DOI: https://doi.org/10.15368/theses.2021.176
Available at: https://digitalcommons.calpoly.edu/theses/2612
Date of Award
9-2021
Degree Name
MS in Electrical Engineering
Department/Program
Electrical Engineering
College
College of Engineering
Advisor
Xiao-Hua (Helen) Yu
Advisor Department
Electrical Engineering
Advisor College
College of Engineering
Abstract
Renewable energies such as wind power have become integral parts of modern power networks. Short-term wind speed prediction is crucial for smart grids, as it can help balance the demand and supply, as well as set the energy price in the market.
In this thesis, we simulate and compare various neural network models for short-term wind speed prediction, including multi-layer feedforward neural networks, convolutional neural networks, long-short term memory networks, and hybrid models such as CNN- LSTM and ConvLSTM. Computer simulation results show that all artificial neural networks are able to provide satisfactory prediction. Among them, the multi-layer feedforward neural networks require less training time and often give reasonable results; while the ConvLSTM networks need longer time to train and implement, but it may provide a slightly better accuracy in some cases.