Global Horizontal Irradiance Forecasting Using Hybrid Method: Convolutional Neural Network-Long Short Term Memory

Along with the growing demand for energy, it has encouraged researchers to develop technologies and methods to optimize renewable energy, especially solar energy. One of the most influential components of the quality of solar energy is Global Horizontal Irradiance. Uncertain characteristics due to meteorological factors are a challenge to establish an optimal method for forecasting, thus solar technology or photovoltaics can produce constant power to meet the power demand. Several methods have been studied, and each method has its own advantages and disadvantages. The combination of several methods is the right step to achieve optimal results in a forecast. In this paper, the author applied a hybrid method CNN-LSTM. The author uses a data set in one year, which is carried out every 15 minutes in a day. However, the data is processed only from 05.00 to 17.00, considering the characteristics of irradiance. The results of optimal forecasting are shown by curves. In addition, error metrics are also calculated to strengthen the evidence that the proposed method is superior. Likewise, the comparison with several other methods such as CNN, LSTM, and CNN-LSTM, presented the superiority of the proposed hybrid method.