The rise of green hydrogen as a clean energy source underscores the need for advanced management of renewable energy inputs, particularly photovoltaic (PV) power generation, which is known for its variability and unpredictability. Effective integration and use of PV power is essential for grid stability and sustainable green hydrogen production, which heavily depends on the availability of excess renewable energy. This study introduces a hybrid deep learning model that combines the N-Beats architecture with a sequential model to enhance the accuracy of PV power generation forecasts. The N-Beats model, known for its ability to capture complex temporal relationships, is used alongside a sequential model that processes additional temporal dependencies. Additionally, this study applies Fourier Analysis to identify significant frequency components in the time variable, revealing key elements like yearly, monthly, and daily cycles. The performance of the hybrid model is compared against traditional algorithms like logistic regression, random forests, support vector machines (SVM), and XGBoost, using metrics such as accuracy, Mean Bias Error (MBE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE). Our results suggest that the hybrid N-Beats model outperforms traditional algorithms in terms of accuracy and stability, making it a promising option for improving efficiency in green hydrogen production.
Although several methodologies, processes, and frameworks are available for constructing sophisticated autonomous multiagent systems organizations, none of them provide techniques for the reliability analysis of multiagent systems designs. This is an important issue when designing a multiagent system because of the nature of the environments where it operates (dynamic, continuous, and partially accessible). Additionally, the multiagent system must be adaptive (self-organized) to adjust its behavior to cope with the dynamic appearance and disappearance of goals (tasks), their given guidelines, and the overall goal of the multiagent system. To address such an issue, we propose a novel approach for computing the reliability, in design time, of organization-based multiagent systems. This process consists of five steps. First, the multi-agent system is designed by adopting a modified version of the OMACS framework. Second, such a design is transformed into a P-graph model to take advantage of the combinatorial nature of the underlying structure. Third, algorithm SSG of the P-graph framework is used to generate all feasible assignment sets, which represents the different ways agents can play roles to achieve goals in the organization. Fourth, for each assignment set, a Markov chain is constructed, which captures the behavior of the system; finally, algorithm R-O is executed on each Markov chain to compute their steady states (either success or failure) for further analysis. The proposed approach is validated through the simulation of two organization-based multiagent systems from the robotics domain.
Global energy demand has been growing over the last decades, having a significant impact on the environment. Alternatives that can mitigate these effects are crucial for the future of our planet. Wastewater Treatment Plants (WWTPs) are a vital infrastructure to manage residual waters, presenting an opportunity for energy production from the released biogas during the anaerobic digestion phase. The present study uses a multivariate recursive forecasting approach to evaluate the performance of Transformer-based models in forecasting electricity production in this context. Transformer-based candidate models were developed, and their hyperparameters were tuned using a grid search. The best Transformer candidate achieved the second-best Root Mean Square Error (RMSE) value of 359.4 kWh, outperforming the Gated Recurrent Unit (GRU) by 9
This study aims to create an artificial intelligence model capable of accurately identifying bronchial segments during broncho-endoscopic navigation. To achieve this, we analyzed 126 videos from bronchoscopic procedures conducted on critically ill patients at a university hospital in Buenos Aires, Argentina. A dataset of consistently annotated videos, captured by bronchoscopists with varied expertise, was established. Inter-annotator agreement for image classification was evaluated using Cohen's kappa coefficient. Images of multiple bronchial segments were used as input to train a convolutional neural network in order to obtain a classification model. This paper presents the annotation schema, labeling guidelines, the developed corpus, and some preliminary results.
Speech entrainment, particularly acoustic-prosodic (a/p) entrainment, has gained significant interest in recent years for its implications on social interaction and dialogue systems. However, measuring a/p entrainment is challenging, mainly due to the scarce consensus on its characterization, and the lack of software libraries readily available to non-expert programmers. This paper presents Entrainment-Metrics, an open-source Python toolkit for quantifying a/p entrainment in spoken dialogue. It allows users to quantify different characterizations of a/p entrainment, measure its occurrence along different a/p features, and choose between different measurement strategies. The toolkit was designed for researchers and practitioners from a wide range of fields, with no need for a strong computational background, who are interested in studying a/p entrainment and its relation to other phenomena. This paper provides a concise introduction to using Entrainment-Metrics and presents an illustrative empirical study that highlights its capabilities.
