Open AccessAn intelligent power flow adjustment method based on reinforcement learning
Author(s) -
Li-Jyun Ye,
Zhongming Xiang,
Yifei Yang,
Yuan Zhu,
C. Zhang,
Haifeng Yao
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.179
H-Index - 26
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/551/1/012009
Subject(s) - computer science , grid , boundary (topology) , power (physics) , convergence (economics) , electric power system , reliability engineering , scale (ratio) , power flow , plan (archaeology) , power flow study , reinforcement learning , industrial engineering , control engineering , distributed computing , engineering , artificial intelligence , mathematics , mathematical analysis , physics , geometry , archaeology , quantum mechanics , economics , history , economic growth
In order to examine the future power grid operation, the operators usually build future power system model files first according to boundary conditions such as infrastructure plans, maintenance plans, power generation plans, load forecasts, and power plan requirements. As these plans are usually prepared independently, and the scale of the power grid is getting larger and larger, the newly built grid model is usually unsolved, ill, or does not meet the actual grid operation requirements. In this situation, the operators normally have to readjust the units power many times, according to their own experience, to obtain a qualified power flow result. A method for power flow adjustments is proposed in this paper. The Q-Learning (QL) method is used to adjust the power flow from nonconvergence to convergence, and then some direction guidance for subsequent adjustments is provided through certain safety and economic indicators of the power grid (such as network loss, output of the balancing machine, etc.). Finally, two examples of IEEE39 nodes system and Zhejiang power grid are presented to verify the effectiveness and practicability of the proposed method.