DefinePK

Abstract

This paper develops and assesses an area-specific prediction model for solar energy generation according to climate parameters on Jeju City, taking advantage of artificial neural networks. The hour-by-hour climate information, open to public in Rep. of Korea, are retrieved from the national weather administration, while the log file from a solar panel monitoring system is interpreted and integratively processed in the MySQL database. The neural network takes temperature, humidity, insolation, and sunlight duration as input, and the amount of solar energy generation is defined for output. Currently, 250 daily records are used for training, with 42 taken for evaluating. The developed model catches the day-by-day change of the power generation and its root mean square error is 16.4 kwh. After all, this model will estimate next day power generation based on the weather forecast, making it possible to build a power generation plan for base energy facilities or an electricity purchase plan from the main power system.