a note on the practical efficiency of using wavelet transform for short-term load forecasting in smart grids

In the smart grid era, load forecasting is the building block of a secure, reliable, and economic power system. therefore, many researchers have spent a lot of time trying different methods to improve load forecasting accuracy. in recent years, one of the rather frequently used methods is the decomposition of load series into high and low-frequency components using wavelet transform, which reportedly has shown impressive results in some articles. in this paper, through several simulations, it’s demonstrated that despite some of the benefits of the wavelet transform, it can produce unrealistic results due to the border distortion problem. in fact, our work investigates the practical efficiency of wavelet transform in the load forecasting task from the viewpoint of a system operator who is forecasting the next day’s load profile every day. to this end, multiple linear regression (mlr) and artificial neural network (ann) models are used with wavelet transform to conduct experiments on new york city electric load dataset.

a note on the practical efficiency of using wavelet transform for short-term load forecasting in smart grids

in the smart grid era, load forecasting is the building block of a secure, reliable, and economic power system. therefore, many researchers have spent a lot of time trying different methods to improve load forecasting accuracy. in recent years, one of the rather frequently used methods is the decomposition of load series into high and low-frequency components using wavelet transform, which reportedly has shown impressive results in some articles. in this paper, through several simulations, it’s demonstrated that despite some of the benefits of the wavelet transform, it can produce unrealistic results due to the border distortion problem. in fact, our work investigates the practical efficiency of wavelet transform in the load forecasting task from the viewpoint of a system operator who is forecasting the next day’s load profile every day. to this end, multiple linear regression (mlr) and artificial neural network (ann) models are used with wavelet transform to conduct experiments on new york city electric load dataset.