Page 498 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 3: Intelligent Transmission Section Flow Control Agent

in Power Systems

Organization: China Southern Power Grid Artificial Intelligence Technology Co., Ltd.

Country: China

Contact Person(s):

Zhen Dai, zhendai07@ foxmail .com or daizhen@ csg .cn

1 Use Case Summary Table

Item Details
Category Intelligent manufacturing
Problem Current dispatcher's empirical adjustments introduce subjectivity, limit
Addressed scenario complexity, prolong response times, and hinder multi-objective
optimization. Growing renewable uncertainties exacerbate these challenges
in accessing elevated or enclosed areas.
Key Aspects of This case applies deep reinforcement learning to enhance power system
Solution dispatching, focusing on reducing transmission section flows while maintain-
ing safety constraints.

Technology Reinforcement learning, Multi-Layer Perceptron (MLP)'s high-dimensional
state modelling, Deep Deterministic Policy Gradient (DDPG)'s continuous
control
Keywords Transmission Section Flow, reinforcement learning

Data Availability Data is private

Metadata (Type Measurements in power system operations (e.g., switch states, active power,
of Data) voltage magnitude, etc) and system parameters (e.g., generation and trans-
mission limits).
Model Training The integration of MLP's high-dimensional state modelling with DDPG's
and Fine-Tuning continuous control enables adaptive section regulation through dynamic
pattern learning while preserving stability constraints.

Testbeds or Deployed in Southwest China's renewable-rich regional grid, the intelligent
Pilot Deploy- agent dynamically adjusts generation outputs in response to real-time fluc-
ments tuations, consistently outperforming human dispatchers in comprehensive
control evaluations.

Code reposito- Not Available
ries

462