1 Introduction
With the world’s energy needs still increasing, renewable energy technologies have attracted great interest because of their potential for clean and sustainable energy solutions. CSP technologies have become an attractive option compared to traditional fossil fuel-based electricity generation because they can collect and store energy from the sun for round-the-clock electricity generation (
Guerrero-Lemus and Martínez-Duart, 2013;
Heller, 2017;
Cavallaro et al., 2019). SPTs have been a promising CSP technology because they can achieve high operating temperatures and efficiency. SPTs use a field of heliostats (mirrors) that track and concentrate the sun’s rays onto a central receiver atop a tower, as shown in Fig. 1. The heat energy absorbed is subsequently utilized to power a thermodynamic cycle, most commonly producing electricity via a steam turbine or other forms of heat engine arrangements (
Guerrero-Lemus and Martínez-Duart, 2013;
Heller, 2017;
Merchán et al., 2022).
Heliostat fields provide the basic energy input to SPTs and account for a significant portion (40%–50%) of the total plant cost. Field layout optimization has an important impact in terms of minimizing optical errors, shadowing and blockage losses, and maximizing overall plant efficiency (
Augsburger et al., 2016;
Maiga et al., 2024). Field configurations and tracking strategies are being further investigated for new field designs in an effort to achieve performance and economic improvement. The receiver is also an essential subsystem tasked with the conversion of concentrated solar radiation to heat. The receivers are exposed to high flux density and temperature gradients; therefore, a trade-off between efficiency, durability, and affordability needs to be balanced (
Frantz et al., 2017;
Merchán et al., 2022). Numerous receiver concepts in the form of tubular or volumetric receivers have been constructed to maximally absorb heat while reducing heat loss.
Heat transfer fluids (HTFs) play an important role in SPT operation by carrying collected thermal energy from the receiver to the power block. Extensively utilized HTFs, like molten salts, enable effective thermal storage and stable operation and hence ensure electricity supply even under adverse solar conditions. An efficient HTF selection is key to ensuring high thermal efficiency and stable operation (
Costa and Lemos, 2015;
Augsburger et al., 2016;
Pérez-Álvarez et al., 2021). For optimum SPT performance and reliability, precise simulation models are required. The models allow for the prediction of the system’s behavior under different environmental and operating conditions, which facilitates the optimization of heliostat fields, receiver designs, and control strategies (
Yu et al., 2012;
Costa and Lemos, 2015;
Frantz et al., 2017).
Various computational methods and computer codes, including the ASTRID code and Monte Carlo ray tracing (MCRT), have been formulated to design and optimize the heat flux distribution onto receivers in order to increase efficiency. The methods improve design accuracy and allow highly efficient SPT systems to be developed (
Cheng et al., 2017;
Frantz et al., 2017). Additionally, sophisticated control systems are important in ensuring the thermal balance of receivers and HTFs. They should successfully regulate the temperature variations because of climatic changes and solar energy intermittency. Effective control mechanisms enhance the reliability and safety of SPT operations, optimum power generation, and less energy losses (
Yu et al., 2012;
Costa and Lemos, 2015).
Simulation software is an essential tool for SPT system design, optimization, and performance analysis by providing the economical means of analyzing complex processes, optimizing designs, and predicting system behavior under diverse conditions(
Khalsa and Ho, 2011;
Ho, 2014;
Noureldin et al., 2021). Simulation software is necessary for design optimization as it allows designers to experiment with numerous configurations and operation strategies in order to identify the best designs (
Scott et al., 2017;
Noureldin et al., 2021). They are also beneficial for performance evaluation, enabling close examination of thermal efficiency, heat transfer, and fluid dynamics for the optimization of system performance (
Khalsa and Ho, 2011;
Ho, 2014;
Patel et al., 2023). In addition, simulation software enables economic evaluation by establishing the financial value of novel control concepts, for instance, spatial DNI nowcasts, that are able to optimize power plant economic return (
Noureldin et al., 2021). In optical modeling, ray tracing techniques imitate the propagation and absorption of solar radiation in order to optimize heliostat field layout in a manner that maximizes the concentration of solar energy (
Barreto et al., 2019;
Ullah, 2019;
Pujol-Nadal and Cardona, 2023). Simulation programs also improve heliostat field layout designs to minimize shading and blocking losses as well as maximize system efficiency (
Noureldin et al., 2021;
Pujol-Nadal and Cardona, 2023). Thermal and computational fluid dynamics (CFD) modeling needs to be used to simulate receiver heat transfer processes, such as radiative, convective, and conductive heat loss, in order to develop designs that optimize thermal efficiency (
Khalsa and Ho, 2011;
Ho, 2014;
Patel et al., 2023). CFD simulations also analyze fluid dynamics to determine the behavior of HTFs and to optimize their performance (
Khalsa and Ho, 2011;
Barreto et al., 2019;
Patel et al., 2023). System-level modeling tools like SolarTherm allow for techno-economic analysis, simulating overall performance, economic feasibility, and impact (
Scott et al., 2017). The tool can also optimize dispatch strategy for cost-effective and efficient attainment of energy demand (
Scott et al., 2017;
Noureldin et al., 2021). Simulation tools in general are important to SPT system design, allowing for important insights and enabling the optimization of major components for better efficiency and economic value. A summary of key research contributions related to the modeling and simulation of SPT systems is presented in Table 1, highlighting the research focus, existing challenges, and contribution of this work.
This review presents a comprehensive and structured analysis of simulation tools for SPTs, covering optical, thermal, and system-level models. While previous studies have evaluated individual aspects, this work offers a holistic comparative assessment and introduces novel solutions to existing challenges. The key contributions of this review are:
1) This review uniquely evaluates optical, thermal, and system-level simulation tools together, highlighting their interdependencies and trade-offs in accuracy, computational cost, and usability.
2) The review systematically analyzes major simulation tools and identifies key challenges such as a lack of multi-physics coupling, computational inefficiencies, and limited real-world validation, which have been overlooked in prior research.
3) This review proposes a novel hybrid simulation framework that integrates ray-tracing optical models with CFD simulations for improved receiver efficiency and system-level models with physics-based simulations for enhanced predictive accuracy.
4) The review introduces AI-powered heliostat field optimization, machine learning-based heat transfer predictions, and digital twin-based real-time SPT simulations to improve accuracy while reducing computational costs.
5) This review provides a decision-making guide for selecting the most suitable simulation tools based on specific SPT applications and outlines future directions such as multi-physics model integration, cloud-based parallel computing, and validation with real-world plant data.
This review serves as a bridge between theoretical models and real-world applications, enabling more accurate, efficient, and scalable SPT simulation frameworks for future advancements in sustainable energy. The structure of this review is as follows: Section 2 discusses optical simulation tools used for heliostat field and receiver design. Section 3 covers thermal and CFD simulation tools for receiver heat transfer analysis. Section 4 provides insights into system-level simulation for techno-economic and operational analysis. Section 5 explores the integration of optical, thermal, and system models, while Section 6 presents the emerging trends in AI and ML for SPT applications. Finally, future directions and conclusions are discussed in Sections 7 and 8, respectively.
2 Methodology for tool selection and comparative framework
To ensure objectivity, transparency, and comprehensiveness in this review, a structured methodology was adopted for the selection, classification, and comparison of simulation tools relevant to SPT systems. The methodology is illustrated in Fig. 2 and consists of three core elements: selection criteria, classification approach, and comparison framework.
Selection Criteria: Tools were selected based on their relevance to one or more key simulation domains of SPT technology, namely optical modeling, thermal-hydraulic performance, computational fluid dynamics (CFD), and system-level techno-economic analysis. Recent usage in peer-reviewed literature, evidence of experimental validation or benchmarking, licensing accessibility (including both open-source and commercial options), and adaptability to emerging modeling needs (such as AI integration, transient conditions, and modularity) were used as guiding principles. This approach ensured the inclusion of both widely adopted tools (such as SolTrace, SAM, and COMSOL) and emerging platforms such as SolarTherm, CombiCSP, OpenModelica, MoSES, and FluxTracer.
Classification Approach: The selected tools were categorized into three simulation domains aligned with the architecture of typical SPT systems: (i) optical tools for ray tracing and heliostat layout; (ii) thermal and CFD tools for energy balance, transient heat transfer, and detailed modeling of fluid flow, convection, and radiation within the receiver; and (iii) system-level tools for annual performance analysis and techno-economic modeling.
Comparison Framework: Tools within each domain were evaluated using consistent criteria to maintain reproducibility. These include the underlying modeling approach and assumptions, supported physical and control features, ease of use and user interface, computational efficiency and scalability, licensing and accessibility, validation status or prior benchmarking, and integration capability with other simulation platforms or experimental data.
This structured and transparent methodology ensures that the review is not biased toward established tools but instead provides a balanced assessment of both conventional and next-generation platforms adopted in the CSP research community.
3 Optical simulation models for heliostat field and receiver design
3.1 Importance of optical simulations in SPTs
Optical simulations have a crucial part in increasing the energy efficiency of SPTs from the perspective of heliostat field layout optimization and ray tracing. Heliostat field layout optimization is essential for improving optical efficiency and reducing energy losses. Numerous strategies like biomimetic layouts, radial-staggered layouts, and hybrid layouts have been investigated for the optimization of efficiency (
Zhang et al., 2016,
2024;
Belaid et al., 2022;
Yang et al., 2024). For example, the optimized biomimetic square hexagonal heliostats have achieved the highest annual weighted optical efficiency, improving receiver efficiency by homogenizing the solar heat flux distribution (
Yang et al., 2024). The optimized geometries also minimize shadowing and blocking losses, which play a very important role in achieving high optical efficiency. A case study on a 5-MW SPT found that an optimized radial-staggered design cut blocking losses by 84.3% and improved yearly optical efficiency by 8.8% (
Zhang et al., 2024). Moreover, utilizing circular heliostats with optimized layouts can even decrease the land area needed and enhance the yearly efficiency and hence decrease the Levelized Cost of Energy (LCOE) (
Belaid et al. 2022).
Ray tracing and cell-based methods are two important optical simulation techniques for SPTs with respective strengths and weaknesses. Ray tracing simulates light interactions, such as reflections and absorptions, and MCRT is one popular statistical accuracy technique (
Cui et al., 2012;
Winter et al., 2015;
Reiners and Blieske, 2018). Ray tracing is extremely accurate and flexible for complicated geometries, but is time-consuming (
Cui et al., 2012;
Hajimirza and Lu, 2017;
Reiners and Blieske, 2018;
Schöttl et al., 2018). The cell-based method, by dividing the domain into small cells, improves computational efficiency and scalability and thus is appropriate for large-scale systems at the cost of introducing approximations that may decrease accuracy for complicated geometries. Ray tracing provides greater accuracy at a greater computational expense, whereas cell-based methods achieve optimum efficiency for large-scale systems. The choice relies on the compromise between computational efficiency and accuracy (
Cui et al., 2012;
Reiners and Blieske 2018;
Schöttl et al., 2018).
Ray tracing is employed to analyze and simulate heliostat field optical performance, and it enables the calculation of shadowing, shading, and blocking effects required for optimizing heliostat layouts and overall system performance (
Liu et al., 2024;
Aslam et al., 2025). The MCRT method is the preferred one since it handles reflectivity and optical imperfections statistically, enabling the modeling of complex mirror fields and receivers with high accuracy (
Lee et al., 2016;
Craig et al., 2016). The detailed steps of the MCRT simulation are provided in Fig. 3. SolTrace, Tonatiuh, and CRS4-2 are a few among a multitude of simulation tools that are widely used to carry out ray tracing and analysis of the optical performance of heliostat fields. These programs are handy for comparing prospective designs and optimizing layouts for peak efficiency (
Jafrancesco et al. 2018). A hybrid MCRT-Gebhart model has been used to simulate solar radiation transfer in cavity receivers and to illustrate the cavity effect in reducing reflection losses and homogenizing solar flux distributions (
Wang et al. 2016).
3.2 Overview of optical simulation software
Optical simulation software plays an essential role in SPT system optimization and design via heliostat field layout simulation, optical efficiency analysis, sunshape, and reflective error. Different simulation software has evolved with time, focusing on higher accuracy and effectiveness of SPT optical modeling. In this section, a comparative study of the most popular optical simulation software is done, and their functionalities, strengths, weaknesses, and recent improvements are discussed.
Optical simulation of SPT systems typically employs either ray tracing or cell-based methods. Ray tracing involves tracing individual light rays through a system and simulating their interaction with reflective and refractive surfaces. This is employed by programs like SolTrace, Tonatiuh, STRAL, and TieSOL and gives very detailed flux distributions and allows accurate loss calculations like cosine losses, shading, blocking, spillage, and atmospheric attenuation. Ray tracing, however, is a computing-intensive and slow process and therefore more appropriate for large-scale optical analysis but not for quick field layout optimization. Cell-based methods, on the other hand, such as DELSOL3 and RCELL (UHC), discretize the heliostat field into fixed zones and numerically approximate the flux distributions using analytical methods. These make the simulation faster and are useful for system planning and optimization at an early stage. However, they are unable to simulate complex reflections and separate individual loss factors. While ray tracing is more precise, cell-based approaches are efficient computationally and therefore ideal for large-scale simulation and rapid optimization.
A number of optical simulation software packages are popular in SPT research and design. NREL’s SOLARPILOT is well known for its complex solar field layout creation and optical performance simulation (
Hamilton et al., 2021,
2022). SunFlower is a web-based application validated against SolTrace with high precision and user-friendliness (
Richter et al., 2018). CombiCSP is an open-source dynamic model of CSP plants, SPT systems included (
Arnaoutakis et al., 2022). Other popular tools are Tonatiuh, SolTrace, Tracer, Solstice, and Heliosim, which offer strong optical modeling features (
Wang et al., 2020).
Every simulation tool has certain capabilities associated with a particular facet of the optical modeling of SPTs. SOLARPILOT facilitates advanced strategies for heliostat aiming and has Python API support for executing automated simulations (
Hamilton et al., 2021,
2022). SunFlower has cross-validation support with SolTrace and guarantees stable simulation accuracy (
Richter et al., 2018). CombiCSP has dynamic system analysis and can predict annual energy yield (
Arnaoutakis et al., 2022). Tools like Tonatiuh, SolTrace, Tracer, and Heliosim are some of the software that enable precise modeling of sunshape and reflector surface slope error, which are critical for precise optical behavior analysis (
Wang et al., 2020).
Though each of these programs offers powerful functionality for SPT modeling, each has a set of strengths and weaknesses. SOLARPILOT is extremely flexible and lends itself to advanced research applications, but at the cost of high learning overhead (
Hamilton et al., 2021,
2022). SunFlower features an easy-to-use web-based interface and high accuracy, but the web-based interface can potentially limit accessibility (
Richter et al., 2018). CombiCSP is an open-source software and has the advantage of new configurations, but is in the initial phase of development and might not have some advanced features (
Arnaoutakis et al., 2022). Tonatiuh, SolTrace, Tracer, and Heliosim are established and validated tools, but have inconsistencies in implementation that can introduce additional errors when simulating large scales (
Wang et al., 2020).
For the selection of an optical simulation tool for SPT applications, accuracy, with validated tools like SOLARPILOT and SunFlower reporting precise results; usability, with user-friendly interfaces and APIs in SunFlower and CoPylot allowing ease of use (
Richter et al., 2018;
Hamilton et al., 2022); and flexibility, as in open-source tools like CombiCSP, which offer opportunities for customization to new CSP configurations (
Arnaoutakis et al., 2022), are primary considerations. Validation against experimental data or model comparisons also improves reliability (
Schöttl et al., 2016;
Richter et al., 2018;
Hamilton et al., 2022). Based on these parameters, researchers can decide which software is best suited to their needs. Table 2 presents a comparative overview of significant optical simulation tools for SPT research.
A comprehensive evaluation of optical simulation tools for SPT systems reveals essential trade-offs between accuracy, computational performance, and flexibility. Wagner and Wendelin (2018) compared DELSOL3 and SolarPILOT for a SPT system with ~8945 heliostats. Both showed identical total efficiency (58.8%) and close agreement in cosine (80.3% vs. 80.6%), blocking (99.3% vs. 99.0%), and intercept efficiency (96.3% vs. 96.0%), as shown in Fig. 4(a). While DELSOL3 uses zonal grouping for speed, SolarPILOT models each heliostat individually and integrates SolTrace, improving spatial accuracy. In a performance-focused comparison, Yellowhair et al. (
2014) compared DELSOL, HELIOS, SolTrace, and Tonatiuh using a shared flat plate receiver and heliostat field. All tools predicted similar total incident power (7.17–7.37 kW), with Tonatiuh and SolTrace reporting the highest (7.37 and 7.34 kW). Average flux ranged from 49.3 kW/m
2 (HELIOS) to 62.4 kW/m
2 (SolTrace). DELSOL produced the highest peak flux (178 kW/m
2), likely due to smoothing from its analytical approximation, compared to 168 kW/m
2 (SolTrace) and 176 kW/m
2 (Tonatiuh) (Fig. 4(b)). Although all tools yielded similar total power predictions, SolTrace and Tonatiuh offered more detailed flux resolution, making them better suited for receiver design and thermal analysis. In contrast, DELSOL’s analytical method, while computationally faster, resulted in smoothed flux profiles and potential overestimation of peak flux. Thus, SolTrace is preferred for high-accuracy studies, while DELSOL is more efficient for layout and optimization tasks.
Salomé et al. (
2013). conducted a validation study comparing SolTrace and HFCAL using the THEMIS solar tower under a central aiming strategy. Both tools predicted comparable peak flux densities (Fig. 4(c)), with 2705 suns for SolTrace and 2603 suns for HFCAL, showing a deviation of just 3.4%. They concluded that while SolTrace, based on MCRT, offered higher spatial resolution, it requires significantly longer computation times. In contrast, HFCAL, using a convolution-based approach, was approximately ten times faster, making it more suitable for iterative or real-time optimization tasks in heliostat field simulations. In another study, Kumar and Krishna (
2016) compared an analytical model with MCRT-based SolTrace using a single 1 m
2 heliostat aimed at a central receiver. Simulations from 9 a.m. to 5 p.m. on April 1st showed solar flux deviations within ± 5% (Fig. 4(d)). At 11:00 a.m., both methods produced closely matching flux values. While the analytical approach offers quick flux estimates, SolTrace better captures optical effects like cosine loss, spillage, and attenuation. Shadowing and blocking were excluded due to the single-heliostat configuration.
In another study, Pujol-Nadal and Cardona (
2023) benchmarked their new open-source OTSun tool against Tonatiuh. They computed the radiation flux using these tools and found remarkable similarity in the results with peak power of 9.81 MW/m
2 and 9.68 MW/m
2 in OTsunWebApp and Tonatiuh, respectively, as shown in Figs. 5(a) and 5(b). The difference in the results was significantly lower at a higher number of rays (Fig. 5(c)). The main disadvantage of OTSunWebApp compared to Tonatiuh is the computational time, which is significantly higher in the case of OTSunWebApp. Moreover, Richter et al. (
2018) introduced SunFlower, a hybrid ray-tracing tool and web-based ray-tracing environment, and validated it against SolTrace. They found that Sunflower may achieve 99.9% accuracy at about 100 times faster compared to the MCRT-based SolTrace (Fig. 5(d)).
In a broader comparison, Jafrancesco et al. (
2018) analyzed Tonatiuh, SolTrace, TracePro, and CRS4-2 using the SSPS-CRS test case. They conducted qualitative and quantitative analyses across three scenarios (S1, S2, S3) defined by solar conditions. Tonatiuh, SolTrace, and CRS4-2 show close agreement in total incident power (e.g., ~2.63–2.65 MW for S1), with differences under 1.5%, while TracePro overestimates power by 10%–13% in S2 and S3, likely due to its non-CSP-specific design (Figs. 5(e)–5(g)). Wang et al. (
2018) found that Tonatiuh and SolTrace produced nearly identical absorbed energy under the Buie sunshape and Gaussian slope error, while SolarPILOT differed by less than 0.2%. However, SolarPILOT showed noticeable flux shape discrepancies due to its simplified Hermite expansion, unlike the more accurate Monte Carlo-based SolTrace and Tonatiuh.
SolarPILOT has also been implemented in large-scale SPT projects, such as the 100 MWe Upington plant, where its integration with tariff-based heliostat selection strategies facilitated comprehensive system-level LCOE analysis (
Pidaparthi et al., 2017). TracePro has been employed in conjunction with MATLAB for time-resolved ray tracing across an SPT receiver throughout the year (
García-Lara et al., 2022). The Campo code, tailored for large collector fields, was scaled up from Gemasolar (2650 heliostats) to Noor III (7400 heliostats), handling over 10,500 heliostats with optimized trimming and staggered layouts in under three minutes (
Collado and Guallar, 2018). Similarly, FluxTracer functions as a post-processing tool for MCRT simulations, enabling spatially resolved flux distribution analysis in 3D receiver regions (
Blanco et al., 2019). Customized ray-tracing approaches have also been proposed using MATLAB for faster computation and enhanced incorporation of slope error maps or deflectometry data, making them suitable for high-resolution flux prediction (
Bravo Gonzalo et al., 2019). These advancements reflect a shift toward modular, highly accurate, and computationally efficient optical modeling tools tailored for experimental integration and optimization. Zemax has been utilized for high-resolution optical modeling and integration with heat transfer analyses (
Wei et al., 2011).
Building on the tool-specific validations above, the optical simulation tools used in SPT studies differ significantly in modeling approach, accuracy, and usability. DELSOL and SolarPILOT are preferred for rapid heliostat field layout and flux optimization using simplified ray tracing methods, though they lack advanced optical modeling. In contrast, Tonatiuh, HFLCAL, and Heliosim employ Monte Carlo or hybrid approaches that better capture sunshape effects and spatial flux distribution, albeit with higher computational demands. Optical engineering tools like TracePro, ZEMAX, and OptiCAD support advanced features such as scattering and thin-film modeling, but are not tailored for CSP field configurations. FRED provides broad physical optics capabilities, though it is commercial and less specialized for SPT applications. Open-source tools like SolTrace and Tonatiuh offer accessible, research-oriented platforms but may lack GUI refinement or automation. Radiance and Solergy extend utility into thermal and daylight modeling at the expense of heliostat-specific accuracy. Ultimately, tool selection depends on the required balance between speed, accuracy, system complexity, and integration needs—DELSOL and SolarPILOT are suited for layout optimization, Tonatiuh and HFLCAL for detailed flux studies, and TracePro or ZEMAX for component-level optical design.
4 Thermal and CFD simulation for receiver heat transfer analysis
4.1 Importance of thermal and CFD modeling
Thermal simulation and CFD modeling are necessary for the optimization of SPT performance via analysis of heat transfer phenomena, fluid dynamics, and receiver efficiency. Various heat transfer fluids (HTFs), such as molten salt, supercritical CO
2, and direct air, are employed in SPT receivers with associated challenges and benefits. Molten salt is the most widely used because of its high specific heat, but it is hindered by freezing and corrosion problems. Supercritical CO
2 has better thermophysical performance but needs to be contained at high pressure, whereas direct air systems are simple but are plagued by poorer heat retention (
Pérez-Álvarez et al., 2019,
2020;
Zheng et al., 2020).
Radiation, convection, and conduction are the primary modes of heat transfer in thermal receivers, as shown in Fig. 6. The dominant mode is radiation, with the concentrated solar energy being absorbed and re-radiated by the receiver surface, which requires sophisticated radiation modeling methods like MCRT. Convective heat transfer takes place both within the working fluid and between the receiver and ambient air, based on parameters like wind speed and turbulence. Conduction, being less than radiation and convection, influences thermal efficiency by receiver wall material and insulation (
Ho, 2014;
Chang et al., 2014;
Marti et al., 2020). Research has proven that the performance of multi-channel volumetric air receivers relies heavily on convective heat transfer enhancement methods (
Jung et al., 2013).
While CFD simulations are an important tool in the design and optimization of SPT receivers, there are different challenges that compromise the accuracy of modeling. It requires high computational power due to the complex interaction between radiation, convection, and turbulence. Mesh refinement is needed for detailed thermal gradient capture, but finer meshes mean increased computational costs and processing time. Additionally, turbulence modeling is challenging since the choice of a suitable model (such as., k-ε, RNG k-ε, or Reynolds Stress Model) requires compromises between accuracy and computational expense (
Marti et al., 2020;
Drikakis and Dbouk, 2022;
Aswini Devi et al., 2024). Future thermal and CFD simulations of SPT systems are required to be enhanced by high-performance computing, hybrid modeling techniques, and artificial intelligence (AI) optimization with respect to accuracy and feasibility.
4.2 Overview of CFD and thermal simulation tools
In the design and optimization of SPT systems, precise simulation and performance analysis are crucial to achieving high-efficiency operation. There are a variety of simulation tools used to evaluate various aspects of SPT plants, ranging from system performance to the individual component level, such as heliostat field layout and thermal storage systems. Table 3 provides a comparison of some of the widely used thermal simulation tools in the SPT industry with different functionalities to model, optimize, and analyze the performance. These tools range from generic simulation software to component-specific software, i.e., heliostat or thermal storage system software.
4.3 Comparative evaluation of CFD-based simulations
Table 4 presents a detailed comparison of three widely used simulation tools: OpenFOAM, ANSYS Fluent, and COMSOL Multiphysics, evaluated based on parameters such as tool overview, strengths and weaknesses, accuracy, ease of use, computational requirements, and their use in SPTs.
Several studies have demonstrated that simulation tools are extensively applied in the development and evaluation of SPT systems across different stages of design and operation. Mishra et al. (
2021) performed a coupled optical-thermal simulation of a cavity linear receiver (CLR) using SolTrace for ray tracing and ANSYS Fluent for thermal CFD analysis. They validated the model against experimental temperature data from a prototype. As shown in Fig. 7(a), the numerical model predicted backside tube temperatures with high accuracy, with a maximum deviation of 13.8% and an average error of 3.59%. This combined modeling approach successfully captured local temperature non-uniformities arising from asymmetric solar flux distributions, which highlights the importance of integrated ray-tracing and CFD methods for reliable thermal predictions in advanced receiver geometries. Zhang et al. (
2014) validated a Dymola-based dynamic model of the 1 MWe Badaling CSP tower using experimental data from actual plant operation. The model captured key thermal parameters, including steam flow rate, drum pressure, and wall temperatures, with deviations mostly under 2.5% (Fig. 7(b)). The predicted trends closely followed measured data during transient conditions, confirming the model’s reliability for dynamic performance analysis.
Scott et al. (
2017) introduced SolarTherm, an open-source Modelica-based simulator for CSP systems, and validated its performance against SAM. For the SPT plant modeled using physical equations and simulated in Dymola, SolarTherm predicted 2.95% higher annual power due to extended operation at lower DNI and improved control strategies (Fig. 7(c)). These comparisons demonstrate SolarTherm’s flexibility for simulating advanced receiver concepts and its consistency with SAM on annual performance metrics (Fig. 7(d)). Flesch et al. (
2017) validated their Dymola-based dynamic receiver model against both CFD and experimental results for a molten salt SPT. The model closely matched CFD-predicted fluid temperatures and heat losses (within 3%), though minor wall temperature deviations occurred due to simplified discretization (Fig. 7(e) and 7(f)). Tagle-Salazar et al. (
2024) developed a transient SPT plant model in OpenModelica with molten salt TES and compared it with SAM. The results demonstrated that the annual energy output differed by less than 1% compared to SAM (853.15 GWh vs. 859.76 GWh). The similar trends were observed for net power with SAM and OpenModelica (OM) (Fig. 7(g)). They concluded that results highlight OpenModelica’s superior capability in simulating real-time TES dynamics and operational behavior in CSP systems.
STAR-CCM + was used by Schmitt et al. (
2019) to assess the thermal and aerodynamic behavior of a receiver dome at the Crescent Dunes facility, finding that the dome reduced convective and radiative losses by up to 1.92 MW. Gentile et al. (
2022) employed SolarReceiver2D, a Modelica-based tool, to model transient thermal behavior in external receivers, capturing circumferential and axial gradients during flux transients. System-level assessments are exemplified by (
Flueckiger et al., 2014b), who integrated DELSOL for optical design, SOLERGY for receiver simulation, and a custom thermocline storage model to simulate a 100 MWe SPT system. Their model achieved a 0.531 annual capacity factor with over 99% storage effectiveness using year-long solar data.
Recent developments have significantly expanded the thermal simulation toolbox beyond traditional platforms. SolarTherm, a flexible Modelica-based simulator, supports dynamic modeling of CSP systems with open-source extensibility and validated agreement with SAM (
Scott et al., 2017). CombiCSP enables year-round performance simulation of solar tower and parabolic trough plants with a modular Python-based framework (
Arnaoutakis et al., 2022). ThermoSysPro, a Modelica library released under open-source terms, provides robust modeling components for solar fields, TES, and heliostats (
Hefni, 2014). Such tools enable tailored CSP plant simulations with enhanced adaptability for novel configurations and multiphysics co-simulation. These platforms are increasingly used for techno-economic evaluation and control strategy testing under transient and hybrid operating conditions.
In addition to commercial tools such as ANSYS Fluent and COMSOL, open-source codes continue to gain traction for high-accuracy CFD simulations of solar receivers. OpenFOAM has been widely used to model porous volumetric receivers, radiative heat transfer, and multiphase particle transport using Euler–Euler and discrete element methods (
Marti et al., 2015;
Barreto et al., 2019). Coupling MCRT with OpenFOAM enables comprehensive multiphysics simulations of central receivers, as demonstrated in the ISTORE and PROTEAS projects (
Votyakov and Papanicolas, 2017). LIGGHTS-PUBLIC, a DEM-based code, has been employed to analyze trickle-flow heat exchanger geometries for particle-based receivers (
Reichart et al., 2021). These developments highlight the potential of open-source tools to simulate complex thermal, radiative, and fluid dynamic interactions in next-generation receiver systems.
Overall, different tools exhibit distinct strengths depending on the modeling scope and application area. Thermolib and SolarReceiver2D offer detailed transient and thermodynamic modeling for direct steam generation (DSG) and receiver-level analysis, but face limitations in validation and full-system integration, respectively. For high-fidelity modeling, COMSOL Multiphysics and STAR-CCM + support advanced computational fluid dynamics (CFD) simulations, capturing convective and structural phenomena, though at the cost of computational load and required expertise. OpenModelica allows flexible thermal storage modeling but requires extensive input handling. Custom-coded models, such as FORTRAN-based approaches, provide deep system-specific insight but lack user-friendliness and standardization. Among CFD platforms, OpenFOAM offers unmatched customization without licensing fees but requires advanced knowledge, limiting industrial usability. ANSYS Fluent enables accurate receiver modeling with advanced turbulence and radiation models, especially with ray-tracing inputs, though it is costly and resource-intensive. COMSOL, with its intuitive interface and multiphysics environment, is well-suited for academic and system-level coupling but requires additional modules for CSP-specific tasks. Ultimately, tool selection should reflect the balance between detail, usability, and simulation objectives, whether targeting full-system performance, detailed receiver physics, or hybrid process integration.
5 System-level simulation for techno-economic and operational analysis
5.1 Importance of system-level modeling
System-level modeling is fundamental to the design, operation, and economic analysis of SPT systems. It has an important role to play in reducing the levelized cost of electricity (LCOE) by optimizing key design parameters, including receiver size, solar multiple, and thermal storage capacity. Research indicates that a solar multiple of 3 with 16 h of thermocline storage capacity leads to an LCOE of 12.2 ¢/kWhe, demonstrating that there is a need for a detailed economic optimization (
Flueckiger et al., 2014a). Additionally, these models aid in making investment decisions by assessing economic metrics like the payback period and internal rate of return, and offer useful feedback for financial planning (
Behar et al., 2020). Sensitivity analysis in these models allows for determining the effect of parameters such as degradation rates and discount rates on overall system feasibility, and scenario-based uncertainty handling guarantees robust performance under diverse operating conditions (
Flueckiger et al., 2013;
Behar et al., 2020). Similar simulation approaches have also been applied in hybrid renewable systems. For instance, Hossen and Shezan (2019) used simulation-based optimization to design a solar-wind-biomass hybrid system tailored for islanded operation, highlighting the versatility of such tools for off-grid contexts.
Optimization of performance is the other critical feature of system-level modeling, as it allows for the assessment of yearly measures like capacity factor and thermal energy rejection to maintain efficient plant operations (
Jiang et al., 2024). These models also make it possible to predict dynamic responses to different environmental conditions, making it easier to pre-emptively make changes in system control mechanisms (
Zhang et al., 2017). Dispatch strategy evaluation in hybrid systems has been demonstrated by Ishraque et al. (
2024), who employed simulation-based optimization for grid-tied hybrid energy systems at institutional scales, reinforcing the role of dispatch-aware models. System-level modeling of hybrid energy systems allows the integration of solar thermal energy into traditional power generation from power plants like coal, enabling it to be more efficient and less fossil fuel-dependent (
Zhang et al., 2022). Thermal energy storage (TES) is particularly critical, as it stabilizes the energy output and enhances peak load regulation, thus enabling SPT plants to be more dependable for support on the grid (
Wang et al., 2023).
On the design front, system-level modeling is used for optimizing heliostat fields and receiver designs for maximum efficiency and component longevity (
Rea et al., 2018). It also enables the creation of modular SPT systems that are dispatchable and scalable and thus cost-competitive with traditional energy sources (
Zhang et al., 2020). Operational analysis is also greatly facilitated by these models since they provide better control measures for the fine regulation of systems. For instance, control of water level in evaporators is required for steady operation with varying input conditions, and this may be optimized by using system models (
Yu et al. 2016). Such simulation frameworks have also been explored for smart residential systems to optimize energy flows (
Rana et al. 2022), demonstrating adaptability across different scales. These simulations also give vital information on peak regulation capacity, which is accountable for grid stability and enhancing the reliability of solar power generation (
Wang et al., 2023). System-level modeling combines techno-economic and operating aspects in the optimization of SPT systems. By merging economic, performance, design, and operating considerations, it offers knowledgeable decision-making for efficiency, cost minimization, and sustainability. Figure 8 shows the workflow for the system-level simulation in SAM.
5.2 Overview of system-level simulation software
Table 5 provides a detailed comparison of three widely utilized simulation tools: TRNSYS, SAM, and Dymola. These tools are used for modeling and simulating SPT systems and their integration with thermal energy storage, economic calculation, and performance modeling.
Table 5 outlines the capabilities and distinctions among the most widely used system-level modeling tools for SPTs. A critical comparison highlights the practical trade-offs and suitability of each tool based on modeling depth, usability, and integration needs. System-level modeling tools such as TRNSYS, SAM, and Dymola vary significantly in scope, specialization, and usability, making tool selection highly dependent on project objectives and user expertise. TRNSYS offers extensive flexibility and is capable of detailed transient simulations, making it suitable for analyzing solar thermal power systems under variable climatic conditions. However, its general-purpose architecture and complex setup often result in longer simulation times and a steeper learning curve, especially when modeling specific CSP configurations. In contrast, SAM, developed by NREL, is purpose-built for renewable energy applications and offers robust financial modeling, dispatch strategy analysis, and an intuitive interface. It also supports thermal storage and system performance modeling with significantly faster simulation times due to its optimized engine. However, its adaptability to detailed physics-based receiver or optical modeling is limited, relying instead on integrated tools like SolarPILOT.
Dymola, which is based on the Modelica language, excels in multi-domain, multiphysics simulation and provides powerful capabilities for modeling complex, dynamic systems involving thermal, fluid, and control components. Its strength lies in high-accuracy modeling and modular integration, making it particularly valuable for advanced research and system prototyping. However, it requires substantial system knowledge, detailed input parameterization, and involves commercial licensing costs, which may limit its accessibility. A key trade-off emerges between model detail and ease of use: TRNSYS offers high customization at the expense of complexity; SAM delivers streamlined performance and cost analysis with less physical detail; and Dymola supports advanced modeling and integration but demands greater user expertise and computational resources. Ultimately, the most appropriate tool should be selected based on specific simulation objectives, although integrating multiple tools can provide a more comprehensive analysis of SPT systems.
To enhance techno-economic modeling, several modular and open-source tools now complement existing platforms like Dymola and SAM. MoSES offers a techno-economic optimization framework for hybrid PV-CSP systems, integrating design assessment and layout benchmarking (
Guccione and Guedez, 2024). OpenModelica has been effectively used for annual performance simulations of molten salt solar towers, with less than 1% deviation from commercial results (
Tagle-Salazar et al., 2024). DWSIM, in co-simulation with SAM, supports design and optimization of CSP-ORC hybrid plants with parametric sensitivity analysis (
Sigue et al., 2023). These platforms offer enhanced modularity, accessibility, and customization, enabling broader research and application beyond licensed environments.
5.3 Comparative evaluation of system-level models
Calle et al. reated a dynamic system-level model of a new SPT plant with liquid sodium as an HTF, phase-change material (PCM) for thermal energy storage, and a supercritical CO
2 (sCO
2) Brayton cycle as a power generation cycle (
De La Calle et al., 2018;
2020). They simulated it in Dymola after modeling it in Modelica to examine the thermodynamic performance and economic viability of the plant. Modelica model and plant layout are provided in Fig. 9. They concluded that a liquid sodium receiver, owing to its capability to operate up to a temperature of 890°C, enables higher thermal efficiency and compact receiver size, which can lower the LCOE compared to traditional molten salt-based SPT plants. Yet, the reactivity of sodium and its low specific heat capacity are disadvantages, which are alleviated by the inclusion of PCM storage with stable high-temperature storage and the reduction of system inertia. The sCO
2 recompression Brayton cycle also enhances efficiency and enables dry cooling, which is highly important for deployment in water-poor areas. The results indicate that the proposed design is lower in LCOE than traditional CST power plants, and 8-h PCM storage is economically the most feasible. They demonstrated through system-level simulation the potential of new-generation HTFs and power cycles to make SPT power plants more economical and put them on a par with traditional and renewable ones.
The TRNSYS modeling of a 1 MW Solar Thermal Tower Power Plant (STTPP) is investigated by Raza et al. (
2018) to identify the optimization of the system parameters for cost minimization with effective power generation. The study predominantly deals with reducing the number of heliostats, which is a significant portion of the plant cost. The schematic of STTPP and the model developed in TRNSYS are shown in Figs. 10(a) and 10(b), respectively. From TRNSYS simulations, the research identified that a minimum of 62 heliostats is needed to maintain a 13000 kg/h steam flow rate, which is needed by the Siemens SST-060 steam turbine. Different HTFs, such as water, CO
2, and molten salt, were experimented on to examine their effects on system performance. Whereas water needed the maximum number of heliostats because of its high specific heat, CO
2 minimized the number of heliostats considerably to 23, although there were storage and handling issues. Molten salt was found to be most appropriate since it allowed thermal energy storage and brought down the heliostat requirement to 26 for a flow rate of 60000 kg/h. Further system optimization showed that, with some fluctuation in steam flow rate and pressure, the heliostat number could be changed, but at the cost of higher input pressure to the turbine and impacting overall viability. The analysis shows that TRNSYS can accurately model transient system behavior and thus allow for meaningful conclusions in solar thermal power plant design and optimization.
Hamilton et al. (
2020) investigated the performance of molten salt-driven Rankine cycles in SPT systems under off-design conditions. They incorporated a validated Rankine cycle model into the SAM to improve off-design performance predictions. The system configuration and the flow of information in the SAM model are shown in Figs. 11(a) and 11(b), respectively. Key parameters, including inlet molten salt temperature, mass flow rate, and ambient temperature, have been examined that exert a significant influence on CSP plant dispatch strategy and efficiency. The results revealed that off-design cycle performance affects optimum subsystem sizing and dispatch decisions, especially in CSP-PV hybrid systems, where CSP has to compensate for photovoltaic power generation. Two operating approaches, sliding pressure and constant pressure, were studied, with the working sliding pressure exhibiting higher part-load efficiency. The research sets down the fact that proper off-design modeling is significant for realistic estimation of performance as well as economic viability analysis of CSP plants. Through dispatch optimization, the research reveals the importance played by the off-design cycling operation in the yearly performance terms, system designing, and the economic results, which illuminates optimizing the integration of CSP in the energy network.
SAM, TRNSYS, and Dymola all have various strengths in SPT system modeling. SAM is strong in economic feasibility, dispatch optimization, and financial modeling, and would be the best one to utilize for LCOE and grid integration strategy determination. TRNSYS would be best for full transient thermal analysis so that thorough HTF, heliostat field design, and energy storage analysis can be performed, but it does not include financial modeling. Dymola, through its object-oriented, high-fidelity dynamic simulations, enables customized system modeling and is suitable for investigating novel concepts such as liquid sodium receivers, sCO2 Brayton cycles, and PCM-based thermal storage. As SAM is most suitable for economic analysis, TRNSYS is for thermal behavior modeling, and Dymola is for new system design flexibility, the tool selection is a function of the study’s purpose.
While simulation tools like SAM, TRNSYS, and Dymola are widely used for techno-economic assessment of SPT systems, their real-world applicability varies significantly depending on how well they incorporate policy, market, and geographical factors. Among these tools, SAM is particularly notable for its built-in support for region-specific financial incentives, electricity pricing structures, and dispatch optimization, making it highly suitable for investment planning and policy analysis (
Blair et al., 2018;
NREL, 2025). In contrast, TRNSYS offers highly flexible thermal and dynamic system modeling capabilities but lacks built-in support for economic or policy parameters, often requiring integration with external tools like Excel or HOMER to evaluate site-specific economic feasibility (
Raza et al., 2018). Dymola, while powerful in modeling complex thermofluid and control systems through the Modelica language, does not provide embedded market or policy modeling features and must be extended with user-defined modules or co-simulation approaches to capture such dynamics (
Dassault Systèmes, 2025). These differences suggest that while TRNSYS and Dymola excel in engineering-level simulations, SAM is more tailored for scenario analysis and decision-making in real-world planning environments. Future tool development should aim to bridge this gap by integrating dynamic pricing models, GIS-based resource data, and policy modules to enhance the practical relevance of techno-economic evaluations.
6 Integration of optical, thermal, and system models: Hybrid approaches
6.1 Challenges in multi-physics coupling
The combination of optical, thermal, and system models in SPT applications is riddled with challenges due to the intricate interaction among various physical domains. In optical models, ray-tracing techniques are needed to simulate solar irradiation and its interaction with receiver surfaces. Such models need to account for properties such as reflectivity, absorptivity, and emissivity, which may be temperature-dependent and surface-property-dependent (
Craig et al., 2016;
Votyakov and Papanicolas, 2017). Thermal models put greater emphasis, however, on heat transfer mechanisms by conduction, convection, and radiation to achieve good performance of units such as solar receivers at high temperatures and flux densities (
Picotti et al., 2020;
Hering et al., 2021;
Shuai et al., 2024). System models incorporate the dynamic response and interaction between devices such as the receiver and heliostat field and provide a more general overview of performance (
Picotti et al., 2020;
Shuai et al., 2024).
The data communication, numerical stability, and computational resource demands make the integration of optical, CFD, and system models in SPT challenging. Among the technical challenges of coupling such models, one of the most important is in data communication. Optical models produce data that need to be precise boundary conditions for CFD models, and effective data exchange mechanisms are thus required (
Gatzhammer et al., 2010;
Craig et al., 2016). Further, mapping and coupling algorithms need to properly couple various spatial and temporal scales, and multiscale problem-solving is especially demanding (
Hegewald et al., 2008;
Blom et al., 2016). Additionally, maintaining modularity while providing frictionless interaction among various models continues to be a key software engineering challenge (
Hegewald et al., 2008;
Gatzhammer et al., 2010). The second essential challenge is the computational cost of comprehensive multi-physics simulations. They require high computational resources and parallel processing capability in order to be feasible for applications on a large scale (
Blom et al., 2016;
Chimakurthi et al., 2018). Numerical stability and scalability of coupled models are also essential to maintain because instability will lead to poor prediction as well as inefficient system operation (
Uekermann et al. 2014;
Blom et al. 2016).
To address these challenges, recent hybrid and modular modeling approaches have been introduced. In partitioned simulation workflows, specialized tools are applied to individual domains (such as SolTrace for optics, Fluent for CFD), and their results are coupled using modular interfaces. For example, MCRT methods are integrated with CFD solvers to simulate solar receivers by importing spatially varying flux distributions into the energy equation, enabling precise modeling of radiation-convection-conduction interactions (
Votyakov and Papanicolas, 2017). Object-oriented modeling platforms such as Dymola and OpenModelica employ equation-based, modular structures that support dynamic co-simulation of thermal, fluid, and control components in a unified framework (
Picotti et al., 2020). Deep learning has also been used to develop surrogate models capable of predicting heat transfer behavior under varying conditions, reducing the reliance on repeated high-fidelity simulations (
Shuai et al., 2024). Commercial platforms like ANSYS System Coupling enable highly integrated multiphysics simulations for fluid-thermal-structural problems, though they typically require high computational resources and licenses (
Chimakurthi et al., 2018).
Building on these developments, several practical workflows have emerged to implement multiphysics coupling in SPT simulation. For instance, ray-tracing outputs from SolTrace or Tonatiuh can be exported as high-resolution flux maps and mapped as boundary heat sources for CFD solvers like ANSYS Fluent or OpenFOAM using mesh projection or interpolation routines. The preCICE framework enables surface coupling and time-step synchronization between solvers with non-matching grids or resolution levels, making it suitable for modular co-simulation involving radiation and fluid dynamics (
Gatzhammer et al., 2010;
Blom et al., 2016). Similarly, object-oriented platforms like OpenModelica allow thermal and control logic to be modeled alongside fluid dynamics in a single acausal structure, supporting real-time plant simulations. AI-based surrogate models, trained on high-accuracy simulation data, are increasingly being used to reduce computational load in optical and thermal analysis, enabling near real-time optimization and control. These practical strategies offer technically feasible and scalable solutions for achieving highly integrated SPT simulation environments.
Figure 12 depicts inputs, outputs, and difficulties in consolidating simulation software for SPT simulations. Multiscale modeling combines optical, thermal, and system models and simulates solar flux, heat transfer, and efficiency. All models possess resolution differences, time scales, and computational complexity. Optical models calculate the solar flux distribution, thermal models examine the heat transfer, and system models analyze the overall performance. Problem-solving with integration increases accuracy, decreases computational costs, and enhances SPT system optimization and reliability.
One of the major barriers to interoperability among simulation tools is the incompatibility of data formats and assumptions across different domains. For instance, optical simulation tools such as SolTrace and Tonatiuh typically output solar flux distributions in custom formats (such as .dat, .csv, or image-based matrices) that are not directly compatible with thermal or CFD tools like ANSYS Fluent or OpenFOAM, necessitating intermediate mapping or manual interpolation routines (
Khalsa and Ho, 2011;
Jafrancesco et al., 2018). Furthermore, discrepancies in geometry representation present significant challenges in model coupling. Optical simulation tools often utilize simplified or idealized planar surfaces, whereas CFD solvers require detailed, three-dimensional discretized geometries, typically in formats such as .stl or .msh (
Votyakov and Papanicolas, 2017). SolarPILOT and Tonatiuh, for instance, generate high-resolution flux maps that must be reformatted into boundary heat flux inputs for CFD solvers, a process prone to interpolation errors or misalignment issues (
Wang et al., 2016). In addition, tools often operate on mismatched temporal resolutions; optical simulations typically assume steady-state irradiance conditions, while thermal and system-level models (such as COMSOL, SAM, or TRNSYS) simulate dynamic behavior over hourly or daily cycles, creating synchronization challenges (
Picotti et al., 2020). In addition, modeling assumptions often vary significantly. For instance, MCRT and discrete ordinates methods may produce different radiation flux profiles, while convective losses, typically ignored in optical simulations, are essential in CFD-based thermal analysis. Finally, lack of standardized interfaces or APIs often forces users to rely on manual data transfer, scripting bridges, or customized coupling approaches. These challenges significantly hinder real-time or automated simulation workflows, increase the burden on researchers, and reduce reproducibility. Recent studies have called for modular, object-oriented platforms or preCICE-based coupling strategies to streamline such multi-physics integration.
While coupling between optical, thermal, and system-level models has been explored through partitioned strategies and co-simulation frameworks, the process remains fragmented, resource-intensive, and error-prone. A compelling future direction is the development of a single, integrated simulation environment that inherently supports ray tracing, CFD-based thermal modeling, and system-level performance analysis in a seamless architecture. Such a tool would eliminate the need for intermediate file conversion, time-step synchronization, and mesh projection routines. Inspired by modular object-oriented environments like Modelica, this platform would feature a common data backbone, unified solver management, and real-time co-simulation capabilities. It could further embed AI-powered surrogate models to reduce computational demand during runtime. This vision, though technically ambitious, represents a transformative shift in CSP research and would significantly improve model accuracy, development speed, and deployment in digital twin environments for solar tower operations.
6.2 Case studies of integrated models
Several studies have demonstrated the integration of ray tracing with CFD for further receiver analysis, particularly for solar power systems. Khalsa and Ho (
2011) demonstrated a procedure wherein the radiance distributions computed on the receiver aperture were converted to radiance boundary conditions to be used in CFD simulations. This strategy enabled the CFD model to concentrate on the cavity receiver domain with complex beam features and interactions with participating media like high-temperature falling solid particles. This strategy was verified with the ANSYS Fluent examples of dish concentrators and heliostat fields. Crocker and Miller (
2012) coupled an in-house MCRT radiation model with ANSYS FLUENT fluid dynamics capability. This combination enabled the use of more realistic geometries, turbulence models, and smaller grid sizing, giving a comprehensive view of the heat transfer and temperature profiles in the receiver. The iterative exchange between the MCRT code and CFD simulation facilitated the correct convergence and realistic solar flux distributions. Table 6 gives some of the studies that employed the hybrid simulation methods for SPT applications.
7 Emerging trends: AI and machine learning (ML) in SPT simulations
7.1 AI for optical and heliostat optimization
Artificial Intelligence (AI) has transformed heliostat field optimization in SPTs due to higher accuracy and efficiency. The application of neural networks and genetic algorithms (GAs) is being utilized more for optimizing heliostat layout and calibration with higher precision and flexibility than conventional techniques. Neural networks, especially deep learning architectures, have also been shown to be high-performing for heliostat calibration. Evidence from research indicates that a combination of Self-Normalizing Neural Networks and transfer learning produces a test accuracy of 0.42mrad, much better compared to conventional regression-based calibration methods (
Pargmann et al., 2021). A different research work using neural networks for heliostat calibration has demonstrated that error levels due to noisy data are less than 0.02mrad, with negligible practical impact (
Sievers et al., 2025). GAs are also essential for the optimization of neural network parameters, including architecture and activation functions, for the minimization of prediction errors. Their combination with neural networks has also been effective in predicting solar power generation, with further system optimization and efficiency (
Sangeetha et al., 2023). Similar optimization techniques have been adapted to embedded systems in other solar applications; for example, Punitha et al. (
2024) implemented an AI-enhanced algorithm for Raspberry Pi Pico controllers in solar tree systems, supporting real-time control with low-cost hardware. The machine learning-based heliostat optimization process follows a systematic approach as shown in Fig. 13.
Wang et al. (
2024) presented an AI-powered method for enhancing heliostat aiming strategies in SPT systems through an improved Particle Swarm Optimization (IUPSO) algorithm. The structure of the real-time aiming strategy optimization and the flowchart of IUPSO are shown in Figs. 14 (a) and 14(b), respectively. They overcame the computational constraints of conventional methods by coupling real-time AI-based decision-making with the utilization of simulation software such as Tonatiuh for flux mapping and a dynamic cloud motion model. IUPSO improved traditional PSO with a population inheritance mechanism according to past optimization states to improve convergence speed and an individual update mechanism according to real-time conditions to adaptively adjust heliostat positions. The algorithm was far superior to GA, bat algorithm (BA), and whale optimization algorithm (WOA) in computational efficiency, robustness, and optimization speed, and it is extremely appropriate for real-time heliostat control simulations.
7.2 AI for heat transfer and performance prediction
ML models are also being used to predict crucial performance metrics in solar power generation systems, including Direct Normal Irradiance (DNI), thermal losses, and heat exchanger performance. Artificial Neural Networks (ANNs) have been effectively utilized to estimate DNI, enabling optimized energy capture for solar irradiance prediction (
Yadav et al., 2021). Furthermore, AI-based techniques like Adaptive Neuro-Fuzzy Inference System (ANFIS) coupled with biogeography-based optimization algorithms have been implemented in the prediction and simulation of thermal losses to provide enhanced thermal efficiency. Heat exchanger performance has also been optimized using AI. Random Forest Regressors have been used effectively for the prediction improvement of power output and dynamic real-time tuning of system parameters (
Khosravi et al., 2021). These AI-driven technologies enhance the effectiveness, cost-savings, and efficiency of solar power plants through cost savings and increased energy production (
Gul et al., 2025). SPT simulations are becoming more accurate, efficient, and economically significant using AI and ML algorithms. This technology opens up a greener path for renewable energy generation, making SPTs feasible in the renewable energy sector. Table 7 summarizes the different AI techniques used in different aspects of SPTs.
7.3 Integration of AI and digital twins in real-time SPT simulation
Digital twins (DTs) are emerging as a promising concept in energy systems, enabling real-time synchronization between physical infrastructure and virtual simulation models through continuous data exchange. In the context of SPTs, a DT would support predictive diagnostics, adaptive control, and performance forecasting by integrating real-time sensor data with simulation platforms. Although a complete, full-scale DT for SPTs has not yet been implemented, foundational components such as real-time system modeling, AI-enhanced prediction, and co-simulation are actively being developed in the broader CSP domain and have been extensively explored in other solar technologies, including PV systems (
Alao et al., 2024). Notably, prototype-level research by Pargmann et al. (
2024) has demonstrated in situ heliostat calibration and flux prediction using differentiable ray tracing and neural networks, confirming the feasibility of key DT functions under real operating conditions.
One enabling technology is SolarTherm, an open-source simulation platform based on Dymola that supports dynamic and modular modeling of CSP systems, including heliostat fields, receivers, and thermal storage. It has been validated against the SAM for annual performance, demonstrating its potential for integration into real-time simulation environments (
Scott et al., 2017). Complementing this, SolarReceiver2D, a Modelica-based tool, allows transient thermal analysis of receiver components, further supporting system-level integration (
Gentile et al., 2022). De la Calle et al. (
2020) also modeled a full SPT system comprising liquid sodium receivers and PCM thermal storage in Dymola, demonstrating techno-economic analysis capabilities that reflect the system-level simulation layer of a DT.
AI plays a key role in operationalizing DTs by enabling data-driven decision-making. For example, deep neural networks have been successfully used for heliostat calibration with sub-milliradian accuracy, enabling real-time adjustment of field performance (
Pargmann et al., 2021;
Sievers et al., 2025). In receiver optimization, Shuai et al. (
2024) applied deep learning models to predict multi-objective thermal performance under varying conditions, paving the way for surrogate models that reduce computational cost during real-time simulation. Similarly, Sarker et al. (
2021) explored ancillary voltage control using simulation-based adaptive tracking strategies in microgrid setups with industrial loads, reinforcing the relevance of AI-integrated control models for real-time power systems. Similarly, the EvoRec framework, based on annual thermal gain assessment and ANN, has been proposed to optimize receiver design dynamically (
Schöttl et al., 2020). A practical architecture for future SPT DTs is shown in Fig. 15.
Recent work also demonstrates real-time AI integration into heliostat control strategies. Wang et al. (
2024) introduced an Improved Particle Swarm Optimization (IUPSO) algorithm that adapts heliostat aiming dynamically using real-time irradiance and cloud motion predictions. While this is not a full DT, it represents a key building block—adaptive, AI-driven control using simulation-informed feedback. Similarly, Johana et al. (
2023) implemented a PLC-SCADA-based simulation framework for an islanded microgrid, emphasizing dispatch optimization and real-time control, principles that are transferable to future SPT DT architectures.
In addition, hybrid AI-physics models have been introduced to mitigate data scarcity and generalization issues common in real-world deployments. Syauqi et al. (
2024) demonstrated that integrating physical models with ML can reduce training data requirements while preserving model accuracy, which is crucial for the practical deployment of DTs in data-sparse CSP environments. This is consistent with findings by Guo et al. (
2019), who developed active learning frameworks for real-time energy applications.
To consolidate these advancements into a practical deployment pathway for SPT DTs, a staged roadmap can be envisioned. First, a robust data acquisition layer should automate heliostat calibration and receiver flux mapping using established camera-target methods and real-time irradiance sensors, minimizing manual intervention. Second, a hybrid simulation core should integrate differentiable ray tracing engines supported by physics-informed neural surrogate models to maintain accurate optical and thermal predictions while allowing rapid updates under changing field conditions. Finally, this DT must interface with the plant’s control systems to enable closed-loop optimization, predictive maintenance, and adaptive dispatch scheduling. This approach builds on the maturity of current simulation tools, proven AI-assisted calibration strategies, and advanced control algorithms, collectively bridging the gap between conceptual DTs and near-term industrial realization for SPTs.
8 Discussion and future research directions
The comparison of optical, thermal, and system-level simulation software emphasizes their key role in SPT system design and optimization. Optical programs like Tonatiuh and SolTrace precisely simulate heliostat field performance with MCRT, but are computationally intensive and require simplifications for large-scale applications. Thermal and CFD software such as OpenFOAM, ANSYS Fluent, and COMSOL Multiphysics provide detailed heat transfer and fluid dynamics data but are computationally costly and difficult to verify. System-level software, such as SAM and TRNSYS, provides robust economic and performance modeling but generally does not communicate with detailed thermal and optical modeling.
The lack of effective integration of multi-physics models, such as optical, thermal, and system-level dynamics, is one of the significant challenges in SPT simulations. Future work must address the creation of hybrid simulation tools that combine ray tracing methods with CFD and system-level economic models. Most of the current models exist independently, which creates errors in predicting system behavior under realistic conditions. The hybrid simulation environments will increase predictive fidelity and allow integrated assessment of SPT performance, optimizing heliostat field geometry, receiver heat transfer, and dispatch strategies within a single workflow.
Although CFD and MCRT methods provide precise results, they are computationally expensive, thus making large-scale simulations infeasible. One possible future research direction is the application of reduced-order models (ROMs) and AI-based simulations. AI-powered surrogate models can potentially decrease the computational time by orders of magnitude without any loss of accuracy. Furthermore, with the availability of cloud computing and high-performance parallel processing, real-time optimization and control of SPT processes are achievable, thereby ensuring improved resource utilization.
Most SPT simulation models have limited validation against real-world data. As outlined in Section 6.3, future research should emphasize pilot-scale testing, real-time data integration, and the progressive development of DT frameworks for accurate, adaptive plant operation. This will enable reliable predictive maintenance, enhance operational efficiency, and increase trust in simulation outcomes.
While AI and ML approaches continue to gain adoption for modeling and optimizing SPT systems, there remain several critical limitations that must be addressed to enable widespread and reliable application. A key challenge is the limited availability of high-resolution, site-specific data, which constrains the training and accuracy of data-driven models. In many cases, these models depend on synthetic or simulated data sets that may not adequately reflect operational uncertainties and variability encountered in real conditions (
Zaim et al., 2023;
Chaudhary et al., 2025). The performance of data-driven models often degrades significantly when working with noisy, incomplete, or limited data sets, which are common in solar forecasting and heliostat calibration scenarios (
Hajj-Hassan et al., 2020;
Sievers et al. 2025). To mitigate this, hybrid AI-physical models have demonstrated potential for reducing data dependency while maintaining prediction accuracy (
Guo et al., 2019;
Syauqi et al., 2024).
Another significant concern is the lack of interpretability in deep learning and other complex models. Many functions as “black boxes,” making it difficult to understand or validate their predictions, especially when safety-critical decisions are involved, such as receiver heat flux control or heliostat tracking (
Khosravi et al., 2021;
Sievers et al., 2025). This lack of transparency limits trust and broader adoption in plant operations. Furthermore, generalization across plant configurations and climatic zones remains a key challenge. Models trained on one system or location often require substantial reconfiguration or retraining before they can be applied elsewhere, reducing their scalability (
Tung, 2024;
Gul et al., 2025). Environmental variability, hardware heterogeneity, and model overfitting contribute to this issue. Despite advances such as ensemble learning and data augmentation, achieving consistent performance across diverse systems remains difficult. Lastly, deploying AI-based models in real-world SPT applications also demands high computational resources and robustness verification, which further restricts their applicability in resource-constrained or safety-critical environments. Addressing these issues through physics-informed learning, transfer learning, and explainable AI frameworks will be crucial for reliable, scalable, and interpretable SPT simulation and control models.
AI and ML applications to SPT simulation are still underexplored. Heliostat field layout optimization, heat transfer modeling, and economic optimization still rely on conventional numerical methods that could be time-consuming and suboptimal. In the future, studies need to investigate the application of reinforcement learning for dynamic heliostat pointing, neural networks for heat flux prediction, and AI-based predictive control systems for energy dispatch optimization. All these developments will make SPT systems smarter and adaptive, more efficient, and reduce operational costs.
In addition, most existing simulation tools are best suited for research-scale studies and are not yet optimized for commercial-scale deployment. Future advances should focus on scalable algorithms that balance accuracy and computational efficiency, along with modular architectures that support seamless interoperability as new technologies emerge.
SPT systems deteriorate through thermal stress, aging of materials, and exposure to the environment. Long-term degradation and maintenance needs cannot be adequately forecast by current simulation models. Further development needs to incorporate degradation modeling methods, introducing material science know-how into the picture, together with long-term operating experience. This will enable predictive maintenance procedures and life-cycle costing optimization to be performed, to improve plant reliability and lower operating expenses. System models are likely to overlook market volatility and grid interdependencies, reducing the economic viability of SPT programs. Subsequent research must combine techno-economic modeling with electricity market simulations to better couple dispatch strategies with real-time-of-day prices, demand variability, and storage.
In summary, advancing SPT simulation demands robust multi-physics integration, efficient and interpretable AI, thorough real-world validation, and comprehensive techno-economic modeling. These developments will drive higher efficiency, lower costs, and more reliable large-scale deployment of SPT technologies as sustainable energy solutions.
9 Conclusions
The advancement of SPT technology is closely tied to the continual evolution of simulation tools that span optical, thermal, CFD, and system-level domains. This review consolidates these domains into a comprehensive perspective, highlighting the strengths, limitations, and interdependencies of current modeling techniques. It shows that while ray-tracing methods like MCRT provide high-fidelity optical predictions, they demand significant computational resources, calling for AI-driven surrogates and hybrid approaches to balance accuracy and efficiency.
Thermal and CFD simulations are indispensable for analyzing heat transfer, fluid flow, and receiver performance, but often face challenges related to turbulence modeling and high computational costs. Integrating these with data-driven models and high-performance computing can overcome existing bottlenecks, enabling real-time design updates and operational insights.
At the system level, tools such as SAM and TRNSYS offer robust techno-economic assessments but generally lack seamless coupling with detailed optical and thermal models. Developing multi-physics co-simulation frameworks and DT architectures can bridge this gap, provide dynamic, real-world validated performance forecasts, and support predictive maintenance strategies.
Future research should focus on advancing hybrid simulation environments that merge ray-tracing, CFD, and system-level economic models within a unified framework. The integration of AI and ML holds promise for optimizing heliostat field layouts, forecasting receiver performance, and enabling adaptive control strategies. Furthermore, coupling these innovations with cloud computing and real-time data acquisition will enhance scalability and practical deployment. Overall, next-generation SPT simulation will increasingly rely on intelligent, interconnected digital ecosystems that combine physical modeling, AI, and operational data, paving the way for more reliable, efficient, and cost-effective renewable power generation.
PDF
(8803KB)
