposed in this paper. Firstly, the difference between the received irradiance and the GHI/DNI is clearly demonstrated. Moreover, the received irradiance of fixed photovoltaic panel is calculated. On this basis, according to the geographical location of the photovoltaic power plant, the received irradiance in a whole year is modelled and used as the input to train BP neural network. The prediction results show that, compared with the conventional prediction methods, the coupled model has a higher accuracy, which can reduce the mean squared error and root mean squared error by about 34% and 16%.
Keywords solar irradiance, BP neural network, photovoltaic power forecasting, coupled model
Copyright ©
Energy Proceedings