Power, Control, and Data Processing Systems

Abstract This paper presents a voltage-aware operational framework for active distribution networks using a mixed-integer linearized optimization model. Unlike conventional approaches that treat voltage limits as hard constraints, a linear voltage deviation penalty is incorporated directly into the objective function, enabling voltage conditions to influence operational decisions, particularly the charging and discharging schedules of energy storage systems. The proposed formulation jointly considers active and reactive power dispatch, storage operation, and network losses within a single computationally tractable framework. Simulation results on the IEEE 33-bus distribution system demonstrate that the proposed approach leads to a smoother voltage profile at critical buses over the scheduling horizon, while maintaining voltages within acceptable limits. Compared to the base case, the total network losses are reduced from 3028.48 kW to 3027.66 kW, corresponding to a relative reduction of approximately 0.03%. In parallel, the modified objective function value reflects the explicit consideration of voltage quality, resulting in an increase of about 9.2% in the objective value, which represents a trade-off between voltage performance and operational cost. The results confirm that incorporating voltage deviation into the optimization objective enables voltage-aware storage operation and provides a transparent mechanism to evaluate its impact on network performance.

Abstract A key challenge in processing electrocardiogram (ECG) signals for personalized biometric authentication systems is the efficient retraining of deep learning models when new user data become available. This study employs a transfer learning approach combined with the VGG16 architecture to investigate the permissible threshold for repeatedly integrating data from a single user—assigned with varying access labels—into a pre-trained base model. The base model was initially trained on 46 users from the MIT-BIH database. For each target user, data were simulated across 100 iterations with random label assignments and incrementally incorporated via gradual fine-tuning. The results reveal that the final model's performance improves with the addition of up to approximately 44 to 46 target models—a quantity comparable to the number of initial users—after which performance declines. The optimal range for adding new data from a single user was identified as 11 to 16 iterations. Within this interval, the model achieves an accuracy exceeding 99% without exhibiting catastrophic forgetting. These findings confirm that while transfer learning offers an efficient mechanism for frequent updates in ECG-based biometric systems, it possesses a finite capacity for accommodating repeated modifications. Determining this threshold provides valuable practical guidance for the development of security and health monitoring systems that must balance functional flexibility with overall model stability.

Abstract Facial expressions are one of the most important techniques of non-verbal communication that humans use to display their emotional states and establish non-verbal communication. In recent research in the field of emotion recognition based on facial expression changes, deep learning networks are used. In this regard, an appropriate network is selected according to the type of data. In the present research, video data is used, and for recognizing facial expressions, recurrent networks are employed. Subsequently, to achieve better results, a hybrid network consisting of Convolutional, Recurrent, and Restricted Boltzmann Machine networks has also been used. Finally, in this paper, using deep learning techniques and the Python language, an algorithm based on a hybrid CNN-LSTM-RBM neural network for emotion recognition through facial expressions in a video is presented. In this method, seven emotional states (happy, sad, disgust, excitement, contempt, fear, and anger) are detected. The CK+48 dataset has been used to train and test the proposed model. The experimental results demonstrate that the proposed Hybrid CNN-LSTM-GRBM model achieves a training accuracy of 99.06%. Crucially, the model attains 91.16% test accuracy and an F1-score of 89.27% on the test set, outperforming existing baseline methods and confirming its robustness in handling spatiotemporal facial features.

Abstract This paper proposes a multi-objective optimization for the optimal energy scheduling of renewable-based energy system. The proposed model simultaneously considers both economic and environmental indices in the operation planning to present a comprehensive scheduling for the energy system. The studied energy system integrates the electrical, heating, and cooling sectors to increase the synergy, reliability, and flexibility of the renewable-based energy system. This integration improves the utilization of renewable energy system because the surplus generation of photovoltaic and wind energy can be converted to other forms of energy to supply the required heat and cooling energy of the system. In addition, Shannon entropy and multi-attribute utility theory (MAUT) are used to determine the relative importance of economic versus environmental criteria and to prioritize the solutions based on the operator’s viewpoint. The proposed model is tested on a standard case study and the numerical results present that the proposed model improves the emission of greenhouse gases by 4.8 % compared to single objective optimization in the normal condition. Also, compared to the scenario where no demand response is present, the demand response programs decreases the operating cost of the multi-energy system by 2.47 %.

Abstract The increasing penetration of wind power has made Optimal Power Flow (OPF) studies more dependent on uncertainty modeling, reserve allocation, and accurate market-price calculation. This paper develops a linearized OPF framework for evaluating Locational Marginal Prices (LMPs) and reserve requirements in wind-integrated power systems. The model distinguishes between flexible ramping reserves and contingency reserves and includes constraints that prevent thermal units from assigning the same unused capacity to both products. To improve the accuracy of the conventional Direct Current Optimal Power Flow (DC-OPF) while preserving a linear structure, transmission losses are represented through a loss-approximated formulation. The proposed model is tested on the IEEE 118-bus system using GAMS and MATPOWER. The results show that reserve separation affects both dispatch decisions and nodal LMPs, especially under wind-power deviations. A sensitivity study also shows that changing the objective from generation-cost minimization to active-power-loss minimization reduces network losses from 69.3 MW to 18.4 MW, at the expense of a higher generation cost.

Abstract Unit commitment (UC) is a key optimization problem in power system operation, but its computational burden increases rapidly when a system includes many identical or similar generating units. The aggregation of identical units can reduce this burden by replacing unit-level binary variables with aggregated integer variables. However, this simplification may reduce the accuracy of the obtained schedule because the operation of individual generating units is not fully represented. This paper proposes a tractable unit commitment (TUC) formulation to improve the computational performance of UC while preserving the accuracy of the unit-level solution. The proposed formulation connects the conventional UC model and the aggregated representation through linking constraints that enforce consistency between unit-level and aggregated variables. The model is tested on scaled IEEE 24-bus and IEEE 118-bus systems under different renewable penetration levels and battery energy storage scenarios. The results show that the aggregation-based formulation provides the shortest CPU time, but it underestimates the operating cost in all cases. In contrast, the proposed TUC formulation keeps the cost error almost equal to zero, with a maximum absolute error of 0.0009%, while reducing CPU time compared with UC. Additional comparisons of generation schedules, commitment decisions, and battery charging and discharging profiles show that TUC closely follows the UC benchmark. These results confirm that the proposed TUC formulation provides a better balance between computational efficiency and solution accuracy than the conventional aggregation-based formulation.