Study Of Solar Photovoltaic Power Forecasting Based On LSTM-NN Model

The use of renewable energy is rapidly increasing as to reduce the effects of climate change and global warming which is badly affecting the environment.A significant amount of electricity is being generated from renewable energy sources since the last decade.Among the major renewable energy,photovoltaic(PV)generation has experienced tremendous growth in electricity generation.The number of PV systems will of course increase rapidly in the future due to major shift in policies of the government and international organizations.However,the variable nature of PV power generation creates negative impacts on the electric grid system,such as the stability,reliability,and operation planning issues.The generation of renewable energy from different natural sources is linked to very dynamic changes due to intermittent nature of resources.It is necessary to improve the prediction accuracy of solar power in order to prepare for the unknown and random conditions in the future.In this work,a model of Long Short Term Memory(LSTM)Neural Network has been developed for PV power forecasting.This model has ability to develop relationship when dealing with big data and can handle the long term dependencies.The main works of this thesis are as follows.Firstly,the concept of Solar PV power forecasting is discussed,with different classifications methods based on time,historical data and different techniques.In addition,major aspects of solar forecasting with standardizing performance measures and the major challenges associated with Grid stability are analyzed.Secondly,LSTM model is proposed and its architecture and principle are discussed in detail.Moreover,the working principle of Artificial Neural Network(ANN)model for PV power forecasting is studied.Thirdly,the results of LSTM-NN model for PV power forecasting are analyzed.Specifically,LSTM model is trained with dataset,and the data included all basic weather parameters.The data collected was preprocessed through normalization and linear interpolation for better accuracy.The RMSE error has also been calculated.Results show that LSTM-NN outperformed than other machine learning models in terms of accuracy in forecasting.