A Date-Driven Voltage Control Strategy for Distribution Network Using Deep Reinforcement Learning

A data-driven voltage control strategy is proposed for real-time optimized distributed voltage control of distribution networks connected to the energy storage system (ESS) and distributed photovoltaics (PVs) with high penetration. Deep reinforcement learning (DRL) methods are used in the control of...

Full description

Saved in:
Bibliographic Details
Published in:2023 IEEE 6th International Electrical and Energy Conference (CIEEC) pp. 77 - 82
Main Authors: Shen, Jiawei, Han, Jing, Wang, Yihan, Dong, Yanhao, Li, Hao
Format: Conference Proceeding
Language:English
Published: IEEE 12.05.2023
Subjects:
data-driven method
Deep learning
distributed voltage control
distribution network
Distribution networks
MATD3
Mathematical models
Photovoltaic systems
Real-time systems
Reinforcement learning
Training
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A data-driven voltage control strategy is proposed for real-time optimized distributed voltage control of distribution networks connected to the energy storage system (ESS) and distributed photovoltaics (PVs) with high penetration. Deep reinforcement learning (DRL) methods are used in the control of voltage in the distribution grid. Firstly, based on the division of distributed networks, the decision-making process of combined distributed voltage control is given, which is mainly based on the reactive power compensation of photovoltaic inverters and supplemented by ESS active power regulation. A multi-agent deep reinforcement learning (MADRL) model is then applied to the distributed voltage control model. The Multi-Agent Twin Delay Deep Deterministic Policy Gradient (MATD3) algorithm is then applied. Finally, offline training and online testing are performed on a 33-bus test network with a branch structure to demonstrate the effectiveness of the proposed method.
DOI:10.1109/CIEEC58067.2023.10167323