Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Correlation and completion of rainfall data in Kampos, Chania using artificial neural network

Kyriakou Nikolaos

Full record

Year 2021
Type of Item Diploma Work
Bibliographic Citation Nikolaos Kyriakou, "Correlation and completion of rainfall data in Kampos, Chania using artificial neural network", Diploma Work, School of Chemical and Environmental Engineering, Technical University of Crete, Chania, Greece, 2021
Appears in Collections


In the context of this thesis, the use and training of artificial neural networks is examined to simulate the rainfall data in Stalos area, which they didn’t recorded cause of damages in the station, processing rainfall data of the whole area of Kampos, Chania, during the period of 9/2018 to 9/2019.In Kampos, Chania, apart from the meteorological station of Stalos, are also located those of the city of Chania, Kounoupidiana (University’s campus), Platanias and Alikianos. Using the rainfall time series data of these five (5) meteorological stations for a period of one year that all operated, a pre-processing of the values ​​was done to isolate the studied values. As the time series resulting from the process of recording rainfall heights are non-linear, the use of Artificial Neural Networks to carry out the work was deemed feasible.Initially, it was essential to create an input table with the rainfall values ​​recorded at the stations of Alikianos, Platanias, the center of Chania and Kounoupidiana, and a target vector in the neural network with the values ​​of the station of Stalos. After the data pre-processing was completed, artificial neural networks were trained with Neural Fitting tool (nftool) of Matlab. The two training algorithms used are Levenberg-Marquardt and Bayesian Regularization. The training of artificial neural networks was based on the above for different parameters each time in terms of hidden nodes, training percentages and training algorithms. During the training of artificial neural networks, an attempt was made to identify the model with the parameters from which the optimal results would be emerged. Selection criteria for selecting the optimal model were the square root of the mean square error and the correlation coefficient. Finally, summarizing our results, a fault of the order 10-2m was achieved using the Bayesian Regularization algorithm.

Available Files