東華大學圖書館 |

Design, Control, and Economic Dispatch of a Nuclear-Solar Integrated Energy System /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Design, Control, and Economic Dispatch of a Nuclear-Solar Integrated Energy System // Aidan Rigby.
Author:	Rigby, Aidan,
Description:	1 electronic resource (187 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-11, Section: B.
Contained By:	Dissertations Abstracts International86-11B.
Subject:	Nuclear engineering. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=32037877
ISBN:	9798314863695

Design, Control, and Economic Dispatch of a Nuclear-Solar Integrated Energy System /
Rigby, Aidan,

Design, Control, and Economic Dispatch of a Nuclear-Solar Integrated Energy System /Aidan Rigby. - 1 electronic resource (187 pages)

Source: Dissertations Abstracts International, Volume: 86-11, Section: B.

Nuclear integrated energy systems (NIES) offer a route to improved operational flexibility enhancing their value to future energy grids dominated by variable sources. However, due to the complexity involved in modeling and designing nuclear systems, finding flexible dispatch strategies that are both economically and safely physically achievable is extremely complex. This research employs deep reinforcement learning (DRL) and dynamic system modeling to develop the methodology for finding safe and profitable dispatch strategies using a novel solar nuclear hybrid design as a case study. The thesis starts by looking at the case study design of a nuclear solar hybrid reactor. Integrating solar-generated heat from parabolic troughs into the feedwater line allows the plant to alleviate turbine bled steam and transiently power boost 15% above nominal power for a nominal small modular reactor cycle. The research presents a parametric study of the design of such a system and provides a full system dynamic model written in the Modelica language to analyze the dispatch in transient load following operation. The control of the system is presented, and the transient analysis is shown to help inform the sizing of the concrete storage and parabolic trough arrays. The trade-off in design between improved degrees of power boosting and system nominal efficiency is investigated with the work suggesting that higher steam generator entry temperatures offer improved opportunities of up to 42% for power boosting flexibility.To enable dispatch analysis with complex methods of optimization, including machine learning, there is a significant need for more weather data than is available. One method to solve this is using synthetic data models such as the auto-regressive moving-average (ARMA) model which has been frequently utilized to create such data. Chapter 3 looks at extending the ARMA algorithm to generate solar components through the use of clearsky detrending, maintaining vector relationships, and leveraging physical relationships. The method for the creation of entirely synthetic weather data files including key weather variables for energy system analysis is presented. Furthermore, a detailed comparison of energy system simulations utilizing both real and synthetic data is made using NREL's System Advisor Model. Whilst good agreement is made for the solar variables, and other weather variables, ARMA methods often fail to capture the standard deviation and skew of annual weather distributions. Vector-ARMA is shown to maintain correlations between variables and thus generate data sets that perform similarly in energy system design. It is shown that the ARMA method fails to preserve day-to-day correlations in weather variables and thus over-predicts optimal energy storage by 21% for a residential solar application.Chapter 4 aims to tackle the control problems exposed in Chapter 2 directly. Current nuclear control systems are typically classical controllers that are often based on PID control. Chapter 4 presents a method of augmenting the existing PID control for difficult transient operations in nuclear power plants by using a reinforcement learning agent derived feedforward signal applied in real time. The feedforward agents are trained on a test thermal load follow problem and are designed to improve steam generator outlet temperature control at a range of fast load follow scenarios covering ramp rates from 9 - 15 %/min. Several reinforcement learning algorithms were initially investigated for the training of the feedforward agents with deep Q-learning (DQN) and proximal policy optimization (PPO) networks, which were found to be the most promising. The DQN controllers utilized discrete actions giving them a better disturbance rejection at steady state but inconsistent response to initial temperature deviations. In contrast, PPO-trained agents taking continuous actions except for a dead zone around zero were shown to have the best combination of high disturbance rejection at steady state and good tracking of the desired value of temperature. The ability of the PPO agent to make real-time decisions was demonstrated with the average time of decision-making found to be on the order of 1 millisecond. The fault properties of the controller under the loss of the reinforcement learning agent feedforward signal were also examined. The controller showed strong performance in situations of "non-signal" faults but was less good at handling "stuck at" faults where the feedforward signal remains at a set value. In both cases, however, the PID was able to successfully maintain stability eventually returning the system to a steady state.Having demonstrated using DRL with dynamic models in the loop allows the culmination of the work in Chapter 5. Previous NIES dispatch methods have failed to close the gap between economic dispatch and physically achievable results validated with dynamic modeling. It is imperative to develop methods that allow the system to remain within the desired NIES operating conditions and perform this based on realistic limited forecasted information. Chapter 5 employs the same deep reinforcement learning (DRL) algorithms as Chapter 4 but now integrates this with the full NIES dynamic system model from Chapter 2. A DRL agent is trained to find a safe and profitable dispatch strategy using correlated synthetic weather and pricing data from Chapter 2. The DRL agent is shown to find a novel dispatch strategy that manages both power ramping and power levels while respecting operational limits. This DRL-based dispatch is compared to other dispatching strategies including an optimal design solution from mixed integer linear programming (MILP). It is found that incorporating the physics of such a tightly coupled NIES limits the profitability of the MILP-based dispatch strategy. As a result, the MILP solution overestimates the design's generated revenue. In contrast, DRL significantly reduces the number of breaches of safe operational conditions by 80% compared to other approaches for energy arbitrage while maintaining profitability. The key takeaway from this thesis is the methodology that proved DRL, utilizing dynamic models, was able to give a new potential route to obtaining dispatch results. This bridges the gap between dynamic system design analysis and economic assessment. It is hoped the new methodology the thesis has demonstrated on the nuclear solar hybrid design paves the way for a more detailed assessment of NIES profitability. The dispatch agent could also be used to aid operator decisions on future NIES projects. The controller design and analysis of the synthetic data algorithms will aid the nuclear industry in ensuring reliable and profitable grids in the future.

English

ISBN: 9798314863695Subjects--Topical Terms:

595435
Nuclear engineering.
Subjects--Index Terms:

Deep reinforcement learning

Design, Control, and Economic Dispatch of a Nuclear-Solar Integrated Energy System /
LDR:08365nmm a22004573i 4500 001 2431842
005 20250624130503.5
006 m o d
007 cr|nu||||||||
008 261011s2025 miu||||||m |||||||eng d
020 $a 9798314863695
035 $a (MiAaPQD)AAI32037877
035 $a AAI32037877
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Rigby, Aidan, $e author. $3 3820838
245 1 0 $a Design, Control, and Economic Dispatch of a Nuclear-Solar Integrated Energy System / $c Aidan Rigby.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2025
300 $a 1 electronic resource (187 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-11, Section: B.
500 $a Advisors: Lindley, Benjamin Committee members: Wagner, Michael; Wilson, Paul; Pacheco Duarte, Juliana; Mikkelson, Daniel.
502 $b Ph.D. $c The University of Wisconsin - Madison $d 2025.
520 $a Nuclear integrated energy systems (NIES) offer a route to improved operational flexibility enhancing their value to future energy grids dominated by variable sources. However, due to the complexity involved in modeling and designing nuclear systems, finding flexible dispatch strategies that are both economically and safely physically achievable is extremely complex. This research employs deep reinforcement learning (DRL) and dynamic system modeling to develop the methodology for finding safe and profitable dispatch strategies using a novel solar nuclear hybrid design as a case study. The thesis starts by looking at the case study design of a nuclear solar hybrid reactor. Integrating solar-generated heat from parabolic troughs into the feedwater line allows the plant to alleviate turbine bled steam and transiently power boost 15% above nominal power for a nominal small modular reactor cycle. The research presents a parametric study of the design of such a system and provides a full system dynamic model written in the Modelica language to analyze the dispatch in transient load following operation. The control of the system is presented, and the transient analysis is shown to help inform the sizing of the concrete storage and parabolic trough arrays. The trade-off in design between improved degrees of power boosting and system nominal efficiency is investigated with the work suggesting that higher steam generator entry temperatures offer improved opportunities of up to 42% for power boosting flexibility.To enable dispatch analysis with complex methods of optimization, including machine learning, there is a significant need for more weather data than is available. One method to solve this is using synthetic data models such as the auto-regressive moving-average (ARMA) model which has been frequently utilized to create such data. Chapter 3 looks at extending the ARMA algorithm to generate solar components through the use of clearsky detrending, maintaining vector relationships, and leveraging physical relationships. The method for the creation of entirely synthetic weather data files including key weather variables for energy system analysis is presented. Furthermore, a detailed comparison of energy system simulations utilizing both real and synthetic data is made using NREL's System Advisor Model. Whilst good agreement is made for the solar variables, and other weather variables, ARMA methods often fail to capture the standard deviation and skew of annual weather distributions. Vector-ARMA is shown to maintain correlations between variables and thus generate data sets that perform similarly in energy system design. It is shown that the ARMA method fails to preserve day-to-day correlations in weather variables and thus over-predicts optimal energy storage by 21% for a residential solar application.Chapter 4 aims to tackle the control problems exposed in Chapter 2 directly. Current nuclear control systems are typically classical controllers that are often based on PID control. Chapter 4 presents a method of augmenting the existing PID control for difficult transient operations in nuclear power plants by using a reinforcement learning agent derived feedforward signal applied in real time. The feedforward agents are trained on a test thermal load follow problem and are designed to improve steam generator outlet temperature control at a range of fast load follow scenarios covering ramp rates from 9 - 15 %/min. Several reinforcement learning algorithms were initially investigated for the training of the feedforward agents with deep Q-learning (DQN) and proximal policy optimization (PPO) networks, which were found to be the most promising. The DQN controllers utilized discrete actions giving them a better disturbance rejection at steady state but inconsistent response to initial temperature deviations. In contrast, PPO-trained agents taking continuous actions except for a dead zone around zero were shown to have the best combination of high disturbance rejection at steady state and good tracking of the desired value of temperature. The ability of the PPO agent to make real-time decisions was demonstrated with the average time of decision-making found to be on the order of 1 millisecond. The fault properties of the controller under the loss of the reinforcement learning agent feedforward signal were also examined. The controller showed strong performance in situations of "non-signal" faults but was less good at handling "stuck at" faults where the feedforward signal remains at a set value. In both cases, however, the PID was able to successfully maintain stability eventually returning the system to a steady state.Having demonstrated using DRL with dynamic models in the loop allows the culmination of the work in Chapter 5. Previous NIES dispatch methods have failed to close the gap between economic dispatch and physically achievable results validated with dynamic modeling. It is imperative to develop methods that allow the system to remain within the desired NIES operating conditions and perform this based on realistic limited forecasted information. Chapter 5 employs the same deep reinforcement learning (DRL) algorithms as Chapter 4 but now integrates this with the full NIES dynamic system model from Chapter 2. A DRL agent is trained to find a safe and profitable dispatch strategy using correlated synthetic weather and pricing data from Chapter 2. The DRL agent is shown to find a novel dispatch strategy that manages both power ramping and power levels while respecting operational limits. This DRL-based dispatch is compared to other dispatching strategies including an optimal design solution from mixed integer linear programming (MILP). It is found that incorporating the physics of such a tightly coupled NIES limits the profitability of the MILP-based dispatch strategy. As a result, the MILP solution overestimates the design's generated revenue. In contrast, DRL significantly reduces the number of breaches of safe operational conditions by 80% compared to other approaches for energy arbitrage while maintaining profitability. The key takeaway from this thesis is the methodology that proved DRL, utilizing dynamic models, was able to give a new potential route to obtaining dispatch results. This bridges the gap between dynamic system design analysis and economic assessment. It is hoped the new methodology the thesis has demonstrated on the nuclear solar hybrid design paves the way for a more detailed assessment of NIES profitability. The dispatch agent could also be used to aid operator decisions on future NIES projects. The controller design and analysis of the synthetic data algorithms will aid the nuclear industry in ensuring reliable and profitable grids in the future.
546 $a English
590 $a School code: 0262
650 4 $a Nuclear engineering. $3 595435
650 4 $a Mechanical engineering. $3 649730
650 4 $a Computer engineering. $3 621879
650 4 $a Alternative energy. $3 3436775
653 $a Deep reinforcement learning
653 $a Dispatch optimization
653 $a Dynamic modelling
653 $a Integrated energy systems
653 $a Nuclear power
653 $a Synthetic data generation
690 $a 0552
690 $a 0548
690 $a 0464
690 $a 0363
710 2 $a The University of Wisconsin - Madison. $b Nuclear Engineering & Engineering Physics. $e degree granting institution. $3 3820839
720 1 $a Lindley, Benjamin $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-11B.
790 $a 0262
791 $a Ph.D.
792 $a 2025
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=32037877