System Identification Methodology of a Gas Turbine Based on Artificial Recurrent Neural Networks (original) (raw)
Related papers
Nonlinear Gas Turbine Modelling : A Comparison of Narmax and Neural Network Approaches
2002
In this paper two nonlinear modelling approaches are employed to derive single nonlinear models for a Rolls Royce aircraft gas turbine. The first approach is based on the estimation of a NARMAX model using conventional structure selection and parameter estimation techniques, and the second approach is based on the use of feedforward Multilayer-Perceptron (MLP) neural networks. The performances of the models derived by the two approaches are demonstrated using a range of engine tests and by analysing their static and dynamic behaviours.
Gas Turbine Identification with Linear and Non-Linear Techniques
2005
The discussion present in this paper explains how identification can be applied on gas turbine dynamics, considering a single-spool gas turbine. This gas turbine have one shaft, with a centrifugal compressor and an axial turbine. Its dynamic behavior have linear and non-linear characteristics, where the non-linear behavior presents some dificulties in construct a complete gas turbine model. Some linear models provide tolerable transient behavior to the identification, such rotor, pressure and temperature dynamics. These dynamics characteristics can be described by linear identification, such as Box-Jenkins and Output-error models. This models can only present linear dynamics, and it’s applicable only to small disturbances around a design point. To realize a non-linear identifcation a Narmax identification is adopted to obtain a wide range model, and a comparison between the linear and non-linear models is important to distinguish the main features in both identifications. The identi...
NARX models of an industrial power plant gas turbine
IEEE Transactions on Control Systems Technology, 2000
This paper reports the experience with the identification of a nonlinear autoregressive with exogenous inputs (NARX) model for the PGT10B1 power plant gas turbine manufactured by General Electric-Nuovo Pignone. Two operating conditions of the turbine are considered: isolated mode and non-isolated mode. The NARX model parameters are estimated iteratively with a Gram-Schmidt procedure, exploiting both Forward and Step-wise regression. Many indexes have been evaluated and compared in order to perform subset selection in the functional basis set and determine the structure of the nonlinear model. Various input signals (from narrow to broadband) for identification and validation have been considered.
Nonlinear identification of aircraft gas-turbine dynamics
Neurocomputing, 2003
Identiÿcation results for the shaft-speed dynamics of an aircraft gas turbine, under normal operation, are presented. As it has been found that the dynamics vary with the operating point, nonlinear models are employed. Two di erent approaches are considered: NARX models, and neural network models, namely multilayer perceptrons, radial basis function networks and B-spline networks. A special attention is given to genetic programming, in a multiobjective fashion, to determine the structure of NARMAX and B-spline models.
Estimating Gas Turbine Internal Cycle Parameters Using a Neural Network
Volume 5: Manufacturing Materials and Metallurgy; Ceramics; Structures and Dynamics; Controls, Diagnostics and Instrumentation; Education; General, 1996
We show that a neural network can be successfully used in place of an actual model to estimate key unmeasured parameters in a gas turbine. As an example we study the combustion reference temperature, a parameter that is currently estimated via a nonlinear model inside the controller and is used in a number of critical mode-setting functions within the controller such as calculating the fuel-split between various manifolds. We show that a feedforward neural network using simple back propagation learning can be built for estimating combustion reference temperature. The neural network matches the accuracy of the current estimate; and it is more robust to errors in its internal parameters. This is advantageous from the point of view of implementation since a number of errors creep in when running the algorithm on a digital controller, and an estimator that is not robust with respect to such errors can degrade the performance of the whole system.
Nonlinear Identification of Process Dynamics Using Neural Networks
Nuclear Technology, 1992
The nonlinear identification of process dynamics encountered in nuclear power plant components is addressed, in an input/output sense, using artificial neural systems. A hybrid feedforward/feedback neural network, namely, a recurrent multilayer perceptron, is used as the model structure to be identified. The feedforward portion of the network architecture provides its well-known interpolation property, while through recurrency and cross-talk, the local information feedback enables representation of temporal variations in the system nonlinearities. The standard backpropagation learning algorithm is modified, and it is used for the supervised training of the proposed hybrid network. The performance of recurrent multilayer perceptron networks in identifying process dynamics is investigated via the case study of a U-tube steam generator. The response of a representative steam generator is predicted using a neural network, and it is compared to the response obtained from a sophisticated computer model based on first principles. The transient responses compare well, although further research is warranted to determine the predictive capabilities of these networks during more severe operational transients and accident scenarios.
This paper explores the application of artificial neural networks for online identification of a multimachine power system. A recurrent neural network has been proposed as the identifier of the two area, four machine system which is a benchmark system for studying electromechanical oscillations in multimachine power systems. This neural identifier is trained using the static Backpropagation algorithm. The emphasis of the paper is on investigating the performance of the variants of the Backpropagation algorithm in training the neural identifier. The paper also compares the performances of the neural identifiers trained using variants of the Backpropagation algorithm over a wide range of operating conditions. The simulation results establish a satisfactory performance of the trained neural identifiers in identification of the test power system.
Several studies has reported the use of neural networks in the dynamic modelling and simulation of heavy duty turbines. However; focus on exhaust gas temperature a key indicator of turbine thermal health is yet to be made. In this paper the modelling of exhaust gas temperature using the non-linear autoregressive network and subsequent multi step prediction with data collected from GT13E2 turbine was embarked upon. Features which were statistically significant for EGT prediction were selected through stepwise regression. One hidden layer was sufficient to approximate the function and The optimal architecture for training was achieved by training the network with a fixed hidden neuron and varying time delay at the inputs and output. It is observed that the optimal performance is realized when the prediction is regressed at tapped input delay of 1 in open loop. 7 hidden neuron and 1 tapped delay is selected for function approximation after series of neurons ranging from 4-15 was tested. The appropriate model was carefully selected by utilizing the method of holdout cross validation, corrected Akaike Informationon Criterion and Schwartz Bayesian information criterion. The final architecture was trained, and converted to close loop NARX network where 100 time steps ahead prediction of EGT was made. Although it was observed that accuracy diminishes as prediction horizon increases, the chosen optimised architecture successfully predicted EGT 100 steps ahead with MAE of 2.9665 and RMSE of 3.9675. Therefore; the dynamic NARX model can be utilized for multistep ahead prediction in incidence of sensor malfunction at the turbine outlet of heavy duty gas turbines.
… Distributed and Grid …, 2012
The paper investigates nonlinear system identification using system output data at various linearized operating points. A feed-forward multi-layer Artificial Neural Network (ANN) based approach is used for this purpose and tested for two target applications i.e. nuclear reactor power level monitoring and an AC servo position control system. Various configurations of ANN using different activation functions, number of hidden layers and neurons in each layer are trained and tested to find out the best configuration. The training is carried out multiple times to check for consistency and the mean and standard deviation of the root mean square errors (RMSE) are reported for each configuration.