Short-term Forecasting For Solar Irradiance Based On Multi-source Data Fusion

With the rapid development of economy and science and technology,human demand for energy is increasing.As a non-polluting,renewable and clean energy source,the development and utilization of solar energy has received increasing public attention.The number and scale of photovoltaic power plants that use solar energy as power source are increasing.The short-term fluctuations in light and periodic changes in line voltage over-limit phenomena gradually increase,resulting in local power transmission section overload or overrun.Therefore,accurately predicting the output power of photovoltaic power generation in the future is of great significance for maintaining the security,stability,and economic operation of the power grid.Solar irradiance is the most important influencing factor of photovoltaic power generation.Whether it can be accurately predicted is crucial to the stable operation of the power grid.In this paper,a seq2 seq model based on the attention mechanism is used to predict the short-term solar irradiance,taking Tianjin as a research area,combining ground weather station data and FY-4 satellite data.The main research results are as follows:(1)Convert the solar irradiance to a clear sky index.Compared with the solar irradiance,the clear sky index contains more information about the atmospheric conditions,which helps to improve the accuracy of solar irradiance prediction.(2)A convolutional neural network was established to extract the features of the satellite images around the site of Xiqing District in Tianjin by 50 * 50 as the model training features,thereby improving the accuracy of solar irradiance prediction.(3)An analysis of the solar irradiance in Xiqing District of Tianjin found that the solar irradiance in this area has obvious seasonal characteristics.The solar irradiance was highest in May,June,and July,and it was suitable for photovoltaic power generation for more than 7months throughout the year.(4)A seq2 seq model based on the attention mechanism was established.The clear sky index were used to classify the weather conditions,and the prediction results of this model were compared with those of the traditional seq2 seq model and the BP neural network model using the national ground observation station of Xiqing District in Tianjin in April 2019 as the test data.Experiments show that compared with the other two algorithms,the seq2 seq model based on the attention mechanism has higher prediction accuracy and has better applicability in predicting solar irradiance.