Yazar "Sarkez, Aimen E. M." seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • Küçük Resim
    Öğe
    SOLAR RADIATION PREDICTION IN SOLAR ENERGY SYSTEMS USING ANN MODELS: A CASE STUDY
    (2020-12-25) Sarkez, Aimen E. M.
    Artificial neural networks (ANN) are a useful option when it comes to analyzing sensor data, detecting faults, setting up performance and reviewing related processes. For its key feature, the present work makes use of ANN modeling to forecast solar radiation in the city of Tripoli in Libya. The simulation is satisfactory and, hence, applicable to solar energy systems (SES), with the main advantage owed to the fact that the research makes use of arbitrary and nonlinear factors. With economizing, SES enhancing, and improving electricity production in mind, the proposed ANN compares the output with the set targets based on weather data obtained from the Libyan State Meteorological Center. The trial-and-error method using MATLAB neural network fitting tool (nftool) helped to determine the best arrangement of functions to achieve the objective herein. The weather data containing various variables is used to come up with a novel approach to determine SR with the highest possible precision. The approach is feed-forward back propagation (BP)-based the function of a tangent sigmoid transfer (tansig) between the input and the hidden layer with a range of neurons within one hidden layer and function of a linear transfer (purelin) at the output node, all employed in MATLAB’s neural network fitting tool (nftool). The Levenberg-Marquardt (LM) algorithm was put to the test with a set of neurons after comparing the predicted figures with actual data from the Center using mean square error (mse) and coefficient of determination (R). The most applicable ANN model was then chosen with LM and ten neurons to forecast SR which yielded promising results with insignificant MSE and R2 above 0.93. To conclude, the suggested approach and ANN model is applicable for SR forecasting and, therefore, a step forward in improving SES-based energy production. Obviously, it goes without saying that certain theoretical architectures based on ANN continue to play a key role in determining the best simulations for yet improved results in the future. ?