Search

This article presents a new algorithm that uses techniques from the field of artificial intelligence to automatically estimate the physical parameters of massive stars from a grid of stellar spectral models. This is the first grid to consider hydrodynamic solutions for stellar winds and radiative transport, containing more than 573 thousand synthetic spectra. The methodology involves grouping spectral models using deep learning and clustering techniques. The goal is to delineate the search regions and differentiate the "species" of spectra based on the shapes of the spectral line profiles. Synthetic spectra close to an observed stellar spectrum are selected using deep learning and unsupervised clustering algorithms. As a result, for each spectrum, we found the effective temperature, surface gravity, micro-turbulence velocity, and abundance of elements, such as helium and silicon. In addition, the values of the line force parameters were obtained. The developed algorithm was tested with 40 observed spectra, achieving 85 % of the expected results according to the scientific literature. The execution time ranged from 6 to 13 min per spectrum, which represents less than 5 % of the total time required for a one-to-one comparison search under the same conditions. [ABSTRACT FROM AUTHOR]

Predicting solar power generation is a complex challenge with multiple issues, such as data quality and choice of methods, which are crucial to effectively integrate solar power into power grids and manage photovoltaic plants. This study creates a hybrid methodology to improve the accuracy of short-term power prediction forecasts using a model called Transformer Bi-LSTM (Bidirectional Long Short-Term Memory). This model, which combines elements from the transformer architecture and bidirectional LSTM (Long–Short-Term Memory), is evaluated using two strategies: the first strategy makes a direct prediction using meteorological data, while the second employs a chain of deep learning models based on transfer learning, thus simulating the traditional physical chain model. The proposed approach improves performance and allows you to incorporate physical models to refine forecasts. The results outperform existing methods on metrics such as mean absolute error, specifically by around 24%, which could positively impact power grid operation and solar adoption. [ABSTRACT FROM AUTHOR]

In recent years, the distribution network in Chile has undergone various modifications to meet new demands and integrate new technologies. However, these improvements often do not last as long as expected due to inaccurate forecasting, resulting in frequent equipment changes and service interruptions. These issues affect project investment, unsold energy, and penalties for poor quality of supply. Understanding the electricity market, especially in distribution, is crucial and requires linking technical quality standards with service quality factors, such as the frequency and duration of interruptions, to understand their impact on regulated distribution to customers. In this context, a comparative study will be carried out between Long Short-Term Memory (LSTM) and transformer architectures, with the aim of improving the sizing of distribution transformers and preventing failures when determining the nominal power of the transformer to be installed. Variables such as voltages and operating currents of transformers installed between 2020 and 2021 in the Valparaíso region, Chile, along with the type and number of connected customers, maximum and minimum temperatures of the sectors of interest, and seasonality considerations will be used. The compilation of previous studies and the identification of key variables will help to propose solutions based on error percentages to optimise the accuracy of transformer sizing. [ABSTRACT FROM AUTHOR]

The quality of verbal communication, understood as the absence of uncertainty in the message transmitted, is a key factor in mission-critical processes. Several processes are handled by direct voice communication between these endpoints and any miscommunication could have an impact in success of the task. For that reason, the quality control of verbal communication is required to ensure that the instructions issued are effectively understood and adequately executed. In this context, it is expected that instructions from the command center are issued once, and that the acknowledgment from the field are minimal. In the present work, the communication between an electrical company control center and factory workers in the field was chosen for analysis. We developed two complementary approaches by using machine learning and deep learning algorithms to assess, in an automatic way, the quality of information transmission in the voice communications. Preliminary results demonstrate that the automatic uncertainty detection is feasible, despite the small number of samples available at the present time. To support further studies, a repository was created in GitHub with the spectrogram and the tokenized words of all audios. [ABSTRACT FROM AUTHOR]