Modeling the superheated steam temperature with a data-driven based approach

Superheated steam temperature is a vital factor that affects the power generation efficiency. A data-driven based approach is proposed to modeling the superheated steam temperature. The ReliefF algorithm is employed to select the input features. In addition, a back propagation neural network (BP) model with parameters optimized by genetic algorithm (GA) is proposed to constructed the prediction model. Experiment results demonstrate that the proposed method can get better forecasting results in comparison with the PSO-BP (particle swarm optimized back propagation neural network), linear regression approach and the MLP (multi-layer perceptron) approach.