Development of Machine-Learning Model for Predicting Solar-Power Generation Based on Deterioration Rate

According to the government’s “11th Basic Plan for Electricity Supply and Demand,” 230.8 TWh of renewable-energy generation and 32.9% of power generation are proposed by 2038, and the proportion of renewable-energy generation, such as solar and wind power, is expected to increase continuously. However, the power generated by renewable-energy sources fluctuates significantly; thus, the power-generation volume must be predicted accurately. In this study, we investigate 2013 data from a 2.4kW demonstration complex located in Seongnam, Gyeonggi-do; analyze the power generation from August 2014 to August 2016 using six machine-learning algorithms; and analyze the correlation indicated by a power-generation prediction model based on the degradation rate, where the difference between actual and predicted values is considered. The prediction distribution is compared and analyzed for a highly accurate model, and the prediction result is confirmed to be within 3%.