Evaluating Solar Prediction Methods to Improve PV Micro-grid Effectiveness Using Nonlinear Autoregressive Exogenous Neural Network (NARX NN)

In recent years, insufficient access to energy and environmental challenges caused the push for clean and sustainable means of power generation. The integration of renewable energy resources into the electricity grid is known to reduce Greenhouse Gas (GHG) emissions and environmental pollution. Solar power has the enormous benefit of availability and can often be accessed cost effectively. However, there is a recurring problem of intermittency and variability in most solar power generation systems. The variability and intermittent nature of solar power sources introduce significant challenges in the planning and scheduling of smart grids. Solar power prediction can mitigate this variability and improve the integration of solar power resources into smart grids. This paper presents an Artificial Neural Network (ANN) model for solar power prediction, and assesses how several weather input variables from Leeds, UK, affect the prediction accuracy. Following this, the Nonlinear Autoregressive Exogenous Neural Network (NARX NN) model performance is compared with Nonlinear Autoregressive Neural Network (NAR NN) model using a time series modelling approach. The result shows that NARX NN model outperformed NAR NN model for the studied geographical location.