| Title |
A Study on the Design of Testable CAM using MTA Code |
| Authors |
김일주(Il-Ju Kim) ; 이송근(Song-Keun Lee) |
| DOI |
https://doi.org/10.5370/KIEEP.2019.68.2.106 |
| Keywords |
Python ; CNN(Convolutional Neural Network) ; Deep Learning Technique ; Solar power generation |
| Abstract |
Recently, due to the depletion of fossil fuel resources and the regulation of CO2 emissions, attention and demand for solar power generation are increasing. The solar power generation has begun to be applied to small-scale power generation, and recently, large- scale power plants have been built and sold to utility companies or power demand has been supplied to cities. In addition, facilities for microgrid for energy self-reliance through renewable energy such as solar power, wind power, and tidal power are being constructed and studied in places where electricity supply is difficult, such as island or inland countryside. The solar power generation is advantageous in that it has no pollution, is easy to maintain, and has a long life. However, the power generation depends on the weather, the installation site is limited, and the initial investment cost and the power generation cost are high. Particularly, since it depends on the weather condition, the generation amount is very intermittent and it is difficult to adjust the power generation amount, and it is difficult to establish a power generation plan in advance. Therefore, high accuracy solar power generation forecasting is essential to reduce the uncertainty of solar power generation and to improve the economical efficiency of solar power generation. Existing solar power generation forecasting has been conducted to predict power generation using Extreme Learning Machine (ELM), Support Vector Regression (SVR), and neural network. However, in this paper, we forecast the solar power generation using the meteorological satellite data from 2011 to 2017 and forecasting method using the CNN (Convolutional Neural Network) which is one of the deep running algorithms. And the feasibility of the proposed method was verified using the meteorological satellite data in 2018. For the forecasting program, we programmed using Python, which is specialized for deep running. |