An Efficient Algorithm for Power Prediction in PV Generation System
Aiming at the existing photovoltaic power generation prediction methods, the modeling is complicated, the prediction accuracy is low, and it is difficult to meet the actual needs. Based on the improvement of the traditional wavelet neural network, a dual-mode cuckoo search wavelet neural network alg...
Saved in:
| Main Author: | |
|---|---|
| Format: | EJournal Article |
| Published: |
Center of Biomass & Renewable Energy, Diponegoro University,
2020-07-15.
|
| Subjects: | |
| Online Access: | Get Fulltext |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Aiming at the existing photovoltaic power generation prediction methods, the modeling is complicated, the prediction accuracy is low, and it is difficult to meet the actual needs. Based on the improvement of the traditional wavelet neural network, a dual-mode cuckoo search wavelet neural network algorithm combined prediction method is proposed, which takes into account the extraction of chaotic features of surface solar radiation and photovoltaic output power. The proposed algorithm first reconstructs the chaotic phase space of the hidden information of each influencing factor in the data history of PV generation and according to the correlation analysis, the solar radiation is utilized as additional input. Next, the proposed algorithm overcomes the limitations of the cuckoo search algorithm such as the sensitivity to the initial value and searchability and convergence speed by dual-mode cuckoo search wavelet neural network algorithm. Lastly, a prediction model of the proposed algorithm is proposed and the prediction analysis is performed under different weather conditions. Simulation results show that the proposed algorithm shows better performance than the existing algorithms under different weather conditions. Under various weather conditions, the mean values of TIC, EMAE and ENRMSE error indicators of the proposed forecasting algorithm were reduced by 43.70%, 45.75%, and 45.41%, respectively. Compared with the Chaos-WNN prediction method, the prediction performance has been further improved under various weather conditions and the mean values of TIC, EMAE and ENRMSE error indicators have been reduced by 25.55%, 27.26%, and 36.83%, respectively. |
|---|---|
| Item Description: | https://ejournal.undip.ac.id/index.php/ijred/article/view/28076 |