Energy consumption prediction of coal-fired unit boiler system based on different operating states

Abstract

ABSTRACT The energy consumption of the coal-fired unit’s boiler system varies significantly when accommodating flexible peak shaving demands. To aid staff in comprehending the boiler’s operational status and optimize its performance, a prediction model of the energy consumption of the boiler system was established based on the different states of the unit operation. First, a dataset of boiler energy consumption under variable load was established based on theory of fuel-specific consumption, and the Mean Impact Value (MIV) algorithm was used to simplify the input features of the model. Second, the Aquila Optimizer (AO) with tent map, adaptive t-distribution, and opposites learning mechanism was introduced to determine the parameters in the prediction model. On this basis, the sliding-window method was used to classify the operating states based on the load of the unit, and the original dataset without operating state distinction, the steady state operating data, the load uplink data, and the load downlink data were used to establish Models 1–4, respectively. The result shows that Model 1 outperforms Model 2 with 24.45% and 18.22% lower aMAE and aRMSE, respectively. compared to Model 3, it shows a decrease of 24.07% and 16.98%. Compared to Model 1, Model 4 shows a reduction of 20.52% and 18.91% in aMAE and aRMSE, respectively. This indicates that distinguishing different operating states to establish boiler energy consumption prediction models can obtain better prediction accuracy.