Forecasting refers to scientifically determining future value(s). Forecasting can be process-driven or data-driven. Process-driven forecasting tries to explicitly model the physical formation process, while data-driven forecasting develops an input-output mapping just based on a long collection of historical data. Data-driven forecasting has been widely adopted recently because of the availability of gauging data, the ever-increasing computational power, the development of advanced modelling theory and software, and the eschewal of explaining the complex physical formation process that hasn’t been well understood. This talk reports our work on data-driven machine learning forecasting models for hydro, solar and wind, including streamflow, evapotranspiration, solar radiation, and wind speed. Machine learning models can address strong nonlinearity inherent in the input-output mapping for forecasting hydro, solar and wind.
Yu Xiang is currently an associate professor with the School of Information Engineering at the Nanchang Institute of Technology, China. He obtained PhD from the Hong Kong University of Science and Technology in 2015. His research interests include metaheuristics, nonlinear modelling, management of water resources and energy systems, and decision support systems. His research has been funded by the National Natural Science Foundation of China, the China Postdoctoral Science Foundation, the Jiangxi Province Department of Education, and the Jiangxi Province Key Laboratory Open Foundation. He has published papers on journals such as Agricultural Water Management, Applied Energy, Applied Mathematics and Computation, Energy Conversion and Management, Information Sciences, International Journal of Electrical Power and Energy Systems, Journal of Hydro-Environment Research, and Journal of Hydrology.
For more information about the ESD Seminars Series, please contact Lin Shao Wei at email@example.com