Method and system for optimizing and analyzing layout scheme of cadmium telluride glass in plant factory

Abstract

The invention provides a plant factory cadmium telluride glass layout scheme optimization analysis method and system. The method comprises the following steps: generating a plurality of potential cultivation surface contour data based on expected plant cultivation data, and determining an equivalent cultivation surface based on each potential cultivation surface contour data; calculating spatial distribution information of daily cumulative illumination quantity formed on the equivalent cultivation surface by the cadmium telluride glass and the gap within the representative time period based on the geometric parameters of the enclosure structure, the first layout parameters and the sun motion trail, and calculating an evaluation index representing the light receiving uniformity of the crop canopy based on the spatial distribution information; and adjusting the first layout parameter by using an optimization algorithm so as to obtain a second layout parameter of which the total generating capacity and the evaluation index both reach a predetermined balance target. According to the invention, while the photovoltaic coverage rate is improved to increase the generating capacity, the illumination intensity and uniformity of crop canopies are ensured to reach the standard, and the crop yield quality and the system energy utilization efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of optimization design technology, and more specifically, to a method and system for optimizing the layout scheme of cadmium telluride glass in plant factories. Background Technology

[0002] Plant factories, as a typical form of facility agriculture, achieve high-efficiency crop production year-round through precise control of light, temperature, humidity, and nutrients. Light is a key environmental factor for plant photosynthesis and morphogenesis. In recent years, to achieve energy self-sufficiency and green sustainable development, researchers have proposed introducing cadmium telluride (CdTe) photovoltaic glass into the enclosure structure of plant factories. This type of glass can absorb some solar radiation for photovoltaic power generation while simultaneously allowing some light to pass through for plant use, thus achieving the dual goals of agricultural and energy production.

[0003] However, existing cadmium telluride glass deployment methods mainly focus on two aspects: first, ensuring crop light exposure by increasing light transmittance or leaving gaps; and second, increasing power generation by increasing the glass coverage area. Because these two objectives are inherently contradictory, existing solutions often struggle to achieve a balance between the light intensity and uniformity required for plant growth and the system's power generation efficiency.

[0004] Furthermore, in structures employing intermittent spacing, the cadmium telluride glass and the light-transmitting gaps create distinct light and dark stripes on the crop canopy. These stripes move across the cultivation surface as the sun's angle changes, directly affecting the light distribution to plants at different positions. However, existing studies typically only estimate the shading ratio at a static moment, failing to systematically consider the dynamic evolution of the light and dark stripes, and lacking analysis of the impact of stripe superposition effects on the uniformity of daily cumulative light intensity (DLI). This can lead to some plants potentially being in low-light areas for extended periods, resulting in inconsistent growth within the plant population and reduced yield and quality.

[0005] In summary, the layout scheme of cadmium telluride glass in plant factories needs further optimization to achieve a balance between power generation efficiency and uniformity of light exposure time in the crop canopy, thereby improving the yield and quality of plant factories. Summary of the Invention

[0006] To address the technical problems mentioned above, this invention provides a method and system for optimizing the layout of cadmium telluride glass in plant factories.

[0007] This invention provides a method for optimizing the layout of cadmium telluride glass in plant factories, comprising the following steps: The geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters of the cadmium telluride glass are determined; wherein, the first layout parameters include at least the glass transmittance, glass width, gap width between adjacent glass panes, and glass laying direction. Based on the expected plant planting data, several potential cultivation surface contour data are generated, and the equivalent cultivation surface is determined based on each potential cultivation surface contour data. Based on the geometric parameters of the enclosure structure, the first layout parameters, and the solar motion trajectory, the spatial distribution information of the daily cumulative light intensity formed by the cadmium telluride glass and gaps on the equivalent cultivation surface during a representative time period is calculated. Based on the spatial distribution information, an evaluation index characterizing the uniformity of light received by the crop canopy is calculated. The first deployment parameters are adjusted using an optimization algorithm to obtain a second deployment parameter in which both the total power generation and the evaluation index reach a predetermined balance target; wherein the total power generation is calculated based on the first deployment parameters and standard solar radiation data, and corresponds to the representative time period.

[0008] This invention also provides a system for optimizing the layout of cadmium telluride glass in plant factories, comprising: The receiving unit is configured to: determine the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters of the cadmium telluride glass; wherein the first layout parameters include at least the glass transmittance, the glass width, the gap width between adjacent glass panes, and the glass laying direction. The equivalent processing unit is configured to: generate several potential cultivation surface contour data based on the expected plant planting data, and determine the equivalent cultivation surface based on each potential cultivation surface contour data; The light evaluation unit is configured to: calculate the spatial distribution information of the daily cumulative light intensity formed by the cadmium telluride glass and gaps on the equivalent cultivation surface during a representative time period based on the geometric parameters of the enclosure structure, the first layout parameters and the solar motion trajectory; and calculate the evaluation index characterizing the uniformity of light received by the crop canopy based on the spatial distribution information. The optimization processing unit is configured to: use an optimization algorithm to adjust the first deployment parameters to obtain a second deployment parameter in which both the total power generation and the evaluation index reach a predetermined balance target; wherein the total power generation is calculated based on the first deployment parameters and standard solar radiation data, and corresponds to the representative time period.

[0009] The beneficial technical effects of this invention are as follows: On the one hand, by constructing an equivalent cultivation surface, the light calculation benchmark for multi-layer / multi-region cultivation scenarios is unified. Combined with the dynamic calculation of the daily cumulative light distribution based on the sun's movement trajectory, the impact of the movement of bright and dark stripes on the uniformity of light is accurately determined, avoiding the problem of long-term low light growth of plants caused by static estimation, and ensuring the light needs of crops. On the other hand, the first deployment parameters are iteratively adjusted using a multi-objective optimization algorithm, and the total power generation and light uniformity evaluation indicators are simultaneously incorporated into the optimization objectives. The final output of the second deployment parameters can increase the photovoltaic coverage rate to increase power generation while ensuring that the light intensity and uniformity of the crop canopy meet the standards. This achieves synergistic optimization of plant factory energy production and agricultural production, and significantly improves crop yield and quality as well as system energy utilization efficiency. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating an optimization analysis method for the layout scheme of cadmium telluride glass in a plant factory, as disclosed in an embodiment of the present invention. Figure 2 This is a schematic diagram of the process for obtaining the second deployment parameters based on a generative adversarial network, as disclosed in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an optimization analysis system for the layout scheme of cadmium telluride glass in a plant factory, as disclosed in an embodiment of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0013] like Figure 1 As shown, an embodiment of the present invention provides a method 10 for optimizing the layout of cadmium telluride glass in a plant factory, comprising the following steps: 100, determine the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters of the cadmium telluride glass; wherein, the first layout parameters include at least the glass transmittance, glass width, gap width between adjacent glass panes, and glass laying direction; In this step, the designer first needs to input the following three types of data into the analysis system of this invention. The specific definitions and functions of each type of data are as follows: (1) Enclosure geometric parameters: refers to the physical dimensional properties of the plant factory roof or cadmium telluride glass mounting surface, including but not limited to roof tilt angle (e.g., 30°, 45°), azimuth angle (e.g., due south, 15° east of south), roof length (e.g., 50m), roof width (e.g., 20m), etc.

[0014] Based on these geometric parameters, a three-dimensional geometric optical model can be constructed. Specifically, by using the roof tilt angle and azimuth angle, the initial angle range of sunlight incident on the installation surface in different seasons and at different times can be determined. Combining the roof length and width, the total area and spatial boundaries of cadmium telluride glass that can be laid can be determined, which is used to subsequently calculate the effective power generation area of ​​cadmium telluride glass. For example, when the roof faces due south and tilts at 30°, the incident angle of sunlight at noon in summer is closer to vertical, and the length of the shadow cast by the glass will be smaller, while in winter the incident angle is smaller, and the shadow will be correspondingly longer.

[0015] (2) Expected plant planting data: refers to the planned crop cultivation information, including crop type (such as lettuce, tomato, strawberry, etc.), cultivation mode (such as flat cultivation, three-layer vertical rack cultivation, bag vertical cultivation, etc.) and expected canopy height range (such as lettuce canopy height 15-20cm, tomato canopy height 80-100cm). Among them, different crops have significantly different light requirements (such as tomatoes need higher daily cumulative light intensity (DLI), while lettuce has relatively lower requirements); and the cultivation mode will affect the spatial distribution of the actual cultivation surface. For example, in multi-layer rack cultivation, the height and spacing of each rack will lead to differences in the light environment of different layers of canopy; the expected canopy height can be used to determine the target plane for light calculation, that is, the main area where crop photosynthesis occurs, to avoid the distortion of light assessment due to the mismatch between the calculation plane and the actual canopy.

[0016] (3) First layout parameters of cadmium telluride glass: refers to the initial layout parameters that need to be optimized. These parameters can be set manually by the designer (e.g., based on experience, the glass width is set to 1m and the gap width to 0.5m) or automatically generated by traditional design software. The specific source is not limited.

[0017] Among them, the light transmittance of glass is an inherent optical property of cadmium telluride glass, which refers to the percentage (e.g., 60%, 70%) of photosynthetically active radiation (PAR, wavelength 400-700nm) that it allows to pass through. This parameter directly determines the light intensity that reaches the crop canopy through the glass. The higher the light transmittance, the more light the crops can utilize, but the power generation efficiency of the glass will decrease accordingly (because some solar radiation needs to be used for light transmission).

[0018] The ratio of glass width to gap width (e.g., glass width 1.2m, gap width 0.3m), i.e., the shading ratio, directly affects the uniformity of light exposure and power generation. Specifically, increasing the glass width and decreasing the gap width can improve photovoltaic coverage and increase power generation, but it will expand the shading area and reduce the uniformity of light exposure; conversely, it can improve the uniformity of light exposure, but it will reduce power generation. The laying direction is divided into east-west and north-south, which determines the movement pattern of the shading shadow on the cultivation surface. Different movement patterns have a significant impact on the uniformity of light exposure to the crop canopy.

[0019] 200, Based on the expected plant planting data, generate several potential cultivation surface contour data, and determine the equivalent cultivation surface based on each potential cultivation surface contour data; In practical applications, plant factories employ diverse cultivation models. If light intensity is calculated separately for each cultivation area (such as each layer of a multi-layer shelf or each planting point in a vertical bag), the evaluation of light uniformity may be distorted due to differences in the calculation benchmarks for each area (such as different layer heights or different locations).

[0020] To address the aforementioned issues, this step extracts the physical contour information of all cultivation areas one by one based on the expected plant planting data. For example, when using a three-tiered vertical rack for lettuce cultivation, the planar dimensions of the first (0.5m above the ground), second (1.2m above the ground), and third (1.9m above the ground) cultivation racks (e.g., each layer is 40m long and 1.5m wide) and the coordinates of the rack edges projected onto the roof need to be obtained separately. If it is planar cultivation, the length (e.g., 50m), width (e.g., 18m), and edge coordinates of the entire cultivation area are directly extracted. If it is bag-type vertical cultivation, the continuous planar contour formed by the canopy after crop maturity (not the irregular shape of the cultivation bags themselves) needs to be extracted based on the canopy plane (e.g., the canopy of each row of cultivation bags forms a rectangular plane with a width of 0.8m and a length of 30m).

[0021] Next, through spatial projection and normalization, multiple dispersed potential cultivation surfaces are integrated into a unified virtual calculation plane. Specifically: using the main incident direction of sunlight (determined based on the azimuth and tilt angle of the enclosure structure, such as the main incident direction of a south-facing roof being south) as a reference, potential cultivation surfaces at different heights and locations are uniformly projected onto a virtual plane parallel to the sun's incident surface. For example, the cultivation surfaces of the 2nd and 3rd layers in a multi-tiered rack are projected onto a virtual plane at the same height as the 1st layer, eliminating the impact of height differences on illumination calculations. Secondly, if there are differences in the area of ​​different potential cultivation surfaces (e.g., each layer of a multi-tiered rack has an area of ​​60㎡, with 3 layers in total, for a total cultivation area of ​​180㎡), proportional conversion is required to ensure that the area ratio of the equivalent cultivation surface is consistent with the actual cultivation area ratio. That is, the equivalent cultivation surface must reflect a total calculated area of ​​180㎡, and the area ratio of the projected area of ​​each layer is the same as the actual area ratio of each layer.

[0022] 300. Based on the geometric parameters of the enclosure structure, the first layout parameters and the solar trajectory, the spatial distribution information of the daily cumulative light intensity formed by the cadmium telluride glass and gaps on the equivalent cultivation surface during a representative time period is calculated. Based on the spatial distribution information, an evaluation index characterizing the uniformity of light received by the crop canopy is calculated. In this step, the inclination and azimuth of the enclosure structure determine the actual angle at which sunlight is incident on the equivalent cultivation surface (the incident angle of the roof needs to be converted into the incident angle of the equivalent cultivation surface); the glass width, gap width, and laying direction in the first layout parameters determine the position and range of the shading area (glass projection area) and the light-transmitting area (gap projection area) on the equivalent cultivation surface at each moment; the solar trajectory data is based on the latitude and longitude of the plant factory and representative time periods (such as 6:00-18:00 on a typical sunny day, or the average sunshine period of a certain season), and the hourly solar altitude angle (such as 45° at 9:00, 70° at 12:00, and 50° at 15:00) and azimuth angle (such as 20° east of south at 9:00, due south at 12:00, and 20° west of south at 15:00) are calculated using astronomical algorithms to simulate the dynamic incident path of sunlight throughout the day.

[0023] Within a representative time period, the instantaneous light intensity of each pixel (e.g., pixel precision 0.1m × 0.1m) on the equivalent cultivation surface is calculated every hour (or every 30 minutes). Specifically: for light-transmitting areas, instantaneous light intensity = instantaneous solar radiation intensity × glass transmittance × incident angle correction factor (the larger the incident angle, the closer the correction factor is to 1, and the smaller the light loss); for shaded areas, instantaneous light intensity = instantaneous solar radiation intensity × glass shading rate (i.e., 1 - transmittance) × scattered light correction factor (considering the supplementary effect of ambient scattered light).

[0024] The instantaneous light intensity calculated hourly is summed to obtain the daily cumulative light intensity (DLI) for each pixel (unit: mol / m²·d). DLI is a core indicator for measuring crop photosynthesis, directly determining the crop's growth rate, yield, and quality (e.g., lettuce requires a DLI of 10-15 mol / m²·d to grow normally; below 8 mol / m²·d will lead to stunted growth). The DLI value of each pixel is then converted into a heatmap, which visually presents the location and extent of high-light areas (e.g., DLI ≥ 15 mol / m²·d, red area in the heatmap), suitable-light areas (e.g., DLI 10-15 mol / m²·d, yellow area), and low-light areas (e.g., DLI < 8 mol / m²·d, blue area) on an equivalent cultivation surface, forming spatial distribution information of the daily cumulative light intensity.

[0025] Furthermore, quantitative indicators are extracted from the aforementioned spatial distribution information to evaluate the uniformity of light exposure in the crop canopy. These quantitative indicators include, but are not limited to: Coefficient of variation (CV): The ratio of the standard deviation to the mean of the DLI values ​​of all pixels on an equivalent cultivation surface. The formula is CV = (standard deviation / mean) × 100%. The smaller the CV value, the more uniform the light distribution. Generally, crop growth requires CV ≤ 15%. If CV > 20%, it indicates significant light unevenness, and some plants may be in low-light areas for a long time. Light compliance rate: This is the percentage of pixels whose DLI value meets the minimum light requirement threshold for crop growth (e.g., 10 mol / m²·d for lettuce based on crop type) out of the total number of pixels in the equivalent cultivation area. A higher compliance rate indicates a wider area that meets the basic light requirements of the crop. If the compliance rate is <80%, the layout parameters need to be adjusted to expand the compliant area. Extreme difference: Calculate the difference between the maximum and minimum DLI values ​​within the equivalent cultivation area. The smaller the difference, the less obvious the area of ​​excessive light (which may cause crop scorching) or insufficient light (which may cause slow growth). For example, a difference of <5 mol / m²·d is ideal, while a difference of >10 mol / m²·d requires optimization.

[0026] 400, The first deployment parameters are adjusted using an optimization algorithm to obtain a second deployment parameter in which both the total power generation and the evaluation index reach the predetermined balance target; wherein, the total power generation is calculated based on the first deployment parameters and standard solar radiation data, and corresponds to the representative time period.

[0027] In this step, the total power generation during a representative time period is first calculated using the following formula: Total power generation (kWh) = Effective power generation area of ​​cadmium telluride glass (m²) × Standard solar radiation data (kWh / m²) × Photovoltaic conversion efficiency × Temperature correction factor. The effective power generation area is calculated based on the roof length and width of the building envelope, as well as the glass width and gap width of the first layout parameters. The standard solar radiation data uses the typical daily total solar radiation of the area where the plant factory is located. The photovoltaic conversion efficiency refers to the inherent power generation efficiency of the cadmium telluride glass (e.g., 18%, 20%). The effect of temperature on efficiency needs to be considered; typically, for every 1°C increase in temperature, the conversion efficiency decreases by 0.3%. Therefore, a temperature correction factor needs to be calculated based on the ambient temperature of the plant factory (e.g., indoor temperature of 30°C in summer, standard test temperature of 25°C) (e.g., correction factor = 1 - 0.003 × (30 - 25) = 0.985).

[0028] Next, multi-objective optimization algorithms such as NSGA-II non-dominated sorting genetic algorithm and particle swarm optimization algorithm are used to iteratively adjust the first layout parameters with the dual optimization objectives of maximizing total power generation and optimizing the evaluation index of uniformity of illumination. The specific optimization process is as follows: (1) Setting parameter adjustment boundaries: Based on the feasibility of engineering construction and material characteristics, set the adjustment range of the first layout parameters. For example, the glass width can be adjusted between 0.8-1.5m (too narrow will increase installation costs, too wide will lead to excessive shadows), the gap width can be adjusted between 0.3-1.0m (too narrow will affect light transmission, too wide will reduce power generation), the laying direction can only be selected as east-west / north-south, and the light transmittance is fixed according to the glass type (60%-80%).

[0029] (2) Each iteration of the optimization algorithm automatically adjusts a set of layout parameters, such as adjusting the glass width from 1m to 1.2m, the gap width from 0.5m to 0.4m, and the laying direction from east-west to north-south. The total power generation and light uniformity index corresponding to this set of layout parameters are obtained in the aforementioned manner (e.g., CV=13%, compliance rate=92%, total power generation=680kWh). (3) Non-dominated solutions are selected through non-dominated sorting, meaning there is no other set of layout parameters that can simultaneously satisfy both higher power generation and more uniform illumination. For example, a set of layout parameters A (glass width 1.2m, gap 0.4m, north-south orientation) has a power generation of 680kWh and CV=13%, while layout parameter B (glass width 1.0m, gap 0.6m, north-south orientation) has a power generation of 620kWh and CV=11%. Both are non-dominated solutions (A has higher power generation but a larger CV, while B has better illumination but lower power generation). Subsequently, combined with the plant factory's predetermined balance target (such as preset power generation ≥650kWh, CV≤15%, and achievement rate ≥90%), the parameter combination that meets the target is selected from the non-dominated solutions, which is the second set of layout parameters to be determined.

[0030] After multiple rounds of iteration and screening, the final output of the second layout parameters includes: the light transmittance of cadmium telluride glass, the glass width and the gap width between adjacent glasses, and the glass laying direction.

[0031] This invention can increase photovoltaic coverage to increase power generation while ensuring that the light intensity and uniformity of crop canopy meet the standards, thereby achieving synergistic optimization of plant factory energy production and agricultural production, and significantly improving crop yield and quality as well as system energy utilization efficiency.

[0032] Furthermore, determining the geometric parameters of the enclosure structure of the plant factory includes: 101. Obtain the BIM model of the plant factory and identify the attribute information of the components in each area related to the cadmium telluride glass layout. In this step, the analysis system reads the BIM design model of the plant factory through a data interface (such as IFC, RVT, etc.). This model is a digital database containing the geometric and non-geometric information of all components of the plant factory, rather than a simple two-dimensional drawing or three-dimensional graphic.

[0033] Then, accurately identify areas where cadmium telluride glass is planned or may be installed, such as roof components or specific exterior wall components. Here, "regional components" refers to the physical objects in the BIM model that represent different parts of the building, such as roofs, exterior walls, and skylights; "attribute information" refers to the parametric attributes defined for each component in the BIM model. These attributes include not only the component's geometric dimensions but also its type, material, and functional identifiers.

[0034] Target components can be identified by searching specific attribute fields, such as all objects categorized as roof or exterior wall. Further filtering can be used to find components whose type names or custom data contain keywords such as photovoltaic, power generation, CdTe, or solar energy. This allows for the automatic targeting of specific roof or facade areas for analysis.

[0035] 102. Based on the attribute information, extract the geometric parameters of the enclosure structure corresponding to the regional components, including at least the roof slope, roof azimuth, roof length, and roof width.

[0036] In this step, the corresponding parameter values ​​are read from the attribute list of the identified area components (such as the roof), or the required parameters are automatically calculated using their geometric data. Specific parameters are extracted, including: Roof slope: The slope or tilt angle parameter value can be directly read from the component properties. For example, the property may explicitly state a slope ratio of 30° or 2:12. This parameter determines the direction of sunlight.

[0037] Roof azimuth: This can be calculated by analyzing the projection of the roof structure's normal vector onto a global coordinate system (such as the world coordinate system). For example, if the calculated roof normal vector points due south, the azimuth is recorded as 0°; if it points 20° east of south, the azimuth is recorded as -20° (or 340°). This parameter determines the direction of sunlight.

[0038] Roof length and width: These are automatically calculated by obtaining the maximum dimensions of the bounding box or outline of the roof components. For example, the maximum span of the component in the east-west direction is read as the roof length, and the maximum span in the north-south direction is read as the roof width. These two parameters together define the total area and layout boundaries of the cadmium telluride glass that can be laid.

[0039] Further, the step of generating several potential cultivation surface contour data based on the expected plant planting data, and determining the equivalent cultivation surface based on each potential cultivation surface contour data, includes: 201. The type, spatial location, and outline dimensions of each cultivation area are derived from the expected plant planting data; In this step, the input expected plant planting data involves the planting plan of the plant factory. By identifying the identification fields or structural features, the type (such as being divided into a planar cultivation area, a three-dimensional rack cultivation area, or a bag-type three-dimensional cultivation area), spatial location (such as being located on the east or west side of the greenhouse), and outline dimensions (such as the length and width of the cultivation bed, the single-layer size and number of layers of the cultivation rack, and the distribution spacing and range of the cultivation columns) of each cultivation area can be automatically extracted.

[0040] 202. For each cultivation area, generate corresponding potential cultivation surface contour data according to its type and contour size; wherein, the type includes planar cultivation area, three-dimensional rack cultivation area, and bag-type three-dimensional cultivation area. In this step, for a planar cultivation area, its cultivation surface can be regarded as a two-dimensional plane. Based on its outline size, a horizontal projection surface that coincides with it is generated as its potential cultivation surface outline data. The height parameter of this surface is set to the typical height value of the canopy of this type of crop.

[0041] For vertical cultivation areas, cultivation occurs at multiple levels at different heights. A separate potential cultivation surface profile is generated for each level of the cultivation rack. Each potential cultivation surface profile includes: the horizontal projection profile of the effective cultivation area of ​​that level (ignoring non-planting structures such as supports); and the specific height value of that level. For example, a five-level vertical cultivation rack will generate five potential cultivation surface profiles located at different heights.

[0042] For bag-type vertical cultivation areas, the crop canopy has an irregular three-dimensional distribution in space. To achieve a simulation that is more in line with agronomical principles, this invention does not use the installation point of the cultivation bag as the basis, but instead uses the continuous overall outer contour naturally formed by the canopy after the crop has grown and matured as the calculation benchmark. The horizontal projection boundary of this overall outer contour is generated, and its average height is calculated, which together constitute the potential cultivation surface contour data of the cultivation area.

[0043] 203. Based on all potential cultivation surface contour data and their spatial locations, the equivalent cultivation surface is calculated using a weighted average or union algorithm.

[0044] In this step, all the potential cultivation surface contour data generated above are integrated, and a unified equivalent cultivation surface is calculated using a weighted average or union method to represent the global illumination calculation. Details are as follows: Weighted average algorithm: Calculate the average height of all potential cultivation surface profiles (using either an arithmetic mean or an area-weighted average). Then, project all cultivation surface profiles onto this calculated average height plane, ultimately forming a single virtual plane that incorporates all profile features.

[0045] The union algorithm performs a Boolean union operation on the horizontal projection profiles of all potential cultivation surfaces, merging them into the largest single projection profile that completely covers all planting areas; the height of the equivalent cultivation surface is set to the lowest value among all cultivation surfaces. This method ensures that even crops in the lowest position are included in the light assessment range, effectively avoiding omissions in low-light areas, and is particularly suitable for applications with extremely high requirements for light uniformity or large differences in cultivation surface height.

[0046] Furthermore, such as Figure 2 As shown, the first deployment parameters are adjusted using an optimization algorithm to obtain second deployment parameters in which both the total power generation and the evaluation index achieve a predetermined balance target, including: 401. Input the first deployment parameters into the pre-trained generative adversarial network model, and the generator in the generative adversarial network model outputs the preliminary deployment parameters that meet the constraints. In this step, intelligent initialization is performed first to generate a high-quality initial solution. Specifically, the first layout parameters are input into a pre-trained Generative Adversarial Network (GAN) model. This GAN model has been fully trained in advance using a large amount of historical excellent layout data or high-fidelity simulation data. Its generator has learned the inherent distribution law of efficient cadmium telluride glass layout schemes and the complex constraint relationship between multi-dimensional parameters, such as the effective combination range of glass width and gap width under a specific roof tilt angle.

[0047] After receiving the first set of parameters, the generator uses them as a guide to generate a new set of parameters that both meet engineering physical constraints (such as parameter value ranges) and are closer to the optimal region in terms of performance, namely the initial set of parameters.

[0048] 402. Using the initial deployment parameters as the initial population, a multi-objective optimization algorithm is used for iterative optimization. In each iteration, the corresponding total power generation and the evaluation index are calculated based on the current deployment parameters. The non-dominated sorting and selection are performed with the common optimization objectives of maximizing the total power generation and optimizing the evaluation index, until the iteration termination condition is reached, and the Pareto optimal solution set is output. In this step, the initial layout parameters generated by GAN are used as the initial population for a multi-objective optimization algorithm (such as NSGA-II non-dominated sorting genetic algorithm), thereby initiating the iterative optimization process.

[0049] Specifically, in each iteration, the following operations are performed: Based on each set of deployment parameters in the current population, the aforementioned illumination and power generation calculation model is invoked to simulate and calculate the corresponding illumination uniformity evaluation index (such as the uniformity coefficient CU) and total power generation. Then, with the common optimization objectives of maximizing total power generation and optimizing the illumination uniformity evaluation index (such as maximizing the CU value), all individuals in the current population are non-dominatedly sorted and selected to identify those superior individuals (i.e., Pareto optimal solutions) that cannot be surpassed by other solutions in both power generation and illumination objectives, and these are then passed on to the next generation.

[0050] Repeat the above iterative process until the preset number of iterations or convergence accuracy is reached, and output a Pareto optimal solution set. Each solution in the solution set represents a different optimal trade-off between power generation efficiency and uniformity of illumination.

[0051] 403. From the Pareto optimal solution set, select the solution that simultaneously satisfies that the total power generation is not lower than the predetermined power generation threshold and that the evaluation index is better than the predetermined uniformity threshold, and use it as the second layout parameter.

[0052] In this step, from the obtained Pareto optimal solution set, all solutions that simultaneously satisfy the following conditions are automatically selected: total power generation is not lower than a predetermined power generation threshold (e.g., daily power generation ≥ 650 kWh) and the uniformity of illumination evaluation index is better than a predetermined uniformity threshold (e.g., uniformity coefficient CU ≥ 0.88). These solutions are a set of solutions that meet both the minimum performance requirements and are globally optimal trade-offs. If multiple solutions exist, one can be selected according to preset rules (e.g., prioritizing the solution with the highest power generation or the solution with the best uniformity) or interactively by the designers. The final selected solution is used as the second layout parameter output by the optimization process.

[0053] In this embodiment, the experience learning capability of artificial intelligence is used to provide a high starting point for the search initial value for subsequent accurate optimization, which can effectively avoid the problem of slow convergence speed or getting trapped in local optima that may be caused by random initialization in traditional optimization algorithms.

[0054] Furthermore, the initial deployment parameters that satisfy the constraints, output by the generator in the generative adversarial network model, include: 4011. Extract multi-dimensional features that affect the optimization process from the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters. Input the multi-dimensional features into a pre-trained optimization difficulty classification model to obtain the optimization difficulty level. The multi-dimensional features include structural complexity features, planting heterogeneity features, and target conflict features. 4012, Based on the optimization difficulty level, the corresponding number of initial deployment parameters is mapped to obtain the number of initial deployment parameters generated, and the generative adversarial network model is controlled to generate the corresponding number of initial deployment parameters; The initial population is then generated based on the initial deployment parameters.

[0055] In traditional optimization algorithms, the initial population is usually randomly generated or generated according to simple rules, and its size is also a fixed value. However, for plant factories with complex structures and diverse planting patterns (such as east-west orientation, multi-slope, and a mixture of vertical racks and bag cultivation), the solution space is very large and complex, and the conflict between objectives is intense. If a small fixed initial population (such as 50) is used, its diversity may be insufficient to fully cover the promising solution region, causing the optimization algorithm to easily get trapped in local optima and fail to find a truly globally optimal trade-off.

[0056] To address the aforementioned technical problems, this invention first performs an intelligent assessment of the optimization difficulty. Specifically, it automatically extracts multi-dimensional features for quantifying the complexity of the optimization process from the geometric parameters of the plant factory's enclosure structure, the expected plant planting data, and the initial layout parameters. These features mainly include: Structural complexity characteristics: These are calculated based on the geometric parameters of the building envelope, such as the difference between the maximum and minimum tilt angles of the roof (for multi-sloped roofs), the deviation angle of the azimuth from true south, and the geometric irregularity index of the roof plan profile (which can be obtained by calculating its shape difference from a standard rectangle). Among these, building envelopes that deviate from the standard true south, have abnormal tilt angles, or have complex shapes will significantly increase the complexity of lighting simulation and parameter optimization.

[0057] Planting heterogeneity characteristics: These are calculated based on expected plant planting data, such as the difference between the maximum and minimum canopy heights of all cultivation areas (planar, vertical, and bagged), the dispersion coefficient of the area under different cultivation patterns, and the total number of cultivation pattern types. Large height differences, uneven area distribution, and the mixing of multiple patterns indicate strong spatial heterogeneity, requiring more constraints to be considered during optimization, thus increasing the difficulty.

[0058] Conflicting objective characteristics: This is derived from the analysis of the first layout parameters, such as the ratio between the cadmium telluride glass coverage rate of the initial scheme and the minimum average light intensity required by the crop based on planting data. If this ratio is much greater than 1, it indicates that the initial design tends to have high power generation but may cause severe shading, and there is a sharp conflict between the two optimization objectives of power generation and light intensity, making the solution difficult.

[0059] Subsequently, the extracted multi-dimensional feature vectors are input into a pre-trained optimization difficulty classification model to derive the initial optimization difficulty level (e.g., easy, medium, or complex) for the set parameters. The optimization difficulty classification model is a machine learning classifier, such as one based on random forest or support vector machine algorithms, which has been pre-trained and labeled using a large amount of historical project data and its corresponding optimization process performance (e.g., the number of iterations required for convergence).

[0060] Next, a dynamic mapping of the initial population size is performed. Specifically, based on the optimization difficulty level obtained above, a preset mapping rule table is queried to determine the number of initial deployment parameters required for the Generative Adversarial Network (GAN). It is understood that the mapping rules follow the principle that the higher the difficulty, the larger the population. For example, if the optimization difficulty level is simple, the number of generated parameters is mapped to N1 (e.g., N1=50); if the optimization difficulty level is medium, the number of generated parameters is mapped to N2 (e.g., N2=100); if the optimization difficulty level is complex, the number of generated parameters is mapped to N3 (e.g., N3=200); and N1... <N2<N3。

[0061] Then, the generator in the generative adversarial network model is controlled to perform forward inference a corresponding number of times (N1, N2, or N3), and outputs a specified number of preliminary deployment parameters that both meet engineering physical constraints and are close to the optimal region in terms of performance.

[0062] Finally, in subsequent optimization steps, the initial population for the multi-objective optimization algorithm is generated based on all the initial deployment parameters generated by the GAN. That is, the size of the initial population is directly equal to the number of parameters generated by the GAN (N1, N2, or N3).

[0063] This embodiment ensures that less computational resources are allocated to simple problems to improve efficiency, while providing a sufficiently large and diverse initial population for complex problems to enhance their global search capabilities and the success rate of finding better solutions.

[0064] Furthermore, using the aforementioned preliminary deployment parameters as the initial population specifically involves: generating an initial population based on all preliminary deployment parameters, including: All the initial deployment parameters generated by the generative adversarial network model are input into the discriminator of the generative adversarial network model to obtain the generation quality score of each initial deployment parameter, and the initial deployment parameters with generation quality scores lower than a predetermined quality threshold are eliminated. The remaining preliminary layout parameters after elimination are subjected to diversity enhancement processing to ensure that their distribution in the solution space meets the diversity requirements. The preliminary layout parameters after diversity enhancement processing are used together as the initial population of the multi-objective optimization algorithm.

[0065] While generators can produce a large number of initial deployment parameters, their output quality may fluctuate, and some initial deployment parameters may perform poorly. This invention addresses this issue by leveraging the powerful discriminator within the same GAN. Specifically, the discriminator, having learned the data distribution characteristics of excellent historical deployment schemes during pre-training, re-inputs all initial deployment parameters generated by the generator (e.g., 200 sets) into the discriminator. The discriminator outputs a generation quality score (e.g., a confidence value between 0 and 1) for each set of initial deployment parameters. This score directly reflects how closely the set of parameters resembles real excellent schemes. All parameter sets with generation quality scores below a threshold (e.g., 0.65) are automatically removed to ensure the high quality and elitism of the initial population baseline.

[0066] Although the remaining initial parameter sets after quality screening are of high individual quality, their distribution in the solution space (i.e., the space composed of all possible parameter combinations) may be too concentrated, exhibiting a clustering phenomenon. Directly using these as the population may cause the optimization algorithm to converge prematurely to a local region, failing to explore the global optimum. To address this issue, this invention enhances the diversity of the remaining parameter sets, specifically as follows: Assessing Distribution Density: Calculate the distribution of all remaining initial parameters in the solution space, identifying regions with excessively dense parameters and sparsely distributed parameters. Reducing Dense Regions: For dense regions with very similar parameter combinations, randomly remove some similar individuals, retaining only a few representative individuals to avoid resource redundancy. Enhancing Sparse Regions: For sparsely distributed or blank regions, use small-scale mutation or crossover strategies to generate new parameter sets to supplement them. For example, apply a small random perturbation to some parameters of existing individuals (such as glass width), or exchange some parameter values ​​between different individuals, thereby generating new solutions that are similar to but different from high-quality solutions, and add these new solutions to the set.

[0067] Finally, all the initial parameters, after the aforementioned quality screening and diversity enhancement processes, are used together as the initial population for the multi-objective optimization algorithm. In this initial population, each individual is certified as a high-quality solution by the discriminator, and the entire population maintains good distribution and diversity in the solution space, which is beneficial for quickly deriving the optimal solution.

[0068] like Figure 3 As shown, this embodiment of the invention also discloses a system 20 for optimizing the layout of cadmium telluride glass in plant factories, comprising: The receiving unit 21 is configured to: determine the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters of the cadmium telluride glass; wherein the first layout parameters include at least the glass transmittance, the glass width, the gap width between adjacent glass panes, and the glass laying direction. The equivalent processing unit 22 is configured to: generate several potential cultivation surface contour data based on the expected plant planting data, and determine the equivalent cultivation surface based on each potential cultivation surface contour data; The light evaluation unit 23 is configured to: calculate the spatial distribution information of the daily cumulative light intensity formed by the cadmium telluride glass and gaps on the equivalent cultivation surface during a representative time period based on the geometric parameters of the enclosure structure, the first layout parameters and the solar motion trajectory; and calculate the evaluation index characterizing the uniformity of light received by the crop canopy based on the spatial distribution information. The optimization processing unit 24 is configured to: use an optimization algorithm to adjust the first layout parameters to obtain a second layout parameter in which both the total power generation and the evaluation index reach a predetermined balance target; wherein the total power generation is calculated based on the first layout parameters and standard solar radiation data, and corresponds to the representative time period.

[0069] Furthermore, the receiving unit 21 is configured as follows: Obtain the BIM model of the plant factory and identify the attribute information of the components in each area related to the cadmium telluride glass layout; Based on the attribute information, extract the geometric parameters of the enclosure structure corresponding to the regional components, including at least the roof slope, roof azimuth, roof length, and roof width.

[0070] Furthermore, the equivalent processing unit 22 is configured as follows: The type, spatial location, and outline dimensions of each cultivation area are derived from the expected plant planting data. For each cultivation area, corresponding potential cultivation surface contour data are generated based on its type and outline size; wherein, the types include planar cultivation areas, three-dimensional rack cultivation areas, and bag-type three-dimensional cultivation areas. Based on all potential cultivation surface contour data and their spatial locations, the equivalent cultivation surface is calculated using a weighted average or union algorithm.

[0071] Furthermore, the optimization processing unit 23 is configured as follows: The first deployment parameters are input into the pre-trained generative adversarial network model, and the generator in the generative adversarial network model outputs preliminary deployment parameters that meet the constraints. Using the initial deployment parameters as the initial population, a multi-objective optimization algorithm is used for iterative optimization. In each iteration, the total power generation and the evaluation index are calculated based on the current deployment parameters. Non-dominated sorting and selection are performed with the common optimization objectives of maximizing the total power generation and optimizing the evaluation index, until the iteration termination condition is reached, and the Pareto optimal solution set is output. From the Pareto optimal solution set, select the solution that simultaneously satisfies that the total power generation is not lower than the predetermined power generation threshold and that the evaluation index is better than the predetermined uniformity threshold, and use it as the second layout parameter.

[0072] Furthermore, the generator in the generative adversarial network model outputs preliminary deployment parameters that meet the constraints, including: Multi-dimensional features influencing the optimization process are extracted from the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters. These multi-dimensional features are then input into a pre-trained optimization difficulty classification model to obtain the optimization difficulty level. The multi-dimensional features include structural complexity features, planting heterogeneity features, and target conflict features. Based on the optimization difficulty level, the corresponding number of initial deployment parameters is mapped to control the generative adversarial network model to generate the corresponding number of initial deployment parameters. The initial population is then generated based on the initial deployment parameters.

[0073] Furthermore, using the aforementioned preliminary deployment parameters as the initial population specifically involves: generating an initial population based on all preliminary deployment parameters, including: All the initial deployment parameters generated by the generative adversarial network model are input into the discriminator of the generative adversarial network model to obtain the generation quality score of each initial deployment parameter, and the initial deployment parameters with generation quality scores lower than a predetermined quality threshold are eliminated. The remaining preliminary layout parameters after elimination are subjected to diversity enhancement processing to ensure that their distribution in the solution space meets the diversity requirements. The preliminary layout parameters after diversity enhancement processing are used together as the initial population of the multi-objective optimization algorithm.

[0074] Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of this application, and the scope of this application is determined by the scope of the appended claims.

Claims

1. A method for optimizing the layout of cadmium telluride glass in a plant factory, characterized in that, Includes the following steps: The geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters of the cadmium telluride glass are determined; wherein, the first layout parameters include at least the glass transmittance, glass width, gap width between adjacent glass panes, and glass laying direction. Based on the expected plant planting data, several potential cultivation surface contour data are generated, and the equivalent cultivation surface is determined based on each potential cultivation surface contour data. Based on the geometric parameters of the enclosure structure, the first layout parameters, and the solar motion trajectory, the spatial distribution information of the daily cumulative light intensity formed by the cadmium telluride glass and gaps on the equivalent cultivation surface during a representative time period is calculated. Based on the spatial distribution information, an evaluation index characterizing the uniformity of light received by the crop canopy is calculated. The first deployment parameters are adjusted using an optimization algorithm to obtain a second deployment parameter in which both the total power generation and the evaluation index reach a predetermined balance target; wherein the total power generation is calculated based on the first deployment parameters and standard solar radiation data, and corresponds to the representative time period.

2. The method for optimizing the layout of cadmium telluride glass in a plant factory according to claim 1, characterized in that: Determine the geometric parameters of the enclosure structure of the plant factory, including: Obtain the BIM model of the plant factory and identify the attribute information of the components in each area related to the cadmium telluride glass layout; Based on the attribute information, extract the geometric parameters of the enclosure structure corresponding to the regional components, including at least the roof slope, roof azimuth, roof length, and roof width.

3. The method for optimizing the layout of cadmium telluride glass in a plant factory according to claim 1, characterized in that: Based on the expected plant planting data, several potential cultivation surface contour data are generated. Based on each potential cultivation surface contour data, an equivalent cultivation surface is determined, including: The type, spatial location, and outline dimensions of each cultivation area are derived from the expected plant planting data. For each cultivation area, corresponding potential cultivation surface contour data are generated based on its type and outline size; wherein, the types include planar cultivation areas, three-dimensional rack cultivation areas, and bag-type three-dimensional cultivation areas. Based on all potential cultivation surface contour data and their spatial locations, the equivalent cultivation surface is calculated using a weighted average or union algorithm.

4. The method for optimizing the layout of cadmium telluride glass in a plant factory according to claim 1, characterized in that: The first deployment parameters are adjusted using an optimization algorithm to obtain second deployment parameters in which both the total power generation and the evaluation index achieve a predetermined balance target, including: The first deployment parameters are input into the pre-trained generative adversarial network model, and the generator in the generative adversarial network model outputs preliminary deployment parameters that meet the constraints. Using the initial deployment parameters as the initial population, a multi-objective optimization algorithm is used for iterative optimization. In each iteration, the total power generation and the evaluation index are calculated based on the current deployment parameters. Non-dominated sorting and selection are performed with the common optimization objectives of maximizing the total power generation and optimizing the evaluation index, until the iteration termination condition is reached, and the Pareto optimal solution set is output. From the Pareto optimal solution set, select the solution that simultaneously satisfies that the total power generation is not lower than the predetermined power generation threshold and that the evaluation index is better than the predetermined uniformity threshold, and use it as the second layout parameter.

5. The method for optimizing the layout of cadmium telluride glass in a plant factory according to claim 4, characterized in that: The generator in the generative adversarial network model outputs preliminary deployment parameters that meet the constraints, including: Multi-dimensional features influencing the optimization process are extracted from the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters. These multi-dimensional features are then input into a pre-trained optimization difficulty classification model to obtain the optimization difficulty level. The multi-dimensional features include structural complexity features, planting heterogeneity features, and target conflict features. Based on the optimization difficulty level, the corresponding number of initial deployment parameters is mapped to control the generative adversarial network model to generate the corresponding number of initial deployment parameters. The initial population is then generated based on the initial deployment parameters.

6. The method for optimizing the layout of cadmium telluride glass in a plant factory according to claim 5, characterized in that: An initial population is generated based on all preliminary deployment parameters, including: All the initial deployment parameters generated by the generative adversarial network model are input into the discriminator of the generative adversarial network model to obtain the generation quality score of each initial deployment parameter, and the initial deployment parameters with generation quality scores lower than a predetermined quality threshold are eliminated. The remaining preliminary layout parameters after elimination are subjected to diversity enhancement processing to ensure that their distribution in the solution space meets the diversity requirements. The preliminary layout parameters after diversity enhancement processing are used together as the initial population of the multi-objective optimization algorithm.

7. A system for optimizing and analyzing the layout of cadmium telluride glass in a plant factory, characterized in that, include: The receiving unit is configured to: determine the geometric parameters of the enclosure structure of the plant factory, the expected plant planting data, and the first layout parameters of the cadmium telluride glass; wherein the first layout parameters include at least the glass transmittance, the glass width, the gap width between adjacent glass panes, and the glass laying direction. The equivalent processing unit is configured to: generate several potential cultivation surface contour data based on the expected plant planting data, and determine the equivalent cultivation surface based on each potential cultivation surface contour data; The light evaluation unit is configured to: calculate the spatial distribution information of the daily cumulative light intensity formed by the cadmium telluride glass and gaps on the equivalent cultivation surface during a representative time period based on the geometric parameters of the enclosure structure, the first layout parameters and the solar motion trajectory; and calculate the evaluation index characterizing the uniformity of light received by the crop canopy based on the spatial distribution information. The optimization processing unit is configured to: use an optimization algorithm to adjust the first deployment parameters to obtain a second deployment parameter in which both the total power generation and the evaluation index reach a predetermined balance target; wherein the total power generation is calculated based on the first deployment parameters and standard solar radiation data, and corresponds to the representative time period.

8. The system for optimizing the layout of cadmium telluride glass in a plant factory according to claim 7, characterized in that: The receiving unit is configured as follows: Obtain the BIM model of the plant factory and identify the attribute information of the components in each area related to the cadmium telluride glass layout; Based on the attribute information, extract the geometric parameters of the enclosure structure corresponding to the regional components, including at least the roof slope, roof azimuth, roof length, and roof width.

9. The system for optimizing the layout of cadmium telluride glass in a plant factory according to claim 7, characterized in that: The equivalent processing unit is configured as follows: The type, spatial location, and outline dimensions of each cultivation area are derived from the expected plant planting data. For each cultivation area, corresponding potential cultivation surface contour data are generated based on its type and outline size; wherein, the types include planar cultivation areas, three-dimensional rack cultivation areas, and bag-type three-dimensional cultivation areas. Based on all potential cultivation surface contour data and their spatial locations, the equivalent cultivation surface is calculated using a weighted average or union algorithm.

10. The system for optimizing the layout of cadmium telluride glass in a plant factory according to claim 9, characterized in that: The optimization processing unit is configured as follows: The first deployment parameters are input into the pre-trained generative adversarial network model, and the generator in the generative adversarial network model outputs preliminary deployment parameters that meet the constraints. Using the initial deployment parameters as the initial population, a multi-objective optimization algorithm is used for iterative optimization. In each iteration, the total power generation and the evaluation index are calculated based on the current deployment parameters. Non-dominated sorting and selection are performed with the common optimization objectives of maximizing the total power generation and optimizing the evaluation index, until the iteration termination condition is reached, and the Pareto optimal solution set is output. From the Pareto optimal solution set, select the solution that simultaneously satisfies that the total power generation is not lower than the predetermined power generation threshold and that the evaluation index is better than the predetermined uniformity threshold, and use it as the second layout parameter.

Images