Nonlinear modelling and state estimation in a real power plant using neural networks and stochastic approximation

The paper deals with two main contributions. The first is the definition of a very accurate model of a section of a real 320 MW power plant. The second is the online tuning of such a model, in accordance with the measures provided by the available sensors of the real system, by suitably using neural networks and stochastic approximation techniques. The proposed approach exhibits very general features so that it can be used for different types of plants. It is therefore possible to design a simulator that can be connected in parallel with the real plant, thus providing the plant technician with information about non-accessible variables that are very useful for supervision purposes. A validation procedure applied to the real power plant and simulation results for the tuning phase show the effectiveness of the approach.