Prediction of Hourly Total Energy in Combined Cycle Power Plant Using Machine Learning Techniques

Abstract

Electricity is a form of energy used around the world to power everything in our daily life. The value of energy and its renewable nature assemble energy as one of the vital topics. The correct approximation of hourly energy created on an exceeding power plant is crucial for producing cost-effective energy. In recent times, Machine Learning (ML) algorithms are widely utilized in predictive analysis of the power plants’ estimated energy production. A Combined Cycle Power Plant (CCPP) refers to a distinctive electrical energy producing station, where energy is generated with the help of the two types of turbines (gas and steam) merged into a single cycle. This study explores and evaluates four ML regression techniques for forecasting the total energy output per hour operated by a CCPP. Our entire set of data is collected from Rural Power Company Limited (RPCL), Mymensingh, Bangladesh, which contains 24 input variables, 8768 observations, and net hourly total energy (MW) as the target variable. The performance evaluation of the following regression techniques: Linear, Lasso, Decision Tree, and Random Forest, shows that Linear Regression performs most efficiently our dataset. The value of R2 for Linear Regression is 0.99910896 (99.91%).