Machine learningbased generic load forecastingmodel for noisy data: LESCO case study with weather influence

Electric load forecasting (LF) involves the projection of peak demand levels and overall energy consumption patterns to support an electric utility’s future system and business operations. Short and mid-range predictions of electricity load allow electricity companies to retain high energy efficiency and reliable operation. Absence of such prior planning results in a current crisis like situation in Pakistan, where power generation is not up-to the mark, its fallout is forced load shedding and voltage instability. To solve the problem of accurate LF, a variety of models is reported in literature. However, the accuracy of modeling techniques is extremely dependent on data quality. Since, the data recording in power systems of Pakistan is manual and it contains abnormalities like missing values, outliers, and duplication of records. Observing all the aforementioned problems, authors got motivation to devise such a LF model that can perform well on noisy data of Pakistan power systems and can handle load affecting parameters of this region. In this paper, a customized LF model formulation is presented, which incorporates machine learning techniques for data preprocessing, analysis, and model development.