Answers are grounded in this paper and cite their sources.
Solar energy (SE) now plays a strategic role in the global energy transition, serving as an essential lever for ensuring competitive and sustainable production on the international market. Fine-tuned and reliable control of SE characteristics is therefore a fundamental prerequisite for optimizing the design, deployment, and management of photovoltaic systems. In this context, this work proposes a major innovation: the development of a predictive approach CNN-RF, based on advanced deep learning techniques, specifically designed to improve the accuracy of solar energy forecasts. This method aims to meet the growing needs for intelligent and secure management of the sector, particularly in terms of proactive maintenance planning for installations, consumption optimization, and the strengthening of energy control and security systems. The originality of this approach lies in the development of an integrated and unified tool capable of providing reliable solar energy (SE) forecasts for all time horizons: in real time, short term, medium term, and long term. This innovative approach combines the identification of the most relevant and decisive meteorological variables in model training, obtained using the Random Forest algorithm, with the predictive power of a convolutional neural network (CNN). It thus offers, within a single system, a solution that combines consistency, accuracy, and computational efficiency. Unlike previous approaches, which required several separate models for each time horizon, this unified solution ensures consistency of results, reduces the risk of human error, and optimizes computing resources. The data used, covering the period 2010-2021, comes from the subtropical region of Pirapora (Brazil). The comparative evaluation of the neural architectures tested revealed the superiority of the convolutional neural network (CNN), which outperformed the multilayer perceptron (MLP), the long short-term memory (LSTM) network, and the gated recurrent unit (GRU), particularly in terms of accuracy and robustness across all forecast scales. To demonstrate the high quality and reproducibility of the work, the key findings were contrasted with those from Salvador in Brazil, an area with a tropical climate. The results observed reinforce the relevance and industrialization potential of this innovation, paving the way for predictive and sustainable management of solar energy on a large scale.
No takes yet. Share an insight, caveat, or question.
Jebli et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: