Contribution à l’optimisation du fonctionnement des réseaux de distribution en présence de génération distribuée

Abstract

this thesis proposes to develop an appropriate time-varying probability load-generation model based on Weibull and Beta probability density functions (PDFs) to estimate the stochastic power output from WTs and SPV arrays, respectively. This model is based on hourly seasonal data, including wind speed, solar irradiance, and ambient temperature, collected over a specified time frame and location. An improved Frilled Lizard Optimization (IFLO) algorithm is developed and proposed for strategic RES planning, aiming to minimize total power losses, improve voltage profiles, and enhance voltage stability while adhering to operational constraints. The IFLO algorithm incorporates three advanced strategies: fitness distance balance, quasi-opposite-based learning, and Cauchy mutation, to strengthen its search capabilities and prevent convergence to local optima traps. The proposed method effectively determines the optimal locations, rated capacities of SPV strings and WTs, and the power factor of WTs. Its performance is validated through simulations on the IEEE 69-bus medium-scale and 85-bus large-scale distribution grids. Simulation results decisively demonstrate that optimal RES allocation significantly improves system performance. Furthermore, the suggested technique outperforms other recent and effective optimization algorithms, including the grey wolf optimizer (GWO), jellyfish search optimizer (JSO), black-winged kite algorithm (BKA), and the original frilled lizard optimization (FLO), in solving the optimal planning problem of RES integration under both deterministic and probabilistic scenarios.