Mauro Tucci

A nonlinear parametric model order reduction approach based on random selection of parameters and hyper-reduction is proposed and applied for the computation of the induced electric field in a subject’s head due to transcranial magnetic stimulation. The resulting reduced order model dramatically decreases computational times when simulating several positions and orientations of the excitation coil in a chosen region of interest. In particular the proposed model allows to obtain field solutions in a faster way with respect to classical solvers as Finite Element Methods without losing accuracy.

Accurate forecasting of Photovoltaic (PV) power generation is essential for integrating renewable energy into power grids, ensuring operational stability, and advancing sustainable energy management. This study examines the influence of training data availability, forecast horizons, and ensemble learning techniques on the accuracy of regional PV power forecasting. A sliding window approach is employed to evaluate model performance over an entire year, capturing the effects of seasonal variations and diverse data conditions. The results show that the amount of available training data, the selected forecast horizon, and the specific combination of base models within ensemble methods substantially influence forecasting accuracy. Crucially, the findings reveal that the effectiveness of an ensemble model is shaped by the learning algorithm and the degree to which its base models are well-suited to the characteristics of the input data and the forecasting timeframe. These insights highlight the need for adaptive ensemble configurations tailored to both data context and prediction objectives, providing practical guidance for the development of more reliable and flexible PV forecasting solutions.

In this contribution the authors propose a hybrid Boundary Element Method – Physics Informed Neural Networks (BEM – PINN) approach, to be used for the resolution of partial differential equations arising when formulating boundary-value problems in electromagnetism. The approach retains both the advantages of integral methods (compact representation and no need to mesh large domains) and differential methods, where the term “differential” refers here to the Automatic Differentiation carried out during the training phase of the PINN. The method is easy to implement and adds an additional flexibility to purely PINN based solution methods.

One of the main goals of the International Energy Agency (IEA) is to manage and utilize clean energy to achieve net zero emissions by 2050. Hydropower plants can significantly contribute to this goal as they are vital components of the global energy infrastructure, providing a clean, safe, and sustainable power source. Accordingly, there is great interest in developing methods to prevent errors and anomalies and ensure full operational availability. With modern hydropower plants equipped with sensors that capture extensive data, machine learning algorithms utilizing these data to detect and predict anomalies have gained research attention. This paper demonstrates that deep learning algorithms are particularly powerful in predicting time series. Three well-known deep learning networks are examined and compared to previous approaches, followed by the introduction of a new, innovative hybrid network. Using real-world data from two hydropower plants, the hybrid model outperforms individual deep learning models by achieving more accurate fault detection, reducing false positives, offering early fault prediction, and identifying faults several weeks before occurrence. These results showcase the hybrid network’s potential to enhance maintenance planning, reduce downtime, and improve operational efficiency in energy systems.

In this study, an in-depth analysis is presented on forecasting aggregated wind power production at the regional level, using advanced Machine-Learning (ML) techniques and feature-selection methods. The main problem consists of selecting the wind speed measuring points within a large region, as the wind plant locations are assumed to be unknown. For this purpose, the main cities (province capitals) are considered as possible features and four feature-selection methods are explored: Pearson correlation, Spearman correlation, mutual information, and Chi-squared test with Fisher score. The results demonstrate that proper feature selection significantly improves prediction performance, particularly when dealing with high-dimensional data and regional forecasting challenges. Additionally, the performance of five prominent machine-learning models is analyzed: Long Short-Term Memory (LSTM) networks, Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), Convolutional Neural Networks (CNNs), and Extreme-Learning Machines (ELMs). Through rigorous testing, LSTM is identified as the most effective model for the case study in northern Italy. This study offers valuable insights into optimizing wind power forecasting models and underscores the importance of feature selection in achieving reliable and accurate predictions.

Predicting electricity production from renewable energy sources, such as solar photovoltaic installations, is crucial for effective grid management and energy planning in the transition towards a sustainable future. This study proposes machine learning approaches for predicting electricity production from solar photovoltaic installations at a regional level in Italy, not using data on individual installations. Addressing the challenge of diverse data availability between pinpoint meteorological inputs and aggregated power data for entire regions, we propose leveraging meteorological data from the centroid of each Italian province within each region. Particular attention is given to the selection of the best input features, which leads to augmenting the input with 1-hour-lagged meteorological data and previous-hour power data. Several ML approaches were compared and examined, optimizing the hyperparameters through five-fold cross-validation. The hourly predictions encompass a time horizon ranging from 1 to 24 h. Among tested methods, Kernel Ridge Regression and Random Forest Regression emerge as the most effective models for our specific application. We also performed experiments to assess how frequently the models should be retrained and how frequently the hyperparameters should be optimized in order to comprise between accuracy and computational costs. Our results indicate that once trained, the model can provide accurate predictions for extended periods without frequent retraining, highlighting its long-term reliability.

In this article, a machine-learning-based model for the regression of cloud-to-ground lightning location and peak current from time-domain waveforms of lightning-induced voltage measurements on overhead transmission lines is presented. A principal component analysis (PCA) procedure is applied for extracting significant features and decreasing the dimension of the input vector. Then, a shallow neural network is trained with the results of the PCA. The obtained results show that the proposed approach can be the base for a tool able to regress lighting location with an accuracy comparable to or even better than traditional methods [i.e., lightning location system (LLS)] and provide a peak current estimate more accurate than LLS and more actual and widespread than direct tower measurements (which are limited to a reduced number of recorded events in some specific regions). Such a tool would also have significant advantages in terms of costs, since it would not require a dedicated instrumentation.