To ensure the long-term normal operation of gas turbine generator sets,there is an urgent need for research on the fault warning method for gas turbine generator sets.A method of gas turbine fault warning based on improved convolutional adversarial network is proposed.Firstly,structural improvements are made to the generator and discriminator in the original structure respectively to enhance the performance of the model and make the model directly output quantized evaluation values.Then,the loss value calculation functions of the generator and discriminator are optimized to extend the effective range.Finally,the test is carried out with an F-type gas turbine generator set as the target.The experimental results show that the method can get rid of the dependence of existing research on calibrated fault data sets,complete the quantitative evaluation conforming to the qualitative estimation for unknown categories of data,and achieve more than 98％fault warning accuracy for multiple categories of unknown data.This research contributes to the greenization and intelligence of cities.
关键词
深度学习/绿色城市/燃气轮机/发电机组/卷积对抗网络/故障告警
Key words
Deep learning/Green City/Gas turbine/Generator set/Convolutional adversarial network/Fault warning
维普
万方数据