.png){kind=icon}
.png){kind=icon}
Return
Sample-Wise Graph-Based Multivariate Short-Term PV Power Forecasting
.png){kind=icon}
PRE
AI
DOI：10.1109/TSTE.2025.3576928.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Reliable short-term photovoltaic (PV) power forecasting is of crucial significance for the rational dispatching of power sources and the effective control of operating costs for the power grid. However, temporal misalignment and regression accuracy imbalance of PV power data pose significant challenges to the reliability of forecast results. In this study, multivariate PV power forecasting is investigated from the perspective of forecast model samples. Firstly, the extent of misalignment of a sample is parameterized by a time-delay vector. Subsequently, the sample-wise graph is defined to relate the time-delay vector with PV power data. Then, the time-delay vector is estimated by minimizing the smoothness metric of the sample-wise graph. Finally, a sample-wise graph-based sample weighting strategy is introduced to address the issue of regression accuracy imbalance. The efficiency of the proposed PV power forecasting scheme is validated through extensive experiments on real-world datasets. Comparison experiments suggest that the proposed scheme can achieve remarkably improved short-term PV power forecasting.
Keywords:
Forecasting
Accuracy
Power generation
Predictive models
Vectors
Training
Learning systems
Probabilistic logic
Power measurement
Photovoltaic systems
Sample-wise graph
sample temporal misalignment
regression accuracy imbalance
time-delay estimation
PV power forecasting
Journal
I
IF:
10
Papers:
210
Citations:
0