Wavelet Transform Based Hybrid Models For Short Term Load Forecasting

Abstract

Abstract newlineThe and#64257;eld of forecasting was and still is one of the primary areas of Scientiand#64257;c newlineinvestigation that is gaining importance in real-world applications. The selection newlineand implementation of a proper forecast methodology has always been an newlineimportant planning and control issue in most of these applications. Electrical newlineload forecasting has received an increasing attention over the past few decades newlineby academicians, industrial researchers and practitioners due to its importance newlinein energy management systems. Forecasts with diand#64256;erent lead times are needed newlinefor diand#64256;erent purposes of the electric industry. Short term load forecasts with newlinelead times ranging from an hour to several days is crucial to economic operation, newlinesecurity analysis, maintenance scheduling and task scheduling for both power newlinegeneration and distribution facilities. Over estimation of the energy demand newlinecan cause over-conservative operation whereas under estimation may result in newlineover-risky operation. Accurate forecasts are necessary for the safe, economic newlineand reliable operation of the power system. newlineThe electric load series is a time-variant, non-linear and volatile signal that is newlinerelated in complex and nonlinear fashion with various factors such as the time newlineof the day, the day of the week, climatic condition and the past usage patterns. newlineThe design of the input data to a forecast engine is an important phase in a newlinemodel and it plays a crucial role in the forecast accuracy. The approach of newlinefeature selection based on and#64257;ltering method is novel technique that has gained newlineimportance over the years. Of the diand#64256;erent available and#64257;lters, wavelet and#64257;lters are newlinewell suited for handling non-stationary and non-linear signals. The wavelet newlinetransform is known to provide a useful decomposition of the time series so newlinexviii newlineAndhra University, Visakhapatnam newlinexix newlinethat faint temporal structures can be revealed and handled by parametric/nonparametric newlinemodels. The transform is known to provide a sound mathematical newlinetechnique for designing and deploying and#64257;lters, which facilitates interpolation, newline