A photovoltaic performance prediction method and system for ternary systems

Abstract

The application discloses a photovoltaic performance prediction method and system for ternary systems, and belongs to the technical field of semiconductor device design. The photovoltaic performance prediction method comprises the following steps: determining a target candidate material set based on multi-dimensional characterization data of a candidate third component material; performing microstructure and electronic property simulation on each material in the target candidate material set to obtain structure description parameters and energy characteristic parameters; constructing a photovoltaic device performance prediction model based on the structure description parameters and the energy characteristic parameters; and inputting matching parameters of each component in the ternary system and device preparation process parameters into the photovoltaic device performance prediction model to obtain corresponding photovoltaic performance prediction results.

Description

Technical Field

[0001] This application relates to the field of semiconductor device design technology, and more specifically, to a method and system for predicting photovoltaic performance in ternary systems. Background Technology

[0002] With the development of organic semiconductor photovoltaic technology, binary semiconductor systems based on donor / acceptor are gradually approaching performance bottlenecks in terms of spectral absorption, energy level matching, and carrier transport. To further improve device performance, introducing a third semiconductor material to construct a ternary system has become an important direction, which can achieve synergistic enhancement by controlling the semiconductor band structure, optimizing interface charge dynamics, and improving morphology. However, ternary semiconductor systems involve complex energy level coupling, carrier transport, and recombination processes among multiple components, and the material ratios are highly coupled with the fabrication process parameters, significantly increasing the difficulty of performance control.

[0003] In existing technologies, photovoltaic performance prediction largely relies on empirical methods or single models, making it difficult to simultaneously characterize the multiple physical processes within semiconductor devices. Even when machine learning methods are introduced, they generally lack effective integration with semiconductor device physical models and lack closed-loop correction mechanisms based on experimental data, resulting in insufficient prediction accuracy and engineering applicability. Furthermore, the screening of third-component semiconductor materials and multi-parameter collaborative optimization still mainly rely on trial and error, which is inefficient.

[0004] In summary, how to achieve ternary photovoltaic performance prediction by integrating multidimensional semiconductor material properties, device physical mechanisms, and data-driven models, and how to achieve efficient coordination of material selection, performance prediction, and parameter optimization, as well as closed-loop correction driven by experimental feedback, has become an urgent technical problem to be solved. Summary of the Invention

[0005] To overcome a series of shortcomings in existing technologies, the purpose of this application is to provide a photovoltaic performance prediction method for ternary systems, comprising the following steps: Based on the multi-dimensional characterization data of the candidate third component materials, the target candidate material set is determined; Microstructure and electronic properties of each material in the target candidate material set are simulated to obtain structural description parameters and energy characteristic parameters; A performance prediction model for photovoltaic devices is constructed based on structural description parameters and energy characteristic parameters; By inputting the proportions of each component in the ternary system and the device fabrication process parameters into the photovoltaic device performance prediction model, the corresponding photovoltaic performance prediction results are obtained.

[0006] In some embodiments, the method for obtaining multi-dimensional representation data is as follows: The candidate third component materials were pretreated, and a unified testing environment and standardized testing parameter system were constructed. Optical performance tests were performed on the candidate third component material to obtain ultraviolet-visible absorption and photoluminescence spectra, and the absorption peak intensity, photoluminescence quantum efficiency and optical band gap were extracted. At the same time, the refractive index dispersion relation and photon trapping ability parameters were calculated. Electrical performance tests were conducted on the candidate third component material to obtain the following electrical characteristic parameters: highest occupied molecular orbital energy level and lowest unoccupied molecular orbital energy level, hole mobility and electron mobility, temperature dependence of conductivity as a function of temperature, dielectric constant, and charge transfer state energy level. Thermal stability and environmental stability tests were conducted on the candidate third component materials to obtain glass transition temperature, melting temperature, thermal decomposition temperature, light stability decay coefficient, and degradation kinetic parameters under water and oxygen conditions. Interfacial properties and molecular interactions of candidate third-component materials were tested to obtain interfacial energy parameters and intermaterial interaction characteristics.

[0007] In some embodiments, the method for determining the target candidate material set is as follows: Based on the multi-dimensional characterization data of the candidate third component materials, a multi-dimensional evaluation index system was constructed, and quantitative evaluation criteria and screening standards were set for each evaluation dimension. Based on multi-dimensional characterization data and quantitative evaluation criteria, parameters were extracted and normalized for each candidate third component material to obtain standardized evaluation values ​​for each evaluation dimension. Determine the weight coefficients for each evaluation dimension and construct a multi-parameter weighted evaluation rule; The comprehensive score of each candidate third component material is calculated based on standardized evaluation values ​​and weighting coefficients, and the comprehensive score is corrected based on the Delphi method. Based on the revised comprehensive score, each candidate third component material is sorted and screened. Candidate third component materials whose comprehensive scores are within the preset range and whose evaluation indicators meet the corresponding threshold conditions are selected to form the target candidate material set.

[0008] In some embodiments, the photovoltaic device performance prediction model employs a hybrid modeling strategy that integrates a semiconductor device physical model and a machine learning model, wherein: The physical model of the semiconductor device is based on the drift-diffusion equation, Poisson equation and carrier continuity equation to construct a numerical solution framework, which is used to describe the carrier generation, transport and recombination process inside the photovoltaic device. The machine learning model is based on a deep neural network structure to construct a nonlinear mapping prediction framework, which is used to establish the mapping relationship between structural description parameters, energy characteristic parameters, ratio parameters, device fabrication process parameters and photovoltaic performance evaluation parameters.

[0009] In some embodiments, the photovoltaic performance prediction method further includes: The photovoltaic performance prediction results under different parameter combinations are evaluated, and it is determined whether they meet the preset performance optimization targets. If the target is not met, the ratio parameters and device fabrication process parameters are adjusted and re-entered into the photovoltaic device performance prediction model for iterative calculation until the performance optimization target is met. If the optimal parameter combination is satisfied, the optimal parameter combination is output, and a ternary photovoltaic device is fabricated based on the optimal parameter combination and its performance is tested to obtain experimental performance data.

[0010] In some embodiments, the method for evaluating photovoltaic performance prediction results under different parameter combinations is as follows: A multi-dimensional evaluation index system is constructed, which includes energy conversion efficiency index, electrical output index, energy loss index, stability index and cost index; Construct a multi-objective comprehensive evaluation function to uniformly represent each evaluation index and define the optimization direction of each evaluation index; The performance of photovoltaic devices under different parameter combinations is calculated and evaluated to obtain corresponding multi-dimensional performance evaluation results; Multi-objective optimization processing is performed on the multi-dimensional performance evaluation results to obtain a set of candidate parameter combinations that satisfy the balance relationship of multiple indicators; The candidate parameter combination set is filtered based on the preset performance constraints to obtain the optimal parameter combination that satisfies both the multi-objective optimization results and the preset performance constraints.

[0011] In some embodiments, the method for adjusting the proportioning parameters and device fabrication process parameters is as follows: Determine the initial value range of the proportioning parameters and device fabrication process parameters, and construct the parameter search space and constraints; The parameter search space is initially screened to identify key influencing parameters, and the correlation between key influencing parameters and device performance is established. Based on the correlation, iterative optimization search is performed on key influencing parameters to determine the parameter optimization direction and search step size; Based on the multi-objective comprehensive evaluation function, the photoelectric conversion efficiency is improved and the radiation loss and non-radiative recombination loss of the charge transfer state are reduced in a coordinated manner, so as to obtain the optimal or near-optimal combination of ratio parameters and device fabrication process parameters that meet the requirements of multi-objective optimization.

[0012] In some embodiments, the method for preparing ternary photovoltaic devices based on optimal parameter combinations is as follows: The conductive substrate is pretreated to obtain a clean and surface-activated substrate structure; A hole transport layer structure is constructed on the substrate structure to form a bottom charge-selective transport interface; An active layer precursor solution is prepared based on an optimal parameter combination. The precursor solution contains a host donor material, a host acceptor material, and a third component material. Under controlled environmental conditions, the active layer precursor solution is prepared into a thin film structure to form an active layer thin film with a target phase separation structure. The active layer thin film is subjected to structural control treatment, and the electron transport layer and electrode structure are constructed sequentially to form a complete ternary photovoltaic device structure.

[0013] In some embodiments, the photovoltaic performance prediction method further includes: The photovoltaic device performance prediction model is corrected based on experimental performance data, and the final photovoltaic performance prediction results and material design guidance information are output based on the corrected photovoltaic device performance prediction model.

[0014] The purpose of this application is also to provide a photovoltaic performance prediction system for ternary systems, used to implement the above-mentioned photovoltaic performance prediction method, including: The materials screening module is used to obtain the target candidate materials set; The microscopic simulation module, connected to the material screening module, is configured to: acquire structural description parameters and energy characteristic parameters, and construct a photovoltaic device performance prediction model; The performance prediction module, connected to the microscopic simulation module, is configured to: acquire the photovoltaic performance prediction results of the ternary system; perform iterative optimization based on the photovoltaic performance prediction results and the preset performance optimization target, and output the optimal parameter combination; The experimental feedback module, connected to the performance prediction module, is configured to: prepare ternary photovoltaic devices based on the optimal parameter combination and test the experimental performance data; revise the photovoltaic device performance prediction model based on the experimental performance data; and output the final photovoltaic performance prediction results and material design guidance information.

[0015] Compared with the prior art, this application has the following beneficial effects: This application achieves high-precision collaborative prediction and adaptive correction of material selection, proportioning process optimization, and device performance in ternary photovoltaic systems by integrating multi-dimensional material characterization data screening, microstructure and electronic property simulation, and closed-loop iterative optimization of physical model and machine learning coupled modeling. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a photovoltaic performance prediction method for ternary systems disclosed in an embodiment of this application.

[0017] Figure 2This is a schematic diagram of the waterfall-type energy level arrangement of the PM6 / BTP-eC9 / L8-BO ternary system in the embodiments of this application.

[0018] Figure 3 This is a schematic diagram of the molecular packing of the PM6 / BTP-eC9 / L8-BO ternary hybrid system in the embodiments of this application.

[0019] Figure 4 This is a schematic diagram of the radial distribution function curve of the PM6 / BTP-eC9 / L8-BO ternary hybrid system in an embodiment of this application.

[0020] Figure 5 This is a schematic diagram of the phase separation morphology of the active layer of the PM6 / BTP-eC9 / L8-BO ternary hybrid system in the embodiments of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some embodiments of this invention, but not all embodiments.

[0022] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0024] like Figure 1 As shown, this embodiment provides a photovoltaic performance prediction method for ternary systems, including the following steps: Based on the multi-dimensional characterization data of the candidate third component materials, the candidate third component materials are comprehensively scored and screened by multiple parameters to determine the target candidate material set; Microstructure and electronic properties of each material in the target candidate material set are simulated to obtain structural description parameters and energy characteristic parameters; A performance prediction model for photovoltaic devices is constructed based on structural description parameters and energy characteristic parameters; The proportioning parameters of each component in the ternary system and the device fabrication process parameters are input into the photovoltaic device performance prediction model to obtain the corresponding photovoltaic performance prediction results. The photovoltaic performance prediction results under different parameter combinations are evaluated, and it is determined whether they meet the preset performance optimization targets. If the target is not met, the ratio parameters and device fabrication process parameters are adjusted and re-entered into the photovoltaic device performance prediction model for iterative calculation until the performance optimization target is met. If the conditions are met, the optimal parameter combination is output, and a ternary photovoltaic device is fabricated based on the optimal parameter combination and its performance is tested to obtain experimental performance data. The photovoltaic device performance prediction model is corrected based on experimental performance data, and the final photovoltaic performance prediction results and material design guidance information are output based on the corrected photovoltaic device performance prediction model.

[0025] The photovoltaic performance prediction method for ternary systems described in this application constructs a third-component material screening mechanism based on multi-dimensional characterization data to achieve multi-parameter comprehensive evaluation of candidate materials and accurate determination of the target material set. By introducing microstructure and electronic property simulation, structural description parameters and energy characteristic parameters are extracted to characterize the intrinsic semiconductor physical properties of the materials. Based on this, a photovoltaic device performance prediction model is constructed by integrating structural and energy characteristics. Multiple input combinations are then used to calculate the photovoltaic performance, combining ternary system proportioning parameters and device fabrication process parameters, enabling quantitative prediction and comparative analysis. Furthermore, the parameter combinations are adjusted through multiple rounds of iterative optimization to gradually approach the performance optimization target. Finally, the prediction model is corrected by combining experimental fabrication and performance testing data, achieving continuous model updates and accuracy improvements. This realizes an integrated collaborative design of ternary photovoltaic material screening, device performance prediction, and process parameter optimization, improving the design efficiency and prediction reliability of photovoltaic devices.

[0026] In some embodiments, the method for obtaining multi-dimensional representation data is as follows: The candidate third component materials were pretreated, and a unified testing environment and standardized testing parameter system were constructed. Optical performance tests were performed on the candidate third component material to obtain ultraviolet-visible absorption and photoluminescence spectra, and the absorption peak intensity, photoluminescence quantum efficiency and optical band gap were extracted. At the same time, the refractive index dispersion relation and photon trapping ability parameters were calculated. Electrical performance tests were conducted on the candidate third component material to obtain the following electrical characteristic parameters: highest occupied molecular orbital energy level and lowest unoccupied molecular orbital energy level, hole mobility and electron mobility, temperature dependence of conductivity as a function of temperature, dielectric constant, and charge transfer state energy level. Thermal stability and environmental stability tests were conducted on the candidate third component materials to obtain glass transition temperature, melting temperature, thermal decomposition temperature, light stability decay coefficient, and degradation kinetic parameters under water and oxygen conditions. Interfacial properties and molecular interactions of candidate third-component materials were tested to obtain interfacial energy parameters and intermaterial interaction characteristics.

[0027] The multi-dimensional characterization data acquisition method described in this application preprocesses candidate third-component materials and constructs a unified testing environment and standardized parameter system to ensure the comparability of data from different materials; it acquires absorption spectra, photoluminescence characteristics, and optical band gaps through optical performance testing, and calculates refractive index dispersion and photon trapping parameters to characterize the material's light absorption capability; it acquires parameters such as energy level structure, carrier mobility, temperature-dependent conductivity, and dielectric constant through electrical performance testing to characterize the material's semiconductor transport properties; it acquires parameters related to thermal decomposition, phase transition, and light / water-oxygen degradation through thermal stability and environmental stability testing to assess the material's service reliability; and it further acquires interface energy and interaction characteristics through interface and molecular interaction testing, achieving quantitative characterization of the material's interface matching capability.

[0028] In some embodiments, the method for determining the target candidate material set is as follows: Based on the multi-dimensional characterization data of the candidate third component materials, a multi-dimensional evaluation index system is constructed, which includes molecular structure characteristics, spectral absorption complementarity, energy level matching, luminescence efficiency, crystallinity compatibility, and molecular packing characteristics. Quantitative evaluation criteria and screening standards are set for each evaluation dimension; Based on multi-dimensional characterization data and quantitative evaluation criteria, parameters were extracted and normalized for each candidate third component material to obtain standardized evaluation values ​​for each evaluation dimension. The weight coefficients of each evaluation dimension were determined using the analytic hierarchy process (AHP), and a multi-parameter weighted evaluation rule was constructed. The comprehensive score of each candidate third component material is calculated based on standardized evaluation values ​​and weighting coefficients, and the comprehensive score is corrected based on the Delphi method. Based on the revised comprehensive score, each candidate third component material is sorted and screened. Candidate third component materials whose comprehensive scores are within the preset range and whose evaluation indicators meet the corresponding threshold conditions are selected to form the target candidate material set.

[0029] For example, the method for obtaining the target candidate material set is specifically explained by screening a third component suitable for the PM6:L8-BO binary matrix system from 20 candidate third component materials.

[0030] Six evaluation dimensions and corresponding quantitative screening criteria are established: molecular structural characteristics (conjugated backbone planarity, dihedral angle ≤15°, molecular weight in the range of 500–1500 Da), and spectral absorption complementarity (absorption peaks located in the range of 500 nm–650 nm to supplement the absorption blind zone of the main system, and molar extinction coefficient of the absorption peak ≥1×10⁻⁶). 5 L·mol -1 ·cm -1), energy level matching (HOMO energy level is between PM6 and L8-BO and meets waterfall arrangement, LUMO and L8-BO difference ≥0.3eV), luminous efficiency (PLQY≥40%), crystallinity compatibility (surface energy difference with host material ≤2mN / m) and molecular packing characteristics (single crystal packing density ≥60% and face-to-face orientation).

[0031] Based on multi-dimensional characterization data, feature parameters corresponding to each evaluation dimension were extracted for each of the 20 candidate materials, and the range normalization method was used to map each parameter to the [0,1] interval, resulting in a standardized evaluation value matrix (20×6) for each material in the six evaluation dimensions.

[0032] Using the analytic hierarchy process (AHP), five field experts were invited to conduct pairwise importance comparisons of the six evaluation dimensions. A judgment matrix was constructed, and the consistency ratio (CR) was calculated to be 0.047 < 0.1, thus passing the consistency test. The final weight coefficients for each dimension were determined as follows: molecular structure characteristics 0.15, spectral absorption complementarity 0.25, energy level matching 0.25, luminescence efficiency 0.15, crystallinity compatibility 0.10, and molecular packing characteristics 0.10.

[0033] Based on standardized evaluation values ​​and weighting coefficients, the comprehensive scores of each candidate material were calculated. Preliminary rankings showed BTP-eC9, Y6-BO, and BTPTT-4F in the top three, with comprehensive scores of 0.823, 0.791, and 0.768, respectively. Subsequently, three rounds of Delphi method expert consultations were organized to adjust the weights based on the practical engineering importance of energy level matching and morphological compatibility. The final adjusted comprehensive scores were 0.841, 0.779, and 0.752, respectively.

[0034] Screening criteria were set as follows: the overall score must be ≥0.75 and all six evaluation indicators must meet their corresponding thresholds. After screening, BTP-eC9, Y6-BO, and BTPTT-4F among the 20 candidate materials simultaneously met the overall score threshold and all individual indicator thresholds, thus forming the target candidate material set.

[0035] The method for determining the target candidate material set described in this application constructs a multi-dimensional evaluation index system covering molecular structure characteristics, spectral absorption complementarity, energy level matching, luminescence efficiency, crystallinity compatibility, and molecular packing characteristics to achieve a unified multi-dimensional characterization of candidate third-component materials. It obtains standardized evaluation values ​​by setting quantitative criteria and screening standards for each evaluation dimension and extracting and normalizing parameters from the multi-dimensional characterization data. Based on this, it uses the analytic hierarchy process (AHP) to determine the weights of each evaluation dimension and constructs a weighted evaluation model to achieve multi-factor collaborative evaluation. Furthermore, it combines the Delphi method with expert correction of the comprehensive score to improve the reliability and engineering applicability of the evaluation results. Finally, it ranks and screens based on the corrected comprehensive score and determines the target candidate material set by combining threshold constraints, thereby achieving high-precision screening and system adaptation optimization of the third-component materials.

[0036] In some embodiments, the method for simulating the microstructure and electronic properties of materials in the target candidate material set is as follows: A first-principles calculation system based on density functional theory was established, and simulation conditions for exchange-correlated functionals, basis set parameters, and solvent environment were set. Based on the first-principles calculation system, the geometric structure of each material in the target candidate material set is optimized to obtain its ground state and excited state equilibrium configuration; Based on the ground state and excited state equilibrium configurations, the dihedral angle of the main chain, the conformation angle of the side chain, and the conjugation length are calculated, and the molecular rigidity and steric hindrance parameters are obtained. Based on quantum chemical calculation methods, the frontier molecular orbital energy level distribution, charge density difference distribution, electrostatic potential distribution, and molecular dipole moment are calculated. The vertical excitation energy, oscillator strength, and exciton binding energy were calculated based on time-dependent density functional theory, and the charge transport recombination energy and charge transfer integral were calculated based on Marcus theory. Based on molecular dynamics simulation, dynamic simulations were performed on a mixed system including donor materials, acceptor materials, and a third component material to obtain the radial distribution function, intermolecular interaction energy, and diffusion coefficient.

[0037] For example, taking the third component material BTP-eC9 in the target candidate material set as an example, the above-mentioned microstructure and electronic property simulation method is specifically explained in conjunction with the host donor PM6 and the host acceptor L8-BO.

[0038] The Gaussian16 software package was selected as the quantum chemical computing platform. The B3LYP exchange-correlation functional was employed, with a basis set of 6-31G(d,p). A continuous medium solvation model (PCM) was introduced to simulate the chlorobenzene solvent environment (dielectric constant). =5.69), thereby constructing a unified quantum chemical computing framework.

[0039] Under the above calculation system, the BTP-eC9 molecule was subjected to ground state analysis. and the first excited state The geometric structure was optimized to obtain the corresponding equilibrium configuration. The optimization results were further verified by vibration frequency analysis, confirming that the obtained configurations were all energy minimum points, that is, there were no imaginary frequencies, thus ensuring the reliability of the structural calculation.

[0040] Based on the optimized molecular equilibrium configuration, key structural parameters were extracted and quantified. The results showed that the dihedral angle of the thiophene-benzodithiazole-thiophene core unit in the BTP-eC9 main chain was 3.2°, indicating high planarity of the main chain; the conformation angles of the alkyl side chains were mainly distributed in the range of 60°±5°; the effective conjugation length was 14 conjugated units; the molecular rotational barrier was 18.3 kJ / mol, and the van der Waals volume was 896 cubic Å, reflecting high rigidity and certain steric hindrance characteristics of the molecule.

[0041] The electronic structure of BTP-eC9 was analyzed under the aforementioned computational system. The calculation results show that its highest occupied molecular orbital (HOMO) level is -5.42 eV, and its lowest unoccupied molecular orbital (LUMO) level is -3.81 eV, corresponding to a band gap of 1.61 eV. The charge density difference distribution indicates that excited-state charges migrate from the donor unit at the molecular center to the acceptor end groups at both ends, exhibiting a clear push-pull electron characteristic. The electrostatic potential distribution results show that the end group region exhibits a strong negative potential (-45 to -55 kcal / mol), which is conducive to the formation of directional intermolecular interactions with the donor material. Furthermore, the molecular dipole moment is 2.37 D.

[0042] The excited-state properties were calculated using time-dependent density functional theory (TD-DFT). The results show that the vertical excitation energy of BTP-eC9 is 1.68 eV, corresponding to an oscillator intensity of 2.14 and an exciton binding energy of 0.21 eV. Further calculations based on Marcus theory yielded a hole transport recombination energy of 0.18 eV, an electron transport recombination energy of 0.16 eV, and a charge transfer integral of 32.5 meV with L8-BO, indicating that the material possesses superior charge transport capabilities.

[0043] A hybrid system consisting of PM6 (40 chains), L8-BO (40 chains), and BTP-eC9 (12 chains, corresponding to a mass fraction of approximately 15%) was constructed. Molecular dynamics simulations were performed using GROMACS software under NPT ensemble conditions (temperature 300 K, pressure 1 atm, simulation time 50 ns). Simulation results show that the radial distribution function between BTP-eC9 and L8-BO exhibits a significant peak at 0.38 nm, indicating a preferential π-π stacking tendency. Simultaneously, the intermolecular interaction energy between BTP-eC9 and PM6 is -3.82 eV / pair, and with L8-BO it is -4.15 eV / pair, indicating a stronger interaction with the acceptor material. Furthermore, the diffusion coefficient of BTP-eC9 in the hybrid system is 1.3 × 10⁻⁶. -7 cm 2 / s indicates that it has moderate migration ability during film formation, which is conducive to forming an ideal phase separation morphology.

[0044] The morphology of the above molecular dynamics simulation results is shown in the figure. Figure 3-5 As shown. Figure 3 The diagram illustrates the molecular packing of the ternary hybrid system: the donor PM6 (40 chains) is arranged in a rectangular sheet-like stacking pattern, indicating its ordered face-to-face orientation; the acceptor L8-BO (40 chains) is shown with an elliptical cross-section, indicating its ordered small molecule packing morphology; the third component BTP-eC9 (12 chains, mass fraction approximately 15%) is marked with a dashed box and distributed in the donor / acceptor interface region, with a triangle symbol indicating its wedge-shaped aggregation structure; the bidirectional dashed arrows between the donor and acceptor phases indicate the π-π preferential packing trend. Figure 4 The radial distribution function g(r) curves between BTP-eC9 and L8-BO, and between BTP-eC9 and PM6 are shown: the solid line corresponds to the BTP-eC9 / L8-BO system, with a significant peak at r=0.38 nm, indicating preferential π-π packing between the two; the dashed line corresponds to the BTP-eC9 / PM6 system, with a lower peak and the peak position slightly shifted towards the direction of larger r; the 0.38 nm peak position is marked by a vertical dashed line in the figure, and the g(r)=1 baseline is marked by a horizontal dashed line. Figure 5 This is a schematic diagram of the phase separation morphology of the active layer: the left side shows the PM6 donor enrichment region (sheet-like regular stacking), the right side shows the L8-BO acceptor enrichment region (elliptical face-to-face stacking), and the middle is the donor / acceptor interface region, where BTP-eC9 preferentially accumulates with a diffusion coefficient of 1.3 × 10⁻⁶. -7 cm 2 / s, the intermolecular interaction energy is -4.15 eV / pair (with L8-BO); the solid arrows indicate the direction of exciton diffusion to the interface and charge separation at the interface, respectively.

[0045] The microstructure and electronic property simulation method described in this application constructs a first-principles calculation system based on density functional theory, and sets exchange-correlation functionals, basis set parameters, and solvent environment conditions to achieve a unified quantum chemical calculation framework for the target candidate materials. Geometric optimization is used to obtain the ground state and excited state equilibrium configurations of the materials, and structural parameters such as main chain dihedral angles, side chain conformation angles, and conjugation lengths are further calculated to characterize molecular rigidity and steric hindrance. Quantum chemical calculations are used to obtain the frontier molecular orbital energy level distribution, charge density difference, electrostatic potential distribution, and molecular dipole moment, achieving a quantitative description of the material's electronic structure and charge distribution. Time-dependent density functional theory is used to calculate the vertical excitation energy, oscillator strength, and exciton binding energy, and Marcus theory is combined to calculate the charge recombination energy and charge transfer integral, achieving a refined characterization of carrier excitation and transport capabilities. Furthermore, molecular dynamics simulations are used to perform dynamic evolution analysis of the donor, acceptor, and third-component mixture system to obtain the radial distribution function, intermolecular interaction energy, and diffusion coefficient, thereby achieving multi-scale synergistic simulation and characterization of material structure, electronic properties, and interface behavior in the ternary photovoltaic system.

[0046] In some embodiments, the photovoltaic device performance prediction model employs a hybrid modeling strategy that integrates a semiconductor device physical model and a machine learning model, wherein: The physical model of the semiconductor device is constructed based on the drift-diffusion equation, Poisson equation, and carrier continuity equation to build a numerical solution framework, which is used to describe the carrier generation, transport, and recombination processes inside the photovoltaic device. Specifically: in the numerical solution framework, the photogenerated carrier generation rate is determined based on the light intensity distribution of each layer and the material absorption coefficient calculated by the optical transfer matrix method, used to characterize the absorption and generation process of incident light in the multilayer structure; the bimolecular recombination process is characterized using the Langevin model, whose recombination rate is related to the dielectric constant and carrier mobility, used to describe the radiative or non-radiative recombination behavior of electrons and holes; the single-molecule recombination process is characterized using the Shockley-Reid-Hall model, whose recombination rate is related to the trap state density, used to describe the defect-state-assisted non-radiative recombination process; the interface energy level alignment is corrected by introducing band bending and Fermi level pinning effects, used to characterize the energy level reconstruction and potential distribution changes at the heterojunction interface. The machine learning model is based on a deep neural network structure to construct a nonlinear mapping prediction framework, used to establish the mapping relationship between structural description parameters, energy characteristic parameters, proportioning parameters, device fabrication process parameters, and photovoltaic performance evaluation parameters. It includes an input layer, four hidden layers, and an output layer. The input features include structural description parameters, energy characteristic parameters, proportioning parameters of the ternary system, and device fabrication process parameters. The structural description parameters and energy characteristic parameters include at least the highest occupied molecular orbital energy level, the lowest unoccupied molecular orbital energy level, photoluminescence quantum efficiency, recombination energy, three-dimensional packing density, molecular orientation ratio, and dielectric constant. The proportioning parameters and device fabrication process parameters include at least the mass ratio of the host donor to the host acceptor, the mass fraction of the third component, solvent type, additive volume fraction, spin coating speed, and annealing conditions. The output layer outputs photovoltaic performance evaluation parameters, which include at least open-circuit voltage, short-circuit current density, fill factor, photoelectric conversion efficiency, and charge transfer state radiation loss. Combined loss with non-radiative ; During model training, an adaptive moment estimation optimization algorithm is used, combined with a cosine annealing learning rate scheduling strategy for parameter updates. At the same time, a composite loss function is introduced for constraint optimization. The composite loss function includes a prediction error term and a physical consistency constraint term, wherein the physical consistency constraint term includes at least energy conservation constraints and charge conservation constraints.

[0047] Specifically, the rate of photogenerated carrier generation The spatial positions are obtained by solving the optical transmission matrix equation. Light intensity distribution at [location] and material absorption coefficient And calculated using the following formula: ,in, Photon energy; bimolecular recombination rate The expression is: ,in, For electron concentration, Hole concentration, This refers to the intrinsic carrier concentration. The composite coefficient is expressed by the formula: ,in, For charge quantity, Where is the dielectric constant. and These are electron and hole mobility, respectively; Single-molecule recombination rate The expression is: ,in, and These are the electron and hole lifetimes, respectively; The electron concentration under thermal equilibrium conditions; The hole concentration under thermal equilibrium conditions; Carrier mobility The expression is: ,in, It characterizes the degree of energy disorder and can be characterized by the Urbach energy; For reference migration rate; Boltzmann's constant; This refers to absolute temperature.

[0048] Specifically, during the training process, the deep neural network architecture incorporates regularization, data augmentation, and ensemble learning strategies to prevent overfitting and improve the model's generalization ability. These strategies include the following: Regularization strategy: A Dropout layer is set after each hidden layer, and the Dropout ratio is adaptively adjusted according to the network layer depth, with 0.2 for shallow layers and 0.4 for deep layers; an L2 regularization term is introduced into the loss function, and the regularization coefficient is determined to be in the range of 0.001 to 0.005 through cross-validation; at the same time, a batch normalization layer is introduced between each fully connected layer and the activation function to accelerate model convergence and reduce sensitivity to parameter initialization. Training process control strategy: An early stopping strategy is adopted. When the validation set loss does not decrease within 25 consecutive training rounds, the training is terminated and rolled back to the model parameters corresponding to the minimum validation set loss. Data augmentation strategies: Introduce small-amplitude Gaussian noise into continuous features to simulate experimental measurement errors; randomly perturb the proportioning and process parameters within physically feasible limits to expand the sample space; and use the SMOTE algorithm to oversample minority class samples to achieve class balance in the dataset. Dataset partitioning strategy: The training dataset, validation dataset, and test dataset are divided into three groups of 70%, 15%, and 15% respectively, and the statistical distribution of ternary material system type, donor / acceptor combination type, third component type, ratio range, and device structure type is kept consistent across different datasets. Ensemble learning strategy: Construct multiple neural network models with different initialization parameters and network structures, and perform weighted averaging or voting fusion of the output results of each model during the prediction stage to improve prediction robustness; Model Performance Evaluation: The model's predictive performance was evaluated on an independent test set, where the mean absolute percentage error of the photoelectric conversion efficiency prediction did not exceed 4%, the open-circuit voltage prediction error did not exceed 0.025V, and the short-circuit current density... The prediction error does not exceed 1.5 mA / cm 2The fill factor prediction error does not exceed 3%, and the charge transfer state radiation loss... Prediction error not exceeding 0.03 eV, non-radiative recombination loss The prediction error should not exceed 0.04 eV, and the Pearson correlation coefficient between the predicted and measured values ​​should not be less than 0.90, in order to meet the accuracy requirements for engineering applications of device performance prediction.

[0049] The photovoltaic device performance prediction model described in this application achieves collaborative modeling of physical mechanisms and data-driven approaches by integrating semiconductor device physical models and machine learning models. A numerical solution framework for the device is constructed using drift-diffusion equations, Poisson equations, and carrier continuity equations. The optical transfer matrix method, Langevin model, and Shockley-Reid-Hall model are combined to characterize photogenerated carrier generation, bimolecular recombination, and monomolecular recombination processes, respectively. An interface energy level correction mechanism is introduced to achieve a physical description of the carrier transport and recombination processes within the device. Based on this, a deep neural network is used to establish a nonlinear mapping relationship between material characteristics, process parameters, and photovoltaic performance. Energy and charge conservation are constrained through a physical constraint loss function, thereby improving prediction accuracy while enhancing the model's physical consistency and generalization ability.

[0050] In some embodiments, the proportioning parameters include the mass ratio of the host donor material, the host acceptor material, and the third component material in the ternary system; wherein the mass ratio of the host donor material to the host acceptor material is 1:1 to 1:1.2; the mass fraction of the third component is 5% to 30% of the total mass of the active layer, and the proportioning adjustment step accuracy is 2.5%; The device fabrication process parameters include: the total solid concentration of the active layer solution is 15 mg / mL to 25 mg / mL; the main solvent is selected from chlorobenzene or o-dichlorobenzene; the additive is selected from 1,8-diiodooctane or 1-chloronaphthalene, with a volume fraction of 0.5% to 3%; the spin coating process adopts a two-stage rotation speed control scheme, wherein the first stage rotation speed is 800 rpm, lasting for 10 s for solution spreading; the second stage rotation speed is 2000 rpm to 2800 rpm, lasting for 40 s to 60 s for forming a uniform film; the thermal annealing temperature is 100℃ to 150℃, and the annealing time is 5 min to 15 min; optionally, solvent vapor annealing is used, wherein the solvent vapor is selected from tetrahydrofuran or carbon disulfide, and the treatment time is 60 s to 180 s; the target thickness of the active layer is 100 nm to 150 nm.

[0051] In some embodiments, the method for evaluating photovoltaic performance prediction results under different parameter combinations is as follows: A multi-dimensional evaluation index system is constructed, which includes energy conversion efficiency index, electrical output index, energy loss index, stability index and cost index. Construct a multi-objective comprehensive evaluation function to uniformly represent each evaluation index and define the optimization direction of each evaluation index; The performance of photovoltaic devices under different parameter combinations is calculated and evaluated to obtain corresponding multi-dimensional performance evaluation results; Multi-objective optimization processing is performed on the multi-dimensional performance evaluation results to obtain a set of candidate parameter combinations that satisfy the balance relationship of multiple indicators; The candidate parameter combination set is filtered based on the preset performance constraints to obtain the optimal parameter combination that satisfies both the multi-objective optimization results and the preset performance constraints.

[0052] For example, the PM6:L8-BO:BTP-eC9 ternary system will be used as an example to illustrate the above evaluation method.

[0053] Based on the requirements for device performance optimization, a multi-dimensional evaluation index system is established, including efficiency, electrical performance, energy loss, stability, and cost. The specific settings are as follows: The energy conversion efficiency index, characterized by photoelectric conversion efficiency (PCE), is set to a target of no less than 18%, with maximization as the optimization direction. The electrical output parameters, including open-circuit voltage Voc, short-circuit current density Jsc, and fill factor FF, are set as follows: Voc ≥ 0.85V, Jsc ≥ 25mA / cm², and FF, respectively. 2 FF≥78%, and both are optimized with maximization as the goal; Energy loss indicators, including and , respectively limited to ≤0.15eV, ≤0.20eV, and minimize it as the optimization objective; The stability index is evaluated by the PCE retention rate after 500 hours of aging at 85℃ / 85%RH, which is set to be no less than 85% and is optimized to be maximized. The cost index is characterized by the mass fraction of the third component, limited to no more than 20%, and minimized as the optimization objective.

[0054] The above evaluation indicators are subjected to dimensionless normalization, and the weights of each indicator are determined based on the analytic hierarchy process (AHP). A comprehensive evaluation function is then constructed based on this determination. : In the above formula, the variable marked with "ˆ" represents the normalized value of the corresponding indicator. This represents the normalized value of the stability index. This represents the normalized value of the cost indicator.

[0055] The 216 constructed parameter combinations were input into the photovoltaic device performance prediction model. The multidimensional performance indicators corresponding to each parameter combination were calculated one by one, and the evaluation result matrix was obtained accordingly. The PCE distribution ranged from 15.2% to 18.6%. The distribution range is 0.15 eV to 0.28 eV, and the comprehensive evaluation function is... The value range is 0.412 to 0.876.

[0056] Based on the above evaluation results, the NSGA-II algorithm was used to solve for the Pareto front. Through 200 iterations, 18 candidate parameter combinations were obtained that lie at the optimal Pareto front. In all of these candidate combinations, the PCE is no less than 17.5%, and... All do not exceed 0.22 eV.

[0057] All preset performance constraints were applied to the 18 candidate parameter combinations for screening, resulting in 3 parameter combinations that met the constraints. Based on these, a comprehensive evaluation function was selected. The combination with the highest value was selected as the optimal parameter combination. The final parameters were determined as follows: third component mass fraction 15%, heat annealing temperature 130℃, annealing time 10 min, additive volume fraction 1%, and spin coating speed 2400 rpm. The corresponding predicted result was: PCE 18.3%. It is 0.12 eV. The voltage was 0.18 eV, and the stability retention rate was 87.2%. It is 0.876.

[0058] The photovoltaic performance evaluation method described in this application under different parameter combinations achieves a unified multi-dimensional characterization of device performance by constructing a multi-dimensional evaluation index system covering energy conversion efficiency, electrical output, energy loss, stability, and cost. It quantitatively evaluates photovoltaic performance under different parameter combinations by establishing a multi-objective comprehensive evaluation function and clarifying the optimization direction of each index. Furthermore, it obtains a set of candidate parameter combinations that satisfy the balance relationship of multiple indices based on a multi-objective optimization method. Finally, it screens the combinations based on preset performance constraints to determine the parameter combination with the optimal overall performance, achieving multi-objective collaborative optimization and precise selection of the optimal solution.

[0059] In some embodiments, the method for adjusting the proportioning parameters and device fabrication process parameters is as follows: Determine the initial value range of the proportioning parameters and device fabrication process parameters, and construct the parameter search space and constraints; The parameter search space is initially screened to identify key influencing parameters, and the correlation between key influencing parameters and device performance is established. Based on the correlation, iterative optimization search is performed on key influencing parameters to determine the parameter optimization direction and search step size; Based on a multi-objective comprehensive evaluation function, the improvement of photoelectric conversion efficiency and the reduction of charge transfer state radiation loss are evaluated. Non-radiative composite loss By reducing the synergistic optimization, the optimal or near-optimal combination of ratio parameters and device fabrication process parameters that meet the requirements of multi-objective optimization can be obtained.

[0060] For example, taking the PM6:L8-BO:BTP-eC9 ternary system as an example, the adjustment methods of the above-mentioned ratio parameters and device fabrication process parameters are explained.

[0061] First, the parameter search space and constraints were constructed. The initial range of the proportioning parameters was set: the mass ratio of the main donor material PM6 to the main acceptor material L8-BO was fixed at 1:1.2; the mass fraction of the third component BTP-eC9 was set to 5%–30% of the total mass of the active layer, discretized in 2.5% increments; for process parameters, the total solid concentration of the active layer was set to 18 mg / mL–22 mg / mL, the spin-coating speed for the second stage was set to 2000 rpm–2800 rpm, the thermal annealing temperature was set to 100℃–150℃, the annealing time was set to 5 min–15 min, and the volume fraction of the additive 1,8-diiodooctane was set to 0.5%–2%. Constraints were also set: the active layer thickness was controlled within the range of 100 nm–150 nm, and all parameter values ​​met the physical feasibility requirements.

[0062] Secondly, preliminary screening of parameters was conducted to identify key influencing parameters. Global sensitivity analysis was performed based on the constructed parameter search space. The results showed that the mass fraction of the third component, the thermal annealing temperature, and the volume fraction of the additives contributed 38.2%, 27.5%, and 19.6% to the photoelectric conversion efficiency (PCE), respectively, with a cumulative contribution exceeding 85%. Therefore, these parameters were identified as key influencing parameters. Based on this, a photovoltaic device performance prediction model was used to establish the correlation between key parameters and device performance indicators, including PCE, open-circuit voltage (Voc), short-circuit current density (Jsc), fill factor (FF), and energy loss term. and The analysis results show that when the mass fraction of the third component is approximately 15%, the PCE is in a locally optimal range; and when the heat annealing temperature is approximately 130℃, it is beneficial to reduce the PCE. .

[0063] Next, iterative optimization search was conducted. Using the mass fraction of the third component, the heat annealing temperature, and the volume fraction of the additive as optimization variables, a Bayesian optimization method was used to conduct an iterative search in the parameter space: In the initial stage, a coarse-grained search was performed on the mass fraction of the third component with a step size of 5%, and the advantageous range for PCE improvement was determined to be 12.5% ​​to 17.5%; then the step size was converged to 2.5%, and a joint fine search was performed by combining the heat annealing temperature (step size of 5℃) and the volume fraction of the additive (step size of 0.25%), completing a total of 48 rounds of iterative calculations to obtain a better combination of parameters.

[0064] Finally, multi-objective collaborative optimization is performed. A multi-objective evaluation function is constructed as follows: ,in, =0.6, =0.2, =0.2. Through multi-objective collaborative optimization calculations, the optimal parameter combination that meets the requirements of performance improvement and energy loss suppression was obtained: the mass fraction of the third component BTP-eC9 is 15%, the heat annealing temperature is 130℃, the annealing time is 10 min, the volume fraction of the additive 1,8-diiodooctane is 1%, and the spin coating second stage speed is 2400 rpm. The corresponding performance prediction results are: PCE is 18.3%, Voc is 0.871V, and Jsc is 26.4 mA / cm. 2 FF is 79.5%. It is 0.12 eV. With a voltage of 0.18 eV, a synergistic optimization effect of improving photoelectric conversion efficiency and reducing energy loss is achieved.

[0065] The method for adjusting the aforementioned proportioning parameters and device fabrication process parameters in this application achieves a standardized definition of the parameter search space by constructing an initial parameter value range and setting constraints; by initially screening the search space, key influencing parameters are identified and their correlation with device performance is established, thereby reducing the optimization dimensionality; further, based on this correlation, iterative optimization search of key parameters is performed to determine the optimization direction and step size, achieving efficient convergence of the parameter space; finally, combined with a multi-objective comprehensive evaluation function, the photoelectric conversion efficiency and energy loss index are improved (…). and Collaborative optimization is performed to obtain the optimal or near-optimal parameter combination that satisfies multi-objective constraints, thereby achieving systematic optimization design of device performance.

[0066] In some embodiments, the method for preparing ternary photovoltaic devices based on optimal parameter combinations is as follows: The conductive substrate is pretreated to obtain a clean and surface-activated substrate structure; A hole transport layer structure is constructed on the substrate structure to form a bottom charge-selective transport interface; An active layer precursor solution is prepared based on an optimal parameter combination. The precursor solution contains a host donor material, a host acceptor material, and a third component material. Under controlled environmental conditions, the active layer precursor solution is prepared into a thin film structure to form an active layer thin film with a target phase separation structure. The active layer thin film is subjected to structural control treatment, and the electron transport layer and electrode structure are constructed sequentially to form a complete ternary photovoltaic device structure.

[0067] The method for fabricating ternary photovoltaic devices based on optimal parameter combinations described in this application optimizes the interface film formation by pretreating the conductive substrate to obtain a clean and surface-activated substrate structure; constructing a hole transport layer on the substrate to form a bottom charge-selective transport interface, thereby improving carrier extraction efficiency; preparing an active layer precursor solution containing a host donor, a host acceptor, and a third component material based on optimal parameter combinations to achieve precise matching of the material system; preparing the precursor solution into a thin film structure under controlled environmental conditions to form an active layer with a target phase separation morphology; further structurally regulating the active layer and sequentially constructing an electron transport layer and electrode structure to complete the overall construction of the ternary photovoltaic device, achieving synergistic optimization of device structure and performance.

[0068] In some embodiments, the method for correcting the photovoltaic device performance prediction model based on experimental performance data is as follows: Obtain the predicted performance data and experimentally measured performance data of photovoltaic devices, and calculate the deviation between the two. Specifically targeting open-circuit voltage, short-circuit current density, fill factor, photoelectric conversion efficiency, and charge-transfer state radiation loss. Non-radiative composite loss Calculate the error statistics separately; The error statistics are compared with the preset error thresholds. When the error of any indicator exceeds the corresponding preset threshold, the model correction process is triggered. Based on the key physical parameters measured in experiments, the physical model of semiconductor devices is corrected to update the physical model parameters; New experimental data samples are introduced into the machine learning model training process, and incremental learning methods are used to update the machine learning model parameters. By integrating the prediction results of the physical model, the prediction results of the machine learning model, and the prediction results of similar cases, an integrated prediction result is obtained, and the parameters of the photovoltaic device performance prediction model are comprehensively corrected. The corrected model is evaluated by cross-validation. The model correction is completed when the preset convergence condition is met; otherwise, iterative correction continues.

[0069] The method described above for correcting the photovoltaic device performance prediction model based on experimental performance data achieves quantitative assessment of model error by obtaining predicted and measured data and calculating the deviation; and by separately analyzing open-circuit voltage, short-circuit current density, fill factor, photoelectric conversion efficiency, and energy loss indicators. , An error statistics system is constructed and compared with a preset threshold to achieve adaptive control of the model correction triggering mechanism. Based on this, the physical model of the semiconductor device is parameter-corrected based on the key physical parameters of the experiment, and the machine learning model is updated by incremental learning in combination with new data samples. Furthermore, the prediction results of the physical model, machine learning model and similar cases are integrated to form an integrated prediction output to achieve multi-model collaborative correction. Finally, the convergence is evaluated through cross-validation to achieve continuous iterative optimization and accuracy improvement of the photovoltaic performance prediction model.

[0070] In some embodiments, the material design guidance information includes suggestions for optimizing the molecular structure of the third component, donor / acceptor energy level matching strategies, screening criteria for the third component material, morphology modulation techniques, device structure improvement schemes, and energy loss suppression mechanisms, wherein: The third component molecular structure optimization suggestions include: selecting benzodithiazole as the electron acceptor unit to utilize its high planarity and electron-vibrational coupling modulation ability; adjusting the alkyl side chain length through side chain engineering to optimize solubility and aggregation behavior; introducing cyano, ester, or thiophene groups through end group modification to regulate energy level structure and luminescence efficiency; and adopting a ternary copolymerization strategy to introduce the benzodithiazole unit into the host donor polymer backbone to improve material compatibility. The donor / acceptor energy level matching strategy is as follows: a cascade-shaped energy level arrangement is constructed, where the donor HOMO is higher than the third component HOMO, the third component HOMO is higher than the acceptor HOMO, and the LUMO exhibits a decreasing gradient distribution. The HOMO energy level difference is controlled within the range of 0.1 eV to 0.3 eV to balance voltage loss and exciton transport driving force; the energy level difference between the third component LUMO and the acceptor LUMO is greater than 0.3 eV to provide charge separation driving force. Taking the PM6:BTP-eC9:L8-BO ternary system as an example, the cascade-shaped energy level arrangement is as follows: Figure 2 As shown; The screening criteria for the third component materials include: having an absorption peak at 550 nm to supplement the absorption of the host system; a photoluminescence quantum efficiency greater than 40% to suppress nonradiative recombination; a surface energy difference of less than 2 mN / m with the host material to ensure good miscibility; a recombination energy of less than 0.2 eV to promote carrier transport; and a single crystal packing density greater than 60% with the thin films arranged face-to-face. The morphology control technique includes: selecting 1,8-diiodooctane or 1-chloronaphthalene as solvent additives; optimizing the thermal annealing conditions within the range of 100℃~150℃ and 5min~15min; using tetrahydrofuran or carbon disulfide solvent vapor annealing to induce molecular rearrangement; and constructing a pseudo-layered structure using a two-step sequential layer-by-layer method to enrich the third component at the bottom interface. Device structure improvement schemes include: optimizing the materials and thickness of the hole transport layer and electron transport layer to regulate interface energy level matching; using an optical microcavity structure to optimize the active layer thickness to enhance light absorption; and introducing a self-assembled molecular layer as an interface modification layer to improve charge extraction efficiency. Energy loss suppression mechanisms include: reducing energy loss related to energy level differences through waterfall-type energy level arrangement; reducing recombination energy by introducing highly planar rigid molecules; suppressing nonradiative recombination loss through high photoluminescence quantum efficiency materials; and reducing energy disorder and trapped state density by optimizing tight three-dimensional packing and face-to-face orientation.

[0071] The material design guidance information in this application proposes a third-component molecular structure optimization scheme to achieve framework design, side-chain engineering, and end-group modification based on benzodithiazole units, thereby controlling the material's solubility, energy level structure, and luminescence performance. By constructing a cascade-type energy level matching strategy, it achieves gradient energy level arrangement between the donor, third component, and acceptor, and limits the HOMO and LUMO energy level difference range to optimize charge separation and transport driving forces. Through establishing multi-dimensional material screening criteria, it systematically screens third-component materials based on spectral absorption, quantum efficiency, interfacial compatibility, recombination energy, and stacking structure. Through morphology control techniques, combined with additive selection, annealing condition optimization, and solvent vapor treatment, it achieves precise control of the active layer phase separation structure and molecular rearrangement. Through device structure improvement schemes, it optimizes the charge transport layer and optical structure, improving carrier extraction and light absorption efficiency. Furthermore, through a multi-mechanism synergistic energy loss suppression strategy, it reduces radiative and non-radiative losses from aspects such as energy level arrangement, recombination process, and molecular stacking, thereby achieving systematic optimization and design guidance for the performance of ternary photovoltaic devices.

[0072] This embodiment also provides a photovoltaic performance prediction system for ternary systems, used to implement the above-mentioned photovoltaic performance prediction method, including: The material screening module is used to perform multi-parameter comprehensive scoring and screening based on the multi-dimensional characterization data of candidate third-component materials to determine the target candidate material set. The microscopic simulation module, connected to the material screening module, is used to simulate the microstructure and electronic properties of each material in the target candidate material set, obtain structural description parameters and energy characteristic parameters, and construct a photovoltaic device performance prediction model based on the structural description parameters and energy characteristic parameters. The performance prediction module, connected to the microscopic simulation module, is used to input the proportioning parameters of each component of the ternary system and the device fabrication process parameters into the photovoltaic device performance prediction model to obtain the photovoltaic performance prediction results, and to evaluate and iteratively optimize the prediction results until the preset performance optimization target is met and then output the optimal parameter combination. The experimental feedback module, connected to the performance prediction module, is used to prepare ternary photovoltaic devices based on the optimal parameter combination and perform performance testing to obtain experimental performance data. Based on the experimental performance data, the module corrects the photovoltaic device performance prediction model and outputs the corrected photovoltaic performance prediction results and material design guidance information.

[0073] The photovoltaic performance prediction system for ternary systems described in this application uses a material screening module to comprehensively score and screen candidate third-component materials based on multi-dimensional characterization data to determine the target candidate material set. A microscopic simulation module simulates the microstructure and electronic properties of the target candidate material set, extracting structural description parameters and energy characteristic parameters, and constructing a photovoltaic device performance prediction model to achieve correlation modeling between intrinsic material properties and device performance. A performance prediction module inputs the proportioning parameters and fabrication process parameters into the prediction model to obtain photovoltaic performance prediction results, which are then evaluated and iteratively optimized to obtain the optimal parameter combination. An experimental feedback module performs device fabrication and performance testing based on the optimal parameter combination, and corrects and updates the prediction model based on experimental data. This forms a closed-loop optimization system of "material screening—mechanism simulation—performance prediction—experimental feedback correction," enabling efficient design and precise optimization of ternary photovoltaic devices.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting photovoltaic performance in ternary systems, characterized in that, Includes the following steps: Based on the multi-dimensional characterization data of the candidate third component materials, the target candidate material set is determined; Microstructure and electronic properties of each material in the target candidate material set are simulated to obtain structural description parameters and energy characteristic parameters; A performance prediction model for photovoltaic devices is constructed based on structural description parameters and energy characteristic parameters; The proportioning parameters of each component in the ternary system and the device fabrication process parameters are input into the photovoltaic device performance prediction model to obtain the corresponding photovoltaic performance prediction results. The method for obtaining multi-dimensional representation data is as follows: The candidate third component materials were pretreated, and a unified testing environment and standardized testing parameter system were constructed. Optical performance tests were performed on the candidate third component material to obtain ultraviolet-visible absorption and photoluminescence spectra, and the absorption peak intensity, photoluminescence quantum efficiency and optical band gap were extracted. At the same time, the refractive index dispersion relation and photon trapping ability parameters were calculated. Electrical performance tests were conducted on the candidate third component material to obtain the following electrical characteristic parameters: highest occupied molecular orbital energy level and lowest unoccupied molecular orbital energy level, hole mobility and electron mobility, temperature dependence of conductivity as a function of temperature, dielectric constant, and charge transfer state energy level. Thermal stability and environmental stability tests were conducted on the candidate third component materials to obtain glass transition temperature, melting temperature, thermal decomposition temperature, light stability decay coefficient, and degradation kinetic parameters under water and oxygen conditions. Interfacial properties and molecular interactions of candidate third-component materials were tested to obtain interfacial energy parameters and inter-material interaction characteristics. The method for determining the target candidate material set is as follows: Based on the multi-dimensional characterization data of the candidate third component materials, a multi-dimensional evaluation index system was constructed, and quantitative evaluation criteria and screening standards were set for each evaluation dimension. Based on multi-dimensional characterization data and quantitative evaluation criteria, parameters were extracted and normalized for each candidate third component material to obtain standardized evaluation values ​​for each evaluation dimension. Determine the weight coefficients for each evaluation dimension and construct a multi-parameter weighted evaluation rule; The comprehensive score of each candidate third component material is calculated based on standardized evaluation values ​​and weighting coefficients, and the comprehensive score is corrected based on the Delphi method. Based on the revised comprehensive score, each candidate third component material is sorted and screened. Candidate third component materials whose comprehensive scores are within the preset range and whose evaluation indicators meet the corresponding threshold conditions are selected to form the target candidate material set.

2. The photovoltaic performance prediction method according to claim 1, characterized in that, The photovoltaic device performance prediction model adopts a hybrid modeling strategy that integrates semiconductor device physical models and machine learning models, wherein: The physical model of the semiconductor device is based on the drift-diffusion equation, Poisson equation and carrier continuity equation to construct a numerical solution framework, which is used to describe the carrier generation, transport and recombination process inside the photovoltaic device. The machine learning model is based on a deep neural network structure to construct a nonlinear mapping prediction framework, which is used to establish the mapping relationship between structural description parameters, energy characteristic parameters, ratio parameters, device fabrication process parameters and photovoltaic performance evaluation parameters.

3. The photovoltaic performance prediction method according to any one of claims 1-2, characterized in that, The photovoltaic performance prediction method also includes: The photovoltaic performance prediction results under different parameter combinations are evaluated, and it is determined whether they meet the preset performance optimization targets. If the target is not met, the ratio parameters and device fabrication process parameters are adjusted and re-entered into the photovoltaic device performance prediction model for iterative calculation until the performance optimization target is met. If the optimal parameter combination is satisfied, the optimal parameter combination is output, and a ternary photovoltaic device is fabricated based on the optimal parameter combination and its performance is tested to obtain experimental performance data.

4. The photovoltaic performance prediction method according to claim 3, characterized in that, The method for evaluating photovoltaic performance prediction results under different parameter combinations is as follows: A multi-dimensional evaluation index system is constructed, which includes energy conversion efficiency index, electrical output index, energy loss index, stability index and cost index; Construct a multi-objective comprehensive evaluation function to uniformly represent each evaluation index and define the optimization direction of each evaluation index; The performance of photovoltaic devices under different parameter combinations is calculated and evaluated to obtain corresponding multi-dimensional performance evaluation results; Multi-objective optimization processing is performed on the multi-dimensional performance evaluation results to obtain a set of candidate parameter combinations that satisfy the balance relationship of multiple indicators; The candidate parameter combination set is filtered based on the preset performance constraints to obtain the optimal parameter combination that satisfies both the multi-objective optimization results and the preset performance constraints.

5. The photovoltaic performance prediction method according to claim 3, characterized in that, The method for adjusting the proportioning parameters and device fabrication process parameters is as follows: Determine the initial value range of the proportioning parameters and device fabrication process parameters, and construct the parameter search space and constraints; The parameter search space is initially screened to identify key influencing parameters, and the correlation between key influencing parameters and device performance is established. Based on the correlation, iterative optimization search is performed on key influencing parameters to determine the parameter optimization direction and search step size; Based on the multi-objective comprehensive evaluation function, the photoelectric conversion efficiency is improved and the radiation loss and non-radiative recombination loss of the charge transfer state are reduced in a coordinated manner, so as to obtain the optimal or near-optimal combination of ratio parameters and device fabrication process parameters that meet the requirements of multi-objective optimization.

6. The photovoltaic performance prediction method according to claim 3, characterized in that, The method for fabricating ternary photovoltaic devices based on optimal parameter combinations is as follows: The conductive substrate is pretreated to obtain a clean and surface-activated substrate structure; A hole transport layer structure is constructed on the substrate structure to form a bottom charge-selective transport interface; An active layer precursor solution is prepared based on an optimal parameter combination. The precursor solution contains a host donor material, a host acceptor material, and a third component material. Under controlled environmental conditions, the active layer precursor solution is prepared into a thin film structure to form an active layer thin film with a target phase separation structure. The active layer thin film is subjected to structural control treatment, and the electron transport layer and electrode structure are constructed sequentially to form a complete ternary photovoltaic device structure.

7. The photovoltaic performance prediction method according to claim 3, characterized in that, The photovoltaic performance prediction method also includes: The photovoltaic device performance prediction model is corrected based on experimental performance data, and the final photovoltaic performance prediction results and material design guidance information are output based on the corrected photovoltaic device performance prediction model.

8. A photovoltaic performance prediction system for ternary systems, used to implement the photovoltaic performance prediction method as described in claim 6, characterized in that, include: The materials screening module is used to obtain the target candidate materials set; The microscopic simulation module, connected to the material screening module, is configured to: acquire structural description parameters and energy characteristic parameters, and construct a photovoltaic device performance prediction model; The performance prediction module, connected to the microscopic simulation module, is configured to: acquire the photovoltaic performance prediction results of the ternary system; perform iterative optimization based on the photovoltaic performance prediction results and the preset performance optimization target, and output the optimal parameter combination; The experimental feedback module, connected to the performance prediction module, is configured to: prepare ternary photovoltaic devices based on the optimal parameter combination and test the experimental performance data; revise the photovoltaic device performance prediction model based on the experimental performance data; and output the final photovoltaic performance prediction results and material design guidance information.

Images