Fault diagnosis of Combined Cycle Power Plant Using ELM

In this study, fault detection and fault diagnosis in the high-pressure tubes of a combined cycle power plant's high pressure steam generator was investigated. Identification and prevention from fault propagation in combined cycle power plants plays the main role in improving the reliability and safety of these systems. In this work, leakage faults detection in the internal, inlet and outlet tubes and their effects on operating parameters of a heat recovery steam generator (HRSG) were studied. In order to determine the system's behaviors, residual generation based on nonlinear autoregressive exogenous networks have been proposed, and appropriate features based on statistical feature generation have been extracted. To achieve more accurate classifying of system's faults, an extreme learning method technique was employed for the training of extracted features. Extreme learning machine method's excellent prediction capabilities are the main advantages of represented method, which could be used for improving the performance and troubleshooting of the power plant's equipment. The performances of the proposed fault diagnosis system were assessed at different leakage fault conditions by performing simulation experiments on an accurate model of power plant. The obtained results show the accuracy and reliability of this method.