Sergio Alejandro Pérez-Rodríguez

Dr. Sergio Alejandro Pérez-Rodríguez

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 July 2024 in IEEE Access
Verified
Journal Article
Open access
SPSergio Alejandro Pérez-Rodríguez
Sergio Alejandro Pérez-Rodríguez
José M. Álvarez-AlvaradoJosé M. Álvarez-Alvarado
José M. Álvarez-Alvarado
JRJulio Alejandro Romero
Julio Alejandro Romero
Marcos AvilesMarcos Aviles
Marcos Aviles
AMAmerica Eileen Mendoza-Rojas
America Eileen Mendoza-Rojas
Juvenal RodriguezJuvenal Rodriguez
Juvenal Rodriguez
In the contemporary world, where the escalating demand for energy and the imperative for sustainable sources, notably solar energy, have taken precedence, the investigation into solar radiation (SR) has become indispensable. Characterized by its intermittency and volatility, SR may experience considerable fluctuations, exerting a significant influence on energy supply security. Consequently, the precise prediction of SR has become imperative, particularly in the context of the potential proliferation of photovoltaic panels and the need for optimized energy management. Several works in the existing literature review the state of the art in SR prediction, focusing on trends identified using machine learning (ML) or deep learning (DL) techniques. However, there is a gap in the literature regarding the integration of optimization algorithms with ML and DL techniques for SR prediction. This systematic review addresses this gap by studying prediction models for SR that leverage metaheuristic optimization algorithms alongside artificial intelligence (AI) techniques, aiming primarily for maximum prediction accuracy. Metaheuristic algorithms such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) have featured in 29% and 12.1% of the analyzed articles, respectively, while intelligent approaches like Convolutional Neural Networks (CNN), Extreme Learning Machine (ELM), and Multilayer Perceptron (MLP) have emerged as the predominant choices, collectively accounting for 43.9% of the studies. Analysis has encompassed studies examining SR across hourly, daily, and monthly intervals, with daily intervals representing 48.7% of the focus. Noteworthy variables including temperature, humidity, wind speed, and atmospheric pressure have surfaced, capturing proportions of 90%, 68.2%, 56%, and 41.4%, respectively, within the reviewed literature.

28 October 2026 in Algorithms
Verified
Journal Article
Open access
PDF file
SPSergio Alejandro Pérez-Rodríguez
Sergio Alejandro Pérez-Rodríguez
José M. Álvarez-AlvaradoJosé M. Álvarez-Alvarado
José M. Álvarez-Alvarado
Diana-Margarita Cordova-EsparzaDiana-Margarita Cordova-Esparza
Diana-Margarita Cordova-Esparza
JRJulio Alejandro Romero
Julio Alejandro Romero
PSPerla Y. Sevilla-Camacho
Perla Y. Sevilla-Camacho
Juvenal RodriguezJuvenal Rodriguez
Juvenal Rodriguez
Deep learning solar forecasters depend strongly on their hyperparameters, yet comparisons of hyperparameter optimization methods usually rely on few runs per optimizer, confounding the optimizer with the random seed. A protocol is proposed that tests whether optimizers are practically equivalent and how reliable published comparison designs are. Seven methods (four metaheuristics, two model-based methods, one of them multi-fidelity, and Random Search) tuned three deep networks for one-hour-ahead forecasting of global horizontal irradiance in Querétaro, Mexico, with public reanalysis data, a shared search space, one hundred evaluations, and thirty paired seeds, in nine hundred and thirty runs across all experiments. The tuned models improved on smart persistence by 4–11% in mean absolute error. With the 10-epoch search objective, no pair of optimizers differed after Holm correction, 62 of 63 pairs were equivalent within 1 W/m2, and the optimizer explained 0.8% of the error variance, compared with 7.9% for the seed. In a higher-fidelity check with 50 epochs, two guided methods improved on Random Search by 0.74 W/m2, significantly for one of them, while remaining equivalent to each other, suggesting that, in this setting, objective fidelity mattered more than the choice among guided optimizers. In emulations of published designs, even with 30 runs, 2 studies agreed on the winner in at most 55% of cases. A validated surrogate benchmark, code, data, and predictions are released.