Mourad Zegrari

Accurate forecasting of indoor environmental variables is essential for occupancy-aware building energy management, yet the interplay between neural network architecture, forecasting strategy, and downstream occupancy inference remains poorly understood. This study presents a comprehensive controlled evaluation framework for long short-term memory (LSTM)-based multihorizon indoor CO ₂ forecasting and downstream occupancy-proxy inference using two years of residential environmental sensor data. Established forecasting approaches are evaluated across six forecast horizons from 1 h to 48 h. A crossed strategy-by-hyperparameter experiment descriptively partitions observed mean absolute error (MAE) variation: hyperparameter configuration accounts for 80.7% at [Formula: see text] h, whereas strategy accounts for 39.1% at [Formula: see text] h, the larger of the two main-effect shares at that horizon. Temporal feature augmentation reduces MAE by 9.3% at [Formula: see text] h and 8.5% at [Formula: see text] h, while increasing MAE at [Formula: see text] h. Finally, we propose an unsupervised occupancy inference framework applying absolute threshold, rate-of-change, and adaptive threshold methods to both measured and LSTM-forecasted CO ₂ . These findings provide operational guidance for deploying neural network forecasters in intelligent building management systems.

Bearing fault diagnosis increasingly relies on multiple sensors, since compound or multi-location faults can produce signatures that are only partially captured by a single sensor. The decision-level fusion (DLF) of independently trained models offers a practical way to combine such complementary information, yet systematic comparisons across DLF techniques remain scarce, particularly for measurements from different sensor locations. This paper benchmarks six DLF strategies, i.e, Max, Average, Majority Voting, Weighted Sum, Dempster–Shafer, and Stacking, on a dual-branch one-dimensional convolutional neural network (1D-CNN) with each branch trained end-to-end on vibration signals from a distinct bearing location. On a two-sensor test bench covering seven health conditions, all methods exceed 99.7% accuracy on clean signals, while Dempster–Shafer fusion proves markedly more robust under noise, retaining up to 84% accuracy at a 5 dB signal-to-noise ratio (SNR). A conflict-coefficient analysis further provides an interpretable account of when fusion succeeds, linking performance to the confidence complementarity between branches.

The integration of electric vehicles (EVs) and their effects on power grids pose several challenges for distribution operators. These challenges are due to uncertain and difficult-to-predict loads. Every electric vehicle charger (EVC) has its specific pattern. This challenge can be addressed by clustering methods to determine EVC energy consumption clusters. Demand side management (DSM) is an effective solution to manage the incoming load of EVs and the large number of EVCs. Considering the challenges of peak consumptions and valleys, the adoption of vehicle-to-grid (V2G) technology requires mastering load clusters to develop energy management systems for distributors. This work used clustering algorithms (K-means, DBSCAN, C-means, BIRCH, Mean-Shift, OPTICS) to identify load curve patterns, and for performance evaluation of algorithms, it worked on metrics like the Silhouette coefficient, Calinski-Harabasz index (CHI), and Davies-Bouldin index (DBI) to evaluate results. C-means achieves the best overall clustering performance, evidenced by the highest Silhouette coefficient (0.30) and a strong Calinski-Harabasz score (543). Mean-Shift excels in the Davies-Bouldin Index (1.13) but underperforms on other metrics. BIRCH provides a balanced approach, delivering moderate results across evaluated metrics.

Given the environmental threats, the adoption of green and clean mobility is crucial for decarbonizing the mobility sector. Green mobility will bring a mass integration of electric vehicle charging stations (EVCSs) to ensure sufficiency for electric vehicle (EVs) users. To achieve this, intelligently distributing the charging load of EVs is essential to prevent stress on local electrical grids. The uneven distribution of EV charging at specific EVCSs leads to load imbalances compared to underutilized stations, necessitating dynamic load-balancing (in real time) mechanisms to optimize grid demands and prevent overloading. To address this challenge, the authors propose an algorithm for balancing EV loads at EVCSs using dynamic charging prices. This algorithm is intended to be integrated into the OCPP. Simulation results indicate that lower pricing at Station A (0.22 $/kWh) attracts more users, reducing congestion at higher-priced Stations B (0.31 $/kWh) and E (0.29 $/kWh). The proposed model encourages users to utilize less crowded stations, achieving a fairer distribution of EV charging demand while providing cost benefits to users selecting those stations.

This study focuses on demand side management (DSM), specifically managing electric vehicle (EV) charging consumption. Power distributors must consider numerous factors, such as the number of EVs, charging station availability, time of day, and EV user behavior, to accurately predict EV charging demand. We utilized machine learning algorithms and statistical modeling to predict the energy required by EV users for a specific charger and compared algorithms like K-Nearest Neighbors, XGBoost, random forest regressor, and ridge regressor. To contribute to the existing literature, which lacks studies on future energy prediction for a specific period, we conducted predictions for the next year 2024 on the energy consumption of electric vehicles for an electric vehicle charging point in a Moroccan city. These predictions can be generalized to other chargers as well. Our results showed that K-nearest neighbors (KNN) outperformed other algorithms in accuracy. This study provides valuable insights for distribution operators to manage energy resources efficiently and contributes to the DSM field by highlighting the effectiveness of KNN in predicting EV charging demand.

This paper focuses on the complex dynamics that concern assembly line balance in the context of mass customization within manufacturing. In fact, the increase in demand for customized products has heightened the complexities associated with achieving optimal efficiency, productivity, product quality, and customer satisfaction. The research proposes a multi-criteria analysis of statistical methods to determine the fluctuation of parameters affecting the state of balance of an assembly line. A 3D matrix model is suggested to analyze the parameters managing the assembly line. This representation is executed using the MATLAB R2024b tool, and a methodology for finding the variability of parameters affecting balance through statistical approaches is proposed. We observed that changes in parameters such as task times, worker efficiency, or material flow led to significant changes in the line’s overall balance. As a result, static balancing becomes inadequate to deal with the complexities introduced by these highly variable parameters. The novelty of this paper consists of the innovative integration of multi-criteria statistical analysis and 3D matrix modeling to detect parameter variability and optimize assembly line balancing. Conventional static approaches are often unable to capture the process-dynamic aspect of modern manufacturing. This work presents a systematic methodology capable of identifying, quantifying, and moderating the variability of key operating parameters. This methodology, carried out using MATLAB-based simulations, is based on principal component analysis (PCA) and correlation analysis to detect critical factors influencing balancing efficiency. By structuring assembly line parameters in a 3D matrix representation, this research gives a holistic, data-based method for improving decision-making in balancing procedures. The research goes beyond theoretical modeling by applying the approach to a real automotive assembly line, validating its effectiveness and demonstrating its practical applicability in industrial conditions.

Traditional multilevel inverter topologies, such FC, NPC, and CHB, have a few significant disadvantages. They need a great number of parts, which raises the complexity, expense, and switching losses. Furthermore, their intricate control schemes make voltage balancing and synchronization challenging. Lastly, under some circumstances, they experience severe harmonic distortion, necessitating the inclusion of expensive filters to enhance signal quality. This paper proposes a novel multilevel converter topology that uses the phase-disposition PWM (PD-PWM) technique to control a 19-level output. This new configuration maintains performance comparable to the CHB-MLI reference while using fewer switches, simplifying control, and reducing costs. Our approach is based on extensive simulations conducted in the MATLAB Simulink environment, with results compared to the CHB-MLI. A low-pass filter is added to improve the output voltage quality, reducing the THD% to 1.33%. This strategy offers several advantages, including simpler control, lower costs, increased reliability, and higher-quality output. The system was replicated using MATLAB Simulink and validated through hardware-in-the-loop (HIL) testing. The HIL method ensures real-world testing without causing damage to the hardware. The integrated system includes sensors and necessary hardware for a comprehensive energy management solution.