Turbine back pressure identification and optimization with learning neural networks

Neural networks have recently generated much interest because of their adaptability, massive parallelism and robustness. The interest is no longer confined to academic research; several applications have been fielded. The features of neural networks could provide desirable characteristics for many applications in the utility industry. We first provide a general overview of neural network technology, focussing on the aspect most relevant for near-term applications - supervised learning. Then we discuss in some detail a particular application of supervised learning neural networks that we are jointly investigating - the identification and optimization of exhaust condenser pressure of steam turbines.