Offshore photovoltaic system output power prediction method and system

Abstract

The invention discloses a method and a system for predicting output power of an offshore photovoltaic system, particularly relates to the technical field of power prediction data processing, and is used for solving the problem of insufficient prediction precision caused by the fact that the dynamic attitude change of a floating photovoltaic platform in waves is not considered in the existing prediction method. The method comprises the following steps: determining a dominant motion mode of a platform and generating a time-varying attitude angle sequence by acquiring real-time wave spectrum data and platform structure parameters of a target sea area; carrying out geometric coupling on the sequence and a sun position vector, and dynamically solving an effective sunlight incident angle cosine value on the surface of the photovoltaic panel; then combining real-time cloud layer monitoring information, comprehensively evaluating geometric attenuation caused by waves and atmospheric attenuation caused by cloud layers, and obtaining a comprehensive radiation attenuation degree; correcting the total solar radiation data of the horizontal plane by using the attenuation degree, and generating an effective solar radiation intensity sequence projected to the surface of the photovoltaic panel; and finally, inputting the sequence into a prediction model and outputting a predicted power value of the offshore photovoltaic system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power prediction data processing, and more particularly, to a method and system for predicting output power of a marine photovoltaic system. BACKGROUND

[0002] As an important direction of renewable energy development and utilization, the accurate prediction of the output power of a marine photovoltaic power generation system is crucial for the stable operation and dispatch of a power system. Existing power prediction methods generally establish a model based on historical meteorological data and photovoltaic system operation data. One of the core links is to calculate the incident angle of sunlight according to the position of the sun and the fixed installation geometric parameters of the photovoltaic panel, and then determine the effective solar radiation intensity received by the photovoltaic panel. This method has a certain prediction reliability when applied to land-based photovoltaic power stations or fixed marine photovoltaic systems with stable installation posture.

[0003] However, when the above prediction method is applied to a floating marine photovoltaic system, there are obvious limitations. The floating photovoltaic platform will produce continuous random attitude changes under the action of waves, causing the actual inclination and azimuth of the photovoltaic panel to deviate from its nominal fixed installation value. Existing prediction models generally consider the installation posture of the photovoltaic panel as a static parameter and fail to take into account the dynamic geometric changes caused by wave disturbance. This results in inherent bias in the calculation of incident angle and effective radiation, making it difficult to meet the actual needs of marine photovoltaic projects in terms of the accuracy of the final power prediction results. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a method and system for predicting the output power of a marine photovoltaic system to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] A method for predicting the output power of a marine photovoltaic system, comprising the following steps:

[0007] S1, obtaining real-time wave spectrum data of a target sea area and platform structure parameters of a floating photovoltaic platform;

[0008] S2, determining the current dominant motion mode of the floating photovoltaic platform according to the real-time wave spectrum data and the platform structure parameters, and generating a time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves based on the motion law corresponding to the dominant motion mode;

[0009] S3, geometrically coupling the time-varying attitude angle sequence with the solar vector determined by the solar position astronomical algorithm to dynamically solve the effective solar light incident angle cosine value of the surface of the photovoltaic panel;

[0010] S4, acquiring real-time cloud layer monitoring information of the target sea area, and comprehensively evaluating the comprehensive radiation attenuation degree under the joint action of the geometric attenuation caused by the wave and the atmospheric attenuation caused by the cloud layer by combining the effective solar light incident angle cosine value and the real-time cloud layer monitoring information;

[0011] S5, correcting the total solar radiation data reaching the horizontal plane based on the comprehensive radiation attenuation degree to generate an effective solar radiation intensity sequence projected onto the surface of the photovoltaic panel;

[0012] S6, inputting the effective solar radiation intensity sequence into a pre-trained photovoltaic power generation power prediction model to output a predicted power value of the offshore photovoltaic system.

[0013] Further, real-time wave spectrum data of the target sea area and platform structure parameters of the floating photovoltaic platform are acquired, including:

[0014] The real-time wave spectrum data of the target sea area is acquired through a marine monitoring satellite or a buoy measuring device;

[0015] The platform structure parameters are acquired from the design data of the floating photovoltaic platform, and the platform structure parameters include platform size, platform mass distribution, and platform draft depth.

[0016] Further, the current dominant motion mode of the floating photovoltaic platform is determined according to the real-time wave spectrum data and the platform structure parameters, and a time-varying attitude angle sequence of the floating photovoltaic platform under the action of the wave is generated based on the motion law corresponding to the dominant motion mode, including:

[0017] The characteristic period and the spectral peak period of the real-time wave spectrum data are extracted;

[0018] The characteristic period and the spectral peak period are matched and compared with the platform roll natural period and the pitch natural period calculated based on the platform size and the platform mass distribution;

[0019] It is determined whether the roll is the dominant motion mode or the pitch is the dominant motion mode according to the matching and comparison result;

[0020] Based on the determined dominant motion mode, a corresponding historical roll attitude angle sequence or a historical pitch attitude angle sequence is selected as a sample;

[0021] The motion law of the sample is fitted through a time series analysis method to extrapolate a time-varying attitude angle sequence of the floating photovoltaic platform under the action of the wave.

[0022] Further, the time-varying attitude angle sequence is geometrically coupled with a solar vector determined by a solar position astronomical algorithm to dynamically solve the effective solar light incident angle cosine value of the surface of the photovoltaic panel, including:

[0023] The direction component of the sun vector in the geographic coordinate system is calculated based on the sun position astronomical algorithm;

[0024] The time-varying attitude angle sequence is converted into the dynamic direction component of the photovoltaic panel normal vector in the geographic coordinate system;

[0025] The effective solar light incident angle cosine value of the photovoltaic panel surface is dynamically solved by calculating the dot product of the sun vector and the photovoltaic panel normal vector.

[0026] Further, the effective solar light incident angle cosine value of the photovoltaic panel surface is dynamically solved by calculating the dot product of the sun vector and the photovoltaic panel normal vector, which is to express the sun vector and the photovoltaic panel normal vector as three-dimensional vectors in the same geographic coordinate system, and directly solve it through the dot product calculation formula of the two three-dimensional vectors.

[0027] Further, real-time cloud monitoring information of the target sea area is obtained, and the effective solar light incident angle cosine value and the real-time cloud monitoring information are comprehensively evaluated to evaluate the comprehensive radiation attenuation degree under the joint action of the geometric attenuation caused by the wave and the atmospheric attenuation caused by the cloud, including:

[0028] The real-time cloud monitoring information of the target sea area is obtained through meteorological satellite remote sensing data, and the real-time cloud monitoring information includes the cloud cover percentage;

[0029] The effective solar light incident angle cosine value is converted into the geometric attenuation coefficient of the photovoltaic panel surface receiving solar radiation caused by the wave;

[0030] The corresponding relationship between the cloud cover percentage and the atmospheric transmission attenuation coefficient caused by the cloud is established;

[0031] The geometric attenuation coefficient and the atmospheric transmission attenuation coefficient are multiplied to obtain the comprehensive radiation attenuation degree.

[0032] Further, the corresponding relationship between the cloud cover percentage and the atmospheric transmission attenuation coefficient caused by the cloud is established, which is to pre-establish a lookup table containing different cloud cover percentage intervals and their corresponding atmospheric transmission attenuation coefficients, and to convert the cloud cover percentage in the real-time cloud monitoring information into a specific atmospheric transmission attenuation coefficient by querying the lookup table.

[0033] Further, based on the comprehensive radiation attenuation degree, the total solar radiation data reaching the horizontal plane is corrected to generate an effective solar radiation intensity sequence projected onto the photovoltaic panel surface, including:

[0034] The total solar radiation data reaching the horizontal plane at each time is multiplied by the comprehensive radiation attenuation degree at the corresponding time;

[0035] The result obtained after multiplication is used as the effective solar radiation intensity projected onto the photovoltaic panel surface at the corresponding time;

[0036] The effective solar radiation intensity of all time points is sequentially aggregated to generate an effective solar radiation intensity sequence.

[0037] Further, the effective solar radiation intensity sequence is input into a pre-trained photovoltaic power generation power prediction model to output a predicted power value of the offshore photovoltaic system, including:

[0038] The effective solar radiation intensity sequence is taken as an input feature of the pre-trained photovoltaic power generation power prediction model;

[0039] The effective solar radiation intensity sequence is processed by the photovoltaic power generation power prediction model to obtain a predicted power value sequence of the offshore photovoltaic system in a future preset time period.

[0040] On the other hand, the present application provides an offshore photovoltaic system output power prediction system, including the following modules:

[0041] The parameter acquisition module is configured to acquire real-time wave spectrum data of a target sea area and platform structure parameters of the floating photovoltaic platform;

[0042] The sequence generation module is configured to determine a dominant motion mode of the floating photovoltaic platform at present according to the real-time wave spectrum data and the platform structure parameters, and generate a time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves based on a motion law corresponding to the dominant motion mode;

[0043] The cosine solution module is configured to geometrically couple the time-varying attitude angle sequence with a solar vector determined by a solar position astronomical algorithm to dynamically solve a cosine value of an effective solar light incident angle on the surface of the photovoltaic panel;

[0044] The degree evaluation module is configured to acquire real-time cloud monitoring information of the target sea area, and comprehensively evaluate a comprehensive radiation attenuation degree under the joint action of geometric attenuation caused by waves and atmospheric attenuation caused by clouds based on the cosine value of the effective solar light incident angle and the real-time cloud monitoring information;

[0045] The data correction module is configured to correct solar total radiation data reaching a horizontal plane based on the comprehensive radiation attenuation degree to generate an effective solar radiation intensity sequence projected onto the surface of the photovoltaic panel;

[0046] The power output module is configured to input the effective solar radiation intensity sequence into a pre-trained photovoltaic power generation power prediction model to output a predicted power value of the offshore photovoltaic system.

[0047] Compared with the prior art, the present application has the following beneficial effects:

[0048] 1. The dynamic attitude change of a floating photovoltaic platform in a real marine environment is introduced into the data processing chain of photovoltaic power prediction. Traditional prediction models consider the installation attitude of photovoltaic panels as a fixed and static parameter. However, the present application couples real-time wave spectrum data with platform structure parameters to first generate a high-time-efficiency time-varying attitude angle sequence, and then dynamically determines the actual incidence angle of sunlight on the surface of the photovoltaic panel at each moment through strict geometric calculation. This fundamentally changes the way the incidence angle parameter is obtained, transforming it from a preset constant to a dynamic variable reflecting the real motion state of the platform. This significantly improves the accuracy of the effective solar radiation intensity, a core input physical quantity, under the operating environment of the offshore photovoltaic system, making the input conditions of the prediction model more realistic.

[0049] 2. A calculation framework is constructed to unify and quantify the attenuation effects of waves and clouds, effectively integrating and evaluating the influence of complex environmental factors. By establishing a product model of "geometric attenuation coefficient" and "atmospheric transmission attenuation coefficient", the platform attitude change caused by waves (geometric effect) and the atmospheric transmission change caused by clouds (physical effect) are normalized into a comprehensive radiation attenuation degree index. This processing method makes the subsequent correction process of horizontal plane solar total radiation data clear and calculable. The final effective solar radiation intensity sequence contains both dynamic geometric and atmospheric physical attenuation information, ensuring that the input data (effective solar radiation intensity sequence) for the prediction model is strictly corrected based on the special marine environment, thereby laying a solid foundation for high-precision power prediction at the algorithm level. The final output of the prediction power value has higher reference value for the fine scheduling of the power system and the stable operation of the offshore photovoltaic power station. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The flowchart of the offshore photovoltaic system output power prediction method of the present application;

[0051] Figure 2 The structural schematic diagram of the offshore photovoltaic system output power prediction system of the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0053] Embodiment 1: Figure 1The application provides a kind of offshore photovoltaic system output power prediction method, it includes the following steps:

[0054] S1, the real-time wave spectrum data of target sea area and the platform structure parameter of floating photovoltaic platform are acquired;

[0055] S2, the current dominant motion mode of floating photovoltaic platform is determined according to real-time wave spectrum data and platform structure parameter, and the time-varying attitude angle sequence of floating photovoltaic platform under the action of wave is generated based on the motion law corresponding to the dominant motion mode;

[0056] S3, the time-varying attitude angle sequence is geometrically coupled with the sun vector determined by the sun position astronomical algorithm, and the effective sun light incident angle cosine value of photovoltaic panel surface is dynamically solved;

[0057] S4, the real-time cloud monitoring information of target sea area is acquired, and the comprehensive radiation attenuation degree under the joint action of geometric attenuation caused by wave and atmospheric attenuation caused by cloud is evaluated by comprehensively considering effective sun light incident angle cosine value and real-time cloud monitoring information;

[0058] S5, the total solar radiation data reaching the horizontal plane is corrected based on the comprehensive radiation attenuation degree, and the effective solar radiation intensity sequence projected to the photovoltaic panel surface is generated;

[0059] S6, the effective solar radiation intensity sequence is input into the photovoltaic power generation power prediction model trained in advance, and the predicted power value of offshore photovoltaic system is output.

[0060] S1, the real-time wave spectrum data of target sea area and the platform structure parameter of floating photovoltaic platform are acquired, and the specific implementation is as follows:

[0061] The acquisition of real-time wave spectrum data is achieved through two parallel and mutually verifiable approaches. The first approach is to use ocean monitoring satellites, such as satellites equipped with synthetic aperture radar, which scan and image the target sea area, and through processing the radar echo signal to obtain sea surface wave height field data, and then generate real-time wave spectrum data containing wave energy distribution information with frequency and direction through spectral analysis technology. This real-time wave spectrum data is usually transmitted in the form of a digital matrix to the data processing center through the satellite ground receiving station, with the rows and columns of the matrix corresponding to the frequency interval and direction interval, and the value of each matrix element representing the wave energy density in the frequency interval and direction interval. The second approach is to use a buoy measurement device deployed in the target sea area, which has an accelerometer, gyroscope and GPS module built-in, by measuring the three-dimensional motion trajectory of the buoy under the action of waves, calculating the wave surface elevation data sequence, and then converting the time domain signal to frequency domain energy distribution through fast Fourier transform, obtaining the real-time wave spectrum data of the buoy measurement point. This real-time wave spectrum data is transmitted in real time to the monitoring platform through a wireless communication network. In practical applications, the wide-area wave spectrum data obtained by satellites and the point wave spectrum data measured by buoys can be spatially interpolated and compared to improve the spatial representativeness and accuracy of the data.

[0062] The acquisition of platform structure parameters requires extracting key static parameters from the engineering design database of the floating photovoltaic platform. Platform size refers to the geometric dimensions of the platform body, including the length, width and depth of the platform, which are directly obtained from the final version of the platform general arrangement drawing and recorded in meters. Platform mass distribution refers to the weight of the platform in the unloaded state and its distribution along the length and width directions, specifically represented by the overall weight of the platform, the coordinates of the center of gravity, and the moments of inertia around the horizontal and vertical axes of the center of gravity. These data are obtained from the platform weight calculation report and stability calculation book. Platform draft refers to the depth of the platform main structure immersed in water in the unloaded and stationary floating state, which is also obtained from the platform stability calculation book. These platform structure parameters are basic physical quantities that characterize the hydrodynamic characteristics of the platform and need to be stored in a structured data format, such as establishing a database table containing parameter name, value and unit, for subsequent calculation process calls.

[0063] It should be noted that real-time wave spectrum data is a dynamic environmental parameter that reflects the current sea state excitation characteristics, while platform structure parameters are relatively fixed physical properties that determine the response characteristics of the platform to wave excitation. Together, they form the basic input data set for subsequent analysis of platform motion behavior. Through the above specific data acquisition approaches and processing methods, the reliability and accuracy of the original data are ensured, laying a solid data foundation for the entire power prediction process.

[0064] S2. Determine the current dominant motion mode of the floating photovoltaic platform based on real-time wave spectrum data and platform structural parameters. Generate a time-varying attitude angle sequence of the floating photovoltaic platform under wave action based on the motion law corresponding to the dominant motion mode. The specific implementation is as follows:

[0065] The specific steps for extracting the characteristic period and peak period from real-time wave spectrum data are as follows. The characteristic period is a representative period obtained by calculating the spectral moments of the wave spectrum. Specifically, it is achieved by first integrating the energy density of all frequency components in the real-time wave spectrum data to obtain the total wave energy, i.e., the zeroth-order spectral moment m0; then integrating the product of frequency and energy density to obtain the first-order spectral moment m1; the characteristic period Tc is calculated as Tc = m0 / m1, reflecting the average energy period of the wave. The peak period Tp is the reciprocal of the frequency corresponding to the maximum energy density in the real-time wave spectrum data. By traversing all frequency points in the wave spectrum matrix, the frequency point fpeak with the highest energy density is found, and then Tp = 1 / fpeak is calculated. These two period parameters characterize the energy distribution features of the wave from different perspectives and need to be calculated accurately and retained to two decimal places, in seconds.

[0066] Calculating the natural periods of rolling and pitching of a platform based on its dimensions and mass distribution requires the application of ship statics principles. The natural period of rolling, Tr, is the natural period of the platform's oscillation around its longitudinal axis. Its formula is Tr = 2π × √[Ixx / (ρ × g × Vv × GMt]], where Ixx is the moment of inertia of the platform about its longitudinal axis, obtained from the platform's mass distribution data; ρ is the density of seawater, taken as a standard value of 1025 kg / m³; and g is the acceleration due to gravity, taken as 9.81 m / s². 2 Vv is the platform's drainage volume, calculated based on the platform's dimensions and draft; GMt is the transverse metacentric height, calculated based on the platform's dimensions and the position of the center of gravity in the platform's mass distribution. The natural period of pitch, Tpitch, is the platform's natural period of swaying back and forth about its transverse axis, calculated as Tpitch = 2π × √[Iyy / (ρ × g × Vv × GMl)], where Iyy is the platform's moment of inertia about its transverse axis, obtained from the platform's mass distribution data; and GMl is the longitudinal metacentric height. These calculations must ensure unit consistency: all length units are meters, mass units are kilograms, and time units are seconds. Results should be rounded to two decimal places.

[0067] The specific logic of matching and comparing the characteristic period and the spectral peak period with the roll natural period and the pitch natural period is as follows. The purpose of matching and comparing is to determine whether the main energy period of the wave is close to a certain natural period of the platform, thereby causing a resonance effect. The absolute difference |Tc - Tr| of the characteristic period and the roll natural period, the absolute difference |Tc - Tpitch| of the characteristic period and the pitch natural period, the absolute difference |Tp - Tr| of the spectral peak period and the roll natural period, and the absolute difference |Tp - Tpitch| of the spectral peak period and the pitch natural period are calculated respectively. A period matching threshold is set to be 20% of the natural period value, for example, if the roll natural period Tr is 10 seconds, the matching threshold is set to 2 seconds. When the absolute value of a certain difference is less than 20% of the corresponding natural period value, it is considered that the period matching is successful. In particular, when one of the characteristic period and the spectral peak period matches the roll natural period and the other matches the pitch natural period, the closeness of the matching needs to be compared, and the matching pair with the smaller difference is selected as the dominant matching.

[0068] The decision rule for determining whether the roll is the dominant motion mode or the pitch is the dominant motion mode according to the matching and comparison results is as follows. If only the roll natural period matches successfully (i.e. at least one of the characteristic period or the spectral peak period has a difference with the roll natural period less than the threshold, and none matches the pitch natural period), it is determined that the roll is the dominant motion mode. If only the pitch natural period matches successfully, it is determined that the pitch is the dominant motion mode. If both natural periods match successfully, the excitation coefficient in the roll direction and the excitation coefficient in the pitch direction are calculated, the excitation coefficient is defined as the sum of the energy proportion of the matched wave period component in the wave spectrum, and the direction with the larger excitation coefficient is selected as the dominant motion mode. If neither natural period matches successfully, the motion mode corresponding to the natural period closer to the frequency band with the largest energy in the wave spectrum (i.e. the frequency band corresponding to the spectral peak period) is selected as the dominant motion mode. This decision-making method based on energy matching and resonance principle can accurately reflect the most likely motion form of the platform in actual sea conditions.

[0069] Based on the determined dominant motion mode, the corresponding historical roll attitude angle sequence or historical pitch attitude angle sequence is selected as the sample. The historical attitude angle sequence is derived from the historical database of the inertial measurement unit installed on the platform, which records the motion data of the platform in the past few months or even years in chronological order. The selection criteria for the sample are as follows: the sea state condition of the historical data (characterized by historical wave spectrum data) is similar to the current real-time wave spectrum data, i.e. the characteristic period, spectral peak period and significant wave height of the historical wave spectrum have a relative error of less than 15% with the current values. From the historical data segments that meet the conditions, a continuous 2-hour historical roll attitude angle sequence or historical pitch attitude angle sequence (selected according to the determined dominant motion mode) with a sampling interval of 1 second is extracted as the modeling sample. The sample data needs to be preprocessed to eliminate abnormal jump points caused by sensor failure and eliminate high-frequency noise through sliding average filtering to retain the main motion trend.

[0070] The specific implementation of fitting the motion law of the sample by time series analysis method and extrapolating to generate the time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves is as follows. The time series analysis uses the autoregressive integrated moving average model. First, the stationarity of the sample sequence is tested, and if it is not stationary, it is differentiated until the sequence is stationary. Then the autocorrelation function and partial autocorrelation function of the sample sequence are calculated, and the autoregressive order p and moving average order q are determined according to the characteristics of function truncation and tailing. The model parameters are estimated using the least squares method, and the residual sequence is tested to see if it is white noise. After the model is established, the last p observations of the sample sequence are used as the initial conditions to recursively predict the attitude angle value at each future sampling time for the purpose of extrapolating the future 1 hour. For the prediction results, a random disturbance term needs to be added, and the strength of the disturbance is determined according to the variance of the sample sequence to simulate the uncertainty caused by the randomness of the waves. The final time-varying attitude angle sequence is a discrete time sequence containing data points every second for the next 1 hour, which reflects the periodicity under the dominant motion mode and contains appropriate random fluctuations, and can truly simulate the motion state of the platform in the future short term. The entire modeling and prediction process needs to be programmed in a computer, and the core algorithms such as model order determination and parameter estimation can be completed by calling standard time series analysis library functions.

[0071] S3, geometrically couple the time-varying attitude angle sequence with the solar vector determined by the solar position astronomical algorithm to dynamically solve the effective solar light incident angle cosine value on the surface of the photovoltaic panel, which is implemented as:

[0072] The direction component of the sun vector in the geographic coordinate system is calculated based on the solar position astronomical algorithm, and the specific implementation adopts the internationally recognized solar position algorithm. The input parameters of the algorithm are the accurate time stamp (in coordinated universal time, accurate to seconds) of the prediction time, the longitude and latitude coordinates (accurate to six decimal places) of the center point of the target sea area, and the altitude (zero value for sea level). The core of the algorithm first calculates the Julian day, and then calculates a series of intermediate astronomical parameters such as the sun's ecliptic longitude, ecliptic inclination, and sun declination angle. Finally, two key angle parameters, the sun elevation angle α and the sun azimuth angle γ, are calculated. The sun elevation angle α is the angle between the sun's light and the horizontal plane, and the sun azimuth angle γ is the angle between the projection of the sun's light on the horizontal plane and the north direction, clockwise for positive. After obtaining the sun elevation angle α and the sun azimuth angle γ, the sun vector in the geographic coordinate system (usually the northeast celestial coordinate system, i.e., the X-axis points east, the Y-axis points north, and the Z-axis points to the zenith) is represented as a three-dimensional vector [Sx, Sy, Sz]. The specific calculation formula of the direction component is: Sx = cos(α) × sin(γ), Sy = cos(α) × cos(γ), and Sz = sin(α). All angle units in the calculation process are in radians, and the angle system needs to be converted to radians for calculation. The final obtained sun vector is a unit vector.

[0073] Converting the time-varying attitude angle sequence into the dynamic directional component of the photovoltaic panel normal vector in the geographic coordinate system requires establishing the conversion relationship between the platform coordinate system and the geographic coordinate system. Assuming that the photovoltaic panel is rigidly fixed on the floating photovoltaic platform, and the platform coordinate system is defined as follows: the origin is located at the center of gravity of the platform, the Xp axis points to the bow along the longitudinal direction of the platform, the Yp axis points to the starboard along the transverse direction of the platform, and the Zp axis is perpendicular to the deck upward. When the platform is stationary and horizontal, the platform coordinate system is aligned with the geographic coordinate system. The normal vector of the photovoltaic panel is initially defined as [0, 0, 1] in the platform coordinate system (i.e., perpendicular to the deck plane upward). When the time-varying attitude angle sequence (mainly including the roll angle θ and the pitch angle φ) acts on the platform, the platform coordinate system rotates relative to the geographic coordinate system. The roll angle θ is the rotation angle of the platform around the Xp axis (the longitudinal axis), and the starboard sinking is positive; the pitch angle φ is the rotation angle of the platform around the Yp axis (the transverse axis), and the bow sinking is positive. Through the rotation matrix, the normal vector of the photovoltaic panel is converted from the platform coordinate system to the geographic coordinate system. The conversion matrix is obtained by multiplying the rotation matrices of roll and pitch, which is: first rotate around the Xp axis by the roll angle θ, then rotate around the Yp axis by the pitch angle φ. The final dynamic directional component [Nx, Ny, Nz] of the photovoltaic panel normal vector in the geographic coordinate system can be calculated by matrix multiplication: Nx = -sin(φ), Ny = sin(θ)×cos(φ), Nz = cos(θ)×cos(φ). Since the time-varying attitude angle sequence changes with time, the calculated [Nx, Ny, Nz] is also a sequence of three-dimensional vectors that changes with time, representing the dynamic direction of the photovoltaic panel normal in space. Attention should be paid to the angle symbol and rotation order in the calculation to ensure correct conversion.

[0074] The cosine value of the effective solar light incident angle of the photovoltaic panel surface is dynamically solved by calculating the dot product of the solar vector and the normal vector of the photovoltaic panel. The mathematical principle is the cosine formula of the angle between two vectors in spatial analytic geometry. At any same time, the solar vector [Sx, Sy, Sz] and the normal vector of the photovoltaic panel [Nx, Ny, Nz] are expressed in the same geographical coordinate system and are unit vectors. The cosine value of the effective solar light incident angle θi of the photovoltaic panel surface at this moment is equal to the dot product (or scalar product) of the two unit vectors. The calculation formula of the dot product is: cos(θi) = Sx × Nx + Sy × Ny+ Sz × Nz. Since [Sx, Sy, Sz] and [Nx, Ny, Nz] are known time-varying sequences, the dot product value can be calculated for each sampling time (for example, every second), thereby obtaining a time sequence corresponding to the effective solar light incident angle cosine value sequence. This cosine value directly reflects the angle between the solar ray and the normal direction of the photovoltaic panel: when the cosine value is 1, it means that the light is perpendicular to the incident; when the cosine value is 0, it means that the light is parallel to the panel; when the cosine value is negative, it means that the sun is located on the back of the photovoltaic panel. In the actual calculation program, this dot product operation can be realized by looping through the vector components at each time point and performing multiplication and summation operations. The final effective solar light incident angle cosine value sequence is a key input parameter for subsequent radiation intensity calculation, and its dynamic characteristics accurately capture the changes in the received solar radiation conditions caused by platform motion.

[0075] S4, obtaining real-time cloud monitoring information of the target sea area, and comprehensively evaluating the comprehensive radiation attenuation degree under the joint action of the geometric attenuation caused by the wave and the atmospheric attenuation caused by the cloud, is implemented as:

[0076] The real-time cloud monitoring information of the target sea area is obtained through meteorological satellite remote sensing data, and its specific implementation depends on the multi-channel scanning radiometer data of geostationary meteorological satellites or polar-orbiting meteorological satellites. The satellite sensor observes the target sea area through the visible light channel and the infrared channel. The reflectivity data of the visible light channel is used to identify clouds during the day, and the brightness temperature data of the infrared channel is used to identify clouds day and night and estimate the cloud top height. The key parameter of cloud cover percentage is calculated by dividing the target sea area into a number of grids (for example, grids with a resolution of one kilometer), performing cloud detection identification on each grid point, and calculating the proportion of the number of grid points covered by clouds to the total number of grid points. The real-time cloud monitoring information is obtained in the form of digital raster data through the satellite data receiving station, and is preprocessed through atmospheric correction and geometric correction to ensure data accuracy. The cloud cover percentage value ranges from zero to one hundred, indicating the proportion of the area covered by clouds in the sky above the target sea area. This data is usually updated every ten minutes to half an hour to meet the real-time requirements of the prediction.

[0077] The conversion of the effective solar incident angle cosine value to the geometric attenuation coefficient of the solar radiation received by the photovoltaic panel surface caused by the wave is based on the Lambert cosine law. This conversion is directly performed because, in the ideal case, the direct solar radiation intensity received by the photovoltaic panel surface is directly proportional to the solar incident angle cosine value. Therefore, at any moment, the geometric attenuation coefficient Kgeo of the solar radiation received by the photovoltaic panel surface caused by the wave is directly taken as the effective solar incident angle cosine value cos(θi) at that moment, i.e. Kgeo = cos(θi). However, in practical applications, an actual situation needs to be considered: when the effective solar incident angle cosine value is less than or equal to zero, it indicates that the sunlight cannot shine on the front of the photovoltaic panel, at this time the geometric attenuation coefficient should be set to zero because the photovoltaic panel cannot receive direct radiation. Therefore, the complete conversion relationship is: Kgeo = max(0, cos(θi)). This conversion directly converts the geometric quantity (angle cosine value) of the platform motion into an efficiency coefficient that affects radiation reception, whose value varies between zero and one, and the larger the value, the more favorable the geometric conditions for radiation reception.

[0078] The corresponding relationship between the cloud cover percentage and the atmospheric transmission attenuation coefficient caused by the cloud layer is established by constructing an empirical lookup table. The establishment of this lookup table is based on the statistical analysis of long-term historical meteorological observation data and radiation observation data. The specific method is to collect long-term historical data of the target sea area or similar sea area with similar climate conditions, including high time resolution cloud cover percentage observation values and synchronous measurement of ground total solar radiation values and atmospheric top solar irradiance values. By calculating the ratio of the ground total solar radiation to the atmospheric top solar irradiance, the actual atmospheric transmission rate is obtained. Then, the cloud cover percentage is divided into several continuous intervals, such as 0%~10%, 10%~30%, 30%~50%, 50%~70%, 70%~90%, and 90%~100%. The average of all corresponding atmospheric transmission rate observation values in each cloud cover percentage interval is calculated to obtain the atmospheric transmission attenuation coefficient Katm represented by the cloud cover interval. For example, through statistical analysis, it may be determined that the cloud cover percentage 0%~10% corresponds to Katm of 0.85, and the cloud cover percentage 90%~100% corresponds to Katm of 0.15. This table containing cloud cover percentage intervals and unique atmospheric transmission attenuation coefficients corresponding to them constitutes a lookup table. In practical applications, according to the real-time obtained cloud cover percentage value, it is determined which pre-defined interval it falls into, and then the Katm value corresponding to this interval is directly read. This method avoids complex real-time atmospheric radiation transmission calculation and ensures evaluation efficiency.

[0079] The geometric attenuation coefficient is multiplied by the atmospheric transmission attenuation coefficient to obtain the comprehensive radiation attenuation degree, and the theoretical basis is that the weakening effect of the geometric attenuation caused by the wave and the atmospheric attenuation caused by the cloud layer on the solar radiation is a continuous process independent of each other. Therefore, the comprehensive radiation attenuation degree Ktotal is obtained by multiplying the geometric attenuation coefficient Kgeo and the atmospheric transmission attenuation coefficient Katm, that is, Ktotal = Kgeo × Katm. The multiplication operation means that the two attenuation effects are in series, and the solar radiation is first attenuated by the cloud layer in the atmosphere, and then is projected onto the surface of the photovoltaic panel due to the dynamic inclination angle and the effective receiving attenuation in geometry. The comprehensive radiation attenuation degree Ktotal is also a dimensionless coefficient between zero and one, and the numerical value comprehensively reflects the attenuation ratio of the actual solar radiation received on the surface of the photovoltaic panel relative to the ideal condition (the photovoltaic panel always faces the sun and there is no atmospheric attenuation) under the joint influence of the platform motion caused by the wave and the cloud coverage at a specific moment. This calculation needs to be performed synchronously for each prediction time, because Kgeo is a sequence that changes with time, and Katm needs to be updated at each time there is new cloud coverage data. Finally, a sequence of comprehensive radiation attenuation degrees that change with time is obtained, which quantitatively and dynamically represents the comprehensive influence of environmental factors on the reception of solar radiation, and provides a key input for the subsequent accurate calculation of the effective solar radiation intensity.

[0080] S5, based on the comprehensive radiation attenuation degree, correcting the total solar radiation data reaching the horizontal plane to generate a sequence of effective solar radiation intensity projected onto the surface of the photovoltaic panel, which is specifically implemented as:

[0081] The acquisition of the total solar radiation data reaching the horizontal plane is the basis for the correction. This data is usually obtained from two sources: one is the real-time measurement of the horizontal plane total solar radiation value by the meteorological station or the radiation monitoring station near the target sea area, and the other is the horizontal plane total solar radiation prediction data derived from the numerical weather prediction model combined with satellite remote sensing data. The total solar radiation includes direct radiation and scattered radiation, and the unit is watt per square meter. This data needs to be a time series data that is completely aligned in time with the sequence of comprehensive radiation attenuation degrees, that is, each data point corresponds to the same time stamp, and the sampling interval is consistent, for example, one data point per minute or one data point per hour. If the time resolution of the data source is inconsistent, time interpolation needs to be performed to ensure time alignment. This data represents the solar radiation energy received on the ideal horizontal plane without the influence of the photovoltaic panel posture and the specific cloud layer attenuation (the cloud layer attenuation here is part of the general atmospheric attenuation, which may be different from the cloud layer attenuation evaluated specifically, and needs to be distinguished or integrated). It is the benchmark value for calculating the radiation received on the surface of the photovoltaic panel.

[0082] The total solar radiation data reaching the horizontal plane at each moment is multiplied by the comprehensive radiation attenuation degree at the corresponding moment, which is the core operation of the correction process. The physical meaning of multiplication is to convert the radiation value of the ideal horizontal plane to the radiation value that the actual floating photovoltaic panel inclined surface can receive after considering the real-time cloud layer effect through a comprehensive attenuation coefficient. The specific operation is to traverse each time point of the time series, read the total solar radiation data value reaching the horizontal plane at that moment, denoted as Ghorizontal(t), and read the comprehensive radiation attenuation degree value at the same moment, denoted as Ktotal(t). Then multiply these two values: Geffective(t) = Ghorizontal(t) × Ktotal(t). This calculation must ensure the correct dimension, the unit of Ghorizontal(t) is watt per square meter, Ktotal(t) is a dimensionless coefficient, so the unit of the calculation result Geffective(t) is also watt per square meter. This operation is usually implemented by looping through the time series array in the computer program to ensure that the data at each moment is processed. It needs to be noted that if the comprehensive radiation attenuation degree Ktotal(t) is zero, it means that the photovoltaic panel cannot receive effective solar radiation at that moment, regardless of how strong the horizontal plane radiation is, the calculation result Geffective(t) is zero. This multiplication operation essentially applies both the geometric receiving efficiency loss caused by wave motion and the atmospheric transmission loss caused by clouds to the reference radiation value.

[0083] The result obtained after multiplication is used as the effective solar radiation intensity projected to the photovoltaic panel surface at the corresponding moment, which is a direct definition and assignment of the calculation result. Geffective(t) obtained after multiplication, its physical meaning has changed to the solar radiation intensity actually projected to the photovoltaic panel surface at a specific moment t, considering the two main influencing factors of floating platform attitude change (caused by waves) and actual cloud coverage (caused by clouds). This value is an instantaneous intensity value, reflecting the real lighting conditions at that moment. It is different from the horizontal plane radiation and the radiation that a static installed photovoltaic panel can receive, its core feature is to include dynamic attenuation effects caused by marine environment. This effective solar radiation intensity is the most direct and key input variable for subsequent photovoltaic power prediction, because it directly determines the strength of the energy source for the photovoltaic panel to produce photoelectric effect.

[0084] The effective solar radiation intensity at all times is aggregated in chronological order to generate an effective solar radiation intensity sequence, which is the process of organizing the instantaneous calculation results into a time series data format that can be used for model input. The aggregation operation refers to arranging the effective solar radiation intensity Geffective(t) corresponding to each time point t calculated above in chronological order into an ordered sequence or array. For example, if the future 24 hours are predicted with a sampling interval of 15 minutes, 96 Geffective(t) values will be obtained, arranged in chronological order into a sequence containing 96 elements. This sequence is the effective solar radiation intensity sequence, which completely describes the continuous change of solar radiation intensity projected onto the dynamically changing surface of the floating photovoltaic panel over a period of time in the future. When generating the sequence, the accuracy and continuity of the time stamp need to be ensured to avoid data missing or time misplacement. The final generated effective solar radiation intensity sequence accurately captures the influence of the unique operating environment of the offshore floating photovoltaic panel, laying a solid foundation for achieving high-precision power prediction.

[0085] S6, input the effective solar radiation intensity sequence into the pre-trained photovoltaic power generation power prediction model, and output the predicted power value of the offshore photovoltaic system, which is implemented as:

[0086] The establishment of the pre-trained photovoltaic power generation power prediction model is a prerequisite for the implementation of this method. The training of this model is based on a historical data set, which contains three key components: the first part is the historical effective solar radiation intensity sequence, which is obtained by processing historical data over a period of time through the aforementioned steps S1 to S5 in this patent method, ensuring the consistency of the training features and the input features during prediction; the second part is the synchronous recorded actual output power value sequence of the offshore photovoltaic system, which is obtained from the monitoring system of the photovoltaic power station, with the same sampling interval as the radiation intensity sequence; the third part is the environmental parameters that may affect the power generation efficiency, such as the temperature of the photovoltaic panel, which can be obtained from the historical records of the temperature sensor installed on the platform. The training process first needs to preprocess the historical data set, including handling missing values, smoothing abnormal fluctuations, and data standardization, so that data of different dimensions are in the same numerical range, facilitating model learning.

[0087] The photovoltaic power prediction model itself can be constructed by selecting one of various machine learning models, such as a linear regression model, a support vector regression model, or a gradient boosting decision tree model. Taking the gradient boosting decision tree model as an example, the training process is as follows: the historical effective solar radiation intensity sequence is taken as the main input feature, and time sequence features (such as hours, whether it is a holiday) and environmental features (such as photovoltaic panel temperature) can be introduced as auxiliary input features. The output target of the model is the actual output power value of the offshore photovoltaic system in a future preset period (for example, the next 24 hours). The model iteratively trains a series of decision trees, each tree learns to correct the prediction error of the previous tree, and finally the prediction results of all trees are weighted and summed to obtain the final power prediction value. During training, the model's hyperparameters need to be set, such as the maximum depth of the tree, the learning rate, the number of iterations, etc. These hyperparameters are adjusted on the validation set through cross-validation techniques to select the optimal combination, with the goal of minimizing the prediction error (such as mean absolute error) of the model on the validation set. The evaluation of the model after training is carried out on an independent test set to ensure its generalization ability.

[0088] Taking the effective solar radiation intensity sequence as the input feature of the pre-trained photovoltaic power prediction model means that when the model is put into actual prediction application, the newly generated effective solar radiation intensity sequence representing the future period is organized according to the format agreed upon during model training and input into the model. The input feature data needs to be standardized in the same way as the training data, that is, the mean and standard deviation calculated during the training phase are used to standardize the new data. In addition, other auxiliary feature data for the same period in the future, such as hour information, date type information, and predicted environmental temperature data, need to be prepared simultaneously. These auxiliary features and the effective solar radiation intensity sequence together constitute the complete input feature vector of the model. Ensuring that the format, dimension, and preprocessing method of the input data are completely consistent with the model training phase is the key to ensuring that the model runs correctly and produces reliable prediction results.

[0089] The effective solar radiation intensity sequence is processed by the photovoltaic power prediction model to obtain the predicted power value sequence of the offshore photovoltaic system in the future preset period, which is the calculation process of the model for forward reasoning. After receiving the organized input features, the internal algorithm of the model calculates the predicted power value at each sampling time point (for example, every 15 minutes in the next 24 hours) in the future preset period according to the mapping relationship learned in the training stage. For the gradient boosting decision tree model, this process is to let the input features pass through each decision tree in turn, make judgments and branch choices according to the feature values on the nodes of the tree, finally reach the leaf nodes and obtain the prediction value of the tree, then add up the prediction values of all trees, and then perform inverse normalization processing (convert the model output value back to the original power dimension) to obtain the final predicted power value. The output of this process is an ordered numerical sequence, i.e., the predicted power value sequence, each value in the sequence represents the model's estimate of the output power of the offshore photovoltaic system at a specific future time, usually in kilowatts or megawatts. The generated predicted power value sequence can ultimately be provided to the grid dispatch center or power station operator for developing power generation plans, conducting power market transactions, or assessing system operation status, thereby maximizing the value of the prediction information.

[0090] Embodiment 2: Figure 2 A structural diagram of a kind of offshore photovoltaic system output power prediction system of the application is given, a kind of offshore photovoltaic system output power prediction system, including the following modules:

[0091] The parameter acquisition module is used to obtain the real-time wave spectrum data of the target sea area and the platform structure parameters of the floating photovoltaic platform.

[0092] The sequence generation module is used to determine the dominant motion mode of the floating photovoltaic platform based on the real-time wave spectrum data and the platform structure parameters, and generate the time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves based on the motion law corresponding to the dominant motion mode.

[0093] The cosine solving module is used to geometrically couple the time-varying attitude angle sequence with the solar vector determined by the solar position astronomical algorithm, and dynamically solve the effective solar light incident angle cosine value on the surface of the photovoltaic panel.

[0094] The degree evaluation module is used to obtain real-time cloud monitoring information of the target sea area, and comprehensively evaluate the comprehensive radiation attenuation degree under the joint action of geometric attenuation caused by waves and atmospheric attenuation caused by clouds based on the effective solar light incident angle cosine value and the real-time cloud monitoring information.

[0095] The data correction module is used to correct the total solar radiation data reaching the horizontal plane based on the comprehensive radiation attenuation degree to generate the effective solar radiation intensity sequence projected onto the surface of the photovoltaic panel.

[0096] The power output module is configured to input the effective solar radiation intensity sequence into a pre-trained photovoltaic power generation power prediction model, and output a predicted power value of the offshore photovoltaic system.

[0097] The calculations involved in the embodiments are all dimensionless numerical calculations, and preset parameters and threshold values in the calculations are set by a person skilled in the art according to actual conditions.

[0098] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0099] Those skilled in the art can appreciate that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and implementation constraints of the technology. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0100] In addition, the functional modules in each of the embodiments of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0101] In the several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the modules is only a logical function division. In actual implementation, another division mode can be used, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the modules shown or discussed can be indirect coupling or communication connection through some interface, device or module, and can be electrical, mechanical or other forms.

[0102] The above description is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0103] Finally, the above description is only the preferred embodiment of the present application, and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for offshore photovoltaic system output power prediction, characterized in that, The method comprises the following steps: S1, obtaining real-time wave spectrum data of a target sea area and platform structure parameters of a floating photovoltaic platform; S2, determining a current dominant motion mode of the floating photovoltaic platform according to the real-time wave spectrum data and the platform structure parameters, and generating a time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves based on a motion law corresponding to the dominant motion mode; S3, geometrically coupling the time-varying attitude angle sequence with a sun vector determined by a sun position astronomical algorithm to dynamically solve the effective sun light incident angle cosine value of the surface of a photovoltaic panel; S4, obtaining real-time cloud layer monitoring information of the target sea area, and comprehensively evaluating the degree of comprehensive radiation attenuation under the joint action of geometric attenuation caused by waves and atmospheric attenuation caused by clouds based on the effective sun light incident angle cosine value and the real-time cloud layer monitoring information; S5, correcting the total solar radiation data reaching the horizontal plane based on the degree of comprehensive radiation attenuation to generate an effective solar radiation intensity sequence projected onto the surface of the photovoltaic panel; S6, inputting the effective solar radiation intensity sequence into a pre-trained photovoltaic power generation power prediction model to output a predicted power value of the offshore photovoltaic system.

2. A method of predicting the output power of a marine photovoltaic system according to claim 1, characterized in that, The method comprises the following steps: Obtaining real-time wave spectrum data of a target sea area and platform structure parameters of a floating photovoltaic platform, comprising: Obtaining real-time wave spectrum data of a target sea area through a marine monitoring satellite or a buoy measuring device; 3. A method of predicting output power of a marine photovoltaic system according to claim 1, characterized in that, Obtaining platform structure parameters from design data of the floating photovoltaic platform, wherein the platform structure parameters include platform size, platform mass distribution, and platform draft depth. Determining a current dominant motion mode of the floating photovoltaic platform according to the real-time wave spectrum data and the platform structure parameters, and generating a time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves based on a motion law corresponding to the dominant motion mode, comprising: Extracting a characteristic period and a spectral peak period of the real-time wave spectrum data; Matching and comparing the characteristic period and the spectral peak period with a platform roll natural period and a pitch natural period calculated based on the platform size and the platform mass distribution; Determining whether the roll is the dominant motion mode or the pitch is the dominant motion mode according to the matching and comparison result; Selecting a corresponding historical roll attitude angle sequence or a historical pitch attitude angle sequence as a sample based on the determined dominant motion mode; 4. The method of claim 1, wherein, Fitting the motion law of the sample by a time series analysis method to extrapolate a time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves. Geometrically coupling the time-varying attitude angle sequence with a sun vector determined by a sun position astronomical algorithm to dynamically solve the effective sun light incident angle cosine value of the surface of a photovoltaic panel, comprising: Calculating the direction component of the sun vector in the geographical coordinate system based on the sun position astronomical algorithm; Converting the time-varying attitude angle sequence into a dynamic direction component of the normal vector of the photovoltaic panel in the geographical coordinate system; Dynamically solving the effective sun light incident angle cosine value of the surface of the photovoltaic panel by calculating the dot product of the sun vector and the normal vector of the photovoltaic panel.

5. A method of predicting the output power of a marine photovoltaic system according to claim 4, characterized in that, The effective solar light incident angle cosine value of the photovoltaic panel surface is dynamically solved by calculating the dot product of the solar vector and the normal vector of the photovoltaic panel, that is, the solar vector and the normal vector of the photovoltaic panel are expressed as three-dimensional vectors in the same geographic coordinate system, and the effective solar light incident angle cosine value is directly solved through the dot product calculation formula of the two three-dimensional vectors.

6. The method of claim 1, wherein, Real-time cloud monitoring information of the target sea area is obtained, and the comprehensive radiation attenuation degree under the joint action of the geometric attenuation caused by waves and the atmospheric attenuation caused by clouds is evaluated by comprehensively considering the effective solar light incident angle cosine value and the real-time cloud monitoring information, including: The real-time cloud monitoring information of the target sea area is obtained through meteorological satellite remote sensing data, and the real-time cloud monitoring information includes the cloud cover percentage; The effective solar light incident angle cosine value is converted into the geometric attenuation coefficient of the photovoltaic panel surface receiving solar radiation caused by waves; A corresponding relationship between the cloud cover percentage and the atmospheric transmission attenuation coefficient caused by clouds is established; The geometric attenuation coefficient and the atmospheric transmission attenuation coefficient are multiplied to obtain the comprehensive radiation attenuation degree.

7. A method of predicting the output power of a marine photovoltaic system according to claim 6, characterized in that, The corresponding relationship between the cloud cover percentage and the atmospheric transmission attenuation coefficient caused by clouds is established by pre-establishing a lookup table containing different cloud cover percentage intervals and their corresponding atmospheric transmission attenuation coefficients, and the cloud cover percentage in the real-time cloud monitoring information is converted into a specific atmospheric transmission attenuation coefficient by querying the lookup table.

8. The method of claim 1, wherein, Based on the comprehensive radiation attenuation degree, the total solar radiation data reaching the horizontal plane is corrected to generate an effective solar radiation intensity sequence projected onto the photovoltaic panel surface, including: The total solar radiation data reaching the horizontal plane at each time is multiplied by the comprehensive radiation attenuation degree at the corresponding time; The result obtained after multiplication is taken as the effective solar radiation intensity projected onto the photovoltaic panel surface at the corresponding time; The effective solar radiation intensities at all times are sequentially collected to generate an effective solar radiation intensity sequence.

9. The method of claim 1, wherein, The effective solar radiation intensity sequence is input into the pre-trained photovoltaic power generation power prediction model to output the predicted power value of the offshore photovoltaic system, including: The effective solar radiation intensity sequence is taken as the input feature of the pre-trained photovoltaic power generation power prediction model; The photovoltaic power generation power prediction model processes the effective solar radiation intensity sequence to obtain a predicted power value sequence of the offshore photovoltaic system in a future preset time period.

10. A system for predicting the output power of a marine photovoltaic system, for implementing a method for predicting the output power of a marine photovoltaic system according to any one of claims 1 to 9, characterized in that, It includes the following modules: A parameter acquisition module is configured to acquire real-time wave spectrum data of a target sea area and platform structure parameters of a floating photovoltaic platform; A sequence generation module is configured to determine a dominant motion mode of the floating photovoltaic platform based on the real-time wave spectrum data and the platform structure parameters, and generate a time-varying attitude angle sequence of the floating photovoltaic platform under the action of waves based on the motion law corresponding to the dominant motion mode; A cosine solving module is configured to geometrically couple the time-varying attitude angle sequence with a solar vector determined by a solar position astronomical algorithm to dynamically solve an effective solar light incident angle cosine value of a photovoltaic panel surface; A degree evaluation module is configured to obtain real-time cloud monitoring information of the target sea area, and evaluate a comprehensive radiation attenuation degree under the joint action of geometric attenuation caused by waves and atmospheric attenuation caused by clouds by comprehensively considering the effective solar light incident angle cosine value and the real-time cloud monitoring information. A data correction module is configured to correct the total solar radiation data reaching the horizontal plane based on the comprehensive radiation attenuation degree, and generate an effective solar radiation intensity sequence projected onto the surface of the photovoltaic panel; A power output module is configured to input the effective solar radiation intensity sequence into a pre-trained photovoltaic power generation power prediction model, and output a predicted power value of the offshore photovoltaic system.

Images