Short-Term Load Forecasting Using AMI Data

  • 2022-05-16 17:04:51
  • Haris Mansoor, Sarwan Ali, Imdadullah Khan, Naveed Arshad, Muhammad Asad Khan, Safiullah Faizullah
  • 0

Abstract

Accurate short-term load forecasting is essential for the efficient operationof the power sector. Forecasting load at a fine granularity such as hourlyloads of individual households is challenging due to higher volatility andinherent stochasticity. At the aggregate levels, such as monthly load at agrid, the uncertainties and fluctuations are averaged out; hence predictingload is more straightforward. This paper proposes a method called Forecastingusing Matrix Factorization (\textsc{fmf}) for short-term load forecasting(\textsc{stlf}). \textsc{fmf} only utilizes historical data from consumers'smart meters to forecast future loads (does not use any non-calendarattributes, consumers' demographics or activity patterns information, etc.) andcan be applied to any locality. A prominent feature of \textsc{fmf} is that itworks at any level of user-specified granularity, both in the temporal (from asingle hour to days) and spatial dimensions (a single household to groups ofconsumers). We empirically evaluate \textsc{fmf} on three benchmark datasetsand demonstrate that it significantly outperforms the state-of-the-art methodsin terms of load forecasting. The computational complexity of \textsc{fmf} isalso substantially less than known methods for \textsc{stlf} such as longshort-term memory neural networks, random forest, support vector machines, andregression trees.

 

Quick Read (beta)

loading the full paper ...