Comparative prediction analysis of 600 MWe coal-fired power plant production rate using statistical and neural-based models

Abstract

Electrical energy is a crucial phenomenon for daily human life & industry, and various types of production plants such as nuclear, thermal, wind, geothermal, hydro, biomass, solar, etc. are developed with distinct advantages and disadvantages recently. Overall production cost is a significant constraint for the power plants and serious optimization problem to be solved in terms of productivity, hence technical, economical and environmental aspects should be considered due to raw material and fuel prices, NOx, SOx and CO2 emission regulations, operational costs and thermodynamic efficiency constraints in the operation. In this research, production rate of a coal-fired thermal power plant is modeled and predicted using Artificial Neural Networks (ANN), Autoregressive Integrated Moving Average (ARIMA) and Multiple Linear Regression (MLR) algorithms selecting appropriate process parameters those effect total amount of generator production output rate. All data was collected from the 600 MWe ICDAS Coal-fired Thermal Power Plant located in the region of Marmara, Turkey during 3 months of operation obtaining the consolidated control system data and reports. The selected performance criterion, regression coefficient and root mean squared error, is found satisfactory after the analysis of computational results and it is seen that the ANN model shows better prediction performance against the MLR and ARIMA approaches in this case study.