Improved prediction method of PV output power based on optimised chaotic phase space reconstruction
Yufei Wang, Yuchao Fu, Hua Xue
Abstract
With a large number of photovoltaic (PV) generation systems connected to power grid, accurate forecasting becomes important, the results can be used to alleviate their impacts on the grid effectively. However, most of existing methods strongly rely on the numerical weather prediction (NWP), accuracy of them under highly volatile weather conditions is poor. In this study, without using meteorological data, an improved prediction method based on optimised chaotic phase space reconstruction is presented. Firstly, chaos theory is introduced to analyse the evolution law of PV output power. Then, in order to decrease the delay effect of original chaotic attractor and the negative effect of each phase space power sequence's fluctuation on forecasting accuracy, the algorithm of ensemble empirical mode decomposition (EEMD) is introduced to perform further analysis, so as to raise the regularity of chaotic attractors and extract the partial fluctuation features. Finally, based on optimized chaotic attractor and genetic algorithm‐back propagation (GA‐BP) neural network, the authors build a combined prediction model and apply it into actual measurement data to verify its validation. Numerical results show that by carrying out the optimised chaotic phase space reconstruction, proposed prediction approach achieves better accuracy than the chaos‐GA‐BP, EEMD‐GA‐BP and NWP‐GA‐BP methods.