Time series forecasting of wind speed and solar radiation for renewable energy sources
File(s)
Author(s)
Sfetsos, Athanasios
Type
Thesis
Abstract
This thesis is focused on the forecasting of wind speed, solar radiation and, additionally, temperature Lime series, as a means to predict: power produced from renewable energy sources. The prediction of power generated from wind turbines and additionally, temperature time series, as a means of predict power produced from renewable energy sources. The prediction of power generated from wind turbines and photovoltaics is crucial for the optimal and economically beneficial operation of a power system with significant penetration levels from these sources. Various linear and nonlinear methodologies are examined for the forecasting of the meteorological time series, and the results are then transformed into power generated from wind and photovoltaic power plants.
A main concern of the research is the development of forecasting schemes, for the production of fast and highly reliable predictions. These are represented by various types of artificial intelligence techniques such as neural networks and adaptive neuro-fuzzy inference systems which have been applied alongside conventional linear and non linear models. Additionally, the application of clustering algorithms in forecasting is examined. For the artificial intelligence approaches, various combinations of past measurements and statistical properties of the time series in consideration are applied as a means of clustering data with similar characteristics. Finally, Neural Logic Networks, are examined as potential forecasting tools. Pre-processing of data with logic rules is proved to be an effective way in determining a better prediction model.
Wind speed and solar radiation time series are forecasted from models based solely on past measurements (univariate) and with the incorporation of available meterological information (multivariable). In addition, the univariate time series include some extra features such as pre-processing through simple and periodic differencing, and a forecasting scheme of the error used for the modification of the initial prediction process.
Artificial Neural Network models are applied for the estimation of the diffuse part of horizontal solar radiation, which is used for the determination of the solar radiation received by a surface of any orientation. The incident solar radiation in combination with temperature and appropriate models are used to estimate the produced power from a photovoltaic plant. For wind turbines a power versus wind speed model is selected.
A main concern of the research is the development of forecasting schemes, for the production of fast and highly reliable predictions. These are represented by various types of artificial intelligence techniques such as neural networks and adaptive neuro-fuzzy inference systems which have been applied alongside conventional linear and non linear models. Additionally, the application of clustering algorithms in forecasting is examined. For the artificial intelligence approaches, various combinations of past measurements and statistical properties of the time series in consideration are applied as a means of clustering data with similar characteristics. Finally, Neural Logic Networks, are examined as potential forecasting tools. Pre-processing of data with logic rules is proved to be an effective way in determining a better prediction model.
Wind speed and solar radiation time series are forecasted from models based solely on past measurements (univariate) and with the incorporation of available meterological information (multivariable). In addition, the univariate time series include some extra features such as pre-processing through simple and periodic differencing, and a forecasting scheme of the error used for the modification of the initial prediction process.
Artificial Neural Network models are applied for the estimation of the diffuse part of horizontal solar radiation, which is used for the determination of the solar radiation received by a surface of any orientation. The incident solar radiation in combination with temperature and appropriate models are used to estimate the produced power from a photovoltaic plant. For wind turbines a power versus wind speed model is selected.
Version
Open Access
Date Issued
1999-02
Date Awarded
1999
Copyright Statement
Attribution-Non Commercial-No Derivatives 4.0 International Licence (CC BY-NC-ND)
Advisor
Coonick, Alun
Publisher Department
Department of Electrical and Electronic Engineering
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)