Posters

Abstract

Wind energy is the most growing source of renewable energy and it requires its accurate future predictions of wind resource potential at a particular site for development of wind farms. The wind speed fluctuations on an hourly and daily basis require fair prediction for managing grid loads. Wind speed, due to its intermittency and dynamic nature, is difficult to forecast in a haphazard fluctuating environment. There are two types of forecasting including short and long-term. The former refers to forecasting done on minutes, hourly or daily basis whereas the latter refers to forecasting done on a daily, monthly or yearly basis. Short-term prediction is essential for reducing scheduling errors and regulating utility-scale operations for better grid reliability. Therefore, accurate forecasts are mandatory for wind energy systems enhanced efficiency. A good quality forecast can also lead to cost reductions and savings, owing to accurate prediction of wind speed seasonality, its dependence on location geographical parameters and climatic conditions. Several researchers have thought out different models for wind speed prediction including physical models, statistical approaches, hybrid physical-statistical models and Artificial Neural Networks (ANN). In the recent era, various studies have led to the advent of the use of artificial intelligence (AI) for predicting variation in metrological parameters such as temperature, humidity, pressure and wind speed. Machine Learning Algorithms (MLAs) are extensive scientific approaches to solving complex non-linear problems. Pressent study discusses application of ANNs in forcasting of wind speed for last week using measured data at four sites of Pakistan. The central objective of this research is to predict wind speed based on various independent variables using a machine learning model known as non-linear autoregressive exogenous (NARX).