신재생에너지모니터링시스템(REMS) 발전량 예측을 위한 클러스터링 방법론 연구 (A Study on Clustering Method for Solar Power Generation Forecasting in Renewable Energy Monitoring System)

서지정보

· 발행기관 : 한국태양에너지학회
· 수록지 정보 : 한국태양에너지학회 논문집 / 44권 / 6호 / 1 ~ 12페이지
· 저자명 : 이영섭, 고아름, 김종규, 유영선, 이건우

초록

A Renewable Energy Monitoring System (REMS) was launched to inform facility owners and local governments of failures. However, as the scale of renewable energy deployment increases and its impact on the national grid grows, it is becoming important to forecast power generation for not only commercial solar facilities but also self-consumption solar facilities. As of April 2024, more than 130,000 solar installations have been connected to REMS. It is not economically efficient to forecast the generation of these small installations individually, and the level of accuracy required for self-consuming solar facilities is not as high. Therefore, suitable monitoring techniques are required. In this study, Chungcheongbuk-do, situated midway in the ranking of REMS-connected municipal solar facilities, was selected as the priority target. The objective is to group 11,268 power plants in the Chungcheongbuk-do region and select a representative power plant for each group. The clustering algorithm employs various factors, including latitude, longitude, insolation, altitude, and power generation history, to group power plants. It utilizes both K-means and Autoencoder techniques. First, the location of each solar facility was converted into latitude and longitude values using the address information of the solar facilities connected to REMS. The altitude and insolation information for each location were then extracted based on the latitude and longitude values for clustering analysis. Second, a methodology is proposed using power utilization data from January to December 2023 to identify normally operating facilities and cluster power generation by targeting them.

영어초록

A Renewable Energy Monitoring System (REMS) was launched to inform facility owners and local governments of failures. However, as the scale of renewable energy deployment increases and its impact on the national grid grows, it is becoming important to forecast power generation for not only commercial solar facilities but also self-consumption solar facilities. As of April 2024, more than 130,000 solar installations have been connected to REMS. It is not economically efficient to forecast the generation of these small installations individually, and the level of accuracy required for self-consuming solar facilities is not as high. Therefore, suitable monitoring techniques are required. In this study, Chungcheongbuk-do, situated midway in the ranking of REMS-connected municipal solar facilities, was selected as the priority target. The objective is to group 11,268 power plants in the Chungcheongbuk-do region and select a representative power plant for each group. The clustering algorithm employs various factors, including latitude, longitude, insolation, altitude, and power generation history, to group power plants. It utilizes both K-means and Autoencoder techniques. First, the location of each solar facility was converted into latitude and longitude values using the address information of the solar facilities connected to REMS. The altitude and insolation information for each location were then extracted based on the latitude and longitude values for clustering analysis. Second, a methodology is proposed using power utilization data from January to December 2023 to identify normally operating facilities and cluster power generation by targeting them.

참고자료

· 없음