.png){kind=icon}
.png){kind=icon}
Return
A Data-Driven Method for Online Constructing Linear Power Flow Model
.png){kind=icon}
PRE
AI
DOI：10.1109/TIA.2023.3287152.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To adapt the data-driven linear power flow (DD-LPF) model for online dispatch in power systems, especially when the size of historical samples or system scale are relatively large, this article proposes an online LPF model construction method. First, the raw measurements are filtered to suppress the adverse impact of measurement noise and outliers. Then, an adaptive moment estimation-based distributed stochastic gradient descent (Adam-DSGD) method is utilized to improve solution efficiency of DD-LPF parameters. Considering the features of online dispatch in power systems, a recursive forgetting factor (RFF) method is further introduced to adapt the LPF model for the present operating condition. Subsequently, the proposed method is applied to an online dispatch-command-tracking problem to demonstrate its practicability. Case studies confirm that the proposed method can achieve 2- to 100-fold less computing time and maintain satisfactory linearization accuracy simultaneously. Moreover, the tracking rate for the optimization objective is also enhanced with applying the proposed method to the real-time optimization problem.
Keywords:
Adaptation models
Pollution measurement
Noise measurement
Computational modeling
Optimization
Mathematical models
Data models
Adaptive moment estimation-based distributed stochastic gradient descent
bad measurements
LPF model
online optimization
recursive forgetting factor
Journal
IF:
4.5
Papers:
1.1W
Citations:
3.5W