Deep Learning Using Genetic Algorithm Optimization for Short Term Solar Irradiance Forecasting
Wadie Bendali, Ikram Saber, Bensalem Bourachdi, Mohammed Boussetta, Y. Mourad
Abstract
The increase in the use of renewable energy sources (RES) has been remarkable in recent years, especially photovoltaic energy, which is one of the most widely used renewable energy sources for electricity generation. In fact, the world has known the installation of a large number of autonomous or grid-connected photovoltaic systems. However the problem with the introduction of PV is the improvised nature of solar energy that can influence the stability of electricity grids and reliability of the grid, Accurate solar forecasting makes it easier to integrate solar generation into the grid by reducing the integration and operating costs associated with intermittent solar power.In this context, the objective of this work is to improve the development of appropriate forecasting models for the prediction of photovoltaic energy production. For that reason, this article presents new hybrid methods to optimize deep learning forecasting by using genetic algorithm based Deep Neural Network. The model is employed to forecast time series of solar irradiation output. Comparisons are made between the performances of three types of neural networks: long short-term memory (LSTM), Gate recurrent unit (GRU), and Recurrent Neural Network (RNN). GA is exploited to find the most appropriate number of window size, and number of units (neurons) in each and all three hidden layers.