A gradient-based optimizer (GBO) is a recently inspired meta-heuristic technique centered on Newton’s gradient-based approach. In this paper, an advanced developed version of the GBO is merged with a crossover operator (GBOC) to enhance the diversity of the created solutions. The merged crossover operator causes the solutions in the next generation to be more random. The proposed GBOC maintains the original Gradient Search Rule (GSR) and Local Escaping Operator (LEO). The GSR directs the search to potential areas and aids in its convergence to the optimal answer, while the LEO aids the searching process in avoiding local optima. The proposed GBOC technique is employed to optimally place and size the distribution static VAR compensator (D-SVC), one of the distribution flexible AC transmission devices (D-FACTS). It is developed to maximize the yearly energy savings via power losses concerning simultaneously different levels of the peak, average, and light loadings. Its relevance is tested on three distribution systems of IEEE 33, 69, and 118 nodes. Based on the proposed GBOC, the outputs of the D-SVCs are optimally varying with the loading level. Furthermore, their installed ratings are handled as an additional constraint relating to two compensation levels of 50% and 75% of the total reactive power load to reflect a financial installation limit. The simulation applications of the proposed GBOC declare great economic savings in yearly energy losses for the three distribution systems with increasing compensation levels and iterations compared to the initial case. In addition, the effectiveness of the proposed GBOC is demonstrated compared to several techniques, such as the original GBO, the salp swarm algorithm, the dwarf mongoose algorithm, differential evolution, and honey badger optimization.

Economic dispatch in combined heat and power (CHP) schemes takes into account the trade-off between generating electricity and producing heat, ensuring that the system operates at the highest level of efficiency and cost-effectiveness. A Mantis Search Algorithm (MSA), in this paper, is developed to address the CHP economic dispatch taking into consideration valve points effect, feasible region constraints for CHP units, and power losses are considered all at once. The MSA draws its inspiration from the unique sexual cannibalism and foraging strategies of praying mantises including three stages of optimization which are hunting, attacking, and cannibalism. Moreover, various recent meta-heuristic optimization algorithms, such as the Aquila Optimizer (AO), Gradient-based Optimizer (GBO), Coot Optimizer (CO), and African Vultures Algorithm (AVO), are utilized for comparative analysis across the two test systems. The developed MSA is demonstrated to be superior across three power and heat loading levels for both a 7-unit system and a larger 84-unit system. In the case of the large 84-unit system, MSA exhibits a significant cost reduction, surpassing AVO, GBO, CO, and AO by 2.017%, 2.511%, 0.817%, and 0.126%, respectively. Simulation results confirm the remarkable efficiency of MSA as a promising solution for economic dispatch in combined heat and power schemes. Testing outcomes indicate that MSA outperforms several previously published methods in the literature, as well as the four adopted optimizers.

This work presents an optimal methodology based on an augmented, improved, subtraction-average-based technique (ASABT) which is developed to minimize the energy-dissipated losses that occur during electrical power supply. It includes a way of collaborative learning that utilizes the most effective response with the goal of improving the ability to search. Two different scenarios are investigated. First, the suggested ASABT is used considering the shunt capacitors only to minimize the power losses. Second, simultaneous placement and sizing of both PV units and capacitors are handled. Applications of the suggested ASAB methodology are performed on two distribution systems. First, a practical Egyptian distribution system is considered. The results of the simulation show that the suggested ASABT has a significant 56.4% decrease in power losses over the original scenario using the capacitors only. By incorporating PV units in addition to the capacitors, the energy losses are reduced from 26,227.31 to 10,554 kW/day with a high reduction of 59.75% and 4.26% compared to the initial case and the SABT alone, respectively. Also, the emissions produced from the substation are greatly reduced from 110,823.88 kgCO₂ to 79,189 kgCO₂, with a reduction of 28.54% compared to the initial case. Second, the standard IEEE 69-node system is added to the application. Comparable results indicate that ASABT significantly reduces power losses (5.61%) as compared to SABT and enhances the minimum voltage (2.38%) with a substantial reduction in energy losses (64.07%) compared to the initial case. For both investigated systems, the proposed ASABT outcomes are compared with the Coati optimization algorithm, the Osprey optimization algorithm (OOA), the dragonfly algorithm (DA), and SABT methods; the proposed ASABT shows superior outcomes, especially in the standard deviation of the obtained losses.

This article presents a modified intelligent metaheuristic form of the Dwarf Mongoose Optimizer (MDMO) for optimal modeling and parameter extraction of solar photovoltaic (SPV) systems. The foraging manner of the dwarf mongoose animals (DMAs) motivated the DMO’s primary design. It makes use of distinct DMA societal groups, including the alpha category, scouts, and babysitters. The alpha female initiates foraging and chooses the foraging path, bedding places, and distance travelled for the group. The newly presented MDMO has an extra alpha-directed knowledge-gaining strategy to increase searching expertise, and its modifying approach has been led to some extent by the amended alpha. For two diverse SPV modules, Kyocera KC200GT and R.T.C. France SPV modules, the proposed MDMO is used as opposed to the DMO to efficiently estimate SPV characteristics. By employing the MDMO technique, the simulation results improve the electrical characteristics of SPV systems. The minimization of the root mean square error value (RMSE) has been used to compare the efficiency of the proposed algorithm and other reported methods. Based on that, the proposed MDMO outperforms the standard DMO. In terms of average efficiency, the MDMO outperforms the standard DMO approach for the KC200GT module by 91.7%, 84.63%, and 75.7% for the single-, double-, and triple-diode versions, respectively. The employed MDMO technique for the R.T.C France SPV system has success rates of 100%, 96.67%, and 66.67%, while the DMO’s success rates are 6.67%, 10%, and 0% for the single-, double-, and triple-diode models, respectively.

This paper proposes a novel innovative version of enhanced artificial rabbit optimization (EARO) algorithm integrating an equilibrium pool (EP) that consists of the best solutions. Furthermore, the detour foraging and hiding mechanisms are modified to amplify the search capability. These modifications enable the algorithm to dynamically focus on exploring various randomized directions emanating from the EP. The proposed EARO is designed to investigate the PV module characteristics identification issue. To obtain the nine parameters of the PV triple diode model (TDM) while taking into account three distinct real‐world PV modules, the novel version of EARO is utilized and evaluated in comparison with to the standard ARO. The novel EARO is tested on three different PV modules: the Ultra 85‐P PV panel, the PVM_752GaAs, and the RTC France. The results corresponding to the novel EARO are compared with respect to several published latest studies. The results of the simulation show that the proposed EARO shows significant overall improvement rates for each of the three modules. A validation of the novel EARO on the common SDM and DDM of the RTC France PV is assessed which illustrates a significant superiority and robustness of the novel EARO over recent published results.

Accurate load forecasting for Electric Vehicle Charging Stations (EVCS) is critical for optimizing energy management and ensuring grid stability amid growing electric vehicle adoption. This study investigates short-term, hourly load forecasting at the station level using a hybrid XGBoost–BiLSTM stacking model (Hybrid 3) with an XGBoost meta-learner. From the Adaptive Charging Network (ACN–Caltech) dataset (April 25, 2018–September 13, 2021), 31,424 raw charging sessions were preprocessed, yielding 14,496 cleaned sessions for modeling. These were split into 80% training and 20% testing sets using a fixed random seed (42) for reproducibility. Hybrid 3 was benchmarked against 24 alternative models spanning statistical (e.g., Persistence, SARIMAX), machine learning (e.g., XGBoost, LightGBM), deep learning (e.g., BiLSTM, CNN), and ensemble methods. On cleaned data, Hybrid 3 achieved an MAE of 2.6870 kWh and R ² = 0.6395—a 3.4% improvement over standalone BiLSTM—while slightly underperforming the top Boosting ensemble (XGBoost + BiLSTM + LightGBM). Robustness was confirmed via five-fold walk-forward validation (mean MAE = 2.5351 kWh, SD = 1.2885). Cross-site evaluation on an independent synthetic dataset (n = 1,965,239 sessions) showed reduced generalization, highlighting site-specific temporal patterns. One-Way ANOVA (p = 0.2073, η ² = 0.2062) indicated no statistically significant but practically relevant differences among top models. Feature importance analysis identified log-transformed charging duration (importance = 0.376) as the dominant predictor, aligning with real-world EV behavior. Overall, Hybrid 3 balances accuracy and complexity effectively, though gradient-boosting ensembles remain preferable for scalable, real-time EVCS forecasting.