Predictive Analytics of Electrical Power Output of Coal-Fired Power Plant Using Machine Learning

The pandemics outbreak of Covid-19 in the world has made society and industrial activities very dynamic. The operating power plant must prepare to fulfil the fluctuating electricity demand from the load dispatcher. Hence, predicting the electrical power output is important to give the accuracy to maximize the profit and minimize losses. This paper discusses and predicts the half-hourly electrical output of Paiton Coal-Fired Power Station Unit 1 by develops many predictive models using five different machine learning regression methods. The five parameters that affect the electrical power output are used in the dataset, such as main steam flow, total coal flow, primary airflow, secondary airflow, and vacuum condenser pressure. These input and target variables as the dataset were collected over one year. The dataset is sorted and observed. Then, the best prediction model is sought for predicting electrical power output. Thus, the best performance of the best subset, which contains a complete set of input variables, has been analyzed using the most accurate machine learning algorithm, which is the random forest, with R-squared of 0.996.