基于ANN-GA集成的ORC混合工质智能筛选与性能优化

doi:10.11949/0438-1157.20251120

摘要/Abstract

摘要：

针对有机朗肯循环（ORC）混合工质筛选耗时长、热物性数据缺失等问题，提出一种基于人工神经网络（ANN）与遗传算法（GA）集成的数据驱动框架，用于ORC系统的快速性能评估与混合工质智能筛选。首先利用Aspen Plus建立严格的热力学模型，采用拉丁超立方采样生成5种纯工质及其二元混合物共1600组样本数据；随后在MATLAB中构建多层前馈ANN模型，实现热效率（ $η$ ）与单位净功（ $W n e t$ ）的高精度预测（R²≥0.999）。最后通过GA结合权重系数法，将多目标优化转化为单目标形式，确定最优工质与最佳操作参数。结果表明，当注重热效率时（ $w 1$ ≤0.6），R123/R601最优（ $η$ =12.76%）；当注重净功时（ $w 1$ >0.6），R601/R245fa最优（ $W n e t$ =61.4 kW），较纯工质提升约26%。该方法显著缩短优化周期，为ORC混合工质设计提供高效工具。

关键词: 有机朗肯循环, 混合工质, 智能筛选, 神经网络, 多目标优化, 计算机模拟, 优化设计

Abstract:

To overcome the challenges of time-consuming screening and incomplete thermophysical property data in selecting mixed working fluids for Organic Rankine Cycle (ORC) systems, this study proposes an integrated data-driven optimization framework combining Artificial Neural Networks (ANN) and Genetic Algorithms (GA). The objective is to enable rapid performance evaluation and intelligent selection of optimal mixed working fluids and operating parameters. A rigorous thermodynamic model of the ORC system was first established in Aspen Plus, and 1,600 datasets were generated using Latin Hypercube Sampling (LHS) for five pure fluids and their binary mixtures. Subsequently, dedicated multilayer feedforward ANN models were developed in MATLAB for each candidate fluid to predict thermal efficiency ( $η$ ) and specific net work output ( $W n e t$ ) with high accuracy (R² ≥0.999). The GA was then applied to perform multi-objective optimization, which was transformed into a single-objective form using the weighting coefficient method to determine the optimal working fluids and operating parameters. Results show that R123/R601 achieves a better thermal efficiency (12.76%) when prioritizing thermal efficiency ( $w 1$ ≤0.6), and when prioritizing net work output ( $w 1$ >0.6), R601/R245fa becomes the optimal mixture, achieving 61.4 kW with a 26% increase than that of pure working fluids. The proposed ANN-GA integrated framework provides a fast, accurate, and generalizable approach for mixed working fluid optimization, effectively reducing the optimization time and offering a practical tool for future ORC system design.

Key words: ORC, mixture working fluid, intelligent screening, neural networks, multi-objective optimization, computer simulation, optimal design

中图分类号:

TQ 021.8

李玲, 庄钰, 刘琳琳, 王超, 都健. 基于ANN-GA集成的ORC混合工质智能筛选与性能优化[J]. 化工学报, DOI: 10.11949/0438-1157.20251120.

Ling LI, Yu ZHANG, Linlin LIU, Chao WANG, Jian DU. ANN-GA integrated framework for intelligent screening of ORC mixture working fluids and performance optimization[J]. CIESC Journal, DOI: 10.11949/0438-1157.20251120.

图/表 9

图1 ANN-GA集成的ORC混合工质筛选与优化技术路线图

Fig.1 Technical Framework of the ANN-GA Integrated Optimization for ORC Working Fluid Screening and Performance Evaluation

图2 有机朗肯循环的结构（a）、热力学示意图（b）

Fig.2 Structure(a)and Thermodynamic Diagram (b)of the Organic Rankine Cycle

表1 五种工质基本参数

Table 1 Basic Parameters of Five Working Fluids

工质	工质类别	摩尔质量/（g mol^-1）	沸点/K	临界温度/K	临界压力/MPa	GWP
R123	HCFC	152.93	300.80	456.80	3.66	77
R601	HC	72.14	309.21	469.70	3.37	11
R601a	HC	72.14	300.98	460.35	3.37	7
R245fa	HFC	134.05	288.29	427.16	3.26	950
R1233zd(E)	HFC	130.50	291.00	436.00	3.62	1

图3 ORC 神经网络示意图

Fig.3 Neural network schematic for ORC

图4 R245fa热效率预测（a）、R245fa单位质量净功预测（b）

Fig.4 R245fa thermal efficiency prediction (a), R245fa net work per unit mass prediction (b)

表2 神经网络预测性能

Table 2 Neural Network Predictive Performa

变量	验证集MSE	测试集MSE	验证集R²	测试集R²
$W n e t$	7.67×10^-4	3.36×10^-4	0.9994	0.9999
$η t h e r m a l$	7.21×10^-4	3.46×10^-4	0.9995	0.9999

表2 神经网络预测性能

Table 2 Neural Network Predictive Performa

变量	验证集MSE	测试集MSE	验证集R²	测试集R²
$W n e t$	7.67×10^-4	3.36×10^-4	0.9994	0.9999
$η t h e r m a l$	7.21×10^-4	3.46×10^-4	0.9995	0.9999

图5 不同权重优化结果及其帕累托非劣解集（R123/R601a）

Fig.5 Optimisation results with different weights and their Pareto non-inferior solution sets

表3 工质的性能指标

Table 3 Performance indicators of the matrix

工质	配比	$m ˙$ (kg·s^-1)	$T e v a p$ （K）	$P e x p$ (Pa)	$T c o n d$ （K）	$P p u m p$ (Pa)	$W n e t$ （KW）	$η t h e r m a l$ (%)
R123	-	2.83	118.82	143.49	52.76	1022.35	26.045	12.35
R601	-	2.81	117.43	132.45	38.12	853.28	48.699	10.56
R601a	-	2.69	101.99	132.54	66.15	853.46	48.701	10.60
R245fa	-	3.97	133.95	182.56	99.78	1181.83	20.419	11.43
R1233zd(E)	-	2.73	111.63	182.62	83.26	1197.16	20.419	11.50
R123/R601	9:1	2.12	105.30	142.82	44.90	1143.88	59.903	12.76
R123/R601a	3:7	3.85	104.27	139.24	82.15	1190.73	56.852	12.56
R123/R245fa	8:2	3.25	109.10	156.73	59.21	1188.16	29.704	12.15
R123/R1233zd(E)	8:2	2.31	107.34	165.98	47.16	1121.50	30.626	12.76
R601/R601a	1:9	2.64	103.08	173.36	74.02	1162.24	58.394	11.54
R601/R245fa	9:1	1.98	103.98	123.51	64.13	1164.72	61.385	12.51
R601/R1233zd(E)	4:6	2.25	110.87	165.54	53.39	1180.81	50.736	11.94
R601a/R245fa	7:3	1.89	133.96	178.68	68.08	1172.72	49.366	11.32
R601a/R1233zd(E)	3:7	2.27	118.00	161.87	72.39	1127.09	53.636	12.61
R245fa/R1233zd(E)	7:3	2.97	109.26	205.36	69.98	1199.82	29.429	10.99

表3 工质的性能指标

Table 3 Performance indicators of the matrix

工质	配比	$m ˙$ (kg·s^-1)	$T e v a p$ （K）	$P e x p$ (Pa)	$T c o n d$ （K）	$P p u m p$ (Pa)	$W n e t$ （KW）	$η t h e r m a l$ (%)
R123	-	2.83	118.82	143.49	52.76	1022.35	26.045	12.35
R601	-	2.81	117.43	132.45	38.12	853.28	48.699	10.56
R601a	-	2.69	101.99	132.54	66.15	853.46	48.701	10.60
R245fa	-	3.97	133.95	182.56	99.78	1181.83	20.419	11.43
R1233zd(E)	-	2.73	111.63	182.62	83.26	1197.16	20.419	11.50
R123/R601	9:1	2.12	105.30	142.82	44.90	1143.88	59.903	12.76
R123/R601a	3:7	3.85	104.27	139.24	82.15	1190.73	56.852	12.56
R123/R245fa	8:2	3.25	109.10	156.73	59.21	1188.16	29.704	12.15
R123/R1233zd(E)	8:2	2.31	107.34	165.98	47.16	1121.50	30.626	12.76
R601/R601a	1:9	2.64	103.08	173.36	74.02	1162.24	58.394	11.54
R601/R245fa	9:1	1.98	103.98	123.51	64.13	1164.72	61.385	12.51
R601/R1233zd(E)	4:6	2.25	110.87	165.54	53.39	1180.81	50.736	11.94
R601a/R245fa	7:3	1.89	133.96	178.68	68.08	1172.72	49.366	11.32
R601a/R1233zd(E)	3:7	2.27	118.00	161.87	72.39	1127.09	53.636	12.61
R245fa/R1233zd(E)	7:3	2.97	109.26	205.36	69.98	1199.82	29.429	10.99

表4 权重灵敏度分析结果

Table 4 Results of weight sensitivity analysis

w₁ (单位净功权重)	w₂ (热效率权重)	最优点	单位净功	热效率	F
0	1	R123/R601	59902.96	12.76	1
		R123/R1233zd(E)	30626.19	12.76	1
0.1	0.9	R123/R601	59902.96	12.76	0.981
0.2	0.8	R123/R601	59902.96	12.76	0.963
0.3	0.7	R123/R601	59902.96	12.76	0.945
0.4	0.6	R123/R601	59902.96	12.76	0.9267
0.5	0.5	R123/R601	59902.96	12.76	0.908
0.6	0.4	R123/R601	61384.86	12.51	0.931
0.7	0.3	R601/R245fa	61384.86	12.51	0.943
0.8	0.2	R601/R245fa	61384.86	12.51	0.954
0.9	0.1	R601/R245fa	61384.86	12.51	0.966
1	0	R601/R245fa	61384.86	12.51	1

参考文献 31

[1]	Dong X, Liao Z W, Sun J Y, et al. Simultaneous optimization for organic Rankine cycle design and heat integration[J]. Industrial & Engineering Chemistry Research, 2020, 59(46): 20455-20471.
[2]	Feng Y Q, Hung T, Zhang Y N, et al. Performance comparison of low-grade ORCs (organic Rankine cycles) using R245fa, pentane and their mixtures based on the thermoeconomic multi-objective optimization and decision makings[J]. Energy, 2015, 93: 2018-2029.
[3]	Gürgen S, Altın İ. Novel decision-making strategy for working fluid selection in Organic Rankine Cycle: a case study for waste heat recovery of a marine diesel engine[J]. Energy, 2022, 252: 124023.
[4]	Gutiérrez-Arriaga C G, Abdelhady F, Bamufleh H S, et al. Industrial waste heat recovery and cogeneration involving organic Rankine cycles[J]. Clean Technologies and Environmental Policy, 2015, 17(3): 767-779.
[5]	Imran M, Park B S, Kim H J, et al. Thermo-economic optimization of Regenerative Organic Rankine Cycle for waste heat recovery applications[J]. Energy Conversion and Management, 2014, 87: 107-118.
[6]	吴维成, 谷志卿, 鞠朋, 等. 工业低温余热回收利用研究进展[J]. 可再生能源, 2024, 42(8): 1096-1103.
	Wu W C, Gu Z Q, Ju P, et al. Research progress in recovery and utilization of low-temperature industrial waste heat[J]. Renewable Energy Resources, 2024, 42(8): 1096-1103.
[7]	Mahmoudi A, Fazli M, Morad M R. A recent review of waste heat recovery by Organic Rankine Cycle[J]. Applied Thermal Engineering, 2018, 143: 660-675.
[8]	Zhao D P, Deng S, Zhao L, et al. Overview on artificial intelligence in design of Organic Rankine Cycle[J]. Energy and AI, 2020, 1: 100011.
[9]	Lecompte S, Huisseune H, van den Broek M, et al. Review of organic Rankine cycle (ORC) architectures for waste heat recovery[J]. Renewable and Sustainable Energy Reviews, 2015, 47: 448-461.
[10]	Sadni F E, Salhi I, Belhora F, et al. Multi-objective optimization of energy and exergy efficiencies in orc configurations using nsga-ii and topsis[J]. Thermal Science and Engineering Progress, 2025, 63: 103606.
[11]	张军辉, 刘娟芳, 陈清华. 有机朗肯循环系统最佳蒸发温度和㶲分析[J]. 化工学报, 2013, 64(3): 820-826.
	Zhang J H, Liu J F, Chen Q H. Optimal evaporating temperature and exergy analysis for organic Rankine cycle[J]. CIESC Journal, 2013, 64(3): 820-826.
[12]	Wang S K, Liu C, Zhang S J, et al. Multi-objective optimization and fluid selection of organic Rankine cycle (ORC) system based on economic-environmental-sustainable analysis[J]. Energy Conversion and Management, 2022, 254: 115238.
[13]	徐建. 中低温余热发电有机朗肯循环系统性能分析及优化研究[D]. 上海: 上海交通大学, 2012.
	Xu J. Performance analysis and optimization research of mid-low- temperature waste heat power generation using organic Rankine cycle[D]. Shanghai: Shanghai Jiao Tong University, 2012.
[14]	Chen Y, Ni W J. Investigation on working fluid selection of organic Rankine cycle for low temperature waste heat utilization[J]. Energy Conservation & Environmental Protection, 2020(4): 60-62.
[15]	Li L, Zhuang Y, Huang Y J, et al. An efficient sequential synthesis model for optimization of steam-organic Rankine cycles and work–heat integration[J]. Industrial & Engineering Chemistry Research, 2025, 64(23): 11459-11473.
[16]	Kang L X, Tang J P, Liu Y Z. Optimal design of an organic Rankine cycle system considering the expected variations on heat sources[J]. Energy, 2020, 213: 118841.
[17]	Huang X J, Lu P, Luo X L, et al. Synthesis and simultaneous MINLP optimization of heat exchanger network, steam Rankine cycle, and organic Rankine cycle[J]. Energy, 2020, 195: 116922.
[18]	Fu C, Gundersen T. Correct integration of compressors and expanders in above ambient heat exchanger networks[J]. Energy, 2016, 116: 1282-1293.
[19]	Huang Y J, Xing Y F, Zhuang Y, et al. An extended superstructure-based model for synthesis of compression-heat-integrated heat exchanger networks coupled with multiple utilities[J]. Industrial & Engineering Chemistry Research, 2024, 63(21): 9456-9470.
[20]	Huang Y J, Zhuang Y, Xing Y F, et al. Multi-objective optimization for work-integrated heat exchange network coupled with interstage multiple utilities[J]. Energy, 2023, 273: 127240.
[21]	王金江, 鲁振杰, 安维峥, 等. ORC发电系统工艺过程预警诊断技术研究与展望[J]. 化工学报, 2025, 76(7): 3137-3152.
	Wang J J, Lu Z J, An W Z, et al. Research and prospect of early warning and diagnosis technology for ORC power generation system process[J]. CIESC Journal, 2025, 76(7): 3137-3152.
[22]	Lee J Y, Chen P L, Xie P S, et al. Design of multi-cycle organic Rankine cycle systems for low-grade heat utilisation[J]. Energy, 2024, 310: 133252.
[23]	刘杰, 陈江平, 祁照岗. 低温有机朗肯循环的热力学分析[J]. 化工学报, 2010, 61(S2): 9-14.
	Liu J, Chen J P, Qi Z G. Thermodynamic analysis of low temperature organic Rankine cycle[J]. CIESC Journal, 2010, 61(S2): 9-14.
[24]	Santos M N P, da Silva Sá Ravagnani M A, Costa C B B. Optimal integration of an Organic Rankine Cycle to a process using a heuristic approach[J]. Brazilian Journal of Chemical Engineering, 2021, 38(4): 653-667.
[25]	Quoilin S, Aumann R, Grill A, et al. Dynamic modeling and optimal control strategy of waste heat recovery Organic Rankine Cycles[J]. Applied Energy, 2011, 88(6): 2183-2190.
[26]	韩中合, 杜燕, 王智. 有机朗肯循环低温余热回收系统的工质选择[J]. 化工进展, 2014, 33(9): 2279-2285.
	Han Z H, Du Y, Wang Z. Medium selection of organic Rankine cycle(ORC) in low temperature waste heat[J]. Chemical Industry and Engineering Progress, 2014, 33(9): 2279-2285.
[27]	Li L, Zheng Y L, Fang S, et al. A two-layer optimization framework for combined heat and power with an enhanced organic rankine cycle[J]. Chemical Engineering Research and Design, 2025, 223: 466-477.
[28]	侯中兰, 魏新利, 马新灵, 等. 循环水流量对ORC余热发电系统性能影响的试验分析[J]. 化工学报, 2019, 70(9): 3283-3290.
	Hou Z L, Wei X L, Ma X L, et al. Experimental analysis of circulating water flow rate on performance of ORC waste heat power generation system[J]. CIESC Journal, 2019, 70(9): 3283-3290.
[29]	Scaccabarozzi R, Tavano M, Invernizzi C M, et al. Comparison of working fluids and cycle optimization for heat recovery ORCs from large internal combustion engines[J]. Energy, 2018, 158: 396-416.
[30]	Permana D I, Fagioli F, De Lucia M, et al. Performance evaluation, prediction analysis and optimization of experimental ORC using artificial neural networks (ANN)[J]. Energy Nexus, 2025, 18: 100383.
[31]	Pavão L V, Costa C B B, Ravagnani M A S S. Multiple utilities targeting in energy integration considering rigorous temperature-enthalpy relations[J]. Chemical Engineering Science, 2023, 276: 118763.