Dual-layered deep learning and optimization algorithm for electric vehicles charging infrastructure planning

•A two-stage deep learning method based on ARIMA and LSTM for the optimal prediction of the EV demands.•A novel multi-objective HAG technique for the optimal placement of EVCSs within the distribution network.•Incorporating total PL and AVDI as objectives within the multi-objective HAG technique. Th...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:International journal of electrical power & energy systems Jg. 166; S. 110545
Hauptverfasser: Sedhom, Bishoy E., Eladl, Abdelfattah A., Siano, Pierluigi, El-Afifi, Magda I.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.05.2025
Elsevier
Schlagworte:
Charging station
Deep learning
Electric vehicle
Hybrid HAG
Load forecasting
Optimal EVs placement
Deep learning
Electric vehicle
Load forecasting
Optimal EVs placement
Hybrid HAG
Charging station
ISSN:0142-0615
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:•A two-stage deep learning method based on ARIMA and LSTM for the optimal prediction of the EV demands.•A novel multi-objective HAG technique for the optimal placement of EVCSs within the distribution network.•Incorporating total PL and AVDI as objectives within the multi-objective HAG technique. This study aims to develop a robust optimization framework for accurately forecasting electric vehicle (EV) loads and strategically placing EV charging stations (EVCSs) within distribution networks to enhance grid stability and efficiency. EVs offer substantial environmental benefits but pose challenges to power grids due to capacity constraints. To address these issues, a dual-layered deep learning method, combining Autoregressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM), is employed for precise EV load prediction. Additionally, a Hybrid Archimedes-Genetic (HAG) algorithm, integrating the Archimedes Optimization Algorithm (AOA) and Genetic Algorithm (GA), is proposed to optimize EVCS locations. This multi-objective approach aims to minimize active and reactive power losses while reducing the average voltage deviation index (AVDI). Two case studies were conducted using MATLAB and Python: the first predicted EV power demands, while the second identified optimal EVCS locations in IEEE 33-node and 69-node test systems. Results demonstrate that the HAG method outperforms GA, Particle Swarm Optimization, and AOA individually. For the 33-bus system, active and reactive power losses decreased by 33 % and 33.01 %, respectively, with an AVDI of 0.004. In the 69-bus system, losses were reduced by 19.38 % and 16.76 %, with an AVDI of 0.0014. These findings highlight the HAG method’s effectiveness in optimizing EVCS placement, improving grid performance, and maximizing the benefits of EV adoption.
ISSN:0142-0615
DOI:10.1016/j.ijepes.2025.110545