TY - GEN
T1 - Data-driven solar generation forecast considering temporal characteristics of data
AU - Shi, Guangyuan
AU - Eftekharnejad, Sara
N1 - Publisher Copyright:
© 2020 IEEE.
PY - 2020/2
Y1 - 2020/2
N2 - Compared to traditional generating resources, distributed Photovoltaic (PV) resources are highly intermittent in nature. Hence, ensuring reliability under high penetration of PV systems would require additional reserves, purchase of power from real-time markets, or renewable curtailments, all of which dramatically increase cost of power generation and transmission. Accurate PV forecasting can eliminate the need for these costly measures and is necessary for seamless grid operations with high levels of solar PV systems. PV forecasting, however, is challenged by the fact that PV generation behavior differs in different regions and under different weather patterns. Hence, it is critical to consider both the spatial and temporal characteristics of the data when developing a data-driven forecast model. This paper seeks to provide a comparison of different methods for forecasting solar irradiance, using data from the Syracuse area, with diverse weather patterns that challenge PV forecasting. For a large dataset, a new approach to forecasting based on Symbolic Aggregate Approximation is introduced which enables increasing the calculation speed and reducing the dimensions for accurate forecasting.
AB - Compared to traditional generating resources, distributed Photovoltaic (PV) resources are highly intermittent in nature. Hence, ensuring reliability under high penetration of PV systems would require additional reserves, purchase of power from real-time markets, or renewable curtailments, all of which dramatically increase cost of power generation and transmission. Accurate PV forecasting can eliminate the need for these costly measures and is necessary for seamless grid operations with high levels of solar PV systems. PV forecasting, however, is challenged by the fact that PV generation behavior differs in different regions and under different weather patterns. Hence, it is critical to consider both the spatial and temporal characteristics of the data when developing a data-driven forecast model. This paper seeks to provide a comparison of different methods for forecasting solar irradiance, using data from the Syracuse area, with diverse weather patterns that challenge PV forecasting. For a large dataset, a new approach to forecasting based on Symbolic Aggregate Approximation is introduced which enables increasing the calculation speed and reducing the dimensions for accurate forecasting.
KW - Artificial Neural Networks
KW - Photovoltaic Forecasting
KW - Renewable Generation
UR - http://www.scopus.com/inward/record.url?scp=85083083786&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85083083786&partnerID=8YFLogxK
U2 - 10.1109/TPEC48276.2020.9042521
DO - 10.1109/TPEC48276.2020.9042521
M3 - Conference contribution
AN - SCOPUS:85083083786
T3 - 2020 IEEE Texas Power and Energy Conference, TPEC 2020
BT - 2020 IEEE Texas Power and Energy Conference, TPEC 2020
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2020 IEEE Texas Power and Energy Conference, TPEC 2020
Y2 - 6 February 2020 through 7 February 2020
ER -