Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Short-Term Photovoltaic Forecasting Model for Qualifying Uncertainty during Hazy Weather

Xuan Yang, Yunxuan Dong, Lina Yang, Thomas Wu

TL;DR

The paper tackles the challenge of short-term PV forecasting under hazy weather, where uncertainty impedes accuracy. It introduces CCRetNet, a hybrid framework that combines Tsallis Entropy by Weighted Permutation Pattern (TEWPP) for uncertainty quantification, median-linkage hierarchical clustering to reduce computation, a CNN front-end with a Retention Network (RetNet) forecaster, and NSGA-II for robust hyperparameter optimization. Key contributions include TEWPP-based clustering, a novel distance/linkage approach for clustering, a multi-form RetNet forecasting architecture, and demonstrated robustness and accuracy gains on Jiangsu 2015 and Beijing 2019 datasets, outperforming several baselines. The approach yields more reliable short-term PV forecasts during hazy conditions, supporting greener, more stable power systems and enabling better grid integration of solar energy.

Abstract

Solar energy is one of the most promising renewable energy resources. Forecasting photovoltaic power generation is an important way to increase photovoltaic penetration. However, the difficulty in qualifying the uncertainty of PV power generation, especially during hazy weather, makes forecasting challenging. This paper proposes a novel model to address the issue. We introduce a modified entropy to qualify uncertainty during hazy weather while clustering and attention mechanisms are employed to reduce computational costs and enhance forecasting accuracy, respectively. Hyperparameters were adjusted using an optimization algorithm. Experiments on two datasets related to hazy weather demonstrate that our model significantly improves forecasting accuracy compared to existing models.

Short-Term Photovoltaic Forecasting Model for Qualifying Uncertainty during Hazy Weather

TL;DR

The paper tackles the challenge of short-term PV forecasting under hazy weather, where uncertainty impedes accuracy. It introduces CCRetNet, a hybrid framework that combines Tsallis Entropy by Weighted Permutation Pattern (TEWPP) for uncertainty quantification, median-linkage hierarchical clustering to reduce computation, a CNN front-end with a Retention Network (RetNet) forecaster, and NSGA-II for robust hyperparameter optimization. Key contributions include TEWPP-based clustering, a novel distance/linkage approach for clustering, a multi-form RetNet forecasting architecture, and demonstrated robustness and accuracy gains on Jiangsu 2015 and Beijing 2019 datasets, outperforming several baselines. The approach yields more reliable short-term PV forecasts during hazy conditions, supporting greener, more stable power systems and enabling better grid integration of solar energy.

Abstract

Solar energy is one of the most promising renewable energy resources. Forecasting photovoltaic power generation is an important way to increase photovoltaic penetration. However, the difficulty in qualifying the uncertainty of PV power generation, especially during hazy weather, makes forecasting challenging. This paper proposes a novel model to address the issue. We introduce a modified entropy to qualify uncertainty during hazy weather while clustering and attention mechanisms are employed to reduce computational costs and enhance forecasting accuracy, respectively. Hyperparameters were adjusted using an optimization algorithm. Experiments on two datasets related to hazy weather demonstrate that our model significantly improves forecasting accuracy compared to existing models.
Paper Structure (29 sections, 30 equations, 11 figures, 5 tables, 2 algorithms)

This paper contains 29 sections, 30 equations, 11 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related works
  3. Problem definition
  4. Proposed model
  5. Model architecture
  6. Clustering method
  7. The tsallis entropy by weighted permutation pattern
  8. Hierarchical clustering
  9. CNN-RetNet
  10. CNN
  11. Retention
  12. The parallel representation
  13. The recurrent representation of retention
  14. The chunkwise recurrent representation of retention
  15. Gated multi-scale retention
  16. ...and 14 more sections

Figures (11)

  • Figure 1: Global capacity of solar energy from 2014 to 2023.
  • Figure 2: Framework of proposed model
  • Figure 3: Three forms of retention
  • Figure 4: Analysis of Air Quality and PM2.5 Concentrations in Jiangsu Province (2015) and Beijing (2019)
  • Figure 5: Comparison of PV power generation forecasting on different models in a specific region of Jiangsu Province, January 2015
  • ...and 6 more figures