Flexible transparent photovoltaic and LED display screen integration display and AI power supply method

Abstract

This invention relates to the field of photovoltaic and display integration technology, and particularly to a method for integrating flexible transparent photovoltaic and LED displays and providing AI power supply. This method collects ambient light and display load information, decomposes the light spectrum to obtain effective spectral components, and determines the power generation of each photovoltaic unit based on the spatial distribution of the photovoltaic system. It then performs pixel-level analysis to calculate the transmittance influence coefficient of the displayed content. Based on the power generation and influence coefficient, a neural network is used to generate a spatial mapping relationship between pixels and photovoltaic units, and pixel positions are reallocated accordingly to achieve a point-to-point power supply topology. Based on the power of each unit in the topology, energy storage scheduling and external power compensation commands are generated, and the compensation effect drives the neural network update. This method achieves adaptive dynamic power supply optimization, improving energy utilization efficiency.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic and display integration technology, and in particular to a flexible transparent photovoltaic and LED display integrated display and AI power supply method. Background Technology

[0002] In existing flexible transparent photovoltaic and LED display integration technologies, the photovoltaic layer is typically laminated to the display layer in the form of a transparent or semi-transparent thin film. The photovoltaic units are distributed in an array, while the display pixels are distributed on the surface of the display screen. The conventional approach is to use the photovoltaic layer as an auxiliary energy source, providing power to the display module through a fixed circuit connection, or to utilize energy storage units for energy buffering to balance the supply and demand difference between photovoltaic power generation and display load. The power generation of the photovoltaic units depends on the ambient light intensity and its spectral transmittance in the area covered by the display layer, while the display load dynamically changes according to the display content requirements.

[0003] This fixed connection method has significant drawbacks. There is a lack of dynamic correlation between the photovoltaic unit and the display pixels. When ambient light distribution is uneven or the displayed content has localized high brightness, the actual output power of the photovoltaic unit cannot effectively match the real-time power consumption of the corresponding pixel, resulting in low energy transfer efficiency. This necessitates additional energy storage or external power compensation, increasing system complexity and cost. Furthermore, the current approach cannot optimize photovoltaic power generation efficiency based on the characteristics of the light spectrum; for example, ineffective spectral components in ambient light are not fully utilized, leading to energy waste.

[0004] Another drawback is that the impact of pixel brightness and color distribution of the displayed content on the transmittance of the photovoltaic layer is not fully considered. Pixel occlusion reduces the light absorption of the underlying photovoltaic units, and existing methods typically use a uniform transmittance assumption, ignoring the influence of pixel-level differences on photovoltaic power generation performance. This leads to deviations in the power estimation of local photovoltaic units, further weakening power supply stability. Such static configurations are difficult to adapt to complex dynamic scenarios, limiting the practical application of integrated systems. Summary of the Invention

[0005] The embodiments of the present invention provide a method for integrating flexible transparent photovoltaic and LED displays and AI power supply, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for integrating flexible transparent photovoltaic and LED displays and powering AI, comprising: Ambient light information and display load demand information are collected, and the ambient light information is spectrally decomposed to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined. The display content in the display load demand information is parsed at the pixel level, the brightness distribution characteristics and color distribution characteristics of each pixel area are extracted, and the influence coefficient of each pixel area on the transmittance of the transparent photovoltaic layer is calculated based on the brightness distribution characteristics and the color distribution characteristics. Based on the transmittance influence coefficient and the available power generation value, a spatial mapping relationship between pixel regions and photovoltaic units is generated through a neural network model, and the pixel positions of the display content are redistributed according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units. Based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology, energy scheduling instructions for the energy storage unit and compensation power supply instructions for the external power supply are generated. The execution effect of the compensation power supply instructions is used as the basis for updating the neural network model, driving the dynamic evolution of the spatial mapping relationship.

[0007] The ambient light information is spectrally decomposed to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined, including: Multi-scale spectral decomposition of ambient light information is performed to obtain multiple sets of spectral components covering different wavelength ranges. The matching degree between each set of spectral components and the photoelectric conversion response characteristic curve of the flexible transparent photovoltaic layer is evaluated. Based on the matching degree evaluation results, spectral components with matching degree higher than a preset benchmark are selected from the multiple sets of spectral components to form effective spectral components. Based on the spatial propagation path and energy attenuation law of the effective spectral components on the surface of the flexible transparent photovoltaic layer, an energy distribution model of the effective spectral components to each photovoltaic unit is established. The effective spectral energy value received by each photovoltaic unit is calculated based on the energy distribution model, and combined with the photoelectric conversion efficiency parameters of each photovoltaic unit, the effective spectral energy value is converted into the initial power generation value of each photovoltaic unit through the energy-power conversion relationship. The initial power generation value is corrected by power coupling across photovoltaic units. The power redistribution process among photovoltaic units is described by constructing a power transfer matrix between photovoltaic units. After the power coupling correction, the available power generation value of each photovoltaic unit is obtained.

[0008] The display content in the display load demand information is analyzed at the pixel level to extract the brightness distribution characteristics and color distribution characteristics of each pixel region. Based on these characteristics, the influence coefficient of each pixel region on the transmittance of the transparent photovoltaic layer is calculated, including: The display content in the display load requirement information is parsed according to pixel units to obtain the position of each pixel area in the display screen and the display state sequence of the pixel area; Based on the display state sequence, a coupled description structure of the brightness of the pixel region in the time evolution dimension and the spatial distribution dimension is constructed. According to the coupled description structure, the brightness change trajectory in the time evolution dimension and the brightness distribution pattern in the spatial distribution dimension are associated and characterized to form brightness distribution features. For the display state sequence of each pixel region, the color component composition and spectral response distribution of each color component of the pixel region are extracted, and the color component composition and the spectral response distribution are associated to form color distribution features; Based on the brightness distribution characteristics and the color distribution characteristics, a dual-path evaluation mechanism is established to assess the influence of pixel regions on the light transmittance of the transparent photovoltaic layer. The dual-path evaluation mechanism is used to quantify the reduction effect of brightness modulation on the total incident light energy and the effect of color filtering on the change of incident spectral structure, respectively. The quantification results of the two paths are then fused to obtain the influence coefficient of each pixel region on the light transmittance of the transparent photovoltaic layer.

[0009] Based on the display state sequence, a coupled description structure of the brightness of the pixel region in the temporal evolution dimension and the spatial distribution dimension is constructed. According to the coupled description structure, the brightness change trajectory in the temporal evolution dimension and the brightness distribution pattern in the spatial distribution dimension are correlated and characterized to form brightness distribution features including: The brightness state of a pixel region at each moment in a continuous display cycle is identified from the display state sequence, and a brightness change trajectory in the time evolution dimension is constructed. At the same time, the brightness state of the pixel region at each spatial position in a single moment is identified, and a brightness distribution pattern in the spatial distribution dimension is constructed. A bidirectional dependency relationship is established between the brightness change trajectory and the brightness distribution pattern. Based on the bidirectional dependency relationship, the brightness evolution state at each moment in the brightness change trajectory is set as the spatial distribution constraint condition of the brightness distribution pattern at the corresponding moment. At the same time, the brightness distribution characteristics at each spatial position in the brightness distribution pattern are set as the temporal evolution constraint condition of the brightness change trajectory at the corresponding spatial position. The temporal evolution dimension and the spatial distribution dimension are mutually locked through the bidirectional constraint relationship to form a coupled description structure. Based on the interlocking relationship in the coupling description structure, the temporal evolution information represented by the brightness change trajectory and the spatial structure information represented by the brightness distribution pattern are jointly encoded. The dynamic change law of brightness in the temporal evolution information and the spatial differentiation law of brightness in the spatial structure information are integrated into a spatiotemporal coupling expression. The brightness distribution characteristics of the pixel region are constituted based on the spatiotemporal coupling expression.

[0010] Based on the transmittance influence coefficient and the available power generation value, a spatial mapping relationship between pixel regions and photovoltaic units is generated through a neural network model. Then, pixel positions are reallocated based on this spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units, including: The transmittance influence coefficient and the available power generation value are input into the neural network model. Based on the matching relationship between the transmittance influence coefficient and the available power generation value, an association mapping mechanism between the spatial layout of the pixel region and the spatial layout of the photovoltaic unit is constructed. Based on the association mapping mechanism, a spatial mapping relationship between the pixel region and the photovoltaic unit is formed. Based on the spatial location information of each pixel region that should be rearranged in the spatial mapping relationship, each pixel region is adjusted from its original spatial location to the target spatial location specified by the spatial mapping relationship, thereby completing the pixel position redistribution of the displayed content; Based on the proximity relationship between the target spatial location of each pixel region and the spatial location of each photovoltaic unit after the pixel position redistribution, a one-to-one power supply pairing relationship is established between each pixel region and each photovoltaic unit, and a point-to-point power supply topology between pixels and photovoltaic units is formed based on the one-to-one power supply pairing relationship.

[0011] The transmittance influence coefficient and the available power generation value are input into the neural network model. Based on the matching relationship between the transmittance influence coefficient and the available power generation value, a correlation mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed, including: The transmittance influence coefficient corresponding to each pixel region and the available power generation value corresponding to each photovoltaic unit are input into the neural network model. The power generation constraint intensity information of the transmittance influence coefficient and the power supply capacity intensity information of the available power generation value are received according to the neural network model. Based on the numerical distribution characteristics between the transmittance influence coefficient and the available power generation value, the matching metric value between the power generation constraint intensity of each pixel region and the power supply capacity intensity of each photovoltaic unit is calculated through the neural network model. Based on the matching metric value, the spatial positional correspondence between the pixel region that makes the matching metric value reach the extreme value and the photovoltaic unit is identified. Based on the spatial location information of the photovoltaic unit that each pixel region should correspond to in the spatial location correspondence, an association mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed.

[0012] Based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology, the energy scheduling command for the energy storage unit and the compensation power supply command for the external power source are generated, including: The real-time output power of each photovoltaic unit and the real-time power consumption of each pixel are obtained from the point-to-point power supply topology. Based on the difference between the real-time output power and the real-time power consumption, the power balance deviation of each power supply channel is calculated. The power supply status of each power supply channel is identified based on the positive or negative attribute of the power balance deviation. Power supply channels with positive power balance deviation are assigned to the power surplus channel set, and power supply channels with negative power balance deviation are assigned to the power deficit channel set. At the same time, the surplus power amplitude of each power supply channel in the power surplus channel set and the deficit power amplitude of each power supply channel in the power deficit channel set are extracted based on the magnitude of the power balance deviation. The total surplus power capacity is calculated based on the surplus power amplitude of each power supply channel in the power surplus channel set, and the total deficit power demand is calculated based on the deficit power amplitude of each power supply channel in the power deficit channel set. The total surplus power capacity, the total power deficit demand, and the releaseable energy status of the energy storage unit are jointly evaluated. Based on the results of the joint evaluation, an energy dispatch command for the energy storage unit is generated. When the sum of the total surplus power capacity and the releaseable energy status can cover the total power deficit demand, an energy release dispatch command for the energy storage unit is generated. When the sum of the total surplus power capacity and the releaseable energy status is insufficient to cover the total power deficit demand, the energy release dispatch command and a compensation power supply command from an external power source are generated.

[0013] A second aspect of the present invention provides a flexible transparent photovoltaic and LED display integrated display and AI power supply system, comprising: The spectral power generation unit is used to collect ambient light information and display load demand information, and to perform spectral decomposition on the ambient light information to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined. The transmittance calculation unit is used to perform pixel-level analysis of the display content in the display load demand information, extract the brightness distribution characteristics and color distribution characteristics of each pixel area, and calculate the transmittance influence coefficient of each pixel area on the transparent photovoltaic layer based on the brightness distribution characteristics and the color distribution characteristics. The mapping generation unit is used to generate a spatial mapping relationship between pixel regions and photovoltaic units through a neural network model based on the transmittance influence coefficient and the available power generation value, and to redistribute pixel positions of the display content according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units. The energy scheduling unit is used to generate energy scheduling instructions for the energy storage unit and compensation power supply instructions for the external power source based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology. The execution effect of the compensation power supply instructions is used as the basis for updating the neural network model to drive the dynamic evolution of the spatial mapping relationship.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] Integrated acquisition and analysis of ambient light information and display load requirements enables precise matching of photovoltaic power generation and display functions in both time and space. Spectral decomposition technology extracts the effective photovoltaic components, and combined with spatial distribution characteristics, assesses the power generation capacity of each unit, significantly improving energy capture efficiency under low light and complex lighting conditions. Pixel-level brightness and color feature analysis and the quantification of their impact on transmittance maximize the light-receiving area of ​​the photovoltaic layer while ensuring display quality, effectively balancing the inherent contradiction between light transmission and power generation.

[0017] The neural network-based spatial mapping between pixel regions and photovoltaic units completely eliminates the mismatch between power generation and display power caused by traditional random layouts. The point-to-point power supply topology ensures that the energy consumption of each pixel directly corresponds to the output of a specific photovoltaic unit, achieving precise localized energy transfer from generation to consumption, significantly reducing transmission losses and intermediate conversion steps. The display content relocation mechanism dynamically adjusts the pixel layout based on real-time load, ensuring that high-brightness areas are preferentially powered directly by high-power photovoltaic units, significantly reducing the overall system's dependence on energy storage and external power sources.

[0018] The compensation scheduling commands between the energy storage unit and the external power supply are generated based on real-time power differences, forming a flexible power supply network with millisecond-level response. The execution effect of the compensation commands serves as the basis for the iterative update of the neural network, driving the mapping relationship to continuously and dynamically evolve with changes in illumination and load, ensuring that the system maintains an optimal matching state under all weather conditions and multiple operating modes. This self-learning and self-optimizing AI power supply method effectively suppresses power fluctuations caused by cloud cover or sudden changes in displayed content, extends the cycle life of energy storage devices, and significantly improves overall energy utilization efficiency and system operational reliability. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the method for integrating flexible transparent photovoltaic and LED displays and providing AI power supply according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the energy scheduling and compensation power supply process based on a point-to-point power supply topology in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] Figure 1 This is a flowchart illustrating the method for integrating flexible transparent photovoltaic and LED displays and providing AI power supply according to an embodiment of the present invention.

[0023] The integrated display and AI power supply methods for flexible transparent photovoltaic and LED displays include: Ambient light information and display load demand information are collected, and the ambient light information is spectrally decomposed to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined. The display content in the display load demand information is parsed at the pixel level, the brightness distribution characteristics and color distribution characteristics of each pixel area are extracted, and the influence coefficient of each pixel area on the transmittance of the transparent photovoltaic layer is calculated based on the brightness distribution characteristics and the color distribution characteristics. Based on the transmittance influence coefficient and the available power generation value, a spatial mapping relationship between pixel regions and photovoltaic units is generated through a neural network model, and the pixel positions of the display content are redistributed according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units. Based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology, energy scheduling instructions for the energy storage unit and compensation power supply instructions for the external power supply are generated. The execution effect of the compensation power supply instructions is used as the basis for updating the neural network model, driving the dynamic evolution of the spatial mapping relationship.

[0024] In one optional implementation, the ambient light information is spectrally decomposed to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined, including: Multi-scale spectral decomposition of ambient light information is performed to obtain multiple sets of spectral components covering different wavelength ranges. The matching degree between each set of spectral components and the photoelectric conversion response characteristic curve of the flexible transparent photovoltaic layer is evaluated. Based on the matching degree evaluation results, spectral components with matching degree higher than a preset benchmark are selected from the multiple sets of spectral components to form effective spectral components. Based on the spatial propagation path and energy attenuation law of the effective spectral components on the surface of the flexible transparent photovoltaic layer, an energy distribution model of the effective spectral components to each photovoltaic unit is established. The effective spectral energy value received by each photovoltaic unit is calculated based on the energy distribution model, and combined with the photoelectric conversion efficiency parameters of each photovoltaic unit, the effective spectral energy value is converted into the initial power generation value of each photovoltaic unit through the energy-power conversion relationship. The initial power generation value is corrected by power coupling across photovoltaic units. The power redistribution process among photovoltaic units is described by constructing a power transfer matrix between photovoltaic units. After the power coupling correction, the available power generation value of each photovoltaic unit is obtained.

[0025] When performing multi-scale spectral decomposition on ambient light information, a combination of continuous wavelet transform and discrete Fourier transform is used to decompose the acquired broadband light signal into multiple spectral components covering the ultraviolet, visible, and near-infrared bands. Each spectral component corresponds to a specific wavelength range. The division of wavelength ranges is determined based on the absorption characteristics of the organic photovoltaic material or perovskite thin film material used in the flexible transparent photovoltaic layer. Typically, the visible light band is subdivided into several sub-bands, and the near-infrared band is processed as a separate group to ensure that the resolution of subsequent matching evaluation is sufficiently fine.

[0026] For each spectral component, its matching degree with the photoelectric conversion response characteristic curve of the flexible transparent photovoltaic layer was evaluated. The photoelectric conversion response characteristic curve describes the external quantum efficiency distribution of the photovoltaic material at different wavelengths, and this curve is obtained through factory calibration or on-site calibration. The matching degree evaluation uses the normalized cross-correlation coefficient as a quantitative index, comparing the energy density distribution of each spectral component with the response characteristic curve point by point within the corresponding wavelength range, and calculating the matching degree coefficient for each spectral component. ,in This is the group index for the spectral components. When Higher than the preset benchmark threshold At that time, the spectral components in this group are included in the effective spectral component set, while components below the threshold are considered insufficient to contribute to photovoltaic power generation and are discarded. Preset benchmark threshold. It can be dynamically adjusted according to the lighting conditions of the actual deployment scenario. In low-light indoor environments, it can be appropriately reduced to retain more usable components, while in strong-light outdoor environments, it can be appropriately increased to filter low-quality components.

[0027] After determining the effective spectral components, an energy distribution model for these components to each photovoltaic unit is established. The flexible transparent photovoltaic layer has a certain curvature and non-uniform light transmission characteristics in its physical structure. Light undergoes refraction, scattering, and interface reflection when propagating on its surface, resulting in differences in the effective irradiance received by photovoltaic units at different locations. The energy distribution model, based on a ray tracing algorithm, establishes a coordinate mapping relationship for the geometric deformation state of the flexible substrate, mapping the irradiance distribution of the incident light field on the photovoltaic layer surface to the light-receiving area of ​​each photovoltaic unit. The energy attenuation law is described by the Beer-Lambert law, and the absorption coefficient and thickness of the flexible transparent encapsulation layer are considered to correct for energy loss along the propagation path. For regions with local curvature, a curvature correction factor is introduced. ,in For the spatial location index of photovoltaic units, It is calculated from the readings of the real-time deformation sensor and is used to correct for changes in the effective light-receiving area caused by bending.

[0028] Based on the energy allocation model, the calculations for each photovoltaic unit are performed. Received effective spectral energy value . The irradiance integral of each effective spectral component on the light-receiving area of ​​the unit and the curvature correction factor A joint decision. Upon obtaining... Then, the photoelectric conversion efficiency parameters of each photovoltaic unit are combined. Perform power conversion. This reflects the actual efficiency of the unit in converting light energy into electrical energy, which is affected by factors such as temperature, aging, and partial shading, and is updated periodically through online calibration. The energy-power conversion equation is: ,in For the first The initial power generation value of each photovoltaic unit, The sampling time window length is denoted by . This formula converts the effective spectral energy received per unit time into the corresponding electrical power output, obtaining an estimate of the initial power generation of each photovoltaic unit at the current moment.

[0029] Calculations relying solely on individual photovoltaic (PV) cells often contain errors due to electrical coupling effects between adjacent PV cells in the flexible transparent PV layer. These effects include changes in the conduction state of bypass diodes, current mismatch in series-parallel configurations, and power redistribution caused by localized hot spots. To accurately describe this cross-cell power transfer behavior, a power transfer matrix between PV cells is constructed. The first of the matrix Line 1 Column elements Indicates the first The power output of the photovoltaic unit affects the first The influence coefficient of the available power generation of each photovoltaic unit, where This is the index of the power transfer source unit. The value is obtained through the joint identification of statistical analysis of historical operating data and physical circuit model. For electrically independent units, Approaching zero; for units that are connected in series, It has a significant non-zero value.

[0030] The power coupling correction process involves adjusting the initial power generation vector. (Each element is) ) and power transfer matrix Matrix multiplication is performed to obtain the corrected available power generation vector. ,Right now ,in The Each component That is, the first The usable power generation value of each photovoltaic unit after coupling correction. The correction process fully considers the mutual influence between photovoltaic units, making the final usable power generation value closer to the actual output capacity under actual electrical operating conditions. In practical applications, the power transfer matrix... It is not fixed. When the flexible display undergoes significant deformation or the working state of the photovoltaic unit changes, the matrix needs to be re-identified to maintain the correction accuracy.

[0031] Following the complete process of multi-scale spectral decomposition, matching degree evaluation, energy allocation modeling, initial power calculation, and power coupling correction, the final usable power generation value of each photovoltaic unit is obtained. This data will serve as the foundation for generating the spatial mapping relationship between subsequent pixel regions and photovoltaic units, supporting the neural network model's dynamic optimization of the point-to-point power supply topology. The entire calculation process is updated once within each sampling period, ensuring that the power supply topology can respond promptly to changes in ambient lighting conditions and maintain stable system operation under different lighting scenarios.

[0032] In one optional implementation, the display content in the display load demand information is analyzed at the pixel level to extract the brightness distribution characteristics and color distribution characteristics of each pixel region. The influence coefficient of each pixel region on the transmittance of the transparent photovoltaic layer is calculated based on the brightness distribution characteristics and the color distribution characteristics, including: The display content in the display load requirement information is parsed according to pixel units to obtain the position of each pixel area in the display screen and the display state sequence of the pixel area; Based on the display state sequence, a coupled description structure of the brightness of the pixel region in the time evolution dimension and the spatial distribution dimension is constructed. According to the coupled description structure, the brightness change trajectory in the time evolution dimension and the brightness distribution pattern in the spatial distribution dimension are associated and characterized to form brightness distribution features. For the display state sequence of each pixel region, the color component composition and spectral response distribution of each color component of the pixel region are extracted, and the color component composition and the spectral response distribution are associated to form color distribution features; Based on the brightness distribution characteristics and the color distribution characteristics, a dual-path evaluation mechanism is established to assess the influence of pixel regions on the light transmittance of the transparent photovoltaic layer. The dual-path evaluation mechanism is used to quantify the reduction effect of brightness modulation on the total incident light energy and the effect of color filtering on the change of incident spectral structure, respectively. The quantification results of the two paths are then fused to obtain the influence coefficient of each pixel region on the light transmittance of the transparent photovoltaic layer.

[0033] When performing pixel-level analysis of display load demand information, the complete display content is first decomposed pixel by pixel according to the grid of physical pixel units. The two-dimensional position information of each pixel region in the display screen coordinate system is recorded, and the display state sequence of the pixel region in the continuous frame sequence is extracted. The display state sequence includes the brightness value, color channel value, and on / off state of the pixel at each frame time, forming a multi-dimensional data stream that changes over time. Since the light transmittance performance of the flexible transparent photovoltaic layer has real-time response characteristics to the display content, analyzing only a single static frame cannot accurately describe the dynamic changes in transmittance. Therefore, it is necessary to establish a descriptive structure in both the temporal evolution dimension and the spatial distribution dimension.

[0034] To construct the brightness distribution features, based on the extracted display state sequence, the trajectory of brightness value changes in each pixel region with the frame sequence is analyzed in the temporal evolution dimension, capturing the upward trend, downward trend, and periodic fluctuation pattern of brightness. In the spatial distribution dimension, the brightness differences and brightness gradient distribution patterns between adjacent pixel regions are statistically analyzed to identify the spatial positional relationships of high-brightness regions, low-brightness regions, and brightness transition zones. The brightness change trajectory in the temporal evolution dimension and the brightness distribution pattern in the spatial distribution dimension are correlated and characterized to form a coupled descriptive structure. This coupled descriptive structure can reflect the overall brightness evolution of a pixel at a certain spatial location over a period of time, as well as the spatial correlation with the brightness distribution of surrounding pixels, thus providing a sufficient information basis for subsequent quantization of the reduction effect of brightness modulation on the total incident light energy.

[0035] In the process of extracting color distribution features, the color component composition of each pixel region is decomposed based on its display state sequence. This typically includes three basic color components: red, green, and blue. The spectral response distribution of each color component is extracted, i.e., its energy distribution curve within the visible light spectrum, reflecting its absorption and transmission characteristics for specific wavelengths of light. Different color components exhibit significant differences in their response distribution to the incident spectrum. The red component primarily absorbs short-wavelength light while transmitting long-wavelength light, the blue component does the opposite, and the green component's response distribution lies in the intermediate wavelength range. By correlating and integrating the color component composition with the spectral response distribution of each component, a color distribution feature is formed. This feature comprehensively describes how and to what extent a pixel region alters the incident spectral structure in the color dimension.

[0036] Based on brightness and color distribution characteristics, a dual-path evaluation mechanism is established to assess the impact of pixel regions on the light transmittance of a transparent photovoltaic layer. This mechanism decomposes the transmittance effect into two independent but ultimately fused quantization paths: the brightness modulation path and the color filtering path.

[0037] In the brightness modulation path, the reduction effect of brightness modulation on the total incident light energy is quantized. Let the... The average brightness of each pixel region is The maximum brightness is Then the brightness normalization coefficient of this pixel region is Higher brightness means greater pixel luminous intensity, resulting in more significant shading and energy consumption of the incident light on the transparent photovoltaic layer by the backlight or self-emission. A time weighting factor is further introduced. The brightness modulation coefficients considering temporal dynamic effects are obtained by weighted integral of the brightness change trajectory of the pixel region in each frame within the observation time window. The calculation method is as follows ,in The brightness is dynamically adjusted based on the fluctuation frequency and amplitude of the trajectory. The higher the fluctuation frequency and the greater the amplitude, the better. The larger the value, the more it reflects the cumulative reduction effect of dynamic brightness on the incident light energy of the photovoltaic layer.

[0038] In the color filtering path, the effect of color filtering on the alteration of the incident spectral structure is quantified, and the denoted color filter is defined as follows: The pixel region in the first The spectral response distribution of each color channel is as follows: ,in This represents the wavelength of light. The spectral response distribution of each color channel selectively affects the transmittance of specific wavelength components in the incident spectrum, thereby altering the spectral structure reaching the photovoltaic layer. The amount of spectral structure change is defined. The integral of the deviation between the weighted superposition of the spectral response distributions of each color channel and the standard white light spectrum reflects the degree of distortion of the overall structure of the incident spectrum by the pixel region. The larger the value, the stronger the color filtering effect of that pixel area, the greater the degree of damage to the effective spectrum of the photovoltaic layer, and thus the more significant the negative impact on photovoltaic power generation efficiency.

[0039] In obtaining the brightness modulation coefficient With the amount of change in spectral structure Then, the quantization results of the two paths are fused to obtain the first... The influence coefficient of each pixel region on the transmittance of the transparent photovoltaic layer The fusion method employs a weighted linear combination, i.e. ,in and These are the fusion weighting coefficients for the luminance path and the color path, respectively, with a sum of 1. These are preset or adaptively adjusted based on the relative importance of brightness variations and color filtering on photovoltaic efficiency in actual application scenarios. When the displayed content primarily consists of a large area of ​​high brightness... The value is relatively large; when the displayed content is mainly composed of areas with high saturation and multiple color combinations. The value is relatively large.

[0040] Transmittance Influence Coefficient The physical meaning of is: the first The degree of overall reduction in the actual incident light energy that the transparent photovoltaic layer can receive, after comprehensively considering the dual effects of brightness modulation and color filtering in each pixel area. The larger the value, the stronger the interference of the pixel region on the light transmittance of the photovoltaic layer, and the more significant the loss of available power generation of the corresponding photovoltaic unit. This coefficient serves as a key input in the subsequent generation of spatial mapping relationships and the construction of point-to-point power supply topology, guiding the neural network model to optimize the pixel position redistribution strategy. This ensures that pixel regions with high light transmittance influence coefficients are relocated to positions with lower contribution to photovoltaic power generation, thereby maximizing the overall power generation efficiency of the photovoltaic layer while ensuring display quality.

[0041] In actual operation, the display state sequence of each pixel area changes continuously with the real-time update of the displayed content. The brightness distribution characteristics and color distribution characteristics need to be dynamically refreshed according to a fixed frame rate cycle to ensure that the transmittance influence coefficient always reflects the real display state at the current moment, providing an accurate input basis for the dynamic evolution of the subsequent power supply topology.

[0042] In one optional implementation, based on the display state sequence, a coupled description structure of the brightness of the pixel region in the temporal evolution dimension and the spatial distribution dimension is constructed. The brightness change trajectory in the temporal evolution dimension and the brightness distribution pattern in the spatial distribution dimension are associated and characterized according to the coupled description structure to form brightness distribution features, including: The brightness state of a pixel region at each moment in a continuous display cycle is identified from the display state sequence, and a brightness change trajectory in the time evolution dimension is constructed. At the same time, the brightness state of the pixel region at each spatial position in a single moment is identified, and a brightness distribution pattern in the spatial distribution dimension is constructed. A bidirectional dependency relationship is established between the brightness change trajectory and the brightness distribution pattern. Based on the bidirectional dependency relationship, the brightness evolution state at each moment in the brightness change trajectory is set as the spatial distribution constraint condition of the brightness distribution pattern at the corresponding moment. At the same time, the brightness distribution characteristics at each spatial position in the brightness distribution pattern are set as the temporal evolution constraint condition of the brightness change trajectory at the corresponding spatial position. The temporal evolution dimension and the spatial distribution dimension are mutually locked through the bidirectional constraint relationship to form a coupled description structure. Based on the interlocking relationship in the coupling description structure, the temporal evolution information represented by the brightness change trajectory and the spatial structure information represented by the brightness distribution pattern are jointly encoded. The dynamic change law of brightness in the temporal evolution information and the spatial differentiation law of brightness in the spatial structure information are integrated into a spatiotemporal coupling expression. The brightness distribution characteristics of the pixel region are constituted based on the spatiotemporal coupling expression.

[0043] After obtaining the display state sequence, a multi-dimensional structured description of the brightness characteristics of the pixel region is required. The display state sequence records snapshots of the brightness state of the pixel region at various moments within several consecutive display cycles. These snapshots contain both dynamic change information in the temporal dimension and implicit distribution pattern information in the spatial dimension. To unify these two types of information into the same descriptive framework, a coupled descriptive structure capable of simultaneously carrying the temporal evolution dimension and the spatial distribution dimension needs to be constructed, so that the temporal change trajectory of brightness and the spatial distribution pattern can be associated and represented in a unified expression system.

[0044] When extracting the brightness state of a pixel region within each display cycle from the display state sequence, the time axis is used as the main axis. The brightness value of each pixel region is read moment by moment, and this sequence of brightness values ​​arranged in chronological order is defined as the brightness change trajectory of that pixel region. Let the total number of pixel regions be... The total number of display cycles is Then the first The brightness change trajectory of a pixel region can be represented as a set of time-indexed... ( The set of brightness states arranged in order ,in Indicates the first The pixel region in the first The brightness value at each display cycle. The brightness change trajectory depicts the dynamic evolution of brightness in a single pixel area over time, such as periodic fluctuations, gradual trends, or abrupt jumps in brightness, and other time-domain behaviors.

[0045] Simultaneously, using spatial coordinates as the main axis, the brightness values ​​of all pixel areas within each display cycle are collected horizontally, and the set of brightness values ​​at each spatial location at the same moment is defined as the brightness distribution pattern. The brightness distribution pattern corresponding to each display cycle can be represented as follows: That is, at a fixed time Below is a snapshot of the distribution of brightness values ​​of all pixel regions in a spatial coordinate system. The brightness distribution pattern describes the spatial pattern of high and low brightness regions in the pixel array at a specific moment, such as the spatial differentiation characteristics of high brightness in the central region and low brightness in the edge region, or the directional distribution law of brightness gradient.

[0046] The trajectory of brightness change and the shape of brightness distribution are not independent of each other; they share the same set of brightness values. The only difference is the organizational dimension. The core of establishing a bidirectional dependency lies in: [the following is a separate point:] each moment in the brightness change trajectory... The brightness evolution state serves as a global constraint on the brightness distribution pattern in the spatial dimension at that moment. Specifically, if the brightness change trajectory is in If the overall brightness shows an upward trend, then the brightness values ​​at each spatial location within the corresponding brightness distribution pattern should be generally high, and the interpretation of the spatial distribution pattern must be carried out within this temporal constraint. Conversely, the brightness distribution characteristics at each spatial location within the brightness distribution pattern also serve as a temporal evolution constraint on the brightness change trajectory at that corresponding spatial location. For example, if a spatial location remains in a high-brightness region for a long period within the brightness distribution pattern, then the brightness change trajectory corresponding to that location should exhibit a sustained high-brightness evolution characteristic on the time axis, rather than random fluctuations. Through this bidirectional constraint, the temporal evolution dimension and the spatial distribution dimension are mutually locked in the descriptive structure; the brightness state of any dimension is constrained and regulated by the other dimension, and the two together constitute a coupled descriptive structure.

[0047] Formal representations of coupled descriptive structures can be achieved using two-dimensional matrices. To carry, where the row index of the matrix corresponds to the spatial location number of the pixel region. The column index corresponds to the time number of the display period. Matrix elements This corresponds to the brightness values. Reading along the matrix columns yields the spatial distribution pattern sequence; reading along the matrix rows yields the temporal variation trajectory sequence. The bidirectional dependency in the matrix structure manifests as a statistical correlation constraint between rows and columns; that is, there is a structural relationship between the value patterns of a certain row and the value patterns of a certain column, rather than random independence. This matrix-form coupled description structure provides a unified data carrier for subsequent joint encoding.

[0048] In the joint encoding stage, temporal evolution information and spatial structure information are integrated into a spatiotemporally coupled representation. Temporal evolution information extracts the dynamic change patterns in the brightness change trajectory, including the amplitude, rate of change, and periodicity of brightness changes over time. Spatial structure information extracts the spatial differentiation patterns in the brightness distribution pattern, including the spatial gradient distribution of brightness, the spatial proportion of bright and low-bright areas, and the spatial continuity of brightness distribution. Joint encoding concatenates or fuses these two types of features in the same feature vector space, so that the final brightness distribution features can reflect both the dynamic evolution of pixel brightness over time and the spatial differentiation pattern of brightness in the pixel array.

[0049] Suppose that the time evolution information is encoded to obtain a feature vector. , dimension Spatial structure information is encoded to obtain feature vectors. , dimension The spatiotemporal coupling representation obtains a joint feature vector by weighted fusion of the two. A spatiotemporal coupling weight matrix is ​​introduced during the fusion process. (dimension is) ,in (For the output feature dimension), the concatenated vector is mapped to a unified spatiotemporally coupled feature space to obtain... The joint eigenvector This refers to the brightness distribution characteristics that constitute the pixel area, which are used for subsequent calculation of the transmittance influence coefficient and generation of spatial mapping relationships.

[0050] In practical processing, displayed content often contains a large number of pixel regions and has a large number of display cycles. Directly constructing a coupling description structure from the entire dataset would result in high computational overhead. Therefore, pixel regions can be divided into blocks. Spatially adjacent pixel regions with similar brightness distribution patterns are grouped into superpixel blocks. The coupling description structure is then constructed using superpixel blocks as units, effectively reducing computational complexity while preserving the main spatiotemporal coupling characteristics. The division of superpixel blocks is based on two criteria: spatial proximity and brightness similarity. This ensures that the pixel regions within a block have high consistency in both the temporal evolution dimension and the spatial distribution dimension, enabling the coupling description structure to accurately reflect the overall brightness behavior characteristics of the pixels within the block.

[0051] Through the above process, the final brightness distribution feature is expressed in the form of spatiotemporal coupling, which integrates the temporal dynamic law and spatial differentiation law of pixel region brightness into a unified feature description, providing sufficient information support for the subsequent accurate calculation of the transmittance influence coefficient and the reasonable construction of point-to-point power supply topology.

[0052] In one optional implementation, based on the transmittance influence coefficient and the available power generation value, a spatial mapping relationship between pixel regions and photovoltaic units is generated through a neural network model, and pixel positions are reallocated according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units, including: The transmittance influence coefficient and the available power generation value are input into the neural network model. Based on the matching relationship between the transmittance influence coefficient and the available power generation value, an association mapping mechanism between the spatial layout of the pixel region and the spatial layout of the photovoltaic unit is constructed. Based on the association mapping mechanism, a spatial mapping relationship between the pixel region and the photovoltaic unit is formed. Based on the spatial location information of each pixel region that should be rearranged in the spatial mapping relationship, each pixel region is adjusted from its original spatial location to the target spatial location specified by the spatial mapping relationship, thereby completing the pixel position redistribution of the displayed content; Based on the proximity relationship between the target spatial location of each pixel region and the spatial location of each photovoltaic unit after the pixel position redistribution, a one-to-one power supply pairing relationship is established between each pixel region and each photovoltaic unit, and a point-to-point power supply topology between pixels and photovoltaic units is formed based on the one-to-one power supply pairing relationship.

[0053] Before inputting the transmittance influence coefficient and available power generation value into the neural network model, both types of data need to undergo unified normalization preprocessing to eliminate the interference of dimensional differences on the mapping learning process. Transmittance influence coefficient This describes the degree to which each pixel region affects the light transmittance of the underlying photovoltaic layer, and can be expressed as a power generation value. This reflects the actual usable output capacity of each photovoltaic unit under the current illumination conditions. The two are concatenated into a joint input feature vector, which is then fed into a neural network model with a multi-layer fully connected structure. The core task of this neural network model is to learn the inherent correlation between the spatial layout of pixel regions and the spatial layout of photovoltaic units; that is, on a two-dimensional plane, which pixel regions have the optimal matching relationship with the power generation capacity of which photovoltaic units.

[0054] When processing the aforementioned joint input, the neural network model models the cross-layer association between pixel regions and photovoltaic units through an attention mechanism. Specifically, for each pixel region, the model calculates the similarity score between its transmittance influence coefficient and the available power generation value of all photovoltaic units, and uses the softmax-normalized score as the association weight to form a soft association distribution from pixel regions to photovoltaic units. Based on this, by introducing positional encoding information, the two-dimensional coordinates of the pixel region and the two-dimensional coordinates of the photovoltaic unit are simultaneously embedded in the feature space. This allows the model to not only perceive the numerical matching relationship between power and transmittance but also capture their geometric proximity in spatial layout. These mechanisms collectively constitute the core of the association mapping mechanism. The final output association mapping mechanism is expressed in matrix form, where each row corresponds to a pixel region, each column corresponds to a photovoltaic unit, and the matrix elements represent the association strength between the pixel region and the photovoltaic unit.

[0055] After the association strength matrix is ​​hard-assigned, the spatial mapping relationship between pixel regions and photovoltaic (PV) units can be obtained. The hard-assignment strategy employs the Hungarian algorithm or a greedy optimal matching algorithm, ensuring that each PV unit is exclusively occupied by at most one pixel region, and determining the pixel-PV pairing with the highest association strength one by one. For cases where the number of PV units and the number of pixel regions are not completely consistent, some PV units are allowed to be unpaired, or some pixel regions are allowed to share the power supply of the same PV unit, but during the topology construction phase, priority is given to matching high-brightness, high-power pixel regions with high-power-generating PV units. The spatial mapping relationship is ultimately stored in the form of an indexed mapping table, recording the target PV unit number assigned to each pixel region and its corresponding target spatial coordinates.

[0056] Based on the spatial location information of each pixel region in the spatial mapping relationship, each pixel region is adjusted from its original spatial location to the target spatial location specified by the spatial mapping relationship. The pixel position reassignment process is divided into two stages: coordinate transformation and rendering remapping. In the coordinate transformation stage, according to the original coordinates and target coordinates recorded in the index mapping table, each frame of the displayed content is spatially rearranged at the pixel level. Let the first... The original center coordinates of the pixel region are The target center coordinates are Then all pixels within the pixel region are translated by the same vector. The entire pixel area is shifted to the target location. During the rendering remapping process, blank areas created by pixel relocation are filled using content-aware fill or background color fill strategies to maintain the visual continuity of the overall display.

[0057] After pixel location redistribution, a clear geometric correspondence exists between the target spatial location of each pixel region and the spatial location of each photovoltaic unit. Based on the proximity relationship between the target spatial location and the photovoltaic unit spatial location, a one-to-one power supply pairing relationship is established between each pixel region and each photovoltaic unit. The proximity relationship is determined based on the Euclidean distance criterion: Let the first... The target center coordinates of the pixel region are , No. The center coordinates of each photovoltaic unit are The spatial distance between the two satisfy Among all candidate photovoltaic units, the following were selected: The smallest photovoltaic unit as the first Each pixel area has a dedicated power supply pairing unit, thus forming a one-to-one power supply pairing relationship.

[0058] Based on a one-to-one power supply pairing relationship, a point-to-point power supply topology is constructed. In this topology, each pixel region forms an independent power supply path with its paired photovoltaic unit. This path achieves directional power transmission and real-time adjustment through a miniature power management unit in a flexible circuit. The point-to-point power supply topology is expressed as a directed graph, where each directed edge points from the photovoltaic unit to its paired pixel region, and the weight of the edge corresponds to the rated transmission power limit of the power supply path. When the real-time power consumption of a pixel region exceeds the real-time output power of its paired photovoltaic unit, the reserved energy storage compensation path or external power compensation path in the topology will be automatically activated to maintain the normal display of that pixel region.

[0059] The dynamic maintenance of point-to-point power supply topology relies on the continuous updating of spatial mapping relationships. When the displayed content changes significantly, causing a significant shift in the distribution of the transmittance influence coefficient, or when changes in lighting conditions cause a major adjustment in the distribution of available power generation values, the neural network model will re-execute the inference process of the association mapping mechanism, outputting updated spatial mapping relationships, thereby triggering the synchronous refresh of pixel position reallocation and one-to-one power supply pairing relationships. The triggering conditions for topology updates are set as follows: when the mean square error between the transmittance influence coefficient vector and the historical value at the last triggering time exceeds a preset threshold, or when the change in the available power generation vector exceeds a set proportion of the rated power, a complete topology reconstruction process is initiated; otherwise, only the local pairing relationships are incrementally adjusted to reduce computational overhead and minimize display screen jitter.

[0060] In practical applications of integrated flexible transparent photovoltaic and LED displays, the bending deformation of the flexible substrate causes dynamic shifts in the spatial position of the photovoltaic units, thus affecting the coordinates of the photovoltaic units. Instead of fixed values, these values ​​are measured and dynamically updated in real time by deformation sensors. Each time the topology is built or updated, the latest coordinate values ​​fed back by the sensors are used as input for determining proximity relationships, ensuring that the point-to-point power supply topology always remains consistent with the actual spatial layout of the physical layer, thereby guaranteeing the accuracy and reliability of the power supply path.

[0061] In one optional implementation, the transmittance influence coefficient and the available power generation value are input into a neural network model. Based on the matching relationship between the transmittance influence coefficient and the available power generation value, a correlation mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed, including: The transmittance influence coefficient corresponding to each pixel region and the available power generation value corresponding to each photovoltaic unit are input into the neural network model. The power generation constraint intensity information of the transmittance influence coefficient and the power supply capacity intensity information of the available power generation value are received according to the neural network model. Based on the numerical distribution characteristics between the transmittance influence coefficient and the available power generation value, the matching metric value between the power generation constraint intensity of each pixel region and the power supply capacity intensity of each photovoltaic unit is calculated through the neural network model. Based on the matching metric value, the spatial positional correspondence between the pixel region that makes the matching metric value reach the extreme value and the photovoltaic unit is identified. Based on the spatial location information of the photovoltaic unit that each pixel region should correspond to in the spatial location correspondence, an association mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed.

[0062] like Figure 2 As shown, the method includes: When inputting the transmittance influence coefficient of each pixel region and the available power generation value of each photovoltaic unit into the neural network model, a unified preprocessing operation is required for these two types of input data. The transmittance influence coefficient reflects the degree to which a pixel region inhibits the light transmission capability of the photovoltaic layer below it; a higher value indicates a stronger constraint on the power generation of the photovoltaic unit by the pixel region under display operation. The available power generation value reflects the maximum power supply capacity that each photovoltaic unit can provide under the current illumination conditions. After normalizing these two types of data to the same dimension range, the neural network model can jointly process them within a unified numerical space, avoiding uneven gradient propagation due to differences in dimensions.

[0063] In the input layer design of the neural network model, the transmittance influence coefficient is formed as a vector on a pixel-by-pixel basis, where each component corresponds to the power generation constraint intensity information of a pixel-by-pixel region. The available power generation value is formed as another vector on a photovoltaic unit basis, where each component corresponds to the power supply capacity intensity information of a photovoltaic unit. These two vectors are concatenated or cross-encoded at the input layer to form a joint input representation. The neural network model extracts features from the joint input representation through a multi-layer fully connected structure or attention mechanism, gradually refining the implicit correlation pattern between the power generation constraint intensity of pixel regions and the power supply capacity intensity of photovoltaic units.

[0064] Based on the numerical distribution characteristics between the transmittance influence coefficient and the available power generation value, the neural network model calculates the matching metric between the power generation constraint intensity of each pixel region and the power supply capacity intensity of each photovoltaic unit in the intermediate layer. Specifically, for the first... The pixel region and the first Each photovoltaic unit defines a matching metric. Used to describe the degree of compatibility between the two. The calculation comprehensively considers the suppressive effect of transmittance on power generation and the power supply margin of the photovoltaic unit itself. When the power generation constraint intensity of the pixel area and the power supply capacity intensity of the photovoltaic unit show a complementary matching relationship in numerical terms, It tends towards a maximum value; when there is a severe power mismatch between the two. It tends towards a minimum. By learning matching patterns from a large amount of historical data, neural network models can quickly output a global matching metric matrix under new input conditions. ,in The Line 1 The column element is .

[0065] Obtaining the global matching metric matrix Subsequently, identifying the spatial correspondence between the pixel regions that cause the matching metric to reach an extreme value and the photovoltaic units is the core step in constructing the association mapping mechanism. Perform an extreme value search operation on each row to find the extreme value that makes the value of the extreme value search result .... Photovoltaic unit index that yields the maximum value That can determine the first The pixel region should be consistent with the first pixel region. A spatial correspondence is established between each photovoltaic unit. This extreme value search process is performed globally using a constrained optimization approach to ensure that each photovoltaic unit is not simultaneously associated with too many pixel regions, thus avoiding excessive fragmentation of power supply capacity. When multiple pixel regions compete for the same photovoltaic unit, the system determines the optimal allocation based on their respective... The relative sizes of the pixels are prioritized, and the pixel regions with lower priority are reassigned to the suboptimal photovoltaic units until the spatial correspondence of all pixel regions is determined.

[0066] During the constraint optimization process, spatial proximity constraints also need to be introduced to prevent pixel regions from being assigned to photovoltaic units that are too far away in spatial location, thereby causing engineering problems such as excessively long power supply lines and increased transmission losses. Spatial proximity constraints are implemented through matching metrics. This is achieved by overlaying a distance penalty term, with the weight coefficient of the distance penalty term being... Used to control the strength of spatial proximity constraints. When When the value is large, the pixel area tends to establish a correspondence with the photovoltaic units in spatial proximity, resulting in a more compact overall layout; when When the value is small, the numerical optimization objective of the matching metric dominates, allowing pairing of pixel regions across larger spatial distances with photovoltaic units. In practical applications, The value is determined comprehensively based on the physical dimensions of the flexible transparent photovoltaic and LED display screen, the maximum allowable length of the power supply line, and the overall energy efficiency target of the system.

[0067] Based on the spatial location information of the photovoltaic units corresponding to each pixel region in the above spatial location correspondence, an association mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed. The association mapping mechanism is stored in the form of a lookup table. Each record in the lookup table contains the spatial coordinate information of the pixel region and its corresponding photovoltaic unit, as well as the matching metric between the two. As a confidence indicator, the correlation mapping mechanism is directly invoked in the subsequent pixel position reallocation step, driving the display content to complete the pixel arrangement adjustment according to the requirements of the point-to-point power supply topology.

[0068] After the correlation mapping mechanism is established, its effectiveness needs to be verified. The verification process evaluates the mapping quality by comparing the difference between the output power of each photovoltaic unit and the actual power consumption of each pixel area during actual operation. If the difference exceeds the preset tolerance range, the local update process of the neural network model is triggered. The matching metric value is recalculated for the pixel area and photovoltaic unit where power mismatch occurs, and the corresponding record in the lookup table is corrected. This online correction mechanism ensures that the correlation mapping mechanism can maintain a high-precision matching state with the dynamic changes of the displayed content and the real-time fluctuations of the lighting conditions, thereby providing reliable spatial layout support for the stable operation of the overall system.

[0069] The neural network model employs a combination of supervised learning and reinforcement learning during the training phase. The supervised learning component uses known, efficient mapping schemes from historical data as labels to guide the model to quickly converge to a reasonable initial parameter state. The reinforcement learning component uses a comprehensive score of overall system energy utilization efficiency and display quality as a reward signal, driving the model to continuously explore better association mapping strategies during actual operation. The synergistic use of these two learning methods enables the neural network model to possess both good initial generalization ability and the ability to continuously accumulate experience and improve mapping quality over long-term operation.

[0070] In one optional implementation, the generation of energy dispatch commands for the energy storage unit and compensation power supply commands for the external power source, based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology, includes: The real-time output power of each photovoltaic unit and the real-time power consumption of each pixel are obtained from the point-to-point power supply topology. Based on the difference between the real-time output power and the real-time power consumption, the power balance deviation of each power supply channel is calculated. The power supply status of each power supply channel is identified based on the positive or negative attribute of the power balance deviation. Power supply channels with positive power balance deviation are assigned to the power surplus channel set, and power supply channels with negative power balance deviation are assigned to the power deficit channel set. At the same time, the surplus power amplitude of each power supply channel in the power surplus channel set and the deficit power amplitude of each power supply channel in the power deficit channel set are extracted based on the magnitude of the power balance deviation. The total surplus power capacity is calculated based on the surplus power amplitude of each power supply channel in the power surplus channel set, and the total deficit power demand is calculated based on the deficit power amplitude of each power supply channel in the power deficit channel set. The total surplus power capacity, the total power deficit demand, and the releaseable energy status of the energy storage unit are jointly evaluated. Based on the results of the joint evaluation, an energy dispatch command for the energy storage unit is generated. When the sum of the total surplus power capacity and the releaseable energy status can cover the total power deficit demand, an energy release dispatch command for the energy storage unit is generated. When the sum of the total surplus power capacity and the releaseable energy status is insufficient to cover the total power deficit demand, the energy release dispatch command and a compensation power supply command from an external power source are generated.

[0071] After the point-to-point power supply topology is established, continuous monitoring and dynamic balance management of the real-time power status of each power supply channel are required. The output power of each photovoltaic unit and the power consumption of its corresponding pixel are read in real-time from the established point-to-point power supply topology. For each power supply channel, a difference calculation is performed, subtracting the real-time power consumption of its corresponding pixel from the real-time output power of the photovoltaic unit to obtain the power balance deviation of that channel. Let the... The power balance deviation of each power supply channel is The real-time output power of the photovoltaic unit is The real-time power consumption of the corresponding pixel is Then there is This differential calculation is performed in parallel on all power supply channels within each control cycle, ensuring the real-time performance and integrity of the power balance status.

[0072] Based on power balance deviation The positive and negative attributes are used to classify and group the various power supply channels. When this occurs, it indicates that the output power of the photovoltaic unit in that channel exceeds the consumption demand of the corresponding pixel, resulting in a power surplus. Such channels are then categorized into the power surplus channel set. ;when If the output power of the photovoltaic unit in that channel is insufficient to meet the consumption requirements of the corresponding pixel, indicating a power deficit, then such channels are categorized into the power deficit channel set. ;when At this time, the channel is in a power self-balancing state and requires no additional scheduling intervention. For those being incorporated... The channel's surplus power amplitude is... Its own positive value; for inclusion The channel has a power deficit amplitude of This refers to the absolute value of the deviation. This classification mechanism enables subsequent energy dispatch decisions to accurately distinguish the power supply status of different channels, avoiding energy waste caused by global unified dispatch.

[0073] After classifying the power supply channels, the power amplitude of the two types of channel sets was summarized and statistically analyzed, and the power surplus channel set was determined. The total surplus power capacity is obtained by summing the surplus power amplitudes of all channels. ,Right now Collect power supply deficiency channels The total power deficit requirement is obtained by summing the deficit power amplitudes of all channels. ,Right now Total surplus power capacity reflects the amount of excess energy that can be output to other channels in the entire photovoltaic array at the current moment, while total deficit power demand reflects the size of the energy gap that needs to be supplemented from external sources. These two aggregate values ​​are the core input parameters for subsequent joint assessment and scheduling decisions.

[0074] After obtaining the total surplus power capacity and total deficit power demand, the releaseable energy state parameters of the energy storage unit are introduced. Conduct a joint assessment. This represents the maximum usable energy converted to power that the energy storage unit can release to the outside under its current state of charge, provided that the minimum charge protection threshold constraint is met. This value is provided in real time by the energy storage unit's state of charge detection module. The core judgment logic of the joint evaluation is: calculation and The size relationship is used to determine the scheduling mode of the energy storage unit and whether external power compensation needs to be activated.

[0075] when This indicates that the sum of the current photovoltaic surplus power and the releaseable energy of the energy storage unit is sufficient to cover the power demand of all deficit channels, eliminating the need to start an external power source. At this point, an energy release scheduling command is generated for the energy storage unit, containing the amount of power the energy storage unit needs to replenish to the deficit channels. . The calculation method is as follows: after prioritizing the allocation of power from the surplus channel to the deficit channel, if there is still a remaining deficit, the energy storage unit will supplement the remaining amount. Specifically, if... In this case, the energy storage unit does not need to release energy, and the excess power from the surplus channel can be used to charge the energy storage unit, with a charging power of [missing value]. ;like and Then the energy storage unit releases power This is to make up for the remaining demand of the deficient channels. The energy release scheduling command also includes a priority sequence of power allocation, with channels that are more deficient being given priority to be supplemented, ensuring that critical areas of the display screen do not experience brightness reduction or color distortion due to insufficient power supply.

[0076] when This indicates that relying solely on the surplus photovoltaic power and the releaseable energy of the energy storage units is insufficient to meet the power demands of all deficit channels. In this case, it is necessary to simultaneously generate energy release scheduling commands for the energy storage units and compensation power supply commands from the external power source. The energy storage units will then release power according to their maximum release capacity. Full output, power required to be compensated by external power supply Determined by the following formula: The external power supply compensation command includes the target value of the compensation power. The system controls the precise intervention of external power supply by setting parameters such as the priority order of compensation channels and the compensation time window, avoiding voltage fluctuations caused by over-compensation or under-compensation. The priority order of compensation power supply commands is determined by comprehensively considering the power deficit magnitude of each deficient channel and the display importance weight of the corresponding pixel, giving priority to the compensation of pixel channels that carry high brightness or critical color areas.

[0077] The execution results of external power supply compensation commands are collected in real time and fed back to the neural network model as training signals for the dynamic evolution of spatial mapping relationships. If the execution effect of the compensation power supply commands shows that some channels are continuously in a deficit state, the neural network model will adjust the spatial mapping relationship between pixel areas and photovoltaic units accordingly, migrating high-power pixel areas to the coverage area of ​​photovoltaic units with stronger power generation capabilities. This reduces the compensation power supply demand in the next scheduling cycle, thereby continuously improving the proportion of photovoltaic self-powered power. The entire power balance management process is executed in a closed loop within each control cycle, ensuring that the flexible transparent photovoltaic and LED display integrated system maintains a highly efficient and stable power supply state under dynamic lighting conditions and changing display content.

[0078] A second aspect of the present invention provides a flexible transparent photovoltaic and LED display integrated display and AI power supply system, comprising: The spectral power generation unit is used to collect ambient light information and display load demand information, and to perform spectral decomposition on the ambient light information to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined. The transmittance meter unit is used to perform pixel-level analysis of the display content in the display load demand information, extract the brightness distribution characteristics and color distribution characteristics of each pixel area, and calculate the transmittance influence coefficient of each pixel area on the transparent photovoltaic layer based on the brightness distribution characteristics and the color distribution characteristics. The mapping generation unit is used to generate a spatial mapping relationship between pixel regions and photovoltaic units through a neural network model based on the transmittance influence coefficient and the available power generation value, and to redistribute pixel positions of the display content according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units. The energy scheduling unit is used to generate energy scheduling instructions for the energy storage unit and compensation power supply instructions for the external power source based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology. The execution effect of the compensation power supply instructions is used as the basis for updating the neural network model to drive the dynamic evolution of the spatial mapping relationship.

[0079] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0080] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0081] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0082] 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flexible transparent photovoltaic and LED display integration display and AI power supply method, characterized in that, include: Ambient light information and display load demand information are collected, and the ambient light information is spectrally decomposed to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined. The display content in the display load demand information is parsed at the pixel level, the brightness distribution characteristics and color distribution characteristics of each pixel area are extracted, and the influence coefficient of each pixel area on the transmittance of the transparent photovoltaic layer is calculated based on the brightness distribution characteristics and the color distribution characteristics. Based on the transmittance influence coefficient and the available power generation value, a spatial mapping relationship between pixel regions and photovoltaic units is generated through a neural network model, and the pixel positions of the display content are redistributed according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units. Based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology, energy scheduling instructions for the energy storage unit and compensation power supply instructions for the external power supply are generated. The execution effect of the compensation power supply instructions is used as the basis for updating the neural network model, driving the dynamic evolution of the spatial mapping relationship.

2. The method of claim 1, wherein, The ambient light information is spectrally decomposed to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined, including: Multi-scale spectral decomposition of ambient light information is performed to obtain multiple sets of spectral components covering different wavelength ranges. The matching degree between each set of spectral components and the photoelectric conversion response characteristic curve of the flexible transparent photovoltaic layer is evaluated. Based on the matching degree evaluation results, spectral components with matching degree higher than a preset benchmark are selected from the multiple sets of spectral components to form effective spectral components. Based on the spatial propagation path and energy attenuation law of the effective spectral components on the surface of the flexible transparent photovoltaic layer, an energy distribution model of the effective spectral components to each photovoltaic unit is established. The effective spectral energy value received by each photovoltaic unit is calculated based on the energy distribution model, and combined with the photoelectric conversion efficiency parameters of each photovoltaic unit, the effective spectral energy value is converted into the initial power generation value of each photovoltaic unit through the energy-power conversion relationship. The initial power generation value is corrected by power coupling across photovoltaic units. The power redistribution process among photovoltaic units is described by constructing a power transfer matrix between photovoltaic units. After the power coupling correction, the available power generation value of each photovoltaic unit is obtained.

3. The method of claim 1, wherein, The display content in the display load demand information is analyzed at the pixel level to extract the brightness distribution characteristics and color distribution characteristics of each pixel region. Based on these characteristics, the influence coefficient of each pixel region on the transmittance of the transparent photovoltaic layer is calculated, including: The display content in the display load requirement information is parsed according to pixel units to obtain the position of each pixel area in the display screen and the display state sequence of the pixel area; Based on the display state sequence, a coupled description structure of the brightness of the pixel region in the time evolution dimension and the spatial distribution dimension is constructed. According to the coupled description structure, the brightness change trajectory in the time evolution dimension and the brightness distribution pattern in the spatial distribution dimension are associated and characterized to form brightness distribution features. For the display state sequence of each pixel region, the color component composition and spectral response distribution of each color component of the pixel region are extracted, and the color component composition and the spectral response distribution are associated to form color distribution features; Based on the brightness distribution characteristics and the color distribution characteristics, a dual-path evaluation mechanism is established to assess the influence of pixel regions on the light transmittance of the transparent photovoltaic layer. The dual-path evaluation mechanism is used to quantify the reduction effect of brightness modulation on the total incident light energy and the effect of color filtering on the change of incident spectral structure, respectively. The quantification results of the two paths are then fused to obtain the influence coefficient of each pixel region on the light transmittance of the transparent photovoltaic layer.

4. The method of claim 3, wherein, Based on the display state sequence, a coupled description structure of the brightness of the pixel region in the temporal evolution dimension and the spatial distribution dimension is constructed. According to the coupled description structure, the brightness change trajectory in the temporal evolution dimension and the brightness distribution pattern in the spatial distribution dimension are correlated and characterized to form brightness distribution features including: The brightness state of a pixel region at each moment in a continuous display cycle is identified from the display state sequence, and a brightness change trajectory in the time evolution dimension is constructed. At the same time, the brightness state of the pixel region at each spatial position in a single moment is identified, and a brightness distribution pattern in the spatial distribution dimension is constructed. A bidirectional dependency relationship is established between the brightness change trajectory and the brightness distribution pattern. Based on the bidirectional dependency relationship, the brightness evolution state at each moment in the brightness change trajectory is set as the spatial distribution constraint condition of the brightness distribution pattern at the corresponding moment. At the same time, the brightness distribution characteristics at each spatial position in the brightness distribution pattern are set as the temporal evolution constraint condition of the brightness change trajectory at the corresponding spatial position. The temporal evolution dimension and the spatial distribution dimension are mutually locked through the bidirectional constraint relationship to form a coupled description structure. Based on the interlocking relationship in the coupling description structure, the temporal evolution information represented by the brightness change trajectory and the spatial structure information represented by the brightness distribution pattern are jointly encoded. The dynamic change law of brightness in the temporal evolution information and the spatial differentiation law of brightness in the spatial structure information are integrated into a spatiotemporal coupling expression. The brightness distribution characteristics of the pixel region are constituted based on the spatiotemporal coupling expression.

5. The method of claim 1, wherein, Based on the transmittance influence coefficient and the available power generation value, a spatial mapping relationship between pixel regions and photovoltaic units is generated through a neural network model. Then, pixel positions are reallocated based on this spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units, including: The transmittance influence coefficient and the available power generation value are input into the neural network model. Based on the matching relationship between the transmittance influence coefficient and the available power generation value, an association mapping mechanism between the spatial layout of the pixel region and the spatial layout of the photovoltaic unit is constructed. Based on the association mapping mechanism, a spatial mapping relationship between the pixel region and the photovoltaic unit is formed. Based on the spatial location information of each pixel region that should be rearranged in the spatial mapping relationship, each pixel region is adjusted from its original spatial location to the target spatial location specified by the spatial mapping relationship, thereby completing the pixel position redistribution of the displayed content; Based on the proximity relationship between the target spatial location of each pixel region and the spatial location of each photovoltaic unit after the pixel position redistribution, a one-to-one power supply pairing relationship is established between each pixel region and each photovoltaic unit, and a point-to-point power supply topology between pixels and photovoltaic units is formed based on the one-to-one power supply pairing relationship.

6. The method of claim 5, wherein, The transmittance influence coefficient and the available power generation value are input into the neural network model. Based on the matching relationship between the transmittance influence coefficient and the available power generation value, a correlation mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed, including: The transmittance influence coefficient corresponding to each pixel region and the available power generation value corresponding to each photovoltaic unit are input into the neural network model. The power generation constraint intensity information of the transmittance influence coefficient and the power supply capacity intensity information of the available power generation value are received according to the neural network model. Based on the numerical distribution characteristics between the transmittance influence coefficient and the available power generation value, the matching metric value between the power generation constraint intensity of each pixel region and the power supply capacity intensity of each photovoltaic unit is calculated through the neural network model. Based on the matching metric value, the spatial positional correspondence between the pixel region that makes the matching metric value reach the extreme value and the photovoltaic unit is identified. Based on the spatial location information of the photovoltaic unit that each pixel region should correspond to in the spatial location correspondence, an association mapping mechanism between the spatial layout of pixel regions and the spatial layout of photovoltaic units is constructed.

7. The method of claim 1, wherein, Based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology, the energy scheduling command for the energy storage unit and the compensation power supply command for the external power source are generated, including: The real-time output power of each photovoltaic unit and the real-time power consumption of each pixel are obtained from the point-to-point power supply topology. Based on the difference between the real-time output power and the real-time power consumption, the power balance deviation of each power supply channel is calculated. The power supply status of each power supply channel is identified based on the positive or negative attribute of the power balance deviation. Power supply channels with positive power balance deviation are assigned to the power surplus channel set, and power supply channels with negative power balance deviation are assigned to the power deficit channel set. At the same time, the surplus power amplitude of each power supply channel in the power surplus channel set and the deficit power amplitude of each power supply channel in the power deficit channel set are extracted based on the magnitude of the power balance deviation. The total surplus power capacity is calculated based on the surplus power amplitude of each power supply channel in the power surplus channel set, and the total deficit power demand is calculated based on the deficit power amplitude of each power supply channel in the power deficit channel set. The total surplus power capacity, the total power deficit demand, and the releaseable energy status of the energy storage unit are jointly evaluated. Based on the results of the joint evaluation, an energy dispatch command for the energy storage unit is generated. When the sum of the total surplus power capacity and the releaseable energy status can cover the total power deficit demand, an energy release dispatch command for the energy storage unit is generated. When the sum of the total surplus power capacity and the releaseable energy status is insufficient to cover the total power deficit demand, the energy release dispatch command and a compensation power supply command from an external power source are generated.

8. A flexible transparent photovoltaic and LED display integrated display and AI power supply system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The spectral power generation unit is used to collect ambient light information and display load demand information, and to perform spectral decomposition on the ambient light information to obtain effective spectral components that can be used for photovoltaic power generation. Based on the effective spectral components and the spatial distribution characteristics of the flexible transparent photovoltaic layer, the available power generation value of each photovoltaic unit at the current moment is determined. The transmittance calculation unit is used to perform pixel-level analysis of the display content in the display load demand information, extract the brightness distribution characteristics and color distribution characteristics of each pixel area, and calculate the transmittance influence coefficient of each pixel area on the transparent photovoltaic layer based on the brightness distribution characteristics and the color distribution characteristics. The mapping generation unit is used to generate a spatial mapping relationship between pixel regions and photovoltaic units through a neural network model based on the transmittance influence coefficient and the available power generation value, and to redistribute pixel positions of the display content according to the spatial mapping relationship to obtain a point-to-point power supply topology between pixels and photovoltaic units. The energy scheduling unit is used to generate energy scheduling instructions for the energy storage unit and compensation power supply instructions for the external power source based on the real-time output power of each photovoltaic unit and the real-time power consumption of each pixel in the point-to-point power supply topology. The execution effect of the compensation power supply instructions is used as the basis for updating the neural network model to drive the dynamic evolution of the spatial mapping relationship.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images