Abstract:
The objective of this research is to study the roadmap of Photovoltaic (PV) power generation trend using neural network. Over the past twelve years, PV power generation has been used increasingly worldwide. The growth in demand of PV power generation is uncertain and it is also nonlinear. Accordingly, it is totally necessary to find the reasonable forecasting method.
Various factors must be taken into account to precisely forecast of PV power generation. In this research, the key factors of input data are global cumulative PV systems installation during period of year 2000 - 2011, oil prices, PV system components cost, the growth of PV industries and a rapid increase in the population of the whole world. They are important input variables to feed the neural network using a particular type of model, known as a rfeed-forward back-propagation networks. The use of Tan-Sigmoid transfer function in a hidden layer and in an output layer is also included.
This research shows that the value of Mean Absolute Percentage Error (MAPE) is only 3.950 compared to Grey]s forecasting method which is 5.035. Finally, it shows that the value of MAPE is lower than Grey]s forecasting method. This research could be valuable enough for government sector that is concerned with national energy policy, reduce of risk of management and decision to investment.