机器学习模型在地热开发水温预测中的应用

董珮瑶; 杜利; 赵磊; 包一凡; 尹茂生

doi:10.19509/j.cnki.dzkq.tb20240063

机器学习模型在地热开发水温预测中的应用

doi: 10.19509/j.cnki.dzkq.tb20240063

董珮瑶^{1, 2,},
杜利^{1, 2},
赵磊^{1, 2},
包一凡^{1, 2},
尹茂生^3, ,

1.
深层地热富集机理与高效开发全国重点实验室，北京 100083
2.
中石化新星（北京）新能源研究院有限公司，北京 100083
3.
南方科技大学环境科学与工程学院，广东深圳 518055

基金项目: 国家自然科学基金青年项目（42102284）；中国石化集团科研项目（JP23087；JP23178）

详细信息

作者简介:
董珮瑶：E-mail：dongpeiyao.xxsy@sinopec.com

通讯作者:
E-mail：myin@eitech.edu.cn

计量
- 文章访问数: 343
- PDF下载量: 25
- 被引次数: 0
出版历程
- 收稿日期: 2024-02-23
- 录用日期: 2024-12-17
- 修回日期: 2024-06-24
- 网络出版日期: 2024-10-22

Application of machine learning models to geothermal groundwater temperature prediction

DONG Peiyao^{1, 2
,},
DU Li^{1, 2},
ZHAO Lei^{1, 2},
BAO Yifan^{1, 2},
YIN Maosheng^{3
, ,}

1.
State Key Laboratory of Deep Geothermal Resources, Beijing 100083, China
2.
Sinopec Star (Beijing) New Energy Research Institute Co., Ltd., Beijing 100083, China
3.
Southern University of Science and Technology, Shenzhen Guangdong 518055, China

More Information

Author Bio:
E-mail: dongpeiyao.xxsy@sinopec.com

Corresponding author: E-mail：myin@eitech.edu.cn

摘要

摘要:
地热作为一种清洁能源具有广阔的应用前景，可持续地开发和利用地热资源中地热水的温度评估是重要的研究课题。人工智能技术已成为矿产和油气资源勘探开发研究的热点和前沿方向，然而在地热资源开发方面相关研究和应用较少。剖析了油气资源开发中大数据与人工智能应用的重要价值，对当前地热资源开发中人工智能技术的应用与探索进行了介绍。以陕西咸阳地热田为例，采用长短期记忆（long short-term memory，简称LSTM）神经网络构建了以灌定采模式下单井水温的时间序列模型；采用随机森林和XGBoost算法，建立了多个井地热水温度的预测模型。研究结果表明，建立的机器学习模型在地热水温度预测方面表现优秀，模型准确度均在95%以上，且速度快。该地区地热水温的首要影响因素是取水段顶深，模型验证了渭北断裂带对热储的重要作用。实例应用验证了机器学习模型在解决地热资源开发复杂难题中的优越性，人工智能技术的合理应用能够为地热资源的高效开发和科学降本提质增效提供更多有效的决策依据。
- 地热开发 /
- 机器学习 /
- 模拟 /
- 地下水 /
- 水温预测
Abstract: <p>Geothermal energy as a kind of clean energy has broad application prospects. The temperature assessment of geothermal water in sustainable development and utilization of geothermal resources is an important research topic.</p></sec> <sec><title>Objective
Artificial intelligence technology has become a hot spot and frontier direction in the exploration and development of mineral and oil and gas fields, but in the field of geothermal field development, there are few relevant studies. This paper first analyzes the important value of large data and artificial intelligence application in oil and gas field development, and then introduces the application of artificial intelligence in geothermal field development at present.
Methods
Taking Xianyang geothermal field in Shaanxi province as an example, the single well geothermal water temperature time series model was constructed by using long and short term memory neural network (LSTM) under the predestined production mode. Random forest and XGBoost algorithm were used to predict the groundwater temperature of multiple geothermal wells.
Results
The accuracy of the three models was above 95%, and the running speed is fast. The depth at the top of the water intake section is the primary influencing factor of geothermal water temperature in this area. The model verifies that the fault zone plays an important role in heat storage.
Conclusion
The application of the example verifies the superiority of machine learning in solving complex problems in geothermal field development, and the reasonable application of artificial intelligence technology can provide more effective decision-making basis for the efficient development of geothermal field and scientific cost reduction, quality improvement and efficiency improvement.
- geothermal development /
- machine learning /
- modeling /
- groundwater /
- hydrothermal prediction

HTML全文

图 1 LSTM神经网络结构原理

A. 一个单元；σ和tanh. 均为激活函数；x_t-1，x_t和x_t+1. 分别为t-1，t和t+1时刻的输入；h_t-1，h_t和h_t+1. 分别为t-1，t和t+1时刻的输出

Figure 1. LSTM neural network structure schematic

下载: 全尺寸图片幻灯片

图 2 陕西咸阳地热田地质构造(a)和地热井分布(b)图

Figure 2. Geological structures (a) and producing well locations (b) of the Xianyang geothermal field, Shaanxi

下载: 全尺寸图片幻灯片

图 3 YGSY1井水温模拟结果和实际观测结果对比

Figure 3. Comparison of groundwater temperature simulation results and observation results in well YGSY1

下载: 全尺寸图片幻灯片

图 4 地热井水温模拟计算结果和实际观测结果对比（随机森林模型）

Figure 4. Comparison between the simulation results of geothermal well groundwater temperature and the observation results (Random Forest Model)

下载: 全尺寸图片幻灯片

图 5 地热井水温模拟计算结果和实际观测结果对比（XGBoost模型）

Figure 5. Comparison between the simulation results of geothermal well groundwater temperature and the observation results (XGBoost Model)

下载: 全尺寸图片幻灯片

图 6 LSTM模型平均绝对误差MAE值随循环迭代次数的变化

Figure 6. Variation of MAE value of LSTM model with the epoch size

下载: 全尺寸图片幻灯片

图 7 根据特征重要性对水温影响因素排序

Figure 7. Rank influencing factors according to feature importance

下载: 全尺寸图片幻灯片

图 8 纬度与中落程流量特征交互SHAP依赖图

灰色区域表示落在该纬度范围内的井样本频数统计，即灰色区域高的部分表示该纬度范围内的井样本多

Figure 8. Feature dependence SHAP plot of the interaction between latitude and medium drawdown flow

下载: 全尺寸图片幻灯片

表 1 咸阳54口地热井试验资料数据内容

Table 1. Content of test data of 54 geothermal wells in Xianyang

序号	特征名称	单位
1	井类型	/
2	经度	°
3	纬度	°
4	井口海拔	m
5	完钻井深	m
6	井斜深	m
7	取水段顶深	m
8	取水段底深	m
9	取水段中深	m
10	静水位	m
11	中落程降深	m
12	中落程流量	m³/h
13	大落程降深	m
14	大落程流量	m³/h
15	小落程降深	m
16	小落程流量	m³/h
17	中落程温度	℃

下载: 导出CSV

表 2 咸阳部分地热井的抽水试验实测数据

Table 2. Pumping test data of some geothermal wells in Xianyang

井名	取水段顶深/m	取水段底深/m	取水段中深/m	静水位/m	中落程降深/m	中落程流量/(m³·h⁻¹)	中落程温度/℃
WR3	2060.9	2724.9	2392.90	12	19	120	82
WR5	1581.5	2961.6	2271.55	−52.5	28	121.03	75
WR6	2003.31	3033.78	2518.55	6	18	130.66	83
WR9	1055.2	1599.13	1327.17	−8	14	102	61
WT2	909.7	1699.93	1304.82	−9	31	120	50

下载: 导出CSV

表 3 LSTM模型参数优选结果

Table 3. LSTM model parameter optimization results

参数	参数取值	优选值
时间步长	5，10，20，40	5
网络层数	1，2	1
隐藏层神经元	128，256	128
学习率	0.01，0.005，0.001	0.001
迭代的次数	50，100，200	200

下载: 导出CSV

表 4 随机森林和XGBoost模型参数优选结果

Table 4. Optimization results of random forest and XGBoost model parameters

模型	参数	参数取值	优选值
随机森林	决策树的个数	40，80，100，200，1000	200
	决策树最大特征数	‘auto’，‘sqrt’，‘log2’	‘auto’
	决策树最大深度	None，10，20，30	10
	叶子节点含有的最少样本	2，3，5	3
	节点可分的最小样本数	9，10，11	9
XGBoost	学习率	0.01，0.05，0.1	0.1
	树的最大深度	2，3，4，5	3
	训练中树的个数	50，100，200	100
	最小叶子节点样本权重和	1，3，5，7	1
	gamma	0，0.1，0.3，0.5	0

下载: 导出CSV

表 5 模型评价指标对比

Table 5. Comparison of model evaluation indexes

评价指标	LSTM	随机森林	XGBoost
决定系数R²	0.9741	0.9558	0.9943
解释方差EV	1.0000	0.9652	0.9960
平均绝对误差MAE	0.0198	0.0216	0.0130
均方根误差MSE	0.0058	0.0017	0.0002
计算时间/s	16.5625	0.1318	0.0928

下载: 导出CSV

参考文献(41)

[1]	汪集旸,胡圣标,庞忠和,等. 中国大陆干热岩地热资源潜力评估[J]. 科技导报,2012,30(32):25-31. doi: 10.3981/j.issn.1000-7857.2012.32.002 WANG J Y,HU S B,PANG Z H,et al. Estimate of geothermal resources potential for hot dry rock in the continental area of China[J]. Science & Technology Review,2012,30(32):25-31. (in Chinese with English abstract doi: 10.3981/j.issn.1000-7857.2012.32.002
[2]	武选民,柏琴,苑惠明,等. 冰岛地热资源开发利用现状[J]. 水文地质工程地质,2007,34(5):129-130. WU X M,BAI Q,YUAN H M,et al. Development and utilization of geothermal resources in Iceland[J]. Hydrogeology & Engineering Geology,2007,34(5):129-130. (in Chinese with English abstract
[3]	姜光政,高堋,饶松,等. 中国大陆地区大地热流数据汇编(第四版)[J]. 地球物理学报,2016,59(8):2892-2910. doi: 10.6038/cjg20160815 JIANG G Z,GAO P,RAO S,et al. Compilation of heat flow data in the continental area of China(4^th edition)[J]. Chinese Journal of Geophysics,2016,59(8):2892-2910. (in Chinese with English abstract doi: 10.6038/cjg20160815
[4]	陈墨香,汪集旸. 中国地热研究的回顾和展望[J]. 地球物理学报,1994,37(增刊1):320-338. CHEN M X,WANG J Y. Review and prospect on geothermal studies in China[J]. Chinese Journal of Geophysics,1994,37(S1):320-338. (in Chinese with English abstract
[5]	文冬光,张二勇,王贵玲,等. 干热岩勘查开发进展及展望[J]. 水文地质工程地质,2023,50(4):1-13. WEN D G,ZHANG E Y,WANG G L,et al. Progress and prospect of hot dry rock exploration and development[J]. Hydrogeology & Engineering Geology,2023,50(4):1-13. (in Chinese with English abstract
[6]	GUO Q H,WANG Y X,LIU W. Hydrogeochemistry and environmental impact of geothermal waters from Yangyi of Tibet,China[J]. Journal of Volcanology and Geothermal Research,2009,180(1):9-20. doi: 10.1016/j.jvolgeores.2008.11.034
[7]	GONG B,LIANG H B,XIN S L,et al. Numerical studies on power generation from co-produced geothermal resources in oil fields and change in reservoir temperature[J]. Renewable Energy,2013,50:722-731. doi: 10.1016/j.renene.2012.07.026
[8]	邹才能,杨智,朱如凯,等. 中国非常规油气勘探开发与理论技术进展[J]. 地质学报,2015,89(6):979-1007. doi: 10.3969/j.issn.0001-5717.2015.06.001 ZOU C N,YANG Z,ZHU R K,et al. Progress in China’s unconventional oil & gas exploration and development and theoretical technologies[J]. Acta Geologica Sinica,2015,89(6):979-1007. (in Chinese with English abstract doi: 10.3969/j.issn.0001-5717.2015.06.001
[9]	张凯,赵兴刚,张黎明,等. 智能油田开发中的大数据及智能优化理论和方法研究现状及展望[J]. 中国石油大学学报(自然科学版),2020,44(4):28-38. doi: 10.3969/j.issn.1673-5005.2020.04.004 ZHANG K,ZHAO X G,ZHANG L M,et al. Current status and prospect for the research and application of big data and intelligent optimization methods in oilfield development[J]. Journal of China University of Petroleum (Edition of Natural Science),2020,44(4):28-38. (in Chinese with English abstract doi: 10.3969/j.issn.1673-5005.2020.04.004
[10]	钟仪华,王淑宁,罗兰,等. 用深度学习挖掘油田开发指标预测模型的知识[J]. 西南石油大学学报(自然科学版),2020,42(6):63-74. ZHONG Y H,WANG S N,LUO L,et al. Knowledge mining for oilfield development index prediction model using deep learning[J]. Journal of Southwest Petroleum University (Science & Technology Edition),2020,42(6):63-74. (in Chinese with English abstract
[11]	王洪亮,穆龙新,时付更,等. 基于循环神经网络的油田特高含水期产量预测方法[J]. 石油勘探与开发,2020,47(5):1009-1015. doi: 10.11698/PED.2020.05.15 WANG H L,MU L X,SHI F G,et al. Production prediction at ultra-high water cut stage via recurrent neural network[J]. Petroleum Exploration and Development,2020,47(5):1009-1015. (in Chinese with English abstract doi: 10.11698/PED.2020.05.15
[12]	谷建伟,周梅,李志涛,等. 基于数据挖掘的长短期记忆网络模型油井产量预测方法[J]. 特种油气藏,2019,26(2):77-81. doi: 10.3969/j.issn.1006-6535.2019.02.013 GU J W,ZHOU M,LI Z T,et al. Oil well production forecast with long-short term memory network model based on data mining[J]. Special Oil & Gas Reservoirs,2019,26(2):77-81. (in Chinese with English abstract doi: 10.3969/j.issn.1006-6535.2019.02.013
[13]	刘想平,汪崎生,汤军. 用神经网络建立自喷井井底流压预测模型[J]. 石油勘探与开发,1997,24(5):92-94. doi: 10.3321/j.issn:1000-0747.1997.05.023 LIU X P,WANG Q S,TANG J. An application of neural network in developing a model for predicting flowing bottomhole pressure of flowing wells[J]. Petroleum Exploration and Development,1997,24(5):92-94. (in Chinese with English abstract doi: 10.3321/j.issn:1000-0747.1997.05.023
[14]	张进平,王煜曦,刘桂宏,等. 通州区地热资源优化开采模式动态研究[J]. 工程力学,2022,39(6):247-256. doi: 10.6052/j.issn.1000-4750.2021.01.0086 ZHANG J P,WANG Y X,LIU G H,et al. Dynamic study on optimal exploitation mode of geothermal resources in Beijing Tongzhou District[J]. Engineering Mechanics,2022,39(6):247-256. (in Chinese with English abstract doi: 10.6052/j.issn.1000-4750.2021.01.0086
[15]	段晓飞,康凤新,吴晓华,等. 基于采灌均衡模拟的砂岩热储合理采灌井距计算方法[J]. 地质科技通报,2024,43(5):170-180. DUAN X F,KANG F X,WU X H,et al. A methodology for determining the optimal well spacing in sandstone geo-thermal reservoirs through production-reinjection equilibrium simulation[J]. Bulletin of Geological Science and Technology,2024,43(5):170-180. (in Chinese with English abstract
[16]	马峰,高俊,王贵玲,等. 雄安新区容城地热田碳酸盐岩热储采灌数值模拟[J]. 吉林大学学报(地球科学版),2023,53(5):1534-1548. MA F,GAO J WANG G L,et al. Numerical simulation of exploitation and reinjection of carbonate geothermal reservoir in Rongcheng geothermal field,Xiongan New Area[J]. Journal of Jilin University (Earth Science Edition),2023,53(5):1534-1548. (in Chinese with English abstract
[17]	赵敬波,刘健,周志超,等. 基于离散裂隙网络模型的地下水流并行模拟方法[J]. 地质科技通报,2023,42(4):55-64. ZHAO J B,LIU J,ZHOU Z C,et al. Parallel groundwater flow simulation method based on a discrete fracture network model[J]. Bulletin of Geological Science and Technology,2023,42(4):55-64. (in Chinese with English abstract
[18]	XU T F,YUAN Y L,JIA X F,et al. Prospects of power generation from an enhanced geothermal system by water circulation through two horizontal wells:A case study in the Gonghe Basin,Qinghai Province,China[J]. Energy,2018,148:196-207. doi: 10.1016/j.energy.2018.01.135
[19]	张娟娟,林建旺,林黎,等. 基于AHP和BP神经网络的深部地热水可持续开发能力评价[J]. 地下水,2008,30(6):46-48. doi: 10.3969/j.issn.1004-1184.2008.06.014 ZHANG J J,LIN J W,LIN L,et al. Evaluation of deep geothermal water sustainable development ability based on AHP and BP neural network[J]. Ground Water,2008,30(6):46-48. (in Chinese with English abstract doi: 10.3969/j.issn.1004-1184.2008.06.014
[20]	董华松,黄文辉. 利用遗传算法优化的小波神经网络实现地热资源预测[J]. 资源与产业,2014,16(3):101-106. DONG H S,HUANG W H. Prediction of geothermal resources by means of wavelet neural network optimized by genetic algorithm[J]. Resources & Industries,2014,16(3):101-106. (in Chinese with English abstract
[21]	赵磊,张海雄,贺婷婷. 基于LSTM神经网络的地热井回灌压力预测方法研究[J]. 当代化工研究,2023(4):119-123. ZHAO L,ZHANG H X,HE T T. Reinjection pressure prediction of geothermal well with LSTM neural network model[J]. Modern Chemical Research,2023(4):119-123. (in Chinese with English abstract
[22]	REN Y Q,KONG Y L,PANG Z H,et al. A comprehensive review of tracer tests in enhanced geothermal systems[J]. Renewable and Sustainable Energy Reviews,2023,182:113393. doi: 10.1016/j.rser.2023.113393
[23]	WU H,FU P C,MORRIS J P,et al. Characterization of flow and transport in a fracture network at the EGS Collab field experiment through stochastic modeling of tracer recovery[J]. Journal of Hydrology,2021,593:125888. doi: 10.1016/j.jhydrol.2020.125888
[24]	HAWKINS A J,FOX D B,KOCH D L,et al. Predictive inverse model for advective heat transfer in a short-circuited fracture:Dimensional analysis,machine learning,and field demonstration[J]. Water Resources Research,2020,56(11):e2020WR027065. doi: 10.1029/2020WR027065
[25]	JIANG G Z,HU S B,SHI Y Z,et al. Terrestrial heat flow of continental China:Updated dataset and tectonic implications[J]. Tectonophysics,2019,753:36-48. doi: 10.1016/j.tecto.2019.01.006
[26]	罗璐,朱霞,何春艳,等. 陕西咸阳地热田地热流体成因研究[J]. 地质论评,2019,65(6):1422-1430. LUO L,ZHU X,HE C Y,et al. Study on the genesis of geothermal fluid in Xianyang geothermal field[J]. Geological Review,2019,65(6):1422-1430. (in Chinese with English abstract
[27]	杨剑锋,乔佩蕊,李永梅,等. 机器学习分类问题及算法研究综述[J]. 统计与决策,2019,35(6):36-40. YANG J F,QIAO P R,LI Y M,et al. A review of machine-learning classification and algorithms[J]. Statistics & Decision,2019,35(6):36-40. (in Chinese with English abstract
[28]	BI Y X,GUAN J W,BELL D. The combination of multiple classifiers using an evidential reasoning approach[J]. Artificial Intelligence,2008,172(15):1731-1751. doi: 10.1016/j.artint.2008.06.002
[29]	CHEN T Q,LI H,YANG Q,et al. General functional matrix factorization using gradient boosting[C]//Anon. Proceedings of the 30^th International Conference on Machine Learning. Atlanta,Georgia,USA:[s. n. ],2013:436-444.
[30]	尹宝才,王文通,王立春. 深度学习研究综述[J]. 北京工业大学学报,2015,41(1):48-59. doi: 10.11936/bjutxb2014100026 YIN B C,WANG W T,WANG L C. Review of deep learning[J]. Journal of Beijing University of Technology,2015,41(1):48-59. (in Chinese with English abstract doi: 10.11936/bjutxb2014100026
[31]	GRAVES A. Supervised sequence labelling with recurrent neural networks[M]. Berlin:Springer,2012.
[32]	LUNDBERG S M,LEE S I. A unified approach to interpreting model predictions[J]. Neural Information Processing Systems,2017:21889700.
[33]	GUIDOTTI R,MONREALE A,RUGGIERI S,et al. A survey of methods for explaining black box models[J]. ACM Computing Surveys,2019,51(5):1-42.
[34]	中华人民共和国国家质量监督检验检疫总局,中华人民共和国建设部. 供水水文地质勘察规范:GB50027-2001[S]. 北京:中国标准出版社,2001. General Administration of Quality Supervision Inspection and Quarantine of the People's Republic of China,Ministry of Construction of the People's Republic China. Standard for hydrogeological investigation of water-supply:GB50027-2001[S]. Beijing:China Standards Press,2001. (in Chinese)
[35]	赵志宏,刘桂宏,王佳铖,等. 城市深层地热能可持续开采多场耦合效应数值模拟研究进展[J]. 煤炭学报,2023,48(3):1126-1138. ZHAO Z H,LIU G H,WANG J C,et al. Coupled multi-field effect on sustainable development of deep geothermal energy in cities[J]. Journal of China Coal Society,2023,48(3):1126-1138. (in Chinese with English abstract
[36]	HUO A D,WANG X F,LYU Y,et al. Simulation research on the reinjection temperature fields of deep geothermal wells based on real-scale experiment[J]. Energy Exploration & Exploitation,2019,37(2):646-662.
[37]	陶晔,杜景林. 基于随机森林的长短期记忆网络气温预测[J]. 计算机工程与设计,2019,40(3):737-743. TAO Y,DU J L. Temperature prediction using long short term memory network based on random forest[J]. Computer Engineering and Design,2019,40(3):737-743. (in Chinese with English abstract
[38]	柯婷婷,黄少鹏,许威,等. 关中盆地沣西地区地热对井采灌开发模式的数值模拟[J]. 第四纪研究,2019,39(5):1252-1263. doi: 10.11928/j.issn.1001-7410.2019.05.17 KE T T,HUANG S P,XU W,et al. Numerical modeling of doublet well system for extracting heat from sandstone geothermal reservoir:A case study of Fengxi area,the Guanzhong Basin,NW China[J]. Quaternary Sciences,2019,39(5):1252-1263. (in Chinese with English abstract doi: 10.11928/j.issn.1001-7410.2019.05.17
[39]	高俊义. 地热温度和生产井深度对岩体水流−传热温度影响机理研究[J]. 地球物理学进展,2020,35(5):1659-1664. doi: 10.6038/pg2020DD0370 GAO J Y. Study on mechanism of the influence of geothermal temperature and production well depth on water flow and heat transfer temperature in rock mass[J]. Progress in Geophysics,2020,35(5):1659-1664. (in Chinese with English abstract doi: 10.6038/pg2020DD0370
[40]	杨国华,李婉露,孟博. 基于机器学习方法的地下水氨氮时空分布规律[J]. 吉林大学学报(地球科学版),2022,52(6):1982-1995. YANG G H,LI W L,MENG B. Spatiotemporal distribution of groundwater ammonia nitrogen based on machine learning methods[J]. Journal of Jilin University (Earth Science Edition),2022,52(6):1982-1995. (in Chinese with English abstract
[41]	肖立志. 机器学习数据驱动与机理模型融合及可解释性问题[J]. 石油物探,2022,61(2):205-212. doi: 10.3969/j.issn.1000-1441.2022.02.002 XIAO L Z. The fusion of data-driven machine learning with mechanism models and interpretability issues[J]. Geophysical Prospecting for Petroleum,2022,61(2):205-212. (in Chinese with English abstract doi: 10.3969/j.issn.1000-1441.2022.02.002