A Sensitivity Study of Machine Learning Techniques Based on Multiprocessing for the Load Forecasting in an Electric Power Distribution System

An accurate and timely forecasting is essential for a utility to manage its finances and maintain a balance between its supply and demand of power. With the broad implementation of intelligent controls on the power grid, there has been an explosion in the amount of data pertaining to energy. It is difficult to deal with an estimate of weight that is based on comparable data, and it requires more time than the amount of time required for a short-term estimate. Thousands of different energy datasets collected by smart meters for day-ahead, periodic, and short-term load forecasting (STLF) were investigated previously. The use of multiprocessing to cut down on the total amount of time required to run the forecasting models is also a method based on some of the research. By sending simultaneous workloads to all of the available processors, it can be accomplished. It is clear that the proposed approach is practical due to the inclusion of ML models, fast execution, and scalability. The idea is to validate by using real energy operational data collected at the distribution motor position within the types of electrical grids. Decision trees outperform other models in terms of their ability to strike a balance between the model's sensitivity and the time required to investigate, according to previous studies.