Robert M. Edwards

State Feedback Assisted Classical Control: An Incremental Approach to Control Modernization of Existing and Future Nuclear Reactors and Power Plants

AbstractA new control concept, state feedback assisted classical control, is described. The concept incorporates a classical output feedback control system as an easily understood inner control loop and casts modern state feedback in the role of a demand signal augmentation to achieve the goals of an optimal control design. The new control configuration is proposed as a method to achieve transparency of control for implementation of optimal control theory for nuclear reactors and power plants. It may find acceptance for incremental modernization of existing plants because it may permit existing control loops to remain in place while new control loops are added to optionally augment demand signals to achieve an optimal control objective. This approach also leads into the design of robust and fault-tolerant control of nuclear power plants.

Improved nuclear reactor temperature control using diagonal recurrent neural networks

A novel approach to wide-range optimal reactor temperature control using diagonal recurrent neural networks (DRNNs) with an adaptive learning rate scheme is presented. The drawback of the usual feedforward neural network (FNN) is that it is a static mapping and requires a large number of neurons and takes a long training time. The usual fixed learning rate based on an empirical trial and error scheme is slow and does not guarantee convergence. The DRNN is for dynamic mapping and requires much fewer neurons and weights, and thus converges faster than FNN. A dynamic backpropagation algorithm coupled with an adaptive learning rate guarantees even faster convergence. The DRNN controller described includes both a neurocontroller and a neuroidentifier. A reference model which incorporates an optimal control law with improved reactor temperature response is used for training of the neurocontroller and neuroidentifier. Rapid convergence of this DRNN-based control system is demonstrated when used to improve reactor temperature performance. >

LQG/LTR robust control of nuclear reactors with improved temperature performance

The authors present the design of a robust controller using the linear quadratic Gaussian with loop transfer recovery (LQG/LTR) for nuclear reactors with the objective of maintaining a desirable performance for reactor fuel temperature and the temperature of the coolant leaving the reactor for a wide range of reactor powers. The results obtained are compared to those for an observer-based state feedback optimal reactor temperature controller. Sensitivity analysis of the dominant closed-loop eigenvalues and nonlinear simulation are used to demonstrate and compare the performance and robustness of the two controllers. The LQG/LTR approach is systematic, methodical, and easy to design and can give improved temperature performance over a wide range of reactor operation. >

Fuzzy-adapted recursive sliding-mode controller design for a nuclear power plant control

In this paper, a multi-input multi-output fuzzy-adapted recursive sliding-mode controller (FARSMC) is designed for an advanced boiling water reactor (ABWR) nuclear power plant, to control reactor pressure, reactor water level and turbine power. The FARSMC is intended to replace the existing conventional controllers for the power range of 70% to 100% rated power. The controller has a recursive form that treats model uncertainty and external disturbances in an implicit way. Thus there is no need to specify uncertainties and disturbances for this controller design in advance. Moreover, the chattering problem common among conventional sliding-mode controllers is completely removed by this recursive sliding-mode control (RSMC) algorithm. The performance of the resulting RSMC is further improved by parameter adaptation using fuzzy logic algorithm, resulting in fuzzy-adapted RSMC (FARSMC). To apply RSMC technique, the original nonlinear plant model is first transformed to a canonical form. Simulations of the simplified ABWR model with the designed FARSMC indicate that FARSMC may result in better performance than the existing PI controllers in that the plant transient responses to the desired output step change have shorter settling time and smaller magnitude overshoot/undershoot. The control actions of the FARSMC are milder than those from the PI controllers. Robustness of the FARSMC with respect to power level variation and capability to reject external disturbances is also achieved. Successful fuzzy adapted sliding-mode controller design and implementation on the model shows that FARSMC may be a practical choice for nuclear power plant control.

A reconfigurable hybrid system and its application to power plant control

This paper presents a reconfigurable approach to implement decision and control systems for complex dynamic processes. The proposed supervisory control system is a reconfigurable hybrid architecture structured into three functional levels of hierarchy, namely, execution, supervision, and coordination. While the bottom execution level is constituted by either reconfigurable continuously varying or discrete-event systems, the top two levels are necessarily governed by reconfigurable sets of discrete-event decision and control systems. Based on the process status, the set of active control and supervisory algorithm is chosen. The underlying concept of the proposed reconfigurable hybrid decision and control is described along with its implementation on the feedwater system at the Experimental Breeder Reactor II of the Argonne National Laboratory. The performance of the hybrid system to accommodate component failures is then evaluated in an in-plant experiment. >

Robust Optimal Control of Nuclear Reactors and Power Plants

The state feedback assisted control (SFAC) uses the concept of state feedback to modify the demand signal for an embedded classical output feedback controller to achieve an optimal control objective. It has been shown that the SFAC concept can improve the performance of primary coolant temperature control in a nuclear reactor. In this paper, how the embedded classical controller assists a state feedback controller in achieving improved performance and stability robustness, which play an important role in implementing optimal control algorithms for reactor control over a wide range of operations, including possible faulted conditions, is demonstrated. While the state feedback component improves system performance, the classical output feedback component enhances stability robustness.

Hybrid feedforward and feedback controller design for nuclear steam generators over wide range operation using genetic algorithm

In this paper, a simplified model with a lower order is first developed for a nuclear steam generator system and verified against realistic environments. Based on this simplified model, a hybrid multi-input and multi-out (MIMO) control system, consisting of feedforward control (FFC) and feedback control (PEC), is designed for wide range conditions by using the genetic algorithm (GA) technique. The FFC control, obtained by the GA optimization method, injects an a priori command input into the system to achieve an optimal performance for the designed system, while the GA-based FBC control provides the necessary compensation for any disturbances or uncertainties in a real steam generator. The FBC control is an optimal design of a PI-based control system which would be more acceptable for industrial practices and nuclear power plant control system upgrades. The designed hybrid MIMO FFC/FBC control system is first applied to the simplified model and then to a more complicated model with a higher order which is used as a substitute of the real system to test the efficacy of the designed control system. Results from computer simulations show that the designed GA-based hybrid MIMO FFC/FBC control can achieve good responses and robust performances. Hence, it can be considered as a viable alternative to the current control system upgrades.

Multivariable robust control of a power plant deaerator

The design of a robust controller for the deaerator of the Experimental Breeder Reactor-II (EBR-II) that uses the linear quadratic Gaussian with loop transfer recovery (LQG/LTR) procedure is described. At present, classical proportional-integral (PI) controllers are used to control the deaerator. When the operating condition changes, the system is disturbed, or a fault occurs, and the PI controllers may fail to maintain the desired performance. A robust controller that can accommodate system faults and obtain a reasonable behavior for a wide range of model uncertainty was designed. The controller provides the desired performance despite a considerable change in the operating condition, accommodates some of the failures that can occur, and provides the choice of penalizing one variable over another. The design is tested for robustness by varying the system operating conditions and simulating a steam valve failure. The set of nonlinear simulations using the modular modeling system and the advanced continuous simulation language is included. >

Reconfigurable control of power plants using learning automata

A deaerating feedwater heater, equipped with a water level controller and a pressure controller, is used to investigate the feasibility of a reconfigurable control scheme for power plants by incorporating the concept of learning automata. The approach uses stochastic automata to learn the current operating status of the plant by dynamically monitoring the system performance and then switching to the appropriate controller on the basis of the observed performance. Simulation results based on a model of the experimental breeder reactor (EBR-II) at the Argonne National Laboratory site in Idaho are presented to demonstrate the efficacy of the scheme. The results show that it is capable of providing a sufficient margin for the net positive suction head at the feedwater pumps under loss of steam flow into the deaerator. Under similar circumstances, the existing controller in the deaerator would be incapable of maintaining the pressure and its decay rate within the safe margins, and so would oblige the plant operator to take additional measures to protect the feedwater pumps.<<ETX>>

Anomaly Detection in Nuclear Power Plants via Symbolic Dynamic Filtering

Tools of sensor-data-driven anomaly detection facilitate condition monitoring of dynamical systems especially if the physics-based models are either inadequate or unavailable. Along this line, symbolic dynamic filtering (SDF) has been reported in literature as a real-time data-driven tool of feature extraction for pattern identification from sensor time series. However, an inherent difficulty for a data-driven tool is that the quality of detection may drastically suffer in the event of sensor degradation. This paper proposes an anomaly detection algorithm for condition monitoring of nuclear power plants, where symbolic feature extraction and the associated pattern classification are optimized by appropriate partitioning of (possibly noise-contaminated) sensor time series. In this process, the system anomaly signatures are identified by masking the sensor degradation signatures. The proposed anomaly detection methodology is validated on the International Reactor Innovative & Secure (IRIS) simulator of nuclear power plants, and its performance is evaluated by comparison with that of principal component analysis (PCA).