A photovoltaic power ultra-short-term prediction method based on cloud field dynamic perception and related device

CN122659882APending Publication Date: 2026-08-28NINGBO ORIENTAL UNIVERSITY OF TECHNOLOGY +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611143827.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-30
Publication Date
2026-08-28

AI Technical Summary

Technical Problem

[0004]本申请针对现有光伏功率超短期预测中云微物理表征不足、短时小尺度云场捕捉滞后且缺少云特性到光伏组件输出功率精细化转换机制的技术问题,提供一种基于云场动态感知的光伏功率超短期预测方法及相关装置

Benefits of technology

本申请提出一种基于云场动态感知的光伏功率超短期预测方法,通过同时获取数值天气预报模式输出的云场初始数据、静止气象卫星的红外亮温序列以及X波段天气雷达的反射率因子数据,并对上述多源数据进行融合,能够在保留数值模式大尺度背景场信息的同时,引入卫星对云顶连续演变的观测能力以及雷达对近场小尺度云体结构的精细捕捉能力,从而提高融合云场对短时、小尺度、快速发展云团的表征精度。进一步地,本申请通过物理感知融合桥接模型对修正后的云场初始数据、红外亮温序列和反射率因子数据进行融合,不仅实现不同来源云场数据的时空对齐,还通过内嵌大气热力学约束,使X波段天气雷达反射率因子所反映的局地云体结构能够转化为对数值天气预报模式云水混合比的修正,从而实现由雷达观测信息到数值模式云场初始数据的物理同化更新。由此,能够降低数值天气预报模式在0-2小时超短期时段内因初始场更新滞后导致的云场位置和强度偏差。在此基础上,本申请基于融合云场构建三维动态云运动矢量场,并结合大气热力学参数构建云微物理结构场,使云体的运动方向、运动速度、空间演变趋势以及云中液态水滴、冰晶等微物理状态能够共同参与后续计算,避免将云场仅作为普通气象因子处理而导致云遮挡过程和辐射影响刻画不足的问题。进一步地,本申请基于三维动态云运动矢量场和云微物理结构场进行辐射传输计算,能够更准确地获得云层对太阳短波辐射衰减、散射和吸收后的地表辐照度数据,从而建立云场动态变化、云微物理结构与地表辐照度之间的物理关联。同时,本申请构建的云微物理结构场包括云中液态水滴和冰晶的相态分布以及有效半径,使云层不再被简化为均匀衰减介质,而是能够基于粒子相态和粒子尺度差异表征云层对太阳短波辐射的非线性调制作用。特别是在突发强对流天气下,云顶冰晶化、粒子有效半径变化以及云体垂直发展会显著影响直射辐照度和散射辐照度,本申请通过将云微物理结构场输入辐射传输模型,能够提高地表辐照度计算对剧烈云遮挡过程的响应能力,从而减少突发强对流天气下光伏功率预测失效的风险。最后,本申请根据地表辐照度数据确定光伏电站的理论发电功率,并利用光伏电站实测数据对理论发电功率进行在线滚动校正,使预测结果能够随最新云场状态和电站实际运行状态持续修正,进而提高光伏功率超短期预测在突发云遮挡、云团快速移动及辐照度剧烈波动工况下的准确性、及时性和稳定性。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122659882A_ABST
    Figure CN122659882A_ABST
Patent Text Reader

Abstract

The application discloses a photovoltaic power ultra-short-term prediction method based on cloud field dynamic perception and related devices, and relates to the technical field of meteorology and new energy power prediction. Cloud field initial data output by a numerical weather prediction model, an infrared brightness temperature sequence of a stationary meteorological satellite, and reflectivity factor data of an X-band weather radar are acquired; the reflectivity factor data is assimilated into a cloud water mixing ratio correction amount by using a physical perception fusion bridging model, so that the cloud field initial data is corrected and a fused cloud field is obtained; a three-dimensional dynamic cloud motion vector field and a cloud microphysical structure field are constructed based on the fused cloud field; the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field are input into a radiation transmission model, and surface irradiance data is calculated; theoretical power generation is determined according to the surface irradiance data, and online rolling correction is carried out in combination with measured data of a photovoltaic power station, and an ultra-short-term photovoltaic power prediction result is output. The accuracy of the photovoltaic power ultra-short-term prediction under a sudden cloud blocking working condition can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application belongs to the field of meteorological and new energy power prediction technology, specifically involving a method and related device for ultra-short-term photovoltaic power prediction based on cloud field dynamic sensing. Background Technology

[0002] With the high proportion of new energy sources connected to the grid, the accuracy of photovoltaic power prediction directly affects grid dispatch and operational safety. This is especially true in coastal areas with frequent cloud cover, fog, and severe convective weather, where photovoltaic output is significantly affected by rapid changes in the cloud field. Changes in cloud phase, particle size, and motion affect the attenuation, scattering, and absorption processes of solar shortwave radiation, thus causing rapid fluctuations in surface irradiance and photovoltaic output power.

[0003] Current photovoltaic (PV) power forecasting typically relies on the fusion of global numerical models and regional convectively resolvable models, combined with deep learning models for weather classification forecasts. It primarily estimates PV output based on numerical weather forecast results or weather conditions. However, existing technologies often treat clouds as ordinary meteorological factors, lacking characterization of cloud microphysical structures such as cloud water, cloud ice phases, and particle size. Pure numerical models have a low update frequency of 0-2 hours, making it difficult to capture sudden small-scale convective clouds. Furthermore, current outputs are biased towards weather conditions, lacking a refined conversion mechanism from cloud characteristics to surface irradiance and then to PV module output power. Summary of the Invention

[0004] This application addresses the technical problems of insufficient cloud microphysical characterization, lagging capture of short-term small-scale cloud fields, and lack of a refined conversion mechanism from cloud characteristics to photovoltaic module output power in existing ultra-short-term photovoltaic power prediction. It provides a method and related device for ultra-short-term photovoltaic power prediction based on dynamic cloud field perception.

[0005] To achieve the above objectives, this application adopts the following technical solution: Firstly, this application proposes a method for ultra-short-term photovoltaic power prediction based on dynamic cloud field sensing, including: Acquire initial cloud field data output from numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radars; The initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data are fused to obtain a fused cloud field; A three-dimensional dynamic cloud motion vector field is constructed based on the fused cloud field, and a cloud microphysical structure field is constructed by combining atmospheric thermodynamic parameters. Radiative transfer calculations are performed based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field to obtain surface irradiance data. The theoretical power generation of the photovoltaic power station is determined based on the surface irradiance data, and the theoretical power generation is corrected online based on the measured data of the photovoltaic power station, and the ultra-short-term photovoltaic power prediction results are output.

[0006] Furthermore, the initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data are fused, including: The reflectivity factor data is converted into a correction value for the cloud-water mixing ratio, and the cloud-water mixing ratio in the initial cloud field data is assimilated based on the correction value to obtain the corrected initial cloud field data. The fused cloud field is obtained by fusing the corrected initial cloud field data, the infrared brightness temperature sequence, and the reflectivity factor data using a physical perception fusion bridging model with an embedded atmospheric thermodynamic constraint neural network. The physical perception fusion bridging model is also used to perform minute-level assimilation updates on the initial cloud field data output by the numerical weather forecast model.

[0007] Furthermore, the construction of a three-dimensional dynamic cloud motion vector field based on the fused cloud field, and the construction of a cloud microphysical structure field in combination with atmospheric thermodynamic parameters, includes: Based on the fused cloud field, the three-dimensional dynamic cloud motion vector field is constructed using optical flow or particle tracking algorithm; Based on the intensity of the vertical updraft output by the physical perception fusion bridging model, determine whether the cloud top height has broken through the freezing layer; If the cloud top height exceeds the frozen layer, the particle phase in the corresponding region of the cloud microphysical structure field will be updated from liquid water droplets to ice crystals, and the effective radius of the particles will be adjusted simultaneously. The cloud microphysical structure field includes the phase distribution and effective radius of liquid water droplets and ice crystals in the cloud.

[0008] Furthermore, radiative transfer calculations are performed based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field, including: inputting the three-dimensional dynamic cloud motion vector field, the cloud microphysical structure field, the real-time solar zenith angle, and the aerosol optical depth into the radiative transfer model to calculate the attenuation, scattering, and absorption effects of the cloud layer on solar shortwave radiation; wherein, the radiative transfer model includes the DISORT radiative transfer model or the 6S model.

[0009] Further, the theoretical power generation capacity of the photovoltaic power station is determined based on the surface irradiance data, including: The theoretical power generation is determined based on the surface irradiance data, the physical parameters of the photovoltaic power station components, and the electrical topology.

[0010] Furthermore, the measured data of the photovoltaic power station includes the measured irradiance collected by the ground-based radiometer and the measured power collected by the photovoltaic power station monitoring system.

[0011] Furthermore, the theoretical power generation is corrected online based on the measured data of the photovoltaic power station, including: correcting the surface irradiance data based on the measured irradiance, and correcting the theoretical power generation based on the measured power.

[0012] Secondly, this application proposes a photovoltaic power ultra-short-term prediction system based on cloud field dynamic sensing, comprising: The data module is used to acquire initial cloud field data output by numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radars. The fusion module is used to fuse the initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data to obtain a fused cloud field. The construction module is used to construct a three-dimensional dynamic cloud motion vector field based on the fused cloud field, and to construct a cloud microphysical structure field in combination with atmospheric thermodynamic parameters; The calculation module is used to perform radiative transfer calculations based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field to obtain surface irradiance data. The prediction module is used to determine the theoretical power generation of the photovoltaic power station based on the surface irradiance data, and to perform online rolling correction of the theoretical power generation based on the measured data of the photovoltaic power station, and output the ultra-short-term photovoltaic power prediction results.

[0013] Thirdly, this application proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for ultra-short-term photovoltaic power prediction based on dynamic cloud field sensing.

[0014] Fourthly, this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for ultra-short-term photovoltaic power prediction based on dynamic cloud field perception.

[0015] Compared with the prior art, this application has the following beneficial effects: This application proposes an ultra-short-term photovoltaic power prediction method based on dynamic cloud field perception. By simultaneously acquiring initial cloud field data from numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radar, and fusing these multi-source data, it can retain the large-scale background field information of the numerical model while introducing the satellite's ability to observe the continuous evolution of cloud tops and the radar's ability to finely capture near-field small-scale cloud structures. This improves the accuracy of the fused cloud field in representing short-term, small-scale, and rapidly developing cloud clusters. Furthermore, this application uses a physical perception fusion bridging model to fuse the corrected initial cloud field data, infrared brightness temperature sequences, and reflectivity factor data. This not only achieves spatiotemporal alignment of cloud field data from different sources but also, through embedded atmospheric thermodynamic constraints, allows the local cloud structure reflected by the X-band weather radar reflectivity factor to be transformed into a correction for the cloud-water mixing ratio of the numerical weather prediction model. This achieves physical assimilation and updating from radar observation information to the initial cloud field data of the numerical model. This reduces the deviations in cloud field position and intensity caused by the lag in initial field updates during the 0-2 hour ultra-short-term numerical weather prediction model. Based on this, this application constructs a three-dimensional dynamic cloud motion vector field by fusing cloud fields and combines atmospheric thermodynamic parameters to construct a cloud microphysical structure field. This allows the cloud's motion direction, velocity, spatial evolution trend, and microphysical states such as liquid water droplets and ice crystals within the cloud to participate in subsequent calculations, avoiding the problem of insufficient characterization of cloud obstruction processes and radiation effects due to treating the cloud field merely as a general meteorological factor. Furthermore, this application performs radiative transfer calculations based on the three-dimensional dynamic cloud motion vector field and cloud microphysical structure field, enabling more accurate acquisition of surface irradiance data after cloud attenuation, scattering, and absorption of solar shortwave radiation, thereby establishing a physical correlation between dynamic changes in the cloud field, cloud microphysical structure, and surface irradiance. Meanwhile, the cloud microphysical structure field constructed in this application includes the phase distribution and effective radius of liquid water droplets and ice crystals in the cloud, so that the cloud layer is no longer simplified as a uniform attenuation medium, but can be characterized based on the differences in particle phase and particle size to represent the nonlinear modulation effect of the cloud layer on solar shortwave radiation. Especially under sudden severe convective weather, cloud top ice crystallization, changes in particle effective radius, and vertical cloud development can significantly affect direct and diffuse irradiance. By inputting the cloud microphysical structure field into the radiative transfer model, this application can improve the response capability of surface irradiance calculation to severe cloud shading processes, thereby reducing the risk of photovoltaic power prediction failure under sudden severe convective weather. Finally, this application determines the theoretical power generation of the photovoltaic power station based on surface irradiance data, and uses measured data of the photovoltaic power station to perform online rolling correction of the theoretical power generation, so that the prediction results can be continuously corrected according to the latest cloud field state and the actual operating state of the power station, thereby improving the accuracy, timeliness, and stability of ultra-short-term photovoltaic power prediction under conditions of sudden cloud shading, rapid cloud movement, and severe irradiance fluctuations.

[0016] This application also proposes a photovoltaic power ultra-short-term prediction system based on cloud field dynamic sensing, an electronic device, and a computer-readable storage medium, which possess the advantages of the aforementioned photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing, as described in this application. Figure 2 This is another flowchart illustrating the ultra-short-term photovoltaic power prediction method based on cloud field dynamic sensing in this application. Figure 3 This is a schematic diagram of the photovoltaic power ultra-short-term prediction system architecture based on cloud field dynamic perception in the embodiments of this application; Figure 4 This is a schematic diagram illustrating the coupling between cloud microphysical parameters and the radiative transfer model in an embodiment of this application; Figure 5 This is a schematic diagram of the logic for calculating surface irradiance driven by multi-source cloud field data in an embodiment of this application; Figure 6 This is a schematic diagram of the photovoltaic power ultra-short-term prediction system based on cloud field dynamic sensing in this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0021] In the ultra-short-term power prediction process for photovoltaic power plants, the generation, movement, dissipation, and microphysical state changes of cloud fields directly affect the solar irradiance reaching the Earth's surface, thus causing rapid fluctuations in the output power of photovoltaic modules. Especially in coastal areas with frequent cloud cover and severe convective weather, the development and movement of short-term, small-scale cloud clusters are characterized by suddenness, significant spatial differences, and rapid evolution. Relying solely on numerical weather prediction models or conventional meteorological elements for prediction is insufficient to accurately and timely characterize the blocking, scattering, and absorption effects of clouds on shortwave solar radiation, and it is also difficult to establish a refined correlation between changes in cloud characteristics and the actual output power of photovoltaic power plants. Therefore, this application integrates multi-source cloud field information from numerical weather prediction models, geostationary meteorological satellites, and X-band weather radar to construct a three-dimensional dynamic cloud motion vector field and cloud microphysical structure field. Combined with radiative transfer calculations and online correction using measured power plant data, this enables ultra-short-term power prediction for photovoltaic power plants.

[0022] like Figure 1 The diagram shown is a flowchart of a photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing, as proposed in this application. This method continuously correlates the dynamic changes of the cloud field, cloud microphysical characteristics, changes in surface irradiance, and the power output process of the photovoltaic power station. This aims to improve prediction errors caused by insufficient cloud field physical characterization, lag in capturing short-term, small-scale cloud fields, and insufficient precision in converting cloud characteristics to module output power. It may include: S101 acquires initial cloud field data output from numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radars.

[0023] The initial cloud field data output by numerical weather prediction models provides large-scale background conditions and initial constraints on the temporal evolution of cloud fields within the prediction area. Infrared brightness temperature sequences from geostationary meteorological satellites reflect the spatial distribution of cloud clusters and thermal variations at cloud tops over continuous time intervals. Reflectivity factor data from X-band weather radars supplements the fine-grained structural information of near-field small-scale clouds and strong convective clouds. By simultaneously incorporating these three types of data, the problem of insufficient response to sudden cloud cluster evolution on a zero- to several-hour timescale can be avoided when relying solely on a single numerical model.

[0024] S102, the initial cloud field data, infrared brightness temperature sequence and reflectivity factor data are fused to obtain a fused cloud field.

[0025] During the fusion process, various data sources are organized temporally and spatially into a unified cloud field representation used for prediction calculations. This allows the fused cloud field to retain background cloud field information provided by numerical weather prediction models, while also incorporating continuous cloud top variation characteristics reflected by satellite observations and local cloud intensity characteristics reflected by radar observations. The resulting fused cloud field provides a more complete cloud field foundation for subsequent cloud motion analysis and radiation impact calculations.

[0026] S103, a three-dimensional dynamic cloud motion vector field is constructed based on the fused cloud field, and a cloud microphysical structure field is constructed by combining atmospheric thermodynamic parameters.

[0027] A three-dimensional dynamic cloud motion vector field is used to characterize the direction, speed, and spatial evolution trend of cloud movement within the prediction area, enabling dynamic characterization of the arrival time and duration of cloud-induced shading of solar power plant radiation. The cloud microphysical structure field, together with atmospheric thermodynamic parameters, describes the physical states within the cloud that affect radiation propagation. This allows clouds to participate in subsequent calculations as physical objects influencing the attenuation, scattering, and absorption of solar shortwave radiation, rather than simply as ordinary meteorological factors in prediction.

[0028] S104. Radiative transfer calculations are performed based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field to obtain surface irradiance data.

[0029] It should be noted that this surface irradiance data reflects the impact of cloud movement and cloud microphysical state on surface solar radiation during the prediction period. It can reflect the rapid changes in surface irradiance when cloud obstruction intensifies or weakens, providing direct radiation input for photovoltaic power prediction.

[0030] S105, determine the theoretical power generation of the photovoltaic power station based on the surface irradiance data, and perform online rolling correction of the theoretical power generation based on the measured data of the photovoltaic power station, and output the ultra-short-term photovoltaic power prediction result.

[0031] The theoretical power generation capacity reflects the potential power output of the photovoltaic power station under the currently calculated surface irradiance conditions. Online rolling correction continuously adjusts the deviation between the theoretical power generation capacity and the actual output using measured data, allowing the prediction results to be updated according to the latest cloud field changes and the actual operating status of the power station. Through this continuous processing, this embodiment forms a complete prediction chain from multi-source cloud field sensing, three-dimensional dynamic cloud field construction, radiative transfer calculation to photovoltaic power rolling correction, thereby improving the response capability of ultra-short-term photovoltaic power prediction to the rapid development and movement of small-scale cloud clusters.

[0032] The present application will be further described in detail below through some more detailed embodiments.

[0033] In some embodiments of this application, such as Figure 2As shown, the ultra-short-term photovoltaic power prediction method based on cloud field dynamic sensing can be executed in the following order: multi-source cloud field data acquisition and fusion, three-dimensional dynamic cloud field construction, surface irradiance calculation, and photovoltaic power conversion and online correction. It is particularly suitable for coastal areas with frequent cloud cover, fog, and severe convective weather. Within the 0-6 hour ultra-short-term prediction period, it reduces photovoltaic power prediction distortion caused by sudden small-scale convective clouds in the summer afternoon by improving the accuracy of expressing cloud field state, cloud movement, and cloud microphysical structure. Specifically: In the multi-source cloud field data acquisition phase, initial cloud field data output from high spatiotemporal resolution numerical weather prediction models are acquired, while infrared brightness temperature sequences from geostationary meteorological satellites and reflectivity factor data from X-band weather radars are simultaneously integrated in real time. Numerical weather prediction models can provide initial cloud field information, background thermal conditions, and large-scale cloud evolution information within the prediction area; infrared brightness temperature sequences from geostationary meteorological satellites are used to continuously reflect changes in cloud top brightness temperature, horizontal cloud expansion, and cloud top cooling processes; and reflectivity factor data from X-band weather radars are used to capture near-field convective clouds, precipitation clouds, or rapidly developing cloud tower structures. Since pure numerical weather prediction models have a relatively low update frequency of 0-2 hours, making it difficult to promptly reflect sudden small-scale convective clouds in the summer afternoon, this application supplements the short-term update capability of numerical models through real-time observations from satellites and radars.

[0034] like Figure 3 As shown, in the multi-source data fusion process, satellite, radar, and model data can be used as data acquisition inputs into the fusion processing link. Specifically, the cloud-water mixing ratio in the initial cloud field data output by the numerical weather prediction model is first assimilated using reflectivity factor data from an X-band weather radar, resulting in corrected initial cloud field data. Through this assimilation process, the local cloud intensity and cloud-water distribution information reflected by the radar reflectivity factor can be fed back into the initial cloud field data of the numerical model, thereby correcting the cloud field simulation bias caused by sea-land breeze circulation in coastal areas. Subsequently, the corrected initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data are fused using a physical sensing fusion bridging model to obtain the fused cloud field. This physical sensing fusion bridging model retains the large-scale background constraints of the numerical weather prediction model during fusion, while introducing continuous observations of cloud top states from satellite infrared brightness temperature and high-resolution observations of local cloud structures from X-band weather radar, enabling the fused cloud field to more accurately express the location, intensity, vertical development characteristics, and short-term evolution trends of clouds within the prediction area.

[0035] In a preferred embodiment of this application, the loss function of the physical perception fusion bridging model during training includes not only the mean squared error loss between the predicted and true values, but also a physical constraint loss term. This physical constraint loss term is constructed based on the atmospheric mass conservation equation or thermodynamic equation, and is used to constrain the cloud-water mixing ratio, cloud thermal state, or cloud field evolution trend output by the model to conform to atmospheric physical laws. As an example, the loss function Loss can be expressed as: Loss=Loss MSE +λ·Loss physics ; Among them, Loss MSE Loss represents the mean squared error loss between the predicted and actual values. physics λ represents the physical constraint loss term, and λ represents the physical constraint weight coefficient.

[0036] Through the aforementioned dual constraint mechanism, the physical perception fusion bridging model, when fusing multi-source cloud field data, not only learns the statistical correspondence between observation data and model data, but is also constrained by atmospheric physical laws, thereby reducing the output of cloud field results that violate physical laws by pure data-driven models in extreme weather or insufficient sample scenarios.

[0037] In the three-dimensional dynamic cloud field construction stage, a three-dimensional dynamic cloud motion vector field is constructed based on the fused cloud field. Specifically, optical flow or particle tracking algorithms can be used to process the fused cloud field at adjacent time points, extracting the displacement changes of cloud clusters or cloud field features in the horizontal and vertical directions to form a three-dimensional dynamic cloud motion vector field that can characterize the direction, speed, and spatial evolution trend of cloud movement. Optical flow is suitable for estimating the overall movement trend of clouds based on continuous cloud field images or grid fields, while particle tracking algorithms are suitable for tracking the short-term migration process of specific cloud clusters, cloud towers, or high-reflectivity areas. Through the three-dimensional dynamic cloud motion vector field, the time, path, and duration of the cloud-shaded area moving towards the photovoltaic power station can be predicted, providing dynamic input for subsequent calculations of abrupt changes in surface irradiance.

[0038] Simultaneously, a cloud microphysical structure field is constructed by combining atmospheric thermodynamic parameters. This field includes the phase distribution and effective radius of liquid water droplets and ice crystals within the cloud, and can be further combined with parameters such as cloud-water mixing ratio, cloud-ice mixing ratio, liquid water content, cloud top height, cloud optical thickness, and water vapor content to characterize the influence of clouds on solar shortwave radiation. Existing prediction methods, when treating clouds as ordinary meteorological factors, struggle to reflect the nonlinear effects of cloud water and ice phases, as well as particle size, on solar radiation scattering and absorption. This application, by constructing a cloud microphysical structure field, enables the cloud phase, particle size, and optical characteristics to be incorporated into the radiative transfer calculation process, thereby enhancing the physical representation of the relationship between cloud characteristics and changes in surface irradiance.

[0039] In one specific embodiment, when constructing the cloud microphysical structure field, the vertical development state of the cloud can be determined based on the intensity of the vertical updraft output by the physical perception fusion bridging model, and the cloud top height can be used to determine whether the cloud top has broken through the freezing layer. If the cloud top height breaks through the freezing layer, the particle phase in the corresponding region of the cloud microphysical structure field is updated from liquid water droplets to ice crystals, and the effective radius of the particles is adjusted synchronously according to the cloud's thermal state and particle growth process. Through this processing, the cloud microphysical structure field can reflect the impact of the transformation of liquid water droplets into ice crystals and the change in particle size on the radiative transfer process during the development of strong convective clouds.

[0040] like Figure 4 As shown, in the coupling process between the cloud microphysical structure field and the radiative transfer model, the three-dimensional dynamic cloud motion vector field, cloud microphysical structure field, real-time solar zenith angle, and aerosol optical depth are input into the radiative transfer model. The radiative transfer model calculates the attenuation, scattering, and absorption effects of cloud layers on solar shortwave radiation based on the cloud motion state and cloud microphysical state, and outputs the total surface irradiance, direct irradiance, and diffuse irradiance. As an optional implementation, the solution process for the radiative transfer model can use the DISORT (Discrete Ordinates Radiative Transfer Program for a Multi-Layered Plane-Parallel Medium) radiative transfer solver or the 6S model shown in the attached figure to calculate the propagation results of solar shortwave radiation in the atmosphere under different cloud and aerosol conditions. Since the solar zenith angle determines the path length of incident solar radiation, aerosol optical depth affects the atmospheric scattering background, and the scale of cloud water, cloud ice, and cloud particles influences the modulation of direct and scattered radiation by clouds, inputting these parameters into the radiative transfer model is beneficial for obtaining surface irradiance data that more closely reflects the actual cloud obstruction process. In one embodiment, the cloud microphysical structure field and the DISORT radiative transfer model can be calculated collaboratively. First, the cloud microphysical structure field is dynamically reconstructed based on the fused cloud field and atmospheric thermodynamic parameters to obtain physical parameters such as cloud-water mixing ratio, cloud-ice mixing ratio, cloud top height, particle phase state, and effective radius. Second, these physical parameters and their corresponding optical constants are input into the DISORT radiative transfer model to calculate the radiation propagation results under different cloud structures. Third, the parameters affecting irradiance changes in the cloud microphysical structure field are corrected based on the radiation field calculation results. Finally, the total surface irradiance, direct irradiance, and scattered irradiance are obtained through multi-scale collaborative calculation.

[0041] In some embodiments of this application, an aerosol-cloud interaction module can be introduced into the radiative transfer model to address the high humidity characteristics of the plum rain season. This aerosol-cloud interaction module is used to correct the influence of aerosol optical thickness on scattered irradiance under high humidity conditions, so that the surface scattered irradiance is no longer determined solely by cloud cover intensity, but also considers the influence of aerosol hygroscopic growth, enhanced scattering, and their coupling with cloud microphysical structures on the irradiance component under high humidity conditions. This module can improve the problem of large estimation errors in scattered irradiance in coastal areas with high humidity and frequent fog.

[0042] After obtaining surface irradiance data, the theoretical power generation is determined based on the surface irradiance data, the physical parameters of the photovoltaic power station's modules, and the electrical topology. Specifically, total surface irradiance, direct irradiance, and diffuse irradiance can be used together as the radiation input for the photovoltaic physical conversion process. Module physical parameters are used to characterize the output response of photovoltaic modules under different irradiance conditions, and the electrical topology is used to characterize the series-parallel relationship of modules, array distribution, and power collection relationship on the power station side. When the surface direct irradiance drops sharply due to cloud obstruction, the theoretical power generation can correspondingly decrease rapidly; when the edge of the cloud passes or the proportion of diffuse irradiance increases, the theoretical power generation can exhibit short-term fluctuations with changes in the radiation component, thus forming a refined physical conversion mechanism from cloud characteristics to photovoltaic module output power.

[0043] During the online rolling correction phase, real-time measured data from the photovoltaic power plant is acquired. This data includes measured irradiance collected by a ground-based radiometer and measured power collected by the photovoltaic power plant monitoring system, which can be a SCADA (Supervisory Control and Data Acquisition) system. The surface irradiance data calculated from radiative transfer is corrected based on the measured irradiance, and the theoretical power generation is corrected based on the measured power, ensuring that the cloud-irradiance mapping parameters and power conversion deviations in the prediction link are updated continuously with the latest observations. This online rolling correction mechanism continuously corrects the relationship between cloud movement speed, cloud obstruction intensity, irradiance calculation deviations, and actual power plant output deviations during the prediction process, ultimately outputting 0-6 hour ultra-short-term photovoltaic power prediction results.

[0044] like Figure 5As shown, in some embodiments of this application, the surface irradiance calculation process driven by multi-source cloud field data can be executed by a radiative transfer calculation engine. Specifically, satellite remote sensing data, meteorological radar data, and numerical model data are used as input data for the radiative transfer calculation engine. Satellite remote sensing data may include cloud top height and cloud optical thickness, used to characterize the vertical development height of clouds and their overall shielding ability against solar shortwave radiation; meteorological radar data may include reflectivity factor and liquid water content, used to characterize the distribution of condensates within near-field clouds and the intensity of local clouds; numerical model data may include aerosol optical depth and water vapor content, used to provide the atmospheric scattering and absorption background required for radiative transfer calculation. The radiative transfer calculation engine includes a cloud microphysical parameter inversion unit and a DISORT radiative transfer solver. The cloud microphysical parameter inversion unit is used to invert cloud microphysical structure parameters based on cloud top height, cloud optical thickness, reflectivity factor, liquid water content, aerosol optical depth, and water vapor content. These cloud microphysical structure parameters include the phase distribution and effective radius of liquid water droplets and ice crystals in the cloud. The DISORT radiative transfer solver calculates the attenuation, scattering, and absorption effects of clouds on solar shortwave radiation based on cloud microphysical structure parameters and atmospheric radiation background parameters, yielding total surface irradiance, direct irradiance, and diffuse irradiance. Furthermore, the radiative transfer calculation engine receives measured irradiance feedback and corrects the cloud microphysical parameter inversion results or radiative transfer calculation results based on this feedback, ensuring that the output total surface irradiance, direct irradiance, and diffuse irradiance are updated with the latest ground observations.

[0045] In another application example of this application, the target of the prediction is a photovoltaic power station in Ningbo. This area is characterized by frequent cloud cover and fog along the coast, and severe convective weather is common in summer. When small-scale convective clouds develop rapidly in the afternoon, the power output of the photovoltaic power station is prone to short-term, drastic fluctuations. For this scenario, cloud field data from the GRAPES (Global / Regional Assimilation and Prediction System) 3km model is acquired. This cloud field data has a spatial resolution of 3km and an update time of 1 hour. Simultaneously, infrared brightness temperature data from the Fengyun-4 geostationary meteorological satellite is accessed, with an interval of 15 minutes and a spatial resolution of 2km. Furthermore, volume scan data from X-band weather radars deployed around the power station is also accessed, with an interval of 1 minute and a spatial resolution of 100m. When a strong convective cloud is developing rapidly, if the vertical updraft intensity output by the physical perception fusion bridging model indicates that the cloud body is continuing to develop and the cloud top height breaks through the freezing layer, then the particle phase state of the corresponding cloud area is updated to the ice crystal state, and the effective radius of the particles is adjusted before being input into the radiative transfer model to reflect the impact of cloud top ice crystallization on direct and diffuse irradiance.

[0046] In this application example, X-band weather radar reflectivity factors are used to identify cumulonimbus cloud towers developing in the afternoon, and infrared brightness temperature data from the Fengyun-4 geostationary meteorological satellite are combined to identify the cloud top ice crystallization process. Radar reflectivity factors can reflect rapidly developing convective cloud towers and locally high reflectivity areas, while infrared brightness temperature sequences can reflect cloud top cooling and ice crystallization processes. After fusing these two types of observations with the cloud field background output from the GRAPES 3km model, they are used to construct a three-dimensional dynamic cloud motion vector field and a cloud microphysical structure field containing the phase distribution and effective radius of liquid water droplets and ice crystals in the cloud.

[0047] Subsequently, the sudden drop in direct irradiance caused by cloud cover was incorporated into the photovoltaic physical conversion process. When the surface direct irradiance drops rapidly due to cumulonimbus clouds passing over the photovoltaic power station, the theoretical power generation also experiences a short-term decrease; when the cloud edge passes by or the proportion of scattered irradiance changes, the theoretical power generation exhibits continuous fluctuations. Therefore, the sawtooth-like fluctuations in photovoltaic power generation under the influence of strong convective clouds can be predicted.

[0048] During the online rolling correction process, the measured power from the previous 15 minutes is used to correct the cloud movement speed parameter for the next hour, and the measured irradiance data collected by the ground-based radiometer is combined to correct the surface irradiance data. Through this rolling correction mechanism, the prediction curve can be updated with the latest measured power and measured irradiance, thereby improving the fit between the ultra-short-term photovoltaic power prediction results for the next hour and even 0-6 hours and the actual power curve.

[0049] like Figure 6 The diagram shown is a schematic representation of a photovoltaic power ultra-short-term prediction system based on cloud field dynamic sensing, which may include: The data module is used to acquire initial cloud field data output by numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radars. The fusion module is used to fuse the initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data to obtain a fused cloud field. The construction module is used to construct a three-dimensional dynamic cloud motion vector field based on the fused cloud field, and to construct a cloud microphysical structure field in combination with atmospheric thermodynamic parameters; The calculation module is used to perform radiative transfer calculations based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field to obtain surface irradiance data. The prediction module is used to determine the theoretical power generation of the photovoltaic power station based on the surface irradiance data, and to perform online rolling correction of the theoretical power generation based on the measured data of the photovoltaic power station, and output the ultra-short-term photovoltaic power prediction results.

[0050] In some embodiments of this application, the aforementioned photovoltaic power ultra-short-term prediction system based on cloud field dynamic sensing can be communicatively connected to a geostationary meteorological satellite receiving terminal, an X-band weather radar, a ground-based radiometer, and a photovoltaic power plant monitoring system. The geostationary meteorological satellite receiving terminal provides satellite infrared brightness temperature sequences, the X-band weather radar provides reflectivity factor data, the ground-based radiometer provides measured irradiance, and the photovoltaic power plant monitoring system provides measured power. After the above data enters the prediction device via the communication connection, cloud field fusion, dynamic cloud field construction, radiative transfer calculation, power conversion, and online correction are completed sequentially. Through this system architecture, a closed-loop prediction process can be formed from remote sensing observation, radar observation, numerical model data to actual power plant feedback, enabling the prediction results to continuously closely reflect the actual operating status of the power plant.

[0051] It should be noted that, in the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of each block is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple blocks may be combined or integrated into another device, or some features may be ignored or not executed. The modules described as separate components may or may not be physically separated. The components shown as modules may be one or more physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs.

[0052] Furthermore, in the various embodiments of the present invention, the modules can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The integrated unit described above can be implemented in hardware or as a software functional unit.

[0053] This application also provides an electronic device, which may include one or more processors, memory and communication interfaces.

[0054] The memory, communication interface, and processor are coupled together. For example, the memory, communication interface, and processor can be coupled together via a bus.

[0055] The communication interface is used for data transmission with other devices. The memory stores computer program code. This computer program code includes computer instructions, which, when executed by the processor, cause the electronic device to perform the steps of the aforementioned photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing.

[0056] The processor can be a central processing unit (CPU) or a controller, such as a CPU, a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this application. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc. The processor can be used to support an electronic device in performing the method steps provided in the above embodiments.

[0057] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. These buses can be categorized as address buses, data buses, control buses, etc.

[0058] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the above-described photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing.

[0059] The computer-readable storage media involved in this application include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage media known in the art.

[0060] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for ultra-short-term photovoltaic power prediction based on dynamic cloud field sensing, characterized in that, include: Acquire initial cloud field data output from numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radars; The initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data are fused to obtain a fused cloud field. A three-dimensional dynamic cloud motion vector field is constructed based on the fused cloud field, and a cloud microphysical structure field is constructed by combining atmospheric thermodynamic parameters. Radiative transfer calculations are performed based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field to obtain surface irradiance data. The theoretical power generation of the photovoltaic power station is determined based on the surface irradiance data, and the theoretical power generation is corrected online based on the measured data of the photovoltaic power station, and the ultra-short-term photovoltaic power prediction results are output.

2. The photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing according to claim 1, characterized in that, The initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data are fused, including: The reflectivity factor data is converted into a correction value for the cloud-water mixing ratio, and the cloud-water mixing ratio in the initial cloud field data is assimilated based on the correction value to obtain the corrected initial cloud field data. The fused cloud field is obtained by fusing the corrected initial cloud field data, the infrared brightness temperature sequence, and the reflectivity factor data using a physical perception fusion bridging model with an embedded atmospheric thermodynamic constraint neural network. The physical perception fusion bridging model is also used to perform minute-level assimilation updates on the initial cloud field data output by the numerical weather forecast model.

3. The photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing according to claim 2, characterized in that, The construction of a three-dimensional dynamic cloud motion vector field based on the fused cloud field, and the construction of a cloud microphysical structure field in combination with atmospheric thermodynamic parameters, includes: Based on the fused cloud field, the three-dimensional dynamic cloud motion vector field is constructed using optical flow or particle tracking algorithm; Based on the intensity of the vertical updraft output by the physical perception fusion bridging model, determine whether the cloud top height has broken through the freezing layer; If the cloud top height exceeds the frozen layer, the particle phase in the corresponding region of the cloud microphysical structure field will be updated from liquid water droplets to ice crystals, and the effective radius of the particles will be adjusted simultaneously. The cloud microphysical structure field includes the phase distribution and effective radius of liquid water droplets and ice crystals in the cloud.

4. The photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing according to claim 1, characterized in that, Radiative transfer calculations based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field include: inputting the three-dimensional dynamic cloud motion vector field, the cloud microphysical structure field, the real-time solar zenith angle, and the aerosol optical depth into a radiative transfer model to calculate the attenuation, scattering, and absorption effects of the cloud layer on solar shortwave radiation; wherein, the radiative transfer model includes the DISORT radiative transfer model or the 6S model.

5. The photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing according to claim 1, characterized in that, The theoretical power generation capacity of the photovoltaic power station is determined based on the surface irradiance data, including: The theoretical power generation is determined based on the surface irradiance data, the physical parameters of the photovoltaic power station components, and the electrical topology.

6. The photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing according to claim 1, characterized in that, The measured data of the photovoltaic power station includes the measured irradiance collected by the ground radiometer and the measured power collected by the photovoltaic power station monitoring system.

7. The photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing according to claim 6, characterized in that, Online rolling correction of theoretical power generation based on measured data of photovoltaic power plants includes: correcting the surface irradiance data based on the measured irradiance, and correcting the theoretical power generation based on the measured power.

8. A photovoltaic power ultra-short-term prediction system based on cloud field dynamic sensing, characterized in that, include: The data module is used to acquire initial cloud field data output by numerical weather prediction models, infrared brightness temperature sequences from geostationary meteorological satellites, and reflectivity factor data from X-band weather radars. The fusion module is used to fuse the initial cloud field data, infrared brightness temperature sequence, and reflectivity factor data to obtain a fused cloud field. The construction module is used to construct a three-dimensional dynamic cloud motion vector field based on the fused cloud field, and to construct a cloud microphysical structure field in combination with atmospheric thermodynamic parameters; The calculation module is used to perform radiative transfer calculations based on the three-dimensional dynamic cloud motion vector field and the cloud microphysical structure field to obtain surface irradiance data. The prediction module is used to determine the theoretical power generation of the photovoltaic power station based on the surface irradiance data, and to perform online rolling correction of the theoretical power generation based on the measured data of the photovoltaic power station, and output the ultra-short-term photovoltaic power prediction results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the photovoltaic power ultra-short-term prediction method based on cloud field dynamic sensing as described in any one of claims 1-7.