Application of Reversed Modeling Method in Prediction of Steam Turbine Heat Rate

To investigate the nonlinear relationship between steam turbine heat rate and the affecting factors, a reversed modeling method was proposed for the heat rate based on adaptive chaotic opposition-based learning gravitational search algorithm (ACOGSA) and least squares support vector machine (LSSVM), following which a heat rate prediction model was established with operation data of a 600 MW supercritical steam turbine unit using LSSVM reversed modeling method. ACOGSA algorithm was used to solve the problem of LSSVM optimal parameters, and the prediction results were compared with those of the GSA-LSSVM model and BP neural network model. Results show that the proposed model has better generalization ability than traditional models, which therefore can predict the heat rate of steam turbine more accurately.