Photovoltaic power prediction model based on parallel dendritic neural model

Abstract

Dendritic neural model (DNM) has characteristics of a simple structure and a fast convergence speed. However, when a single DNM is applied to a scene with a large data set, the number of branch layers often needs to be increased, which makes the structure of DNM larger and leads to a poor prediction accuracy. From this perspective, this paper proposes a parallel-structure based DNM with multiple sub-networks, which uses a fuzzy C-means clustering (FCM) algorithm to divide the data set. The FCM algorithm can effectively reduce the amount of data required for the training of each sub-network. Consequently, actual photovoltaic data simulation results verify that the accuracy of the photovoltaic power prediction model can be further improved, and the proposed model is effective and efficiency.