Short term prediction of wind speed and wind power using hybrid signal processing and machine learning techniques

Abstract

newline vi newlineABSTRACT newlineThe demand for renewable energy has increased worldwide to deal with electrical newlineenergy crisis. Renewable energy resources are very attractive because it is environment newlinefriendly and clean in nature. Wind energy is one of the most attractive and promising newlinerenewable energy resources (RESs) to fulfill the global power demand for the last decade. newlineThe rapid growth of wind power penetration into modern power grids in a micro or smart newlinegrid environment, wind speed and wind power forecasting plays a significant role in newlineplanning the integration of wind power to the power grids for efficient energy management newlinetasks like the unit commitment, capacity planning, load balancing, power quality newlineimprovement and frequency regulation, etc. However, the wind speed is one of the most newlinecomplex natured weather parameter due to its dependence on different parameters like newlinerotation of world, topographical properties of earth, temperature and pressure, etc. newlineTherefore, it is difficult to forecast the wind speed and wind power accurately. Further, newlineaccurate short-term wind speed and wind power forecasting is necessary for the system newlineoperators to make the decisions of power generation schedules and dispatch at the newlineconventional power plant and to find out the reserve power. Therefore, prediction models newlineare designed to produce accurate forecasting result in short term basis. newlineThe main aim of the thesis study is to develop advance, reliable, fast, and accurate newlinehybrid models for short-term prediction of wind speed and wind power. In this thesis we newlinemainly focus on two types of short-term forecasting for wind speed and wind power, we newlinestart with the classical point forecasting and then move towards probabilistic forecasting. In newlinethis study various types of hybrid models are proposed for both point an