Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Medium term solar radiation and wind speed prediction based on hourly time series data

Efstathopoulos Nikolaos

Full record

Year 2020
Type of Item Diploma Work
Bibliographic Citation Nikolaos Efstathopoulos, "Medium term solar radiation and wind speed prediction based on hourly time series data", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2020
Appears in Collections


It is a fact that in recent years the demand for electricity has been steadily increasing, while at the same time reserves of fossil fuels are gradually declining. Therefore, the integration of renewable energy sources (RES) on a large scale in the Electricity Systems as well as the restructuring of electricity markets are of a major concern to global energy policy. Also, if we take into account the fact that they are environmentally friendly, in contrast to conventional energy sources, then it is obvious that renewable energy sources will play a decisive role in meeting energy requirements in the near future. Both wind and solar energy are among the most widely used alternative forms of energy, as evidenced by their growing global power. However, there is intense variability both in solar and even more in wind energy, which makes it difficult for them to be introduced in electricity networks. Therefore, the forecasting of wind and solar power are very important issues, both for the safe operation of the system and the management of RES, as well as for the provision of high quality power at the lowest possible cost.Given the relationship between solar radiation and solar energy, as well as wind speed and power generated by wind turbines, it is necessary to create models that can better predict the above atmospheric variables.As a result of all the above, in the framework of the present dissertation, models of medium-term wind speed and solar radiation prediction were constructed. These models fall into both the category of artificial intelligence and machine learning, as well as advanced mathematical prediction methods. Specifically, regression trees, neural networks, deep neural networks and regression models with ARMA errors were made. The data which were used as input to the models, were recorded from the island Dia (uninhabited area), which is located north of Heraklion, Crete. These data were divided by month and then followed by the process of training the models and the medium-term forecast. The medium-term forecast was basically done using the Recursive Multi-step Forecast method, in all individual models. However, the Multiple Output Forecast Strategy was tested and compared with the basic prediction method mentioned above. In addition, data were used, which were recorded within the area of Chania, specifically in Chalepa. This was done in order to highlight the value of data in the training process and consequently in the forecasting of time series. Finally, the errors from the application of the models are evaluated, as well as conclusions and ideas for future research are recorded.

Available Files