Multi-Agent Deep Reinforcement Learning-Based Distributed Voltage Control of Flexible Distribution Networks with Soft Open Points DOI Creative Commons
Liang Zhang, Fan Yang, Dawei Yan

et al.

Energies, Journal Year: 2024, Volume and Issue: 17(21), P. 5244 - 5244

Published: Oct. 22, 2024

The increasing number of distributed generators (DGs) leads to the frequent occurrence voltage violations in distribution networks. soft open point (SOP) can adjust transmission power between feeders, leading evolution traditional networks into flexible (FDN). problem be effectively tackled with control SOPs. However, centralized method for SOP may make it difficult achieve real-time due limitations communication. In this paper, a is proposed FDN SOPs based on multi-agent deep reinforcement learning (MADRL) method. Firstly, framework proposed, which updating algorithm intelligent agent MADRL expounded considering experience sharing. Then, Markov decision process multi-area coordinated where areas are divided electrical distance. Finally, an IEEE 33-node test system and practical Taiwan used verify effectiveness It shows that while ensuring better effect.

Language: Английский

Optimized Coordination of Distributed Energy Resources in Modern Distribution Networks Using a Hybrid Metaheuristic Approach DOI Open Access
Mohammed Alqahtani, Ali S. Alghamdi

Processes, Journal Year: 2025, Volume and Issue: 13(5), P. 1350 - 1350

Published: April 28, 2025

This paper presents a comprehensive optimization framework for modern distribution systems, integrating system reconfiguration (DSR), soft open point (SOP) operation, photovoltaic (PV) allocation, and energy storage (ESS) management to minimize daily active power losses. The proposed approach employs novel hybrid metaheuristic algorithm, the Cheetah-Grey Wolf Optimizer (CGWO), which synergizes global exploration capabilities of Cheetah (CO) with local exploitation strengths Grey Optimization (GWO). model addresses time-varying loads, renewable generation profiles, dynamic network topology while rigorously enforcing operational constraints, including radiality, voltage limits, ESS state-of-charge dynamics, SOP capacity. Simulations on 33-bus demonstrate effectiveness across eight case studies, full DER integration (DSR + PV SOP) achieving 67.2% reduction in losses compared base configuration. By combining CO GWO, CGWO algorithm outperforms traditional techniques (such as PSO GWO) avoids premature convergence preserving computational efficiency—two major drawbacks standalone metaheuristics. Comparative analysis highlights CGWO’s superiority over algorithms, yielding lowest (997.41 kWh), balanced utilization, stable profiles. results underscore transformative potential coordinated enhancing grid efficiency reliability.

Language: Английский

Citations

0
A two‐stage reactive power optimization method for distribution networks based on a hybrid model and data‐driven approach DOI Creative Commons
Ghulam Abbas, Zhi Wu, Aamir Ali

et al.

IET Renewable Power Generation, Journal Year: 2024, Volume and Issue: 18(16), P. 3967 - 3979

Published: Aug. 28, 2024

Abstract The uncertainty of distributed energy resources (DERs) and loads in distribution networks poses challenges for reactive power optimization control timeliness. computational limitations the traditional algorithms development artificial intelligence (AI) based technologies have promoted advancement hybrid model‐data‐driven algorithms. This article proposes a two‐stage method (DNs) on approach. In first stage, topology line parameters DN, as well forecasts renewable outputs, mixed‐integer second‐order cone programming (MISOCP) algorithm is used to on‐load tap changer (OLTC) positions an hourly day‐ahead basis. second leveraging deep learning technology, real‐time output photovoltaics (PV) wind units controlled at 5‐min time scale throughout day. Specifically, using solvers, global optimal PV determined first, corresponding various load scenarios. Then, neural are trained map node outputs units, capturing complex physical relationships. For transformer network framework with self‐attention mechanism multi‐head attention training applied uncover intrinsic spatial relationships among high‐dimensional features. proposed tested modified IEEE 33‐bus system multiple sources. case study results demonstrate that effectively coordinates controls devices, achieving model‐free Compared (DNNs) convolutional (CNNs), provides superior results.

Language: Английский

Citations

2
Research on a Three-Stage Dynamic Reactive Power Optimization Decoupling Strategy for Active Distribution Networks with Carbon Emissions DOI Creative Commons
Yuezhong Wu, Yujie Xiong, Xiaowei Peng

et al.

Energies, Journal Year: 2024, Volume and Issue: 17(11), P. 2774 - 2774

Published: June 5, 2024

The reactive power optimization of an active distribution network can effectively deal with the problem voltage overflows at some nodes caused by integration a high proportion distributed sources into network. Aiming to address limitations in previous studies dynamic using cluster partitioning method, three-stage decoupling strategy for networks considering carbon emissions is proposed this paper. First, emission index based on intensity, and mathematical model established minimum loss, deviation, as satisfaction objective functions. Second, order satisfy requirement all-day motion times discrete devices, around medoids clustering algorithm proposed. Finally, taking improved IEEE33 PG&E69-node systems examples, linear decreasing mutation particle swarm was used solve model. results show that all indicators throughout day are lower than those other methods, which verifies effectiveness algorithm.

Language: Английский

Citations

1
Distributed reactive power optimization of flexible distribution network based on probability scenario-driven DOI

Junxiao Chang,

Junda Zhang, Xiaobing Liao

et al.

Energy Reports, Journal Year: 2024, Volume and Issue: 13, P. 68 - 81

Published: Dec. 8, 2024

Language: Английский

Citations

1
Replacement of the tie-switches on the radial distribution network by optimal sitting and sizing of soft open points DOI
Thuan Thanh Nguyen

Electric Power Systems Research, Journal Year: 2024, Volume and Issue: 241, P. 111336 - 111336

Published: Dec. 10, 2024

Language: Английский

Citations

1
Dual-Layer Voltage and Var Control for Power Distribution Systems DOI
Gaurav Yadav, Yuan Liao, Nicholas P. Jewell

et al.

2021 IEEE Power & Energy Society General Meeting (PESGM), Journal Year: 2024, Volume and Issue: unknown, P. 1 - 5

Published: July 21, 2024

Language: Английский

Citations

0
Multi-Agent Deep Reinforcement Learning-Based Distributed Voltage Control of Flexible Distribution Networks with Soft Open Points DOI Creative Commons
Liang Zhang, Fan Yang, Dawei Yan

et al.

Energies, Journal Year: 2024, Volume and Issue: 17(21), P. 5244 - 5244

Published: Oct. 22, 2024

The increasing number of distributed generators (DGs) leads to the frequent occurrence voltage violations in distribution networks. soft open point (SOP) can adjust transmission power between feeders, leading evolution traditional networks into flexible (FDN). problem be effectively tackled with control SOPs. However, centralized method for SOP may make it difficult achieve real-time due limitations communication. In this paper, a is proposed FDN SOPs based on multi-agent deep reinforcement learning (MADRL) method. Firstly, framework proposed, which updating algorithm intelligent agent MADRL expounded considering experience sharing. Then, Markov decision process multi-area coordinated where areas are divided electrical distance. Finally, an IEEE 33-node test system and practical Taiwan used verify effectiveness It shows that while ensuring better effect.

Language: Английский

Citations

0