The growth of Renewable Energy Communities (RECs) across Europe has increased the need for reliable tools to assess their technical performance and economic feasibility. In Italy, the implementation of the RED II Directive has made the design of RECs particularly relevant, especially under collective self-consumption schemes, where performance depends on the temporal alignment between photovoltaic generation and demand. This study proposes an integrated framework combining physics-based modeling and data-driven techniques to analyze and optimize energy sharing within RECs. Using real consumption profiles from ARERA and photovoltaic production simulated through PVGIS for a 3 kW system, multiple community configurations are evaluated to identify the structure that maximizes shared energy. Energy flows are modeled through a physically consistent formulation linking production, self-consumption, injected energy, and shared energy. After identifying the optimal REC configuration through combinatorial analysis, a machine learning approach is applied to model the relationship between energy variables and the economic value of shared energy, focusing on the best-performing configuration. The results show that the system is governed by a constrained relationship driven by the minimum between injected energy and demand. The superior performance of the linear model suggests that the problem is highly structured and largely predictable. Overall, the framework combines physical interpretability and predictive modeling, providing a practical and transparent tool for REC design and optimization.
We present an analytical series solution for steady magnetoconvection of an electrically conducting micropolar fluid in a vertical parallel plate channel under the Oberbeck-Boussinesq approximation. The fully developed model yields a coupled boundary value system for the axial velocity, microrotation, induced magnetic field, and temperature. Adomian’s decomposition method is applied directly to the coupled second order formulation, producing compact truncated series that satisfy the wall conditions and make the dependence on the governing nondimensional parameters explicit. The resulting expressions are used to examine how the Hartmann number, coupling number, micropolar parameter, and buoyancy parameter shape the profiles of V(y), W(y), and H(y). The analysis highlights departures from the Newtonian limit as micropolar coupling strengthens, while increasing magnetic effects damp the axial motion and modify magnetic induction.
Academic fraud refers to various dishonest behaviors that violate standards of integrity and transparency. Unethical activities include research misconduct, plagiarism, peer review manipulation, and data fabrication or falsification. Because academic fraud undermines the credibility of scientific research, it is crucial to investigate it. In this paper, we analyze suspicious data contained in retracted papers about the use of chloroquine for COVID-19 treatment. To accomplish this, we use Benford's law, a well-known mathematical model employed to describe the first significant digit distribution in several collections of numbers. In fact, it is considered a useful tool in data analysis to detect frauds, identify financial statement manipulation and anomalies in income reporting, assess the integrity of experimental data, and verify the accuracy of reported health statistics. A typical way to check if a statistical population adheres to Benford's law is through hypothesis testing. We report some tests to check the Benfordness of data and apply them to a real dataset.
When we aim to estimate a density function with bounded support, the naive kernel density estimation becomes strongly biased in the boundary region due to the well-known estimated density overflow problem. The reflection method appears to be an efficient way to alleviate this issue. We propose a simple domain transformation that could favor an elegant application of such a principle. A small numerical experiment and a real case study involving suicide data end the work.
We develop an ensemble machine learning framework integrating environmental, ecological, and socioeconomic variables to enable prospective hantavirus risk prediction. Trained on 689 laboratory-confirmed cases from three United States jurisdictions, the ensemble of random forest, gradient boosting, support vector machine, and logistic regression classifiers achieves area under the receiver operating characteristic curve of 0.92 on independent test data. Shapley additive explanations identify precipitation variability, land-use patterns, and rodent species richness as dominant predictors, with substantial contributions from socioeconomic determinants.
