PV Power Prediction Based on ABC-FCM Optimized SVM

Abstract

The safe operation of photovoltaic (PV) grid-connected requires high accuracy of PV power prediction. Although support vector machine (SVM) has advantages in solving the non-linearity of prediction data, it has many problems, such as slow convergence, easy to fall into local optimal solution, dimension disaster, and the selection of penalty factor c and kernel function width σ, which have a deep impact on prediction accuracy. In this paper, the artificial bee colony(ABC) algorithm combined with fuzzy C-means clustering (FCM) algorithm is proposed to optimize SVM for PV power forecasting. FCM is used to calculate the value of the fuzzy membership degree to mark the normality of the samples to generate the fuzzy samples, Optimizing the selection process of SVM parameters using ABC algorithms. The fuzzy samples are input into SVM optimized by ABC algorithm for training, and finally used for PV power prediction. The experimental results show that the resolvable coefficient is close to 1, the root mean square error and relative error of the predicted results decrease, the prediction accuracy of ABC-FCM SVM has been improved, where can be seen through predicted curve and the optimization ability and fitting ability have also been enhanced. It can provide some scientific reference for the development of PV power prediction technology.