SAGA-FCM-LSSVM model-based short-term power forecasting of photovoltaic power plants

2018 International Conference on New Energy and Future Energy System

Peng, Z.N.^1,2 ; Lin, P.J.^1,2 ; Lai, Y.F.^1,2 ; Chen, Z.C.^1,2 ; Wu, L.J.^1,2 ; Cheng, S.Y.^1,2

College of Physics and Information Engineering, And Institute of Micro-Nano Devices and Solar Cells, Fuzhou University, Fuzhou

350116, China^1

Jiangsu Collaborative Innovation Center of Photovoltaic Science and Engineering, Changzhou

213164, China^2

关键词: Back-propagation neural networks;    Fuzzy C means clustering;    Initial clustering centers;    Least Square Support Vector Machine (LS-SVM);    Non-linear relationships;    Photovoltaic power plant;    Prediction performance;    PV power generation systems;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/188/1/012079/pdf
DOI  :  10.1088/1755-1315/188/1/012079

来源: IOP

【 摘 要 】

With the rapid development of social economy, vigorously developing solar energy has become a powerful means to solve energy and environmental problems. However, the instability of weather condition makes the output of PV power have strong randomness, fluctuations and intermittence. Thus accurate photovoltaic (PV) power forecast eliminates the negative impacts of the grid connection of PV power generation systems, which are very meaningful for effectively integrating the PV power systems into the grid. The paper presents a Simulate Anneal and Genetic Algorithm (SAGA), fuzzy c-means clustering (FCM) and least square support vector machine (LSSVM) (SAGA-FCM-LSSVM) model-based power short-term forecasting of PV power plants approach. The experimental effect of the proposed prediction method is verified by employing large datasets from the Desert Knowledge Australia Solar Center (DKASC) website. In this work, the FCM clustering algorithm is adopted to cluster the historical power datasets, and the LSSVM technique maps the multivariate meteorological factors and power data nonlinear relationship. The SAGA method is applied to improve the initial clustering centers of the FCM clustering algorithm to obtain a higher prediction performance. The prediction result of the method in this paper is contrasted with back propagation neural network (BPNN) and LSSVM models, and reveals excellent effect in improving the accuracy of prediction.