A Hybrid Machine Learning Technique for Short-Term Electric Load Forecasting for Residential Users

Abstract

With the advancement of technology, people’s reliance on energy is rapidly increasing. Calculating a country’s power consumption is one technique to estimate its productivity and rate of development . Recently, countries are using renewable energy and smart grids to store and distribute energy. Traditional techniques to assume that how much load will be required for consumers are not applicable. To overcome this problem researchers have utilized short term load forecasting techniques. Developed countries already using many effective short term load forecasting techniques. However, developing countries, such as Pakistan, have been deprived of this. Recently, PRECON dataset was introduced in Pakistan. In my research, I tried to make use of the best machine learning algorithms such as Xgboost and LightGBM for forecasting. This research collectively uses hand crafted and extracted features and then important features are extracted from them using Xgboost and LightGBM. Final forecast results from our models show the our work has the best mape from the previous studies.