Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Ugolini; Yoshikawa, Shinji; Ozawa, Kenji
PNC TN9410 95-253, 13 Pages, 1995/10
This report presents the implementation of the a model reference adaptive control system based on the artificial neural network technique (MRAC
) in a fast breeder reactor (FBR) building block type (BBT) simulator representing the Monju prototype reactor. The purpose of this report is to improve the control of the outlet steam temperature of the three evaporators of the Monju prototype reactor. The connection between the MRAC system and the BBT simulator is achieved through an external shared memory accessible by both systems. The MRAC system calculates the demand for the position of the feedwater valve replacing the signal of a PID controller collocated inside the heat transport system model of the Monju prototype reactor. Two series of simulation tests havc been performed, one with one loop connected to the MRAC system (leaving the remaining two connected to the original PID controller), and the other with three loops connected to the MRAC system. In both simulation tests the MRAC system performed better than the PID controller, keeping the outlet steam temperature of the evaporators closer to the required set point value through all the transients.
Ugolini; Yoshikawa, Shinji; Ozawa, Kenji
PNC TN9410 95-210, 11 Pages, 1995/09
The proper control of the outlet steam temperature of the evaporator is of major importance for improving the overall performance of the balance of plant of a nuclear power reactor. This report presents a predictive and an anticipatory control algorithms based on the artificial neural network (ANN) technique. The two control algorithms are embedded on a model reference adaptive control system based on the ANN technique, defined as MRAC. It has already been illustrated that nonlinear dynamical systems such as the evaporator of a nuclear power plant can be controlled by an MRAC system. However, little attention has been devoted on exploiting the forecasting potential of the ANN technique for enhancing the accuracy and improving the efficacy of the control action of the MRAC system. The improved MRAC system has been tested to simulate the behavior of a fast breeder reactor (FBR) evaporator and to control its outlet steam temperature. The simulation results indicate that the performance of the MRAC system substantially improves when the predictive and the anticipatory control algorithms are activated.
Yoshikawa, Shinji; Okusa, Ryoichi; Ozawa, Kenji
PNC TN9410 95-035, 19 Pages, 1995/03
This report discusses a method to equip a multi layer neural network(NN) with a calculational function to derive differential values of the output parameters against the input parameters. Multi layer NNs have been applied in various domains of engineering, because of easy construction, flexible interpolation of nonlinear multi-input functions, and some other preferable features. However, derivatives of those output parameters have been approximately calculated by interpolating between two different output values. And new methods to guarantee the accuracy of the derivatives have been desired. We payed their attention at sigmoid functions, which are commonly used to realize the nonlinear characteristics of nodes in NNs, and at one of important features of this function type that the derivative is represented by a polinomial of itself. And, we developed a method to add a calculational function to derive differentiated values of the output parameters to multi layer NNs, whose CPU cost is smaller than the original NNs.
