Repository logo

Browsing by Issue Date, starting with "2022-07-15"

Filter results by year or month
Now showing 1 - 1 of 1
  • Smart irrigation platform based on machine and deep learning
    Publication . Martins, Pedro Jorge Migueis Vaz; Schutz, G.; Cardoso, Pedro J. S.
    Due to climate change, the hydrological drought is assuming a structural character with a tendency to worsen in many countries. In fact, the frequency and intensity of droughts is predicted to increase, particularly in the Mediterranean region and Southern Africa. Since a fraction of the fresh water that is consumed is used to irrigate urban fabric green spaces, which are typically made up of gardens, lanes and roundabouts, it is urgent to implement water waste prevention policies. Reference evapotranspiration (ETo) is a measurement that can be used to estimate the amount of water being taken up or used by plants, allowing a better management of the watering volumes but, the exact computation of the evapotranspiration is not possible without using complex and expensive sensor systems. As an alternative to the devoted sensor solutions, weather parameters can be used to estimate ETo. However that also raises some problems, such as the fact that it requires the use of a dedicated weather station or of data available on the internet. In both contexts, solar radiation (SR) is not commonly available, and since evapotranspiration is dependent on solar radiation, the need to develop ETo prediction models that do not require it as an input parameter arises. This thesis presents some high accuracy reference evapotranspiration and solar radiation prediction models, both having as input a set of limited meteorological features, namely, temperature, humidity and wind, which exclude the need for solar radiation as a parameter. Two approaches were explored for deriving such models: (i) the use of machine learning algorithms like linear regression (OLS, Ridge, Lasso), k-Nearest Neighbors (kNN), Support Vector Machine (SVM), Decision Tree and Random Forest, and (ii) the use of neural networks such as Artificial Neural Networks (ANN), Long Short Term Memory Networks (LSTM), Gated Recurrent Unit Networks (GRU), Recursive Neural Networks (RNN), and the development of hybrid neural network models such as LSTM-ANN, RNN-ANN, and GRU-ANN. Using experimental data collected from a weather station located in Vale do Lobo, south Portugal, and using the machine learning approach mentioned in the previous paragraph, (i), the best performing ETo model gave a coefficient of determination (R2) of 0.975, and the best SR model gave an R2 of 0.831, over the test dataset. When using the neural networks approach, (ii), the best performing ETo model gave an R2 of 0.977, and the best SR model gave an R2 of 0.833. As a final notice, the limited meteorological input parameters were carefully selected so that they are compatible with online freely available weather forecast services.
    2022-07-15Master thesis Open access Show more