Industry: Energy Region: Cyprus Transaction price: US$ 1 million Transaction method: equity investment
Project introduction:
The photovoltaic energy prediction project of Cyprus University of Science and Technology is led by the Department of Chemical Engineering and jointly conducted by the research teams of Eindhoven University of Technology and Twente University. The project focuses on developing a dynamic data assimilation model for short-term solar radiation prediction, aiming at coping with the power fluctuation caused by solar photovoltaic power generation. The project team pays special attention to using artificial neural network (ANN), especially long-term and short-term memory network (LSTM), to predict solar energy in the near future. The research team has used photovoltaic data for model training and testing for four years, of which 80% of the data is used for training and 20% for testing.
The results show that the single-layer LSTM network is equivalent to the more complex five-layer network in prediction performance, and its normalized root mean square error (nRMSE) is 10.7%, which indicates that the simple LSTM network is also effective in solar energy prediction.
Cooperation mode:
The company is looking for technical cooperation with photovoltaic parks, inverter manufacturers or power grid technology enterprises.