Ultra-short-term photovoltaic (PV) power forecasts are vital for secure grid operation as solar penetration rises. We propose a two-stage hybrid framework, WDT–CRMABIL–Fusion. In Stage 1, we apply a three-level discrete wavelet transform to PV power and key meteorological series (shortwave radiation and panel irradiance). We then forecast the approximation and detail sub-series using specialized component predictors: a 1D-CNN with dual residual multi-head attention (feature-wise and time-wise) together with a BiLSTM. In Stage 2, a compact dense fusion network recombines the component forecasts into the final PV power trajectory. We use 5-min data from a PV plant in Milan and evaluate 5-, 10-, and 15-min horizons. The proposed approach outperforms strong baselines (DCC+LSTM, CNN+LSTM, CNN+BiLSTM, CRMABIL direct, and WDT+CRMABIL direct). For the 5-min horizon, it achieves MAE = 1.60 W and RMSE = 4.21 W with R2 = 0.943 and CORR = 0.973, compared with the best benchmark (MAE = 3.87 W; RMSE = 7.89 W). The gains persist across K-means++ weather clusters (rainy/sunny/cloudy) and across seasons. By combining explicit multi-scale decomposition, attention-based sequence learning, and learned fusion, WDT–CRMABIL–Fusion provides accurate and robust ultra-short-term PV forecasts suitable for storage dispatch and reserve scheduling.

The resilience of electricity networks to extreme weather is an essential component of sustainable energy systems, as power interruptions can affect essential services, economic activities, public safety, and the integration of renewable energy sources. However, this assessment can be challenging due to the inherent differences and complexities between weather and grid data. To address this issue, we developed a data pipeline to acquire, prepare, and integrate power grid outage records, meteorological measurements, and lightning strike data into a unified relational database. The framework was applied to the Italian power grid, with detailed regional analyses conducted for Piedmont and Campania. The results show that, for more than 90% of the matched weather-related failures, at least one relevant weather condition was identified in the integrated datasets. Rainfall showed the most widespread association with failures throughout the year, while lightning was particularly relevant during summer and snowfall during colder months. The proposed framework provides a reproducible basis for monitoring weather-related infrastructure vulnerability and can support evidence-based maintenance prioritization, climate adaptation, and resilience-oriented grid planning. By facilitating the identification of regional risk patterns, the approach contributes to the development of more reliable and sustainable electricity infrastructure.

Accurate photovoltaic power forecasting is essential for the reliable integration of solar energy into the electrical grid. This work presents a high-resolution dataset and acquisition framework that integrates electrical measurements, environmental variables, and solar position data into a unified relational database, suitable for PV power prediction across all temporal horizons. Using this dataset, we focus on very-short-term forecasting and propose a comprehensive forecasting framework that combines wavelet-based feature extraction with advanced deep learning techniques. The framework is evaluated across forecasting horizons from 5 to 30 min, achieving nMAE values between 0.73% and 4.64%, nRMSE between 1.65% and 7.98%, and PICP ranging from 62.4% to 74.7%. Robustness is assessed by simulating realistic cloud-induced perturbations in the input data. A hybrid approach that combines the deep learning model with a gradient boosting regressor to correct residual errors reduces the overall nMAE from 4.72% to 3.89% and nRMSE from 9.52% to 6.83%, effectively mitigating large errors caused by abrupt power fluctuations. These results demonstrate the framework’s ability to provide accurate and reliable probabilistic forecasts under both standard and perturbed conditions, offering a solid foundation for future PV prediction research and practical applications.

Lithium-ion batteries are widely used across diverse applications due to their high energy density, long cycle life, and fast charging capabilities. As battery-powered systems become increasingly critical, accurate estimation of the Remaining Useful Life (RUL) is essential for ensuring reliability, safety, and effective maintenance planning. This work investigates Evidential Deep Learning (EDL) for data-driven RUL estimation and introduces a novel risk-aware loss function designed to enhance both predictive accuracy and uncertainty quantification in the End-of-Life (EoL) region, where precise and trustworthy predictions are most needed. Using a publicly available dataset of lithium iron phosphate (LFP) cells, we benchmark the proposed approach against a baseline Conv-LSTM model, Monte Carlo (MC) Dropout, and Deep Ensembles. The results show that integrating the risk-aware loss into the EDL framework substantially improves the calibration of predictive uncertainty while achieving state-of-the-art accuracy near EoL. Unlike MC Dropout and Deep Ensembles, which exhibit increasing or unstable uncertainty as degradation accelerates, the proposed EDL model demonstrates a consistent reduction in uncertainty and significantly higher reliability in late-stage predictions. The findings indicate that the risk-aware evidential framework offers a reliable and computationally efficient solution for battery RUL estimation, enabling more informed decision-making in both safety-critical and consumer-oriented applications.

This paper presents an integrated approach to power quality monitoring across Italian transmission and distribution systems. The proposed methodology combines datasets from the Italian transmission system operator, Terna, and from the QuEEN monitoring system developed by RSE, which collects power quality information in terms of voltage events and variations on medium voltage busbars in primary substations. The integration enables a unified SQL-based database architecture that supports advanced statistical analysis and cross-layer event correlation. Two case studies are presented considering a 14-years observation period: i) the evaluation of the selectivity of the protection systems carried out by analyzing the number of load disconnections per failure; ii) the investigation of the correlation between high voltage outages and medium voltage interruptions recorded by QuEEN measurement units. Results highlight a gradual decreasing trend in system selectivity and a significant match between high and medium voltage events, highlighting the benefits brought by an integrated monitoring for reliability assessment and predictive maintenance.

In recent years, batteries, particularly lithium-ion batteries, have become essential to a wide range of applications, from everyday technologies like portable electronics and electric vehicles to industrial applications. Estimating a battery's Remaining Useful Life (RUL), i.e., the number of charge and discharge cycles it can perform before needing replacement, is a critical aspect of predictive maintenance. Accurately determining a battery's RUL, especially as it approaches its End of Life (EoL), enhances system reliability, improves maintenance practices, and helps reduce costs. Furthermore, in industrial environments, providing clear explanations of Artificial Intelligence (AI) model outputs is essential to building trust in AI, guaranteeing safety, and facilitating smoother decision-making, as these outputs directly influence operational processes. In this paper, we propose a novel set of features based on temperature and capacity data, features commonly found in public datasets, to predict RUL. Additionally, we introduce the Monte Carlo dropout technique during inference to enrich the model's output. This approach provides not only the predicted RUL values but also the standard deviation and distribution of the predictions, making the decision process more transparent and reliable.

Accurately estimating the Remaining Useful Life (RUL) of lithium-ion batteries using data-driven models requires large amounts of clean, diverse data. However, publicly available datasets are often collected under varying and limited testing conditions, which poses challenges for model training and generalization. To address this, we present a pipeline for cleaning, transforming, and integrating two distinct datasets covering four battery specifications and a wide range of test scenarios. We apply filtering and normalization techniques to address noise, outliers, and inconsistencies, and organize the cleaned data into a relational database. Building on this integrated dataset, we train a Conv-LSTM neural network and evaluate its performance against models trained on individual battery groups. Our results show that multi-source training improves model generalization, particularly for smaller or more heterogeneous datasets, and can help enhance prediction accuracy in the later stages of battery life.

Power grids play a fundamental role in modern society, providing the energy that sustains all daily operations. Given their critical importance, understanding and analyzing outages in these systems is essential to guarantee consistent and reliable energy delivery. While the Italian transmission system operator publishes outage data and information on user sites connected to the grid, they are often provided in formats that can complicate access. Additionally, the structure of the tables frequently changes over the years, making integration a challenging task. To address this, we developed a data pipeline to acquire, prepare, and integrate these data into a unified relational database, with the ultimate goal of facilitating reliability analyses for the Italian energy network. In this work, we focused on designing the database, particularly its conceptual and logical schema, and integrating all power grid data while maintaining high data quality throughout the pipeline. Finally, we conducted a failure trend analysis, using the extensive data from our final database, to evaluate how the reliability of the grid has changed over time. From this analysis, we observed that the number of outages and their duration has increased over the years, exhibiting an overall upward trend.

Predictive maintenance in industrial machinery relies on the timely detection of component faults to prevent costly downtime. Rolling bearings, being critical elements, are particularly prone to defects such as outer race faults and ball spin defects, which manifest as characteristic vibration patterns. In this study, we introduce a novel bearing vibration dataset collected on a testbench under both constant and variable rotational speeds (0–5000 rpm), encompassing healthy and faulty conditions. The dataset was used for failure classification and further enriched through feature engineering, resulting in input features that include raw acceleration, signal envelopes, and time- and frequency-domain statistical descriptors, which capture fault-specific signatures. To quantify prediction uncertainty, two different approaches are applied, providing confidence measures alongside model outputs. Our results demonstrate the progressive improvement of classification accuracy from 87.2% using only raw acceleration data to 99.3% with a CNN-BiLSTM (Convolutional Neural Network–Bidirectional Long Short-Term Memory) ensemble and advanced features. Shapley Additive Explanation (SHAP)-based explainability further validates the relevance of frequency-domain features for distinguishing fault types. The proposed methodology offers a robust and interpretable framework for industrial fault diagnosis, capable of handling both stationary and non-stationary operating conditions.

Lithium-ion batteries (LiBs) are widely used in diverse applications due to their high energy and power density, efficiency, and long cycle life. However, their performance varies over time due to a combination of reversible effects and irreversible aging mechanisms, driven by factors such as temperature, state of charge (SOC), and current rate. Accurate estimation of key battery parameters—including internal resistance—is essential for assessing the state of health (SOH), SOC, and state of power, which are critical for reliable operation and battery management. This work proposes a machine learning approach for estimating the low-frequency internal resistance of LiBs, leveraging prior estimation of the high-frequency component. The method builds on previous research and aims to provide accurate predictions across varying SOC, temperature, and aging conditions, while avoiding the need for complex hardware or intensive computations. The results show that, although estimating low-frequency resistance is inherently challenging, the use of a neural network—especially when incorporating high-frequency resistance as an input—reduces the estimation error to below 3%. The proposed model represents a promising solution for practical, real-time resistance estimation in advanced battery management systems.

Advances in Industry 4.0 and the emergence of Industry 5.0 are driving the development of intelligent, sustainable manufacturing systems, where embedded sensing and real-time health diagnostics play a critical role. However, implementing robust predictive maintenance in production environments remains challenging due to the variability in machine operations and the lack of access to internal control data. This paper introduces a lightweight, embedded-compatible framework for health status signature extraction based on empirical mode decomposition (EMD), leveraging only data from a single triaxial accelerometer. The core of the proposed method is a cycle-synchronized segmentation strategy that uses accelerometer-derived velocity profiles and cross-correlation to align signals with machining cycles, eliminating the need for controller or encoder access. This ensures process-aware decomposition that preserves the operational context across diverse and dynamic machining conditions to address the inadequate segmentation of unstable process data that often fails to capture the full scope of the process, resulting in misinterpretation. The performance is evaluated on a challenging real-world manufacturing benchmark where the extracted intrinsic mode functions (IMFs) are analyzed in the frequency domain, including quantitative evaluation. As results show, the proposed method shows its effectiveness in detecting subtle degradations, following a low computational footprint, and its suitability for deployment in embedded predictive maintenance systems on brownfield or controller-limited machinery.

Predictive maintenance is crucial in modern industrial settings, as it minimizes unexpected failures and optimizes operational costs. Rolling bearings, essential components in rotating machinery, are particularly prone to wear and faults, necessitating continuous real-time monitoring. This paper introduces a novel dataset comprising vibrational data collected from bearings under varying operational conditions on a test bench. A Convolutional Neural Network (CNN) is employed for fault classification, leveraging domain-specific feature selection, including raw acceleration signals and their envelope extractions. Additionally, Monte Carlo (MC) dropout is applied to quantify prediction uncertainty, enhancing trust in AI-driven fault detection. Results demonstrate that incorporating feature engineering and dataset balancing significantly improves classification accuracy, while MC dropout provides valuable uncertainty estimates. The proposed approach contributes to more reliable and interpretable machine learning models for industrial fault diagnosis.

In today’s electronics manufacturing, Printed Circuit Board (PCB) component welding faults are a critical issue that can significantly impact the reliability and functionality of devices. Accurate detection of anomalies in welded components not only improves the quality and efficiency of the manufacturing process but also helps reduce economic waste.This paper presents a data pipeline that combines classical machine learning and deep learning techniques with computer vision to identify potential soldering defects in PCB assemblies. The proposed method involves cleaning the data obtained from Automated Optical Inspection (AOI) systems to create a dataset suitable for machine learning tasks, such as classification and anomaly detection.We then use this cleaned data to develop two classification models: a Random Forest (RF) model and a Convolutional Neural Network (CNN). These models are applied to perform both binary classification—distinguishing between defective and non-defective components—and multi-class classification to identify specific types of defects. The models are also tested on noisy data to assess their reliability when exposed to data differing from the training set.The results show good accuracy across almost all tests, with the CNN model demonstrating greater resilience to noise in the test set.

Machine tools are critical to modern manufacturing, yet their high energy consumption and vulnerability to faults present significant operational challenges. While predictive models can enhance energy optimization and fault diagnosis, their performance is often constrained by the scarcity of high-quality training data. To address this gap, this study presents a real-time digital twin (DT) framework that integrates OPAL-RT HIL simulation with OPC-UA-based cloud communication. The system enables both energy monitoring and synthetic fault data generation under diverse machining conditions. The DT operates in a bidirectional loop with a cloud-based data acquisition layer, allowing real-time parameter input and retrieval of simulated outputs. Model fidelity is verified by aligning simulation results with real-world CNC machine measurements and further confirmed through pattern-based external validation. The framework is applied to analyze energy consumption across varying machining parameters — such as electrospindle speed, feed rate, tool length, and depth of cut — and to simulate bearing fault scenarios for evaluating their impact on power consumption. These simulations produce labeled datasets suitable for future diagnostic and predictive maintenance applications. This work delivers a validated, closed-loop DT framework that unites high-fidelity OPAL-RT simulation, real-time OPC-UA data exchange, and synthetic data generation, extending predictive maintenance capabilities beyond those of prior modeling or diagnostic approaches. The proposed methodology offers a scalable foundation for energy-aware machining and real-time fault detection, contributing to sustainable manufacturing practices and operational resilience in smart industrial systems.

Power grid failures represent a significant challenge to grid reliability, with far-reaching impacts on both the economy and society. In Italy, thousands of these failures occur annually, differing widely in severity, the number of affected user sites, and associated costs. Gaining a detailed understanding of these failures at the component level and identifying their underlying causes is crucial for enhancing grid resilience. This paper focuses on overhead transmission lines (OTL), the most expansive and critical components of the power grid. Due to their large geographical coverage, OTL are highly vulnerable to environmental conditions, aging infrastructure, and accidental damage. This study offers a detailed analysis of failure severity and reliability for Italian OTL, utilizing different metrics categorizing failures by transmission line voltage levels and causes. A refined severity factor is introduced to quantify and generalize the impact of failures, enabling the identification of trends and vulnerabilities. The findings highlight the most vulnerable components within OTL and emphasize the importance of implementing targeted maintenance strategies to address failures driven by environmental factors and infrastructure aging. By examining failure causes and their consequences, this work provides valuable insights for enhancing grid reliability, optimizing resource allocation, and supporting a more resilient and efficient power transmission system.

Evaluating the quality of the machining process annotated by experts on the floor in case of developing a silent anomaly is a challenging task. Components wear, wrongly labeled processes, or highly imbalanced data are some examples of real-world difficulties that may prevent the reliability of machine learning algorithms in the manufacturing environment. Since human experts may face several challenges while annotating such high-frequency data, this letter evaluates effective health indexes using time-frequency analysis to extract reliable patterns or vibration signatures assigned to the process quality or bearing health status. A benchmark dataset for process monitoring of Brownfield milling machines over two years is utilized in this letter where the resulting process is evaluated by experts in a gauging station. Vibration signals are collected from three different computer numerical control (CNC) using a triaxial accelerometer, which is mounted on the rear side of the machines. Considering a single operation, the extracted vibration signature is validated on two test CNC machines. As results show, the overall energy level in the frequency range of 0-1 kHz while considering only radial axes gives effective insight into the quality of the process and degradation pattern.

Grid-connected photovoltaic (PV) systems are commonly designed for maximum energy production. However, as their presence grows, revised grid regulations increasingly require these systems to partially adhere to the primary frequency regulation (PFR) by reducing their power production when the grid frequency exceeds the nominal value. Nevertheless, full participation in PFR of PV systems, without battery storage, would require operating with a dedicated active power reserve to be utilized in the event of underfrequency conditions. This paper presents a model-based (MB) flexible power point tracking (FPPT) algorithm for PV systems. Results of a microgrid simulation show that the proposed algorithm enables a PV system with a nominal power of 100 kW to maintain a 20 kW active power reserve under various irradiation and temperature conditions, with a relative error lower than 3%. Furthermore, a simulation of the PV system assisting in primary frequency regulation is presented. Thanks to the proposed strategy, when employing the algorithm, the frequency deviation is reduced both in the event of underfrequency and overfrequency conditions. By enabling operation with a constant active power reserve, as well as a quick adjustment of the produced active power, in the event of frequency imbalance, the proposed MB algorithm is an interesting candidate for achieving complete PFR capabilities of PV systems.

Lithium-ion batteries (LiBs) are extensively used in numerous applications, with electric vehicles being one of the most important. Consequently, significant research efforts have been directed towards developing battery models that can predict battery behavior, increase efficiency, enhance safety, and reduce degradation. One of the key state parameters for this purpose is the state of charge (SOC). Many estimation methods in the literature rely on understanding the open circuit voltage (OCV)-SOC relationship, which is affected by battery temperature and can be modeled using different approaches, such as table-based, analytical, physical-based, and machine learning (ML). ML approaches are gaining increasing popularity and interest, although they require extensive experimental data and the identification of the most informative features. Specifically, given the nature of the problem, we proposed an ML algorithm based on deep neural networks capable of estimating the SOC of an LiB for electric vehicles using only a single measurement of the actual OCV and battery temperature. Finally, the proposed algorithm was validated through an extensive experimental campaign.

The concept of the Digital Twin (DT) has risen to prominence in parallel with the ongoing digitization initiatives across industries and the rise of Industry 4.0. This innovative approach harnesses a diverse array of cutting-edge technologies, including Artificial Intelligence (AI), Big Data analytics, and the Internet of Things (Io'T), to smoothly merge virtual representations with their physical counterparts. Through this integration, organizations can conduct in-depth analysis and make real-time decisions, all backed by data. This transformative fusion of digital and physical fields not only enhances operational efficiency but also opens up new avenues for innovation and value creation within various sectors.