Optimized forecasting of photovoltaic power generation using hybrid deep learning model based on GRU and SVM

The growing integration of renewable energy sources and the rapid increase in electricity demand have posed new challenges in terms of power quality in the traditional power grid. To address these challenges, the transition to a smart grid is considered as the best solution. This study reviews deep learning (DL) models for time series data management to predict solar photovoltaic (PV) power generation. We first summarized existing deep learning models in the literature. We also developed PV power prediction models such as support vector machine (SVM), gate recurrent unit (GRU), feed forward neural network (FFNN), and long short-term memory (LSTM) for efficient prediction applied to many datasets. We characterized each site based on data variability, which refers to the direct influence of weather data on power generation. To achieve more accurate forecasting, we proposed an innovative approach based on a parallel hybrid SVM-GRU method, optimized by the ant colony optimization (ACO) algorithm. This method combines the advantages of both SVM and GRU models by merging their predictions, and the ACO algorithm is used to optimize the fusion function parameters to obtain the best prediction performance. The model achieved a standard correlation coefficient of 0.9986 which demonstrate the outperformance of the proposed method. In our future work, we plan to explore even more advanced fusion methods to further improve prediction accuracy. The ultimate goal is to achieve accurate and reliable real-time prediction of solar PV power generation, which will contribute to better integration of renewable energy sources into the power grid.