Predicating photovoltaic power generation using an improved hybrid heuristic method

Due to the rapid growth of energy consumption and environmental issue, the use of photovoltaic (PV) power, has attracted many attentions since such energy is renewable and clean. This study proposes a simple yet effective method for PV power prediction in which a dividing-stage strategy is investigated and a hybrid heuristic, denoted as GAPSO (genetic algorithm and particle swarm optimization) is applied to optimize the BP (Back-Propagation) neural network. In comparison with GABP neural network, the hybrid GAPSOBP neural network achieves a higher accuracy for PV power prediction. Moreover, by incorporating the dividing-stage strategy into GAPSOBP neural network, the prediction accuracy is improved even further.

I. Introduction

In recent years, photovoltaic (PV) becomes a popular source of renewable energy because of its easy-accessibility and environmental friendliness [1]. A PV system typically consists of solar cell array, controller, inverter and storage battery. It is believed that the PV power generation can be widely used in various fields in the near future. However, the output of PV power generation is often criticized as intermittent and instable, because of many meteorological factors, such as the radiation intensity and the external temperature [2]. Also, a large number of PV systems may increase the complexity of the power system, and make the whole system easy to be collapsed. However, by an accurate PV power prediction, an effective power scheduling plan can be made, and thus, improving the stability of PV system. This study is then focus on introducing a simple yet effective PV power prediction method.