A Data Science Approach Towards Dynamic Modelling of Chemical Process

Nagalla, Sree Harsha and Mitra, Kishalay (2019) A Data Science Approach Towards Dynamic Modelling of Chemical Process. Masters thesis, Indian institute of technology Hyderabad.

Text
Mtech_Thesis_TD1491_2019.pdf
Restricted to Repository staff only until 10 July 2020.
Download (2MB) | Request a copy

Abstract

Accurate forecasting of solar irradiance is helpful in monitoring and control of a solar plant. It will be beneficial in maximizing solar power harvesting through efficient utilization of the resource. This issue has motivated the researchers in finding novel techniques for efficient solar energy harvesting. Statistical forecasting methods have been used in modelling solar irradiance time series data. However, the conventional methods of time series modelling fail to model data with long term dependencies. In this work, Long Short Term Memory (LSTM) networks are implemented in solar irradiance forecasting. LSTMs are a kind of deep neural networks which are known for efficiently modelling nonlinear data with long term dependencies. LSTM networks are implemented in Keras, a neural network library in Python for modelling Solar irradiance data. The obtained results of the LSTM network are compared with system Identification tools such as wavelet networks, and nonlinear auto-regressive exogenous model. Finally, the LSTMs are utilized for forecasting day ahead solar irradiance at 2 locations.

[error in script]

IITH Creators: