Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
5.01: Operation and Performance of Nuclear Power Plants
3:30pm - 5:00pm
Session Chair: Gyunyoung Heo, Kyung Hee University, Korea, Republic of (South Korea) Session Chair: Spencer Brown, FANR, United Arab Emirates
Optimizing NPP Staffing
Charles T. Goodnight
Goodnight Consulting, Inc.
Optimizing NPP staffing does not mean that the lowest number of people should be applied. In many cases, low NPP staffing levels have led to poor operational performance and regulatory compliance challenges. This paper will discuss three sets of staffing approaches, and their consequences. The first is a lean approach, the second a moderate approach, and the third a highly staffed approach.
From the study of NPP staffing from more than three decades, specific examples will be provided for each of these approaches, with resulting consequences. For example, lean staffing approaches leave smaller margins to compensate for operating condition abnormalities, technical challenges, or external events. Highly staffed approaches typically devolve into situations of loss of direct accountability and the ineffective division of labor. Moderate staffing levels are not necessarily in the statistical middle of the other approaches, and must be carefully evaluated.
Finally, this paper will discuss the staffing approaches seen in the Top Performers in the US nuclear power industry. US Top Performers are those NPPs with a combination of three major performance characteristics: 1) lower than industry average staff/MWe, 2) above industry average 3-year capacity factor, and 3) above average 3-year regulatory reviews (US NRC ROP scores). These Top Performers have balanced labor costs (i.e., the staffing approach relative to the plant output), plant performance, and regulatory compliance for a sustained period. Top Performer NPPs have a number of common staffing approaches which can be applied by existing and planned NPPs to help optimize costs, performance, and compliance. These common, effective approaches will be described.
Optimized NPP staffing levels can be recognized by overall excellence in performance. When costs, plant performance, and regulatory compliance conditions are simultaneously balanced, NPP staffing will be properly balanced. More than thirty years of industry experience shows how this can be achieved.
Data-Driven Uncertainty-Aware Nuclear Power Plant Sensor Modeling
Matthew Stanley, Michael Bowman, Xiaoli Zhang, Jeffrey King
Colorado School of Mines
The monitoring of a nuclear power plant is important to ensure it is operating within the conditions specified in the reactor’s licensing and safety documentation. Data collection from sensors is critical for monitoring the status of a reactor, for providing real-time feedback signals in closed-loop control, and for predicting a reactor’s future conditions. However, the challenging operational environment in a nuclear reactor core often results in high sensor uncertainties with significant amounts of noise. To overcome these environmental challenges, an uncertainty-aware data-driven model is developed to filter the noise and reduce the sensor uncertainties. The methods were tested on trials from the Transient Reactor Test Facility (TREAT); however, this approach can be generalized such that it can be used in different reactors with various sensors. The resulting model significantly improves the sensor output from during steady-state and pulsed operations.
Adaptive Neural Network Algorithm for Power Control in Nuclear Power Plants
Husam Fayiz Al Masri
National Research Nuclear University MEPhI
The aim of this paper is to design, test and evaluate a prototype of an adaptive neural network algorithm for the power controlling system of a nuclear power plant. The task of power control in nuclear reactors is one of the fundamental tasks in this field. Therefore, researches are constantly conducted to ameliorate the power reactor control process. Currently, in the Department of Automation in the National Research Nuclear University (NRNU) MEPhI, numerous studies are utilizing various methodologies of artificial intelligence (expert systems, neural networks, fuzzy systems and genetic algorithms) to enhance the performance, safety, efficiency and reliability of nuclear power plants. In particular, a study of an adaptive artificial intelligent power regulator in the control systems of nuclear power reactors is being undertaken to enhance performance and to minimize the output error of the Automatic Power Controller (APC) on the grounds of a multi-functional computer analyzer (simulator) of the Water-Water Energetic Reactor known as Vodo-Vodyanoi Energetichesky Reaktor (VVER) in Russian. In this paper, a block diagram of an adaptive reactor power controller was built on the basis of an intelligent control algorithm. When implementing intelligent neural network principles, it is possible to improve the quality and dynamic of any control system in accordance with the principles of adaptive control. It is common knowledge that an adaptive control system permits adjusting the controller's parameters according to the transitions in the characteristics of the control object or external disturbances. In this project, it is demonstrated that the propitious options for an automatic power controller in nuclear power plants is a control system constructed on intelligent neural network algorithms.
Comparison of detection time for abnormality diagnosis model using two-stage gated recurrent units
Jae Min Kim, Junyong Bae, Seung Jun Lee
Ulsan National Institute of Science and Technology
Abnormal conditions occurring in nuclear power plants can be divided into abnormal and emergency conditions. Criterion that separates the two conditions is whether or not a reactor trip occurs. In the event of an abnormal event, recovery is possible to maintain normal operation before without a reactor trip until maintenance period. However, abnormal events may also cause reactor trip if the situation is not successfully mitigated in the proper amount of time and becomes too severe to be handled with abnormal operating procedures (AOPs). In the case of the Advanced Power Reactor 1400, operators are required to select an appropriate AOP from among 82 AOPs with 224 total sub-procedures. This large number of AOPs and corresponding plant parameters for diagnosis make identifying the appropriate AOP difficult. To support this task for operators, an abnormality detection model is suggested in the previous research. In this paper, the time taken for the diagnostic model to select an appropriate AOP was compared with the time taken for generating the sufficient number of alarms that required for the operators to determine the AOP. The model has two stages using the gated recurrent units to determine an AOP and detailed cause of the corresponding event, respectively. Principal components analysis was conducted on the entire dataset for the main algorithm and the partial datasets for the sub-algorithms to enhance accuracy of the model. By analyzing operating parameters directly, operators can be informed in advance of possible abnormalities depending on the current plant conditions. This will help operators in selecting AOPs in abnormal situations and further contribute to improving plant safety.
Nuclear Operator Performance Estimation by Facial Expression Analysis Using Machine Learning Based Prediction System
Cho Woo Jo, Young Ho Chae, Poong Hyun Seong
Korea Advanced Institute of Science and Technology
Previous studies have applied various performance estimation methods for managing human error of nuclear operators with the goal of reducing one of the main causes of nuclear accidents, human error. The aim of this research was to study the use of an automatic facial expression analysis system as a performance estimation method, specifically for nuclear operators. We examined whether a machine learning based prediction system can be employed to estimate nuclear operator performance; by analyzing the real time facial expression data; and by developing a machine learning model to estimate human performance from facial expressions. The experiments were conducted in unexpected nuclear accident situations when operators make quick accident diagnosis. The measurements were shown to reliably determine the accident diagnosis performance with an accuracy of 83.3%. In addition, from the statistical analysis of facial expressions, we found that the high error group’s facial expressions were related to engagement more frequently than the low error group. Furthermore, the results demonstrated that facial expression can distinguish the high error from low error group better than conventional surveys do. The value of our research lies in its practicality for predicting performance in real time, without affecting operator performance, which is suitable to nuclear power plant workplace.