Autonomous Nuclear Power Plants with Artificial Intelligence

Preface
Acknowledgements
Contents
Acronyms
1 Introduction
	1.1 Background
	1.2 A Framework of Autonomous NPPs
	References
2 Artificial Intelligence and Methods
	2.1 Definitions of AI, Machine Learning, and Deep Learning
		2.1.1 AI
		2.1.2 ML
		2.1.3 DL
	2.2 Classification of ML Methods Based on Learning Type
		2.2.1 Supervised Learning
		2.2.2 Unsupervised Learning
		2.2.3 Reinforcement Learning (RL)
	2.3 Overview of Artificial Neural Networks (ANNs)
		2.3.1 History of ANNs
		2.3.2 Overview of ANNs
	2.4 ANN Algorithms
		2.4.1 Convolutional Neural Networks (CNNs)
		2.4.2 Recurrent Neural Networks (RNNs), Long Short-Term Memory Networks (LSTMs), and Gated Recurrent Units (GRUs)
		2.4.3 Variational Autoencoders (VAEs)
		2.4.4 Graph Neural Networks (GNNs)
		2.4.5 Generative Adversarial Networks (GANs)
	2.5 Model-Based and Data-Based Approaches
	References
3 Signal Validation
	3.1 Sensor Fault Detection Through Supervised Learning
		3.1.1 Sensor Fault Detection System Framework with Supervised Learning
		3.1.2 Case Study
	3.2 Signal Validation Through Unsupervised Learning
		3.2.1 Signal Behaviour in an Emergency Situation
		3.2.2 Signal Validation Algorithm Through Unsupervised Learning for an Emergency Situation
		3.2.3 Validation
	3.3 Signal Generation with a GAN
		3.3.1 GAN
		3.3.2 GAN-Based Signal Reconstruction Method
		3.3.3 Experiments
	References
4 Diagnosis
	4.1 Diagnosis of Abnormal Situations with a CNN
		4.1.1 Raw Data Generation
		4.1.2 Data Transformation
		4.1.3 Structure of the CNN Model
		4.1.4 Performance Evaluation Metrics
		4.1.5 Experimental Settings
		4.1.6 Results
	4.2 Diagnosis of Abnormal Situations with a GRU
		4.2.1 Characteristics of Abnormal Operation Data
		4.2.2 PCA
		4.2.3 GRU
		4.2.4 Two-Stage Model Using GRU
		4.2.5 Experimental Settings
		4.2.6 Results
	4.3 Diagnosis of Abnormal Situations with an LSTM and VAE
		4.3.1 Methods
		4.3.2 Diagnostic Algorithm for Abnormal Situations with LSTM and VAE
		4.3.3 Implementation
	4.4 Sensor Fault-Tolerant Accident Diagnosis
		4.4.1 Sensor Fault-Tolerant Diagnosis System Framework
		4.4.2 Comparison Results
		4.4.3 Considerations for Optimal Sensor Fault Mitigation
	4.5 Diagnosis of Multiple Accidents with a GNN
		4.5.1 GNN
		4.5.2 GNN-Based Diagnosis Algorithm Representing System Configuration
		4.5.3 Experiments
	4.6 Interpretable Diagnosis with Explainable AI
		4.6.1 Need for Interpretable Diagnosis
		4.6.2 Explainable AI
		4.6.3 Examples of Explanation Techniques
		4.6.4 Application to an Abnormal Event Diagnosis Model
	References
5 Prediction
	5.1 Real-Time Parameter Prediction
		5.1.1 Multi-step Prediction Strategies
		5.1.2 Plant Parameter Prediction Model with Multi-step Prediction Strategies
		5.1.3 Case Study with Data from an NPP Simulator
		5.1.4 Operator Support System with Prediction
	References
6 Control
	6.1 Autonomous Control for Normal and Emergency Operations with RL
		6.1.1 Case Study 1: Power-Increase Operation
		6.1.2 Case Study 2: Emergency Operation
	References
7 Monitoring
	7.1 Operation Validation System Through Prediction
		7.1.1 CIA System Framework
		7.1.2 Step 1 Filtering: PCC Module
		7.1.3 Step 2 Filtering: COSIE Module
		7.1.4 CIA System Prototype
		7.1.5 Case Study
		7.1.6 Summary and Scalability of the Operation Validation System
	7.2 Technical Specification Monitoring System
		7.2.1 Identification of Functional Requirements
		7.2.2 Conceptual Design of the TSMS
		7.2.3 Implementation of the TSMS
		7.2.4 TSMS Prototype
	References
8 Human–Autonomous System Interface and Decision-Making
	8.1 Human–Autonomous System Interface
	8.2 Decision-Making
	References
9 Conclusion
Index