AI-Driven Spatiotemporal Mapping and Grid Optimization for Solar and Wind Energy

Renewable energy sources are essential for energy production and energy transfer systems. In this paper, we explore a novel approach that combines natural and social sciences through a machine learning (ML) technique, which integrates environmental and geographical information systems (GISs) and aligns with the United Nations' 17 Sustainable Development Goals (SDGs). With the help of a dataset, we derived one hundred regional observations that covered solar irradiation, wind energy, temperature, relative humidity, and altitude. Then, we enhanced this dataset using GIS information (latitude and longitude) and available energy production information at historical timestamps. This dataset was used for training the neural network. With the help of TensorFlow's Sequential Application Programming Interface (API), we used dropout regularisation and dense layers for overfitting prevention. The resulting model was a deep learning architecture that was capable of preventing overfitting. The dataset was standardised and split (80-20) for training and testing. Overfitting prevention deep learning architecture was used with a batch size of 16, trained for 50 epochs using the Adam optimiser, mean squared error (MSE) loss, and an initial training loss of 98,273.70 was obtained. The model trained loss was reduced to 16,651.12 while stabilising validation loss, indicating strong generalisation. Validation shows that the model GIS visualisations for energy generation aid in providing spatial dependence of energy interdependence for grid improvement and energy generation interdependence in spatial systems for grid planning. The proposed approach is, therefore, an integration of GIS and deep learning with the aim of obtaining spatially informative energy potential estimates.