Forecasting photovoltaic power using bagging feed-forward neural network

Abstract

ABSTRACT This paper presents a forecast model of the active power of a photovoltaic (PV) power generation system. In this model, a feed-forward neural network (FNN) is combined with bootstrap aggregation techniques using the Box–Cox transformation, seasonal and trend decomposition using Loess, and a moving block bootstrap (MBB) technique. An analysis is conducted using the data provided by the active power of the PV power generation system; the data are collected every 30 min for 12 months. The FNN method combined with MBB techniques consistently outperformed the original FNN in terms of forecasting accuracy based on the root mean squared error, on the forecast from one day of anticipation. The results are statistically significant as demonstrated through the Ljung–Box test, which verifies that the forecast errors are not correlated, thereby validating the proposed model.