Solar Power Forecasting using CNN LSTM, RNN, ANN

I worked under Hong Kong University of Science and Technology (HKUST)  Assistant Professor of Envrionment and Sustainability, Professor Lu using solar output data and weather features from the HKUST Supersite to develop a machine learning model including to predict solar power output multiple hours ahead. I achieved about 6% forecasting error on one hour ahead and less than 10% on multiple hours ahead.

My research focuses on the integration of solar energy into the power grid by forecasting solar power ahead of time using a time series forecasting problem formulation. The aim is to help reduce the impact of fossil fuels on the environment. The two main objectives are namely to forecast photovoltaic (PV) power one timestep ahead and to forecast PV power multiple timesteps ahead. Data has been taken from the Hong Kong University of Science and Technology. My research compares machine learning techniques such as CNN LSTM, RNN LSTM, Dense Neural Networks, and Convolutional Neural Networks, to find the best predictor of PV output. The tuning of hyperparameters such as the learning rate, regularization parameter, activation function, number of iterations, etc. is also an essential part of the discussion. The long short-term memory (LSTM) and dense feed-forward layers, as well as convolutional layers, help to introduce new important features from the original features.

​

Impact

The research work helped improve upon existing prediction algorithms by 30% for solar panels at the HKUST supersite and this can help to increase the share of renewable energy in the HK energy grid from 0.6% to more than 5% of the energy grid.

​

Recognition

This paper has been published in the SSRN Energy Journal, Research Gate, and has been accepted in the HKAGE Academy's Got Talent development programme for submission to international competitions such as ISEF Regeneron and International Exhibition of Inventions Geneva.