Framework of model selection criteria approximated genetic programming for optimization function for renewable energy systems
Introduction
Physical understanding, prediction and optimization of complex renewable energy systems is a difficult problem. To address this problem, an evolutionary system identification method, such as genetic programming (GP), can be applied to develop mathematical models/functional relationships between the process parameters. The obtained process parameters can be further optimized to obtain the approximation state of the system.
Among the complex renewable energy systems, one such system includes solar collectors. One of the most common collectors is the traditional flat surface collector, which absorbs solar energy through black or spectral selective solid surfaces. The absorber transfers heat to the collector through conduction and then transfers energy to the circulating fluid through convection. However, the thermal performance of these types of solar collectors is limited by the absorption capacity of the black surface and how heat is effectively transferred to the workflow. In order to overcome the shortcomings of traditional solid surface collectors, several methods have been proposed. One proposed method is a collector in which solar radiation is absorbed directly by the working fluid instead of a thin layer on the surface. Therefore, in direct absorption solar collector (DASC), the thermal resistance of converting solar energy into thermal energy is reduced. The concept of direct absorption originated in the 1980s to simplify the design of surface absorbers for fluid plates and potentially improve thermal performance through direct radiation absorption [1]. Typical heat carrier fluids used in solar collectors (i.e., water, ethylene glycol, propylene glycol and heat-conducting oil) have been proven to have poor thermophysical properties and deficient absorption properties in solar spectra. Among the four abovementioned liquids, water is considered as the best absorbent of solar energy. However, it only absorbs about 13% of the energy in previous tests, which implies it is still a weak absorbent [2]. Therefore, the improvement of the thermal performance of the DASC system by improving the ability of the liquid to absorb solar energy has become popular. The thermal performance of DASC systems can be improved by implanting small particles in the matrix which absorbs solar energy. Some researchers introduced the use of liquids with black particles (dispersing Indian ink into a mixture of ethylene glycol and water), to absorb the solar radiation [1].
In recent years, several types of nanoparticles have been synthesized and dispersed in the typical matrix. The SU-Pension of nanoparticles is more stable than that of micron or larger particles. Besides, for suspensions containing nanoparticles, blockage and fouling are found to have little effect. Many studies have shown that the radiation properties of alkaline liquids can be improved by adding an appropriate number of nanoparticles (volume less than 1%). Spectral transmittance and absorptivity of various nanofluids are reported [3], [4], [5], [6], [7]. Researchers use mathematical theories, such as the Rayleigh scattering method [7,8], Maxwell Garnett effective medium theory [7], and Mie scattering theory [9], [10], [11].
The experimental and numerical studies on the direct absorption collector show that the collection efficiency can be improved by using Nano-fluids as the fluid medium. In prior works, the performance of the non-centralized direct absorption solar collector is theoretically studied and compared with that of a typical flat plate solar collector [1]. The mixtures of water and aluminum nanoparticles are 10% more efficient than flat collectors. The experimental results of solar collectors based on carbon nanotubes, graphite and silver nanofluids show that the efficiency can be increased by 5% [12,13]. The feasibility of using graphite nanoparticles dispersed in thermosensitive alcohol VP-1 to concentrate DASC in the laboratory is studied. When the volume fraction of graphite is 0.125%, the efficiency of the base solution is increased by about 11%. Lenert and Wang used Graphite-Coated cobalt to enrich the solar absorption column in thermosensitive alcohol VP-1 nanofluids. The experimental results show that under the optimum conditions, the system efficiency exceeds 35% [14]. Lee et al. employed numerical simulation in the prediction of DASC efficiency. Their results indicated that at 0.05% particle volume concentration, the application of gold-plated silica nanoparticles is able to increase the efficiency of DASC by 70%. Bandarra Filho et al. studied the photothermal conversion ability of water-based silver nanofluids under practical conditions. His-studies concluded that when the concentration of nanoparticles is 6.5 ppm, the stored heat energy can be increased by 144% [15]. Luo et al. performed experiments on a tex-atherm oil-enriched direct absorption column. His-studies show that the photo-thermal efficiency of 0.01 Vol. graphite and 0.5 Vol. Al2O3 nanofluids are 22.7% and 17.5% higher than that of conventional coated collectors [16]. The theoretical study of copper-water nanofluids direct absorption solar collector from Pavin et al. [17] shows that the heat transfer performance and entropy production of the collector is improved when the volume fraction is less than 3% [17].
It can be seen that most of the previous studies emphasized the photothermal conversion efficiency of nanofluids in pre-designed DASC without the incorporation of the geometry of the collector. The system complexity is attributed to interaction among the design variables influencing the outputs. There are also uncertainties in the system due to random and unknown variations in the design and response variables. In a study conducted by Gorji and Ranjbar [1], the effect of different geometric shapes on the overall thermal performance of DASC based on nanofluids is evaluated. The fluent simulation in ANSYS and response surface methodology (RSM) is used. The principle functionality of RSM is based on the statistical assumptions and fits well for the data given in the design domain range. In addition, the models formulated from these methods may not work well for the extrapolation/interpolation cases. Artificial neural network is thought to be a powerful method for numerical model analysis. However, the relatively fixed model structure and form may affect the flexibility of model fitting. Meanwhile, over-fitting may occur in the model generation, which negatively affects the calculation and prediction performance of ANN models [18]. In this situation, an evolutionary system identification method, such as GP, can be applied to develop mathematical models/functional relationships between the process parameters.
The evolutionary search in GP is guided by objective functions used in its framework. Based on preliminary studies, it is found that the GP suffers from the high complexity structure of its solutions and non-optimal convergence, which would lead to poor fitness values. The reasonable selection of mathematical criteria can effectively guide the evolutionary search towards global fitness and penalize the highly complex models and ensures satisfactory accuracy simultaneously. To address this uncertainty/problem, the framework based on model selection criteria approximated genetic programming (MSC-GP) is proposed for the formulation of geometry optimization function for the DASC system. In the proposed method of MSC-GP, the four mathematical model selection criteria are used as an approximation for objective functions in GP framework for evaluation of the fitting accuracy and structure of the model. The proposed and formulated model is used for studying the factors affecting the overall heat dissipation performance of the DASC system.
In general, the geometric design shall affect the thermal behavior and performance of the DASC system. Furthermore, the selection of the objective function is the vital factor for the improvement of performance of the generated model. Objective function comprises both empirical error term and complexity term, which has a greater influence on the performance of models. Therefore, this paper proposes a method combined simulation experiment and artificial intelligence (MSC-GP) to study the effect of geometric design on the thermal behavior of DASC systems, which provides a novel strategy for investigating the performance of DASC system. Findings from parametric interaction analysis and optimization in this research are useful in the optimization of functioning of DASC system. Thus, more studies can be done in future to evolve new measures of defining its complexity to improve the MSC-GP method to improve the efficiency of researches on DASC systems.
The rest of the paper is arranged as follows. Section 2 provides a general view of the research problem. Section 3 describes methodology proposed in formulation of model development. Section 4 discusses the results: the performance of GP models by applying different objective functions, fitting performance comparison of models generated by GP, ANN and RSM, evaluation and selection of generated models, sensitivity analysis and parametric analysis, and the optimization settings for the DASC system. Section 5 concludes the paper and recommendations for future work.
Section snippets
Research problems statement
This work involves two primary contributions: the development of MSC-GP method and the formulation of geometry design based thermal efficiency and entropy generation optimization function of nano-fluids direct absorption solar collector system. The formulated model is then used to study the influence of geometry design parameters of DASC on the thermal performance of the system through post-model analysis. The research problem can be subdivided into two parts:
- (a)
Primary problem: The primary
Basis for genetic programming
The process of using experimental data for establishing mathematical models to reflect the correlation between the input and output variables is known as system identification (SI) [25,26]. In this research, to carry on the numerical analysis and optimization work, a numerical model that reflects the correlation between inputs and outputs is necessary. Artificial intelligence (AI) methods, such as artificial neural network (ANN) [27], [28], [29], machine learning [30], deep learning [31],
Results and discussion
The models generated are utilized for optimization of geometric design of DASC system. A proper selection of these models is vital to ensure the high accuracy and reliability of the numerical analysis. Therefore, MSC-GP is a useful approach in this research. The obtained values of each criterion are the main indexes to select the target models generated by using the selected function as an objective function. The fitting accuracy and the prediction reliability of selected MSC-GP models are
Conclusions
In this paper, the framework based on model selection criteria approximated genetic programming (MSC-GP) is proposed for the formulation of geometry optimization function of the direct absorption solar collector (DASC) system. The four mathematical model selection criteria are used as an approximation for objective functions in GP framework for evaluating the fitting accuracy and structure of the model. The proposed approach and formulated model is used for studying the factors affecting the
CRediT authorship contribution statement
Akhil Garg: Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review & editing. Shaosen Su: Formal analysis, Investigation, Software, Validation, Visualization, Writing - original draft. Fan Li: Software, Validation, Visualization, Writing - review & editing. Liang Gao: Formal analysis, Funding acquisition, Investigation, Project administration, Resources, Software, Supervision, Writing - review &
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the Program for HUST Academic Frontier Youth Team [grant number 2017QYTD04], the Program for HUST Graduate Innovation and Entrepreneurship Fund [grant number 2019YGSCXCY037].
References (57)
- et al.
ScienceDirect Geometry optimization of a nanofluid-based direct absorption solar collector using response surface methodology
Sol. Energy
(2015) - et al.
Optical properties of liquids for direct absorption solar thermal energy systems
Sol. Energy
(2009) - et al.
Enhancement of solar radiation absorption using nanoparticle suspension
Sol. Energy
(2011) - et al.
International Journal of Heat and Mass Transfer Evaluation of the effect of nanofluid-based absorbers on direct solar collector
Int. J. Heat Mass Transf.
(2012) - et al.
Optimization of nanofluid volumetric receivers for solar thermal energy conversion
Sol. Energy
(2012) - et al.
International Journal of Heat and Mass Transfer Performance improvement of a nanofluid solar collector based on direct absorption collection (DAC) concepts
Int. J. Heat Mass Transf.
(2014) - et al.
International Journal of Heat and Mass Transfer Heat transfer and entropy generation through nanofluid filled direct absorption solar collector
Int. J. Heat Mass Transf.
(2014) - et al.
Advances in engineering software assessment of artificial neural network and genetic programming as predictive tools
Adv. Eng. Softw.
(2015) - et al.
Fast and accurate measurement of entropy profiles of commercial lithium-ion cells
Electrochim. Acta
(2015) - et al.
Entropy measurement of a large format lithium ion battery and its application to calculate heat generation
Electrochim. Acta
(2019)
Analyses of exergy efficiency and pumping power for a conventional flat plate solar collector using SWCNTs based nanofluid
Energy Build.
An application of evolutionary system identification algorithm in modelling of energy production system
Measurement
Nonlinear system identification and model reduction using artificial neural networks
Comput. Chem. Eng.
Machine learning estimates of plug-in hybrid electric vehicle utility factors
Transp. Res. Part D
A review of deep learning for renewable energy forecasting
Energy Convers. Manag.
Prognostics of lithium-ion batteries based on Dempster – Shafer theory and the Bayesian Monte Carlo method
J. Power Sources
Using Gaussian distribution to construct fitness functions in genetic programming for multiclass object classification
Pattern Recognit. Lett.
Evaluation of batteries residual energy for battery pack recycling : proposition of stack stress-coupled-AI approach
J. Energy Storage
Unbalanced breast cancer data classification using novel fitness functions in genetic programming
Expert Syst. Appl.
SMT-GP method of prediction for ground subsidence due to tunneling in mountainous areas
Tunn. Undergr. Sp. Technol. Inc. Trenchless Technol. Res.
Multidimensional genetic programming for multiclass classification
Swarm Evol. Comput.
Statistical genetic programming for symbolic regression
Appl. Soft Comput. J.
Comparative evaluation of optimization algorithms at training of genetic programming for tensile strength prediction of FDM processed part
Procedia Mater. Sci.
Sensitivity analysis of energy inputs in crop production using artificial neural networks
J. Clean. Prod.
Carbon nanohorns-based nanofluids as direct sunlight absorbers
Opt. Exp.
Absorption and scattering properties of carbon nanohorn-based nanofluids for direct sunlight absorbers
Nanoscale Res. Lett.
Thermal properties of carbon black aqueous nanofluids for solar absorption
Nanoscale Res. Lett.
Nanofluid optical property characterization : towards efficient direct absorption solar collectors
Nanoscale Res. Lett.
Cited by (4)
A comprehensive review of automatic programming methods
2023, Applied Soft ComputingThermo-economic modeling and optimization of a solar network using flat plate collectors
2022, EnergyCitation Excerpt :They suggested that there would be no benefit in using a Kalina cycle instead of a Rankine cycle when generating power from heat taken directly from the solar receiver [18]. To understand the complex of renewable energy systems, an evolutionary system identification procedure such as genetic programming (GP) can be employed to improve mathematical models/functional relationships between the process variables [19]. Santiago et al. [20] suggested a novel accurate Micro Genetic Algorithm GA) for multi-objective optimization problems.
Data-driven hybrid approaches for renewable power prediction toward grid decarbonization: Applications, issues and suggestions
2021, Journal of Cleaner ProductionCitation Excerpt :The classification and regression approaches consist of support vector machine (SVM) (Pan et al., 2020) and random forest (RF) (Prasad et al., 2019) while the time series method covers autoregressive integrated moving average (ARIMA) (Kaytez, 2020) and seasonal autoregressive integrated moving averages (SARIMA) (Mao et al., 2018). The optimization algorithms have been employed extensively in renewable power prediction, which comprises fruit fly optimization algorithm (FOA) (Wang et al., 2020), genetic algorithm (GA) (Garg et al., 2020), backtracking search algorithm (BSA) (Mehmood et al., 2019), coral reefs optimization (CRO) (Durán-Rosal et al., 2018), and particle swarm optimization (PSO) (Alshammari and Asumadu, 2020). The deep learning methods are popular due to their strong self-learning capabilities and high accuracy, which contain long short-term memory (LSTM) (Yang et al., 2020) and deep belief network (DBN) (Wang et al., 2018).
Selection Criteria for the Implementation of a Power Generation Cycle Using Carbon Dioxide as Working Fluid and Solar Thermal Energy
2022, International Journal on Energy Conversion