基于SEQ2SEQ与ARIMA组合预测模型的小型模块化压水堆瞬态运行预测技术

成以恒; 李桐; 谭思超; 王博; 田瑞峰; 何正熙; 沈继红

doi:10.13832/j.jnpe.2024.070066

基于SEQ2SEQ与ARIMA组合预测模型的小型模块化压水堆瞬态运行预测技术

doi: 10.13832/j.jnpe.2024.070066

成以恒^{1, 2, 3,},
李桐^{1, 2, 3},
谭思超^{1, 2, 3, ,},
王博^{1, 2, 3},
田瑞峰^{1, 2, 3},
何正熙⁴,
沈继红⁵

1.
哈尔滨工程大学船舶热能动力全国重点实验室，哈尔滨，150001
2.
哈尔滨工程大学黑龙江省核动力装置性能与设备重点实验室，哈尔滨，150001
3.
哈尔滨工程大学核安全与先进核能技术工信部重点实验室，哈尔滨，150001
4.
中国核动力研究设计院，成都，610213
5.
哈尔滨工程大学数学科学学院，哈尔滨，150001

基金项目: 中核集团领创科研项目（CNNC-LCKY-202251）

详细信息

作者简介:
成以恒（2000—），男，硕士研究生，现主要从事小型模块化压水堆与人工智能方面的研究，E-mail: chengyiheng@me.com

通讯作者:
谭思超，E-mail: tansichao@hrbeu.edu.cn

中图分类号: TL333
计量
- 文章访问数: 14
- HTML全文浏览量: 6
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2024-07-31
- 修回日期: 2024-08-31
- 刊出日期: 2025-08-15

Prediction Technology for Transient Operations of Small Modular PWR Based on SEQ2SEQ and ARIMA Hybrid Prediction Model

Cheng Yiheng^{1, 2, 3
,},
Li Tong^{1, 2, 3},
Tan Sichao^{1, 2, 3
, ,},
Wang Bo^{1, 2, 3},
Tian Ruifeng^{1, 2, 3},
He Zhengxi⁴,
Shen Jihong⁵

1.
State Key Laboratory of Ship Thermal Power, Harbin Engineering University, Harbin, 150001, China
2.
Heilongjiang Provincial Key Laboratory of Nuclear Power System & Equipment, Harbin Engineering University, Harbin, 150001, China
3.
Key Laboratory of Nuclear Safety and Advanced Nuclear Energy Technology, Ministry of Industry and Information Technology, Harbin Engineering University, Harbin, 150001, China
4.
Nuclear Power Institute of China, Chengdu, 610213, China
5.
School of Mathematical Sciences,Harbin Engineering University, Harbin, 150001, China

摘要

摘要: 为确保海洋条件下反应堆运行的安全可靠运行，提升海洋条件下的热工运行参数长期预测准确性，本文基于IP200的海洋条件下小型模块化压水堆一维仿真模型的热工运行数据，提出序列到序列（SEQ2SEQ）与自回归差分移动平均模型（ARIMA）的组合预测模型，首先利用ARIMA进行数据的特征提取，随后利用SEQ2SEQ预测振荡值。反应堆在海洋条件下运行时易造成系统内部液面的晃荡，进而导致其他运行参数发生波动。对稳压器压力、冷却剂流量、蒸汽发生器蒸汽出口流量三种不同振荡特征的热工运行参数的预测结果表明：较单独使用ARIMA、SEQ2SEQ模型与传统长短期记忆网络（LSTM）模型相比，预测精度提升约一个数量级。本研究提出的ARIMA和SEQ2SEQ组合预测模型具有计算速度快、预测精度高的特点，为海洋条件下小型模块化压水堆的潜在故障预测提供了一种有效方法。
Abstract: To ensure the safe and reliable operation of reactors under ocean conditions, the long-term prediction accuracy of thermal operation parameters under ocean conditions is improved. Based on the thermal operation data of the one-dimensional simulation model of a small modular PWR IP200 under ocean conditions, this study proposes a prediction model combining sequence to sequence (SEQ2SEQ) and autoregressive integrated moving average (ARIMA). First, ARIMA is used to extract the features of the data, and then SEQ2SEQ is used to predict the oscillation value. When the reactor is operating under ocean conditions, it is easy to cause the sloshing of the liquid level inside the system, which in turn causes oscillation in other operating parameters. The prediction results of three thermal operation parameters with different oscillation characteristics, namely, pressurizer pressure, coolant flow, and steam generator steam outlet flow, show that the prediction accuracy is improved by about one order of magnitude, compared with that of using ARIMA model, SEQ2SEQ model and traditional Long Short-Term Memory (LSTM) model alone. The prediction model combining ARIMA and SEQ2SEQ proposed in this study has features of fast calculation speed and high prediction accuracy, which provides an effective method for the prediction of potential failures of small modular PWR under ocean condition.

HTML全文

图 1 系统节点划分示意图

Figure 1. Schematic Diagram of System Node Division

下载: 全尺寸图片幻灯片

图 2 归一化稳压器压力

Figure 2. Normalized Pressurizer Pressure

下载: 全尺寸图片幻灯片

图 3 SEQ2SEQ算法结构

Figure 3. SEQ2SEQ Algorithm Structure

下载: 全尺寸图片幻灯片

图 4 组合模型算法结构

Figure 4. Hybrid Model Structure

下载: 全尺寸图片幻灯片

图 5 归一化冷却剂流量预测情况

Figure 5. Normalized Coolant Flow Rate Prediction

下载: 全尺寸图片幻灯片

图 6 归一化稳压器压力预测情况

Figure 6. Normalized Pressurizer Pressure Prediction

下载: 全尺寸图片幻灯片

图 7 引入随机噪声归一化数据预测情况

Figure 7. Normalized Prediction with Noise Data Introduced

下载: 全尺寸图片幻灯片

表 1 反应堆基本热工运行参数

Table 1. Basic Thermal Operation Parameters of the Reactor

参数名	参数值
一回路压力/MPa	15.5
冷却剂流量/(kg·s⁻¹)	809
冷却剂入口温度/K	555
冷却剂出口温度/K	600
反应堆加热功率/MW	220
二回路压力/MPa	3.0
二回路给水流量/(kg·s⁻¹)	83.7
二回路给水温度/K	373
出口蒸汽温度/K	554

下载: 导出CSV

表 2 SEQ2SEQ模型超参数

Table 2. Hyper-parameters of SEQ2SEQ Model

超参数	设置情况
LSTM层数	2
初始学习率	0.005
输入序列长度	Study Samples
输出序列长度	Prediction Length
迭代次数	200（组合预测）/800（单独预测）
目标函数	均方误差（MSE）
训练优化算法	SGD

下载: 导出CSV

表 3 冷却剂流量预测MSE

Table 3. Prediction MSE of Coolant Flow Rate

预测时长/s	各模型的MSE
预测时长/s	LSTM	ARIMA	SEQ2SEQ	ARIMA与SEQ2SEQ组合
10	0.0055	0.0030	0.0030	0.0013
30	0.0068	0.0052	0.0051	0.0050
50	0.0085	0.0079	0.0079	0.0078

下载: 导出CSV

表 4 一回路压力预测MSE

Table 4. Prediction MSE of Primary Circuit Pressure

预测时长/s	各模型的MSE
预测时长/s	LSTM	ARIMA	SEQ2SEQ	ARIMA与SEQ2SEQ组合
10	0.0056	0.0043	0.0041	1.8×10^–5
30	0.0044	0.0064	0.0031	9.9×10^–5
50	0.0038	0.0084	0.0022	1.4×10^–4

下载: 导出CSV

表 5 蒸汽发生器出口蒸汽流量引入噪声数据预测MSE

Table 5. Prediction MSE of Steam Outlet Flow Rate with Noise Data Introduced

预测时长/s	各模型的MSE
预测时长/s	LSTM	ARIMA	SEQ2SEQ	ARIMA与SEQ2SEQ组合
10	0.0084	0.025	0.014	0.0025
30	0.0180	0.038	0.021	0.0088
50	0.0280	0.051	0.026	0.0130

下载: 导出CSV

参考文献(12)

[1]	WANG C Y, ZHANG H N, LI X B, et al. A data-driven online prediction method for surface-deformed liquid level in vessels under ocean conditions[J]. Nuclear Engineering and Design, 2023, 414: 112571. doi: 10.1016/j.nucengdes.2023.112571
[2]	KO Y, SEO H, CHO H K, et al. SBLOCA analysis of a floating nuclear reactor under ocean conditions using MARS-KS moving reactor model[J]. Progress in Nuclear Energy, 2024, 168: 105017. doi: 10.1016/j.pnucene.2023.105017
[3]	魏天一,张彪,李东阳,等. 海洋条件下小型堆稳压器液位智能预测研究[J]. 哈尔滨工程大学学报,2023, 44(9): 1590-1596. doi: 10.11990/jheu.202201031
[4]	KIM S, KO W, NAM H, et al. Statistical model for forecasting uranium prices to estimate the nuclear fuel cycle cost[J]. Nuclear Engineering and Technology, 2017, 49(5): 1063-1070. doi: 10.1016/j.net.2017.05.007
[5]	朱少民,夏虹,吕新知,等. 基于ARIMA和LSTM组合模型的核电厂主泵状态预测[J]. 核动力工程,2022, 43(2): 246-253.
[6]	侯秦脉,朱伟,邹象,等. 基于ARIMA-LSTM模型的核电厂运行事件变化分析及预测[J]. 核技术,2022, 45(12): 120602. doi: 10.11889/j.0253-3219.2022.hjs.45.120602
[7]	XU S W, WANG Y J, XU X L, et al. A multi-step wind power group forecasting seq2seq architecture with spatial–temporal feature fusion and numerical weather prediction correction[J]. Energy, 2024, 291: 130352. doi: 10.1016/j.energy.2024.130352
[8]	ZHOU G Y, HU G F, ZHANG D X, et al. A novel algorithm system for wind power prediction based on RANSAC data screening and Seq2Seq-Attention-BiGRU model[J]. Energy, 2023, 283: 128986. doi: 10.1016/j.energy.2023.128986
[9]	STEFENON S F, SEMAN L O, AQUINO L S, et al. Wavelet-Seq2Seq-LSTM with attention for time series forecasting of level of dams in hydroelectric power plants[J]. Energy, 2023, 274: 127350. doi: 10.1016/j.energy.2023.127350
[10]	ZHANG Y, LI Y T, ZHANG G Y. Short-term wind power forecasting approach based on Seq2Seq model using NWP data[J]. Energy, 2020, 213: 118371. doi: 10.1016/j.energy.2020.118371
[11]	刘建阁,彭敏俊,蒋立国. 多用途一体化轻水堆初步设计方案和安全分析[J]. 原子能科学技术,2009, 43(S2): 210-214.
[12]	程坤,谭思超,陈莹莹,等. 海洋条件反应堆热工水力系统分析程序开发及验证[J]. 哈尔滨工程大学学报,2017, 38(8): 1223-1230.