Climate change is intensifying extreme temperature events, especially heatwaves, which affect both energy demand and the performance of photovoltaic systems in the metropolitan region of São Paulo, particularly in the municipality of Santo André. The central problem lies in the difficulty of accurately estimating solar generation during these extreme events, as photovoltaic panels exhibit complex behavior. Although heatwaves are associated with higher solar radiation due to reduced cloud cover, high temperatures significantly reduce module efficiency, potentially causing losses of up to half a percent for each degree Celsius above standard conditions. This study presents a machine-learning-based methodology using the Extreme Gradient Boosting (XGBoost) algorithm to estimate photovoltaic generation while explicitly considering the effects of heatwaves and future climate projections. Using historical data from the Brazilian DailyWeather Gridded Data dataset, combined with extreme temperature event indices calculated with Climpact software and projections from four global climate models under the SSP2-4.5 and SSP5-8.5 scenarios, the developed model estimated photovoltaic energy generation. During the validation, the XGBoost model achieved a coefficient of determination greater than 0.996. The distinguishing contribution lies in the explicit integration of heatwave indices as predictive variables and systematic analysis of future climate scenarios up to 2100. Results from the ensemble of four climate models indicate that annual mean generation decreases by 0.33% under SSP2-4.5, and 0.37% under SSP5-8.5 over 2021–2100, with monthly reductions reaching up to 3.92% under the high-emission scenario, providing quantitative tools for energy planning in systems with high solar radiation incidence.
1 January 2025 in International Journal of Engineering
The steady increase in energy demand results in under-voltage problems and an increase in the active power losses in electrical distribution networks. The optimal placement of capacitor banks in distribution systems is essential for enhancing the voltage profile and reducing power losses. This paper proposes a modified particle swarm optimization (PSO) algorithm using an adaptive inertia constant to expand the search space and determine the optimal locations and sizes of capacitor banks. Two loss sensitivity indices (LSI) were applied to identify candidate buses. This methodology was applied to the IEEE 33-bus radial distribution system considering two scenarios: one with fixed capacitors (Case 1) and the other with switched capacitors (Case 2). The results demonstrate that the proposed algorithm effectively reduces system losses by 31.30% and 31.38% in Cases 1 and 2, respectively. In addition, in both cases, the voltage profiles were improved and maintained within allowable limits. Therefore, the proposed methodology is expected to work well in larger electrical distribution networks.
The increasing growth in electricity demand leads to higher voltage drops and power losses in distribution networks. The integration of capacitor banks (CBs) and photovoltaic generators (PVGs) helps improve network performance; however, their effectiveness depends on their optimal sizing and placement within the electrical network. This study proposes a hybrid metaheuristic algorithm to determine the optimal locations and sizes of CBs and PVGs in distribution networks under hourly AC power flow, hourly dynamic switching of capacitor banks, and an economic model with tariffs differentiated by consumer type. The proposed PSO–WOA algorithm combines the fast convergence capability of Particle Swarm Optimization (PSO) with the local refinement capability of the Whale Optimization Algorithm (WOA). In addition, to guide the initialization of the population, loss sensitivity indices based on the derivatives of active power losses with respect to voltage and reactive power are employed. The hybrid algorithm begins with the PSO phase to intensively exploit the search space and subsequently transfers the population without reinitialization to the WOA phase to refine promising solutions and improve the final convergence. The application of the proposed PSO–WOA algorithm to the IEEE 33-bus distribution system achieved a 51.92% reduction in active power losses, increased the minimum voltage from 0.9163 p.u. to 0.9500 p.u., and reduced the total annual cost by 50.05% (from 2.573 to 1.285 MUSD/year). A comparison with similar optimization algorithms, namely PSO, WOA, the reverse hybrid WOA–PSO, and the Genetic Algorithm (GA), showed that the proposed PSO–WOA achieved the best overall performance in terms of feasibility and statistical stability, with a mean solution quality comparable to that of PSO but with lower dispersion and a better worst case across independent runs. GA and WOA showed the greatest difficulty in satisfying the operational constraints, achieving only 60% and 79% feasible solutions, respectively, whereas the proposed algorithm achieved a 100% feasibility rate.