Reliability design method of water surface photovoltaic foundation under extreme environmental load

Abstract

This invention discloses a reliability design method for a water surface photovoltaic (PV) base under extreme environmental loads, comprising: acquiring time-series data of environmental driving variables of the target water area, including at least wind speed time-series data and operating water level time-series data; based on the time-series data, deterministically generating wave characteristic parameters using a shallow water wave forecasting physical model; constructing a dynamic response model based on the wave characteristic parameters and time-series data to obtain the multi-degree-of-freedom mechanical response of the base and mooring system; and performing Monte Carlo simulation accelerated by a surrogate model based on a pre-constructed joint probability distribution and multi-degree-of-freedom mechanical response to obtain reliability indices. This invention achieves physical dimensionality reduction of environmental variables, improving the accuracy of probabilistic assessment of the mechanical system and the efficiency of engineering calculations.

Description

Technical Field

[0001] This invention relates to the field of engineering design technology for water surface photovoltaic bases, and in particular to a reliability design method for water surface photovoltaic bases under extreme environmental loads. Background Technology

[0002] In the design and construction of floating photovoltaic power plants, the base structure, as the key carrier supporting the photovoltaic modules, directly determines the safe operation and service life of the entire power generation system. Load assessment and mechanical response prediction of the floating photovoltaic base help improve the project's disaster prevention and mitigation capabilities, optimize structural selection, and control project costs.

[0003] Currently, the design and evaluation of surface photovoltaic (PV) bases typically draws upon traditional marine engineering standards and methods. In practice, wind loads, wave loads, and current loads are often treated as independent stochastic environmental factors. Design parameters are generated for each based on its respective extreme value distribution model, and then combined and superimposed using preset empirical constants (such as fixed correlation coefficients). For dynamic calculations, classical fluid dynamics equations, such as the Morrison equation, are typically used to evaluate the stress on slender tubular components based on fixed design water depth boundary conditions, and direct finite element simulation is employed to assess the structural safety margin.

[0004] However, under extreme environments and other special physical boundary conditions, traditional design methods face problems such as systematic biases in load prediction and limited efficiency in engineering optimization iterations. Therefore, further research and innovation are needed to solve these problems in existing technologies. Summary of the Invention

[0005] Purpose of the invention: To provide a reliability design method for water surface photovoltaic bases under extreme environmental loads, in order to solve the above-mentioned problems existing in the prior art.

[0006] Technical solution: Firstly, a reliability design method for water surface photovoltaic bases under extreme environmental loads, comprising:

[0007] Acquire time-series data of environmental driving variables for the target water area. The time-series data of environmental driving variables should include at least wind speed time-series data and operating water level time-series data.

[0008] Based on wind speed time series data and operating water level time series data, wave characteristic parameters are deterministically generated using a shallow water wave forecasting physical model.

[0009] Based on wave characteristic parameters, wind speed time series data, and operating water level time series data, a dynamic response model of the water surface photovoltaic base is constructed to obtain the multi-degree-of-freedom mechanical response of the base and mooring system.

[0010] Based on the joint probability distribution of time-series data of environmental driving variables and the multi-degree-of-freedom mechanical response of the base and mooring system, Monte Carlo simulation accelerated by surrogate model is performed to obtain the reliability index of the water surface photovoltaic base.

[0011] An apparatus comprising at least a water-surface photovoltaic base, which is arranged and formed using the construction method described in any of the first aspects.

[0012] Beneficial effects: The independent random variables of wind and waves are reduced to a deterministic physical coupling of wind speed and water level, superimposed with a forced truncation mechanism of shallow water wave breaking and time constraints. This helps to suppress the systematic blind spots caused by artificially set empirical correlation coefficients, and improves the prediction of wave loads under extreme conditions from a rough statistical fitting to a physical inference that conforms to the causal laws of hydrodynamics, thus enhancing the physical authenticity of the source environment input. Attached Figure Description

[0013] Figure 1 A flowchart illustrating a reliability design method for a water surface photovoltaic base under extreme environmental loads, as provided in this application embodiment.

[0014] Figure 2 The flowchart provided in this application describes the deterministic generation of wave characteristic parameters based on wind speed time series data and operating water level time series data, using a shallow water wave forecasting physical model.

[0015] Figure 3 A flowchart illustrating the physical boundary verification mechanism provided in this application embodiment.

[0016] Figure 4 This is a flowchart for verifying wave characteristic parameters by using the shallow water wave breaking limit, as provided in an embodiment of this application.

[0017] Figure 5 A flowchart illustrating the cascade mechanical analysis of the anchoring component provided in this application embodiment.

[0018] Figure 6 A flowchart of a reliability design method for a water surface photovoltaic base under extreme environmental loads is shown. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] To address the aforementioned issues, the applicant conducted in-depth searches and analyses, and discovered:

[0022] Existing technologies fail to adequately characterize the endogenous dynamic relationships of environmental factors during the formation of enclosed water bodies, and neglect the reshaping effect of wide fluctuations in environmental water levels on the structural stress characteristic boundaries. Existing methods are prone to systematic biases in load prediction under variable working conditions.

[0023] When dealing with complex structures, direct numerical simulation evaluation methods struggle to balance computational cost with multi-dimensional state discrimination, thus limiting the efficiency of engineering optimization iterations.

[0024] To solve these problems, combined with Figures 1 to 5 The present invention will be specifically described through the following embodiments.

[0025] In this invention, some parameters are explained as follows:

[0026] For example, the correlation join function can be implemented using the Copula join function;

[0027] The Copula function can connect multiple random variables with arbitrary marginal distributions into a multidimensional joint distribution while preserving the nonlinear dependency structure between variables.

[0028] For example, the nonlinear smooth transition function can be implemented using a continuously differentiable function, such as the sigmoid function or a higher-order polynomial transition function; the appropriate form of the transition function can be selected according to the actual state switching boundary width and numerical stability requirements.

[0029] In the following text, d _local This represents the actual local water depth, which can be uniformly represented by d after dimensionless conversion. _bar It indicates dimensionless water depth.

[0030] The location of the photovoltaic base on the water surface is also known as the base installation location.

[0031] Floating photovoltaic bases typically take the form of rectangular or polygonal floating boxes with a large horizontal span.

[0032] The target water area refers to reservoirs, lakes, and sea areas where photovoltaic power plants are planned or have already been built.

[0033] Environmental driving variable time series data are continuous environmental observation records or hydrological and meteorological forecast data carrying timestamp sequences.

[0034] The shallow water wave forecasting physical model has a built-in computational logic module for the shallow water wave generation and evolution dynamics equations.

[0035] Wave characteristic parameters refer to representative physical quantities that can fully describe the statistical characteristics of irregular waves, including at least the significant wave height and the spectral peak period; the significant wave height reflects the magnitude of the wave energy, and the spectral peak period reflects the frequency component where the wave energy is most concentrated.

[0036] The following are multiple degrees of freedom, such as swaying, horizontal swaying, vertical swaying, horizontal rocking, vertical rocking, and head rocking.

[0037] The candidate sample pool represents a numerical set containing a large number of discrete combinations of environmental states.

[0038] The Kriging surrogate model corresponds to an interpolation algorithm based on Gaussian process regression.

[0039] Active learning evaluation criteria are used to selectively identify samples from the candidate sample pool that are valuable for characterizing the failure boundary classification.

[0040] Parameter sensitivity is the numerical partial derivative or gradient vector of the limit state response value with respect to each physical design parameter.

[0041] The location parameter reflects the central location of the extreme value distribution and is set to evolve linearly over time.

[0042] The scale parameter reflects the degree of dispersion of the extreme value distribution. To ensure the non-negativity of this parameter due to physical constraints, it is set to evolve exponentially over time.

[0043] The Akaike Information Criterion is used to sum the negative logarithm of the model fit likelihood with a penalty term for the number of model parameters.

[0044] As an optional approach, the wave breakage attenuation coefficient can be limited to a range of 0.45-0.78, which can be specifically set by the on-site engineer based on the actual working conditions, and no limitation is imposed on it. Other numerical examples follow the same principle.

[0045] Firstly, a reliability design method for water surface photovoltaic bases under extreme environmental loads mainly includes the following steps:

[0046] Step 101: Obtain time-series data of environmental driving variables for the target water area. The time-series data of environmental driving variables shall include at least wind speed time-series data, wind direction data, and operating water level time-series data.

[0047] Accordingly, this embodiment breaks away from the model of inputting multiple independent environmental factors such as wind, waves, and currents separately. Instead, it reduces the environmental driving variables to two dimensions—wind speed and water level—based on the physical characteristics of the target water body, whether on land or at sea. In the target water body, waves are mainly directly generated by local winds, and their growth and decay are strongly constrained by the local water depth.

[0048] Furthermore, wind speed time-series data is typically obtained through meteorological stations or wind measurement towers located near the target water area, or through inversion based on satellite remote sensing. After data acquisition, basic quality control and height correction are usually required.

[0049] For example, when the installation height of the anemometer is not the standard 10m height, the wind speed at the observation height can be converted to the wind speed at the standard height using the logarithmic wind speed profile principle.

[0050] For example, the operating water level time series data is usually provided by the hydrological monitoring station of the reservoir management department, or obtained by real-time measurement by water level gauges;

[0051] This data reflects the process of absolute elevation changes in reservoirs under the influence of scheduling rules, such as flood discharge during the flood season, water storage during the dry season, or natural hydrological cycles.

[0052] The data can be organized in comma-separated CSV tables or the common network data format NetCDF. The sampling frequency and duration must meet the requirements of subsequent reliability extreme value analysis, such as covering daily or hourly data sequences of several years.

[0053] Step 102: Based on wind speed time series data and operating water level time series data, wave characteristic parameters are deterministically generated using a shallow water wave forecasting physical model.

[0054] In this step, hydrodynamic physical constraints can be used to establish a deterministic mapping between wind and waves.

[0055] In this embodiment, the shallow water wave forecasting physical model uses wind speed time series data as the energy source driving wave growth, and uses the local water depth calculated from the operating water level time series data as the energy dissipation boundary limiting wave growth.

[0056] By inputting two time-series variables, the model internally performs the solution of the wave energy balance equation or the derivation based on empirical physical formulas.

[0057] Furthermore, generating wave characteristic parameters can transform wave height and period, which originally needed to be input as independent random variables, into dependent variables of wind speed and water level, reducing the random dimension of the environmental system and facilitating subsequent probability assessment.

[0058] Step 103: Based on wave characteristic parameters, wind speed time series data, and operating water level time series data, construct a dynamic response model of the water surface photovoltaic base and obtain the multi-degree-of-freedom mechanical response of the base and mooring system.

[0059] Alternatively, based on wave characteristic parameters, wind speed time series data, and operating water level time series data, a dynamic response model of the water surface photovoltaic base is constructed, and the multi-degree-of-freedom mechanical response of the base and mooring system is obtained based on the dynamic response model.

[0060] After obtaining the wind speed, water level, and wave parameters derived from them, it is necessary to evaluate the specific effects of the load on the physical structure.

[0061] Among them, the dynamic response model is a numerical analysis framework used to describe the interaction between fluids and structures, as well as the stiffness and damping characteristics of the structure itself.

[0062] In practical applications, this model can be constructed based on potential flow theory solvers or computational fluid dynamics methods.

[0063] In this model, wind speed time series data is converted into aerodynamic loads acting on the photovoltaic panels and the freeboard portion of the base; wave characteristic parameters, combined with the current operating water level, are converted into hydrodynamic loads acting on the underwater portion of the base.

[0064] Among them, the multi-degree-of-freedom mechanical response of the base and mooring system corresponds to the motion parameters of the base body in space with multiple degrees of freedom, such as displacement, velocity, and acceleration, which are excited by the combined load, as well as the structural internal forces or tensions borne by the various components of the mooring system that are moored on the bottom of the water to limit the drift of the base.

[0065] By combining the three types of input parameters—wave, wind, and water—the model can reproduce the severe stress state of a structure under conditions such as low water levels and strong winds and waves, and output the corresponding numerical response results.

[0066] Step 104: Based on the joint probability distribution of pre-constructed environmental driving variable time series data and the multi-degree-of-freedom mechanical response of the base and mooring system, perform Monte Carlo simulation accelerated by surrogate model to obtain the reliability index of the water surface photovoltaic base.

[0067] Directly calling the dynamic response model to perform full-scale finite element or hydrodynamic calculations for every possible combination of extreme loads would be computationally prohibitive in engineering.

[0068] Correspondingly, the pre-constructed joint probability distribution, which is obtained by pre-fitting historical long-sequence data in the offline stage, is used to reflect the statistical dependence structure of wind speed and water level, for example, strong winds are often accompanied by high water levels during the flood season.

[0069] Monte Carlo simulation is a numerical method for assessing the probability of structural failure by performing large-scale random sampling in the joint probability space.

[0070] Optionally, this step employs a surrogate model acceleration mechanism to address the computational cost issue, specifically including:

[0071] An approximate model, i.e., a proxy model, is constructed using a small number of samples obtained from calculations of the dynamic response model based on actual calls.

[0072] This proxy model is used to replace the real physical model for rapid prediction of Monte Carlo sampling points.

[0073] During the iteration process, the real dynamic response model is called to calculate and the surrogate model is corrected only for a very small number of key samples that the surrogate model considers to have the highest classification uncertainty and the most valuable characterization of the failure surface.

[0074] Based on this, the output reliability index can be characterized as failure probability or reliability index, which is used to quantify the ability of the water surface photovoltaic base to resist extreme environmental damage during its design life, thereby providing a basis for judging the compliance of engineering design specifications.

[0075] In one possible implementation, before deterministically generating wave characteristic parameters based on wind speed time-series data and operating water level time-series data using a shallow water wave forecasting physical model, the following steps are also included:

[0076] Step 201: Obtain the pre-stored digital elevation model of the target water area.

[0077] Correspondingly, a digital elevation model is a set of digitized matrices containing three-dimensional spatial elevation data of the target water basin or reservoir.

[0078] The model is pre-mapped and stored in a database readable by computing devices during the offline phase. Its spatial resolution is usually determined based on the physical area of ​​the target water body to reflect the real topographic relief.

[0079] Step 202: Based on the time series data of the operating water level and the digital elevation model, extract the dynamic water surface boundary corresponding to the current operating water level.

[0080] In this embodiment, the absolute elevation of the target water area will fluctuate due to artificial scheduling or natural hydrological changes. When the operating water level drops, the boundary between land and water shrinks towards the center of the reservoir area.

[0081] Among them, the action of extracting dynamic water surface boundaries is to traverse and search the set of contour line coordinates in the digital elevation model whose elevation values ​​are equal to the current operating water level, in order to capture the time-varying physical characteristics of the water surface area.

[0082] Step 203: Based on the dynamic water surface boundary and the pre-configured base installation position, calculate the distance along the wind direction ray corresponding to the wind direction data to obtain the length of the dynamic effective wind zone, specifically including:

[0083] Multiple discrete rays are emitted from the base installation location as the origin and the prevailing wind direction determined by wind direction data as the center, within a preset angle range on both sides.

[0084] The base installation position is represented by a two-dimensional coordinate point (x, y) in a plane absolute coordinate system;

[0085] The prevailing wind direction can be determined by the azimuth angle of the wind direction configured from the same source as the wind speed time series data;

[0086] The preset angle range can be set to an extended water area symmetrical on both sides of the prevailing wind direction.

[0087] In some alternative implementations, the angular range of the extended region can be limited to ±45° from the prevailing wind direction. Within this preset angular range, multiple radial ray paths are generated according to a fixed angular step size.

[0088] Obtain the actual projected distance of each discrete ray to the dynamic water surface boundary.

[0089] Furthermore, the geometric intersection points of each discrete ray and the dynamic water surface boundary are obtained within the computational domain.

[0090] For example, the Euclidean distance from the base installation position along each ray direction to the corresponding geometric intersection is calculated; this distance, in fluid mechanics, characterizes the physical length of the airflow that can continuously blow across the water surface and transfer energy to the water body under a specific wind direction branch.

[0091] Furthermore, by using a preset angle weighting algorithm related to the azimuth angles of each of the multiple discrete rays, the multiple actual projected distances are weighted and fused to obtain the dynamic effective wind zone length.

[0092] Accordingly, considering that the energy contribution of water areas deviating from the prevailing wind direction to wave growth is physically attenuated, it is advisable to use the distribution function to perform discrete weighted summation calculation of the actual projected distances of multiple rays.

[0093] In some embodiments, F _eff =∑(r _i ×cos(θ _i -θ _w ) 2) / ∑(cos(θ _i -θ _w ) 2 );

[0094] Among them, F _eff The dynamic effective wind zone length quantifies the nonlinear physical effect of water level drop causing local water surface contraction, which in turn inhibits wave growth. ∑ is a mathematical operator that performs discrete summation on all ray indices within a preset angle interval, and r _i Let θ be the actual projected distance of the i-th discrete ray. _i Let θ be the azimuth angle of the i-th discrete ray. _w The azimuth of the prevailing wind direction.

[0095] In some embodiments, wave characteristic parameters are deterministically generated based on wind speed time-series data and operating water level time-series data using a shallow water wave forecasting physical model, specifically including:

[0096] Optionally, the local water depth at the base installation location can be determined based on the operating water level time series data and the digital elevation model.

[0097] Accordingly, the absolute bottom elevation data of the digital elevation model at the base installation location is read, and the current operating water level value is subtracted from the absolute bottom elevation data to calculate the physical water column depth below the float, i.e., the local water depth.

[0098] Furthermore, the wind speed time series data is converted into the corresponding wind stress factor;

[0099] Among them, the wind stress factor is used to correct the nonlinear drag enhancement phenomenon of the air-sea interface in real physical scenarios.

[0100] For example, since the higher the wind speed, the higher the sea surface roughness, the conversion process satisfies the following exponential relationship:

[0101] U _A =0.71×(U _10 ) 1.23 ;

[0102] Among them, U _A U is the wind stress factor, 0.71 is an empirical constant. _10 The wind speed time series data is corrected to a standard height of 10m for the observation height, and 1.23 is a dimensionless exponential parameter.

[0103] In another example, the formula calculation of the present invention can also use only dimensionless numerical values ​​for derivation, without introducing physical unit constraints.

[0104] According to one aspect of this application, in the process of inputting wind stress factor, local water depth, and dynamic effective wind zone length into a shallow water wave forecast physical model for nonlinear physical coupling calculation to obtain wave characteristic parameters, a physical boundary verification mechanism is also included, namely:

[0105] Accordingly, the duration of wind speed corresponding to the wind speed time series data is obtained;

[0106] Among them, wind speed duration corresponds to the time statistics extracted from meteorological records, which is used to characterize the duration for which a specific extreme wind speed event maintains stable blowing in the atmospheric boundary layer.

[0107] Furthermore, based on the wind stress factor and the length of the dynamic effective wind zone, the shortest development time required for waves to reach a steady state is determined.

[0108] The wave energy needs to undergo a physical evolution process to reach absorption saturation in a given scale of water. The logic for calculating this time boundary satisfies the following formula:

[0109] t _min =(U _A / g)×62.38×(g×F _eff / (U _A ) 2 ) 0.667 ;

[0110] Among them, t _min To achieve the shortest development time, U _A Where is the wind stress factor, g is the gravitational acceleration constant, 62.38 is a constant coefficient, and F... _eff 0.667 represents the dynamic effective wind zone length, and 0.667 is the dimensionless exponential coefficient.

[0111] In some scenarios, when the duration of wind speed is less than the shortest development time, the equivalent wind zone length is calculated using the duration of wind speed, and the equivalent wind zone length is used to replace the dynamic effective wind zone length in the shallow water wave forecast physical model.

[0112] When the duration of wind speed is greater than or equal to the shortest development time, the length of the dynamic effective wind zone remains unchanged.

[0113] After obtaining the wave characteristic parameters, the shallow water wave breaking limit is determined based on the local water depth;

[0114] The wave characteristic parameters were truncated and verified using the shallow water wave breaking limit, resulting in truncated wave characteristic parameters.

[0115] In this embodiment, the extreme wind speeds in the target water area are mostly induced by short-duration severe convective weather. Assuming that after dimensionless normalization, the duration of the current dimensionless wind speed is less than the calculated minimum dimensionless development time, the current operating condition is determined to be time-constrained. At this time, the flow field cannot fully capture wind energy, and a shorter equivalent wind zone length needs to be calculated based on the actual blowing duration.

[0116] Accordingly, t _min The formula is inversely solved by substituting the actual wind speed duration into the equivalent wave development time constraint to calculate the corresponding equivalent wind zone length. The calculation of the equivalent wind zone length satisfies the following formula:

[0117] F _eq =((U _A ) 2 / g)(t _actual ×g / (62.38×U _A )) 1 / 0.667 ;

[0118] In the formula, F _eq t is the equivalent wind zone length. _actual This represents the actual duration of the wind speed.

[0119] Conversely, if the duration of wind speed is not less than the shortest development time, the calculated dynamic effective wind zone length is directly retained.

[0120] Furthermore, the wind stress factor, local water depth, and dynamic effective wind zone length are input into the shallow water wave forecast physical model for nonlinear physical coupling calculation to obtain wave characteristic parameters, which include at least the meaningful wave height and spectral peak period.

[0121] Correspondingly, the physical model for shallow water wave forecasting integrates a two-way coupled attenuation mechanism. Specifically, the reduction in local water depth directly exacerbates energy frictional dissipation at the bottom of shallow water and leads to a shortening of the dynamic effective wind zone length.

[0122] For example, the model can use the hyperbolic tangent function to perform an asymptotic mapping calculation on the input parameters, that is:

[0123] H _s =((U _A ) 2 / g)×0.283×tanh(0.530×(d _bar ) 0.75 )×tanh(0.00565×(F _bar ) 0.5 / tanh(0.530×(d _bar ) 0.75 ));

[0124] Among them, H_s Let d be the meaningful wave height, tanh be the hyperbolic tangent operator, and d be the tangent function. _bar F is the dimensionless water depth obtained based on the transformation of local water depth. _bar The lengths are dimensionless wind zone lengths derived from the transformation of dynamic effective wind zone length. 0.283, 0.530 and 0.00565 are all dimensionless constant coefficients.

[0125] Based on this, the spectral peak period is calculated synchronously using a unified physical dimensionality reduction logic framework.

[0126] In some embodiments, the wave characteristic parameters are truncated and verified using the shallow water wave breaking limit, specifically including: extracting the design wave height under extreme conditions based on the meaningful wave height and pre-configured statistical exceedance probability in the wave characteristic parameters.

[0127] Furthermore, extreme value response indices are extracted from the structural reliability assessment. Based on the Rayleigh statistical distribution law, the meaningful wave height is multiplied by the coefficient factor calculated from the statistical exceedance probability.

[0128] For example, when the statistical exceedance probability is set to 1%, the extracted design wave height is approximately 1.52 times the meaningful wave height.

[0129] Furthermore, the shallow water wave breaking limit is set as a physical cutoff threshold that is strictly positively correlated with the local water depth;

[0130] Accordingly, when the design wave height is less than or equal to the physical cutoff threshold, the design wave height is kept unchanged to obtain a truncated design wave height for use as load input in the dynamic response model. This physical cutoff threshold establishes a mandatory upper energy limit boundary in shallow water flow fields.

[0131] As an example, the calculation process can be as follows:

[0132] H _threshold =γ _b ×d _local ;

[0133] Among them, H _threshold γ is the physical cutoff threshold. _b d is the wave breaking attenuation coefficient. _local This refers to the local water depth.

[0134] Accordingly, when the design wave height is greater than the physical cutoff threshold, the design wave height is updated to the physical cutoff threshold.

[0135] It should be understood that the above mechanism prevents the physical model from outputting giant wave data that violates the limits of hydrodynamics under extreme conditions of low water level and strong wind.

[0136] Based on the above embodiments, a dynamic response model of the water surface photovoltaic base is constructed to obtain the multi-degree-of-freedom mechanical response of the base and mooring system, including:

[0137] Step 301: Based on the operating water level time series data and the pre-stored bottom elevation data of the location of the water surface photovoltaic base, calculate the local water depth at the location of the water surface photovoltaic base.

[0138] That is, by combining the absolute bottom elevation data of the target water area, the local water depth is obtained by subtracting the absolute bottom elevation data from the current operating water level value.

[0139] Local water depth not only limits the growth limit of preceding waves, but also directly determines the boundary of the kinematic properties of water particles in fluid dynamics calculations.

[0140] Step 302: Based on the large-size flat structure characteristics represented by the pre-configured structural geometric model of the water surface photovoltaic base, the boundary element method is used to solve the diffraction potential and radiation potential of the surrounding flow field.

[0141] Correspondingly, when the ratio of the characteristic width of the base to the incident wavelength is greater than a preset ratio, the obstruction and reflection effects of the structure on the wave flow field cannot be ignored. At this time, the flow field velocity potential is decomposed into incident potential, diffraction potential, and radiation potential generated by the base motion.

[0142] For example, the preset ratio can be set to 0.2 here.

[0143] Correspondingly, a low-order constant boundary element method can also be used to discretize the wet surface of the photovoltaic substrate into multiple planar quadrilateral elements, with a constant source intensity set at the center of each element. Furthermore, by simultaneously solving the linear equations for the source intensity under the continuous boundary conditions of the normal velocity of each element, the diffraction potential and radiation potential at each frequency can be obtained.

[0144] In another embodiment, when the water surface photovoltaic base adopts a slender floating tube splicing structure and the ratio of the characteristic width to the incident wavelength is less than or equal to 0.2, the fluid diffraction effect decays.

[0145] As an alternative, instead of solving for the diffraction and radiation potentials, the drag force and inertial force on a unit length fluid element can be calculated directly using a preset empirical formula.

[0146] Step 303: Calculate the six-degree-of-freedom wave excitation force, including the vertical force component, based on the diffraction potential, radiation potential, local water depth, and wave characteristic parameters.

[0147] That is, the total velocity potential is integrated over discrete elements on the wetted surface using the fluid dynamics governing equations. The calculation results not only include the horizontal sway and roll excitation forces, but also directly output the vertical wave force and the yaw, roll, and pitch moment components, which have an important impact on the overturning stability of large-area flat structures.

[0148] According to one aspect of this application, a cascade mechanical analysis of the anchoring components is specifically provided as follows:

[0149] Accordingly, based on the operating water level time series data, the pre-stored anchor bottom elevation data, and the pre-configured draft of the photovoltaic base on the water surface, the vertical span between the anchor cable suspension end and the bottom anchor end is calculated.

[0150] The draft of the photovoltaic base on the water surface is determined only by the ratio of the total mass of the system to the surface area of ​​the waterline, and does not change physically with the change of the absolute water level.

[0151] Therefore, by subtracting the constant draft from the calculated local water depth, the vertical span from the anchor cable suspension point to the bottom of the water is obtained. The vertical span fluctuates nonlinearly with the rise and fall of the reservoir water level.

[0152] Accordingly, based on the geometric relationship between the vertical span and the total length of the pre-configured anchor cable, it is determined whether the anchor cable is in a catenary state with a bottom contact section or in a tensioned state without a contact section.

[0153] In this step, the pre-configured physical cable length parameters are extracted and compared numerically with the vertical span. When the total pre-configured anchor cable length is greater than the vertical span, it is determined that the anchor cable has a section touching the reservoir bottom and is in catenary operation. When the total pre-configured anchor cable length is close to or less than the vertical span expansion caused by extremely high water levels, it is determined that the anchor cable is completely detached from the underwater suspension and is in a tensioned state.

[0154] Accordingly, based on the determined working state, and combined with the vertical span and the pre-configured total length of the anchor cable, the restoring force of the anchor cable is calculated using the corresponding nonlinear force mechanism.

[0155] Under catenary operation, the horizontal tension of the anchor cable and the vertical span constitute a nonlinear transcendental equation relationship. In specific calculations, an iterative solution algorithm is used to obtain the horizontal tension, and the component of the top restoring force acting at the connection point of the photovoltaic base is calculated in combination with the arc length model of the suspended section.

[0156] Furthermore, to avoid the divergence of the derivative matrix of the subsequent surrogate model caused by the state switching point, a nonlinear smooth transition function is introduced in the geometric critical region between the catenary and the tensioned state.

[0157] In some embodiments, T _next =T _current -f(T_current ) / f _prime (T _current );

[0158] Among them, T _next For the horizontal tension in the next iteration, T _current For the current iteration's horizontal tension, f(T) _current Let f'(T) be the residual function constructed by combining the cantilever arc length and the vertical span. _current ) is the first derivative of the residual function with respect to the current horizontal tension.

[0159] According to one aspect of this application, the process of obtaining the multi-degree-of-freedom mechanical response of the base and mooring system also includes time-varying calculations of wind loads considering attitude coupling, specifically including:

[0160] Optionally, the windward projected area of ​​the structure above the water surface is dynamically updated based on the pre-configured installation tilt angle of the photovoltaic panel and the pitch angle response of the base under wave excitation.

[0161] In other words, during the time-step solution of the dynamic response model, the windward projected area of ​​the structure above the water surface is dynamically updated based on the pre-configured installation tilt angle of the photovoltaic panel and the pitch angle response of the base under wave excitation before the current time step.

[0162] Accordingly, photovoltaic panels have a fixed pre-configured installation tilt angle to receive solar radiation. When wave dynamics cause the base to sway, the actual wind-blocking posture of the overall structure changes in real time.

[0163] The area update process satisfies the following formula:

[0164] A _exp =A _panel ×sin(α _p +θ _pitch )+B _total ×D _f ;

[0165] In the formula, A _exp For the windward projected area, A _panel α represents the total physical area of ​​the photovoltaic panel. _p For pre-configured installation tilt angle, θ _pitch For the pitch angle response of the base, B _total D is the total width of the windward side of the base. _f This refers to the freeboard height of the base.

[0166] Optionally, by combining wind speed time-series data with dynamically updated windward projected area, the time-varying wind load acting on the photovoltaic base at the current time step is calculated, and the time-varying wind load is used as a synchronous excitation source input into the dynamic response model.

[0167] Based on this, the extracted wind speed time series data is substituted into the quasi-steady-state aerodynamic calculation logic, and aerodynamic drag is extracted by combining dynamic windward characteristics.

[0168] For example, F _wind =0.5×ρ _a ×C _s ×A _exp ×(U _10 ) 2 ;

[0169] Among them, F _wind For time-varying wind load, ρ _a C is the density constant of air. _s A is the wind shape coefficient. _exp For the windward projected area, U _10 This is time-series wind speed data.

[0170] For example, the six-degree-of-freedom wave excitation force is introduced into the dynamic response model to solve for the multi-degree-of-freedom mechanical response of the base and mooring system.

[0171] Another example is to input the restoring force as a displacement constraint boundary condition into the dynamic response model to obtain the multi-degree-of-freedom mechanical response of the base and mooring system.

[0172] In this embodiment, the two examples above represent a joint physics process that is solved synchronously in the same fluid-structure interaction kinematics matrix.

[0173] Optionally, the six-degree-of-freedom wave excitation force and time-varying wind load can be used as external environmental excitation vector inputs, while the restoring force can be injected into the boundary conditions in the form of a nonlinear constraint matrix.

[0174] Based on this, a multi-degree-of-freedom numerical integration algorithm is used to solve the dynamic response model, and the time-series mechanical data such as displacement, rotation angle and anchor cable tension of the photovoltaic base within the analysis time window are extracted and output.

[0175] Based on the above embodiments, Monte Carlo simulations accelerated by a surrogate model are further performed to obtain the reliability indicators of the water surface photovoltaic base. This can be done through the following steps:

[0176] Step 401: Generate a candidate sample pool based on the joint probability distribution of pre-constructed time series data of environmental driving variables.

[0177] In this step, random sampling or Latin hypercube sampling is performed on the joint probability distribution to generate environmental input samples on the order of hundreds of thousands to millions. Since the dimensionality reduction mechanism removes the random dimension of wave height, the column dimension of the candidate sample pool is compressed, thereby reducing the sparsity problem of high-dimensional space filling.

[0178] Step 402: Select a portion of sample points from the candidate sample pool, and use the obtained multi-degree-of-freedom mechanical response of the base and mooring system to calculate the corresponding limit state response value and construct the initial training set.

[0179] In this step, a small sample set is formed by extracting a small number of initial design points from the candidate sample pool. For each initial design point, the dynamic response model is called to solve the problem. The limit state response value is calculated by subtracting the structural resistance threshold from the obtained multi-degree-of-freedom mechanical response value.

[0180] When the response value is less than or equal to 0, it indicates that the structure is in a failure state under the sample load.

[0181] The above provides data support for constructing a Gaussian process regression model.

[0182] Step 403: Based on this (initial training set), construct the Kriging surrogate model to predict the remaining sample points in the candidate sample pool and obtain the predicted response value and prediction uncertainty.

[0183] Accordingly, the model is used to perform calculations on a large number of remaining sample points that have not undergone physical simulation. The output results not only include the predicted response values ​​of each limit state equation corresponding to each sample point, but also the prediction uncertainty reflecting the confidence level of the prediction results.

[0184] In this context, prediction uncertainty is usually quantified using the prediction variance or standard deviation parameters of the Gaussian process.

[0185] Optionally, the limit state response value covers multiple independent failure modes, which include at least stress failure mode, overturning failure mode and displacement failure mode.

[0186] In the model building phase, the Kriging agent model is constructed, specifically including:

[0187] For multiple independent failure modes, an initial training set is shared to simultaneously build a multi-output Kriging surrogate model;

[0188] Furthermore, the photovoltaic base on the water surface faces the risk of combined failure under extreme wind and waves.

[0189] Among them, stress failure mode can usually be defined as the maximum equivalent stress of the floating frame or connecting node exceeding the yield strength of the material;

[0190] The overturning failure mode can be defined as the instantaneous maximum pitch or roll angle of the photovoltaic base exceeding the critical tilt angle that allows the photovoltaic panel to avoid being submerged by water under the combined action of wave force and wind tilting moment.

[0191] Displacement failure mode can be defined as the deformation of the mooring system causing the overall horizontal drift of the float to exceed the maximum safety margin allowed by the spacing between the power station arrays.

[0192] Optionally, for multiple independent failure modes, an initial training set can be shared to simultaneously build a multi-output Kriging surrogate model.

[0193] In this embodiment, the intrinsic correlation between different mechanical response quantities from the same fluid-structure interaction physical simulation is explored. Using the same initial training set as input, multiple limit state function features are extracted, thereby constructing a multi-output Kriging surrogate model that can simultaneously and in parallel predict stress, overturning and displacement states.

[0194] Step 404: Based on this (predicted response value and prediction uncertainty), construct an active learning evaluation criterion, select the target sample point with the highest uncertainty from the candidate sample pool to obtain the true limit state response value, add the target sample point to the initial training set to iteratively update the model (Kriging surrogate model) until the preset convergence condition is met.

[0195] For example, the convergence condition can be set to the active learning evaluation index of all uncalculated sample points in the candidate sample pool being greater than a preset convergence threshold.

[0196] For example, U _x =abs(G _x ) / σ _G_x ;

[0197] Among them, U _x For active learning evaluation metrics, abs is the absolute value function, and G... _x To predict the response value, σ _G_x To predict uncertainty.

[0198] Furthermore, this criterion is used to traverse the candidate sample pool and select U. _x The sample point with the smallest value is selected as the target sample point. Where, U _x The smaller the value, the closer the point is to the boundary and the greater the prediction variance.

[0199] For example, select any normalized sample point in the candidate sample pool, predict the mean and standard deviation of the normalized limit state of the sample point through the surrogate model, and then calculate the corresponding active learning evaluation index.

[0200] If the active learning evaluation index is greater than the preset convergence threshold, the sample point is determined to have high state classification certainty.

[0201] Conversely, if the active learning evaluation metric is less than the preset convergence threshold, the real physical response model is invoked.

[0202] Obtain the true response value of the extreme state corresponding to the sample point and add it to the initial training set to update the hyperparameters of the surrogate model;

[0203] The iterative process continues until all uncomputed sample points in the candidate sample pool are included in the U-shaped sample pool. _x If all values ​​exceed the preset convergence condition, the iteration and addition of points should be stopped.

[0204] In some scenarios, during the judgment phase, the proportion of invalid samples is statistically analyzed, specifically including:

[0205] Accordingly, based on the serial logic model, when any failure mode meets the failure determination condition, the overall structure is determined to have failed at that sample point.

[0206] After the multi-output prediction is completed, the corresponding stress, overturning and displacement prediction results are extracted for each specific environmental sample input.

[0207] For example, the serial logic model is represented by a logical OR operation rule. If the structure exceeds the corresponding safety threshold condition in any physical evaluation dimension, it is determined that the entire water surface photovoltaic base structure loses its function under the input load.

[0208] Step 405: The candidate sample pool is evaluated using the converged model (Kriging surrogate model), the proportion of failed samples is statistically analyzed, and the reliability index of the water surface photovoltaic base is obtained.

[0209] After iterative convergence, the surrogate model completes the state classification calculation for the entire candidate sample pool. The process of calculating the reliability index satisfies the following formula:

[0210] P _f =N _fail / N _total ;

[0211] Among them, P _f To calculate the proportion of invalid samples, N _fail To determine the total number of samples that have failed, N _total This represents the total number of samples contained in the candidate sample pool.

[0212] Based on this, the calculated statistical failure sample ratio can be converted into a numerical reliability index through the inverse cumulative distribution function of the standard normal distribution, which can then be used as the final evaluation metric.

[0213] In other scenarios, the method of this embodiment can also be:

[0214] A candidate sample pool is generated based on the joint probability distribution of pre-constructed time series data of environmental driving variables.

[0215] Select a subset of sample points from the candidate sample pool, and use the dynamic response model to generate wave characteristic parameters and obtain multi-degree-of-freedom mechanical responses for the selected sample points, calculate the corresponding limit state response values, and construct an initial training set.

[0216] A Kriging surrogate model is constructed based on the initial training set to predict the remaining sample points in the candidate sample pool and obtain the predicted response value and prediction uncertainty.

[0217] An active learning evaluation criterion is constructed based on the predicted response value and the predicted uncertainty. The target sample point with the highest uncertainty is selected from the candidate sample pool. The true limit state response value of the target sample point is obtained by using the dynamic response model. The target sample point and its true limit state response value are added to the initial training set to iteratively update the Kriging surrogate model until the preset convergence condition is met.

[0218] The candidate sample pool was evaluated using the converged Kriging surrogate model, and the proportion of failed samples was statistically analyzed to obtain the reliability index of the water surface photovoltaic base.

[0219] According to one aspect of this application, after obtaining the reliability indicators of the water-surface photovoltaic base, the method is also used to guide the dynamic optimization of structural parameters, specifically including:

[0220] The reliability indicators are compared with the preset target safety thresholds;

[0221] The target safety threshold is a mandatory lower limit of safety control values ​​pre-set by structural engineering design codes.

[0222] After obtaining the calculation results of the reliability index, perform a comparison operation on the numerical values;

[0223] If the reliability index does not meet the target safety threshold requirement, the physical design parameters of the water surface photovoltaic base shall be adjusted according to the parameter sensitivity. The physical design parameters shall include at least the geometric dimensions or material properties.

[0224] When the calculated reliability index is lower than the preset value requirement, the structural optimization closed-loop mechanism is triggered.

[0225] Calculate the sensitivity of each parameter and extract the key physical parameters that contribute the most to the failure probability;

[0226] Based on the sensitivity gradient direction, the cross-sectional thickness of the relevant floating body components is increased or the structural materials with higher yield strength are replaced. During this numerical adjustment process, a step size control strategy is adopted to limit the physical increment of a single configuration parameter update, so as to avoid the structural entity falling into an abnormal design state of over-strengthening.

[0227] Based on the adjusted physical design parameters, the Monte Carlo simulation of the acquisition of multi-degree-of-freedom mechanical response and the acceleration of the surrogate model is re-executed until the updated reliability index meets the target safety threshold requirements.

[0228] The mechanical state of the water surface photovoltaic base is reconstructed based on the updated extracted physical parameters, driving the environmental load mapping and probability evaluation framework into a new round of calculation.

[0229] It should be understood that the above adopts a closed-loop feedback configuration, which uses quantitative probability assessment data to directly guide and complete the optimization iteration of the physical engineering parameters.

[0230] According to another aspect of this application, the above method can also be implemented in the following manner:

[0231] The reliability index is compared with the preset target safety threshold to obtain the reliability comparison conclusion;

[0232] If the reliability comparison results indicate that the reliability index has met the target safety threshold requirements, then the current physical design parameters and reliability index are output as the final design result.

[0233] If the reliability comparison results indicate that the reliability index does not meet the target safety threshold requirements, then the pre-configured physical design parameters of the water surface photovoltaic base are adjusted based on the parameter sensitivity calculated by the partial derivative of the multi-degree-of-freedom mechanical response with respect to each physical design parameter. The physical design parameters include at least geometric dimensions or material properties.

[0234] Based on the adjusted physical design parameters, the Monte Carlo simulation of the acquisition of multi-degree-of-freedom mechanical response and the acceleration of the surrogate model are re-executed until the updated reliability index meets the target safety threshold requirements.

[0235] In one possible implementation, the joint probability distribution of pre-built time-series data of environment-driving variables is obtained through the following offline construction steps:

[0236] Step 501: Extract extreme environment sample points from the historical environmental data sequence.

[0237] In other words, we acquire historical environmental data sequences and extract extreme environmental sample points from them.

[0238] Correspondingly, historical environmental data sequences are long-term continuous observation data sequences collected from meteorological and hydrological stations in the target water area, typically spanning the past 30-50 years.

[0239] In some embodiments, it is necessary to separate data samples representing extreme operating conditions from a large amount of daily observation data.

[0240] For example, the extraction of sample points in extreme environments can be achieved using the annual maximum value method or the peak value exceeding the threshold method.

[0241] For example, when using the annual maximum value method, historical wind speed data and operating water level data are sliced ​​according to the Gregorian calendar year, and the maximum wind speed value and water level value recorded in each calendar year are calculated and retained, thereby generating an independent set of extreme environment sample points.

[0242] Step 502: A non-stationary generalized extreme value distribution model containing a time trend term is used to statistically fit the extreme environment sample points. The location parameters and scale parameters of the non-stationary generalized extreme value distribution model are set to change over time to obtain the extreme value edge distribution of the target design life.

[0243] In other embodiments, this step may also be:

[0244] A non-stationary generalized extreme value distribution model incorporating a time trend term was used to statistically fit extreme environment sample points corresponding to wind speed time series data and operating water level time series data, respectively. The location and scale parameters of the non-stationary generalized extreme value distribution model were set to change over time to obtain the extreme value marginal distributions of wind speed time series data and operating water level time series data under the pre-configured target design life.

[0245] Traditional reliability design typically assumes that the statistical characteristics of extreme weather events do not change over time. However, the design life of a floating photovoltaic base is usually 25-30 years, during which climate change can cause a systematic shift in the frequency and intensity of extreme wind speeds or abnormal water levels.

[0246] Based on this, the non-stationary generalized extreme value distribution model adopted in this embodiment extends the characteristic parameters of the distribution function from static constants to dynamic functions with time as the independent variable, thereby capturing the above-mentioned time evolution trend.

[0247] During the fitting process, the shape parameters are kept constant to control the tail type of the extreme value distribution.

[0248] For example, μ(t) = μ _0 +μ _1 ×t;

[0249] In the formula, μ(t) is the position parameter that varies with time. _0 μ is the initial position constant. _1 t is the time trend coefficient of the location parameter, where t is the time variable.

[0250] For example, σ(t) = exp(σ _0 +σ _1 ×t);

[0251] In the formula, σ(t) is a time-varying scaling parameter, exp is an exponential function operator with the natural constant as its base, and σ _0 σ is the initial scale constant. _1 t is the time trend coefficient of the scale parameter, where t is the time variable.

[0252] In this case, the maximum likelihood estimation method is used to optimize the model parameters. Optionally, the design life end time of the target surface photovoltaic power station can be substituted into the non-stationary generalized extreme value distribution model to infer the extreme value marginal distribution that takes into account the climate deterioration trend; the extreme load recurrence level given by this marginal distribution is usually greater than the value estimated under the stationary assumption, thereby improving the engineering redundancy of the long-term service base.

[0253] Optionally, a statistical significance test can also be performed after the parameter fitting is completed.

[0254] That is, when the historical environmental data sequence of the target water area shows no significant climate evolution trend after verification, the calculated μ... _1 With σ _1 When the value approaches 0, the non-stationary generalized extreme value distribution model automatically degenerates.

[0255] As an alternative, a stationary Gumbel distribution can be directly used to fit and extract data from extreme environment sample points.

[0256] Step 503: Use a preset correlation connection function to couple and correlate the extreme value marginal distributions of the corresponding wind speed time series data and the operating water level time series data to obtain the joint probability distribution of the pre-constructed environmental driving variable time series data.

[0257] Accordingly, multiple alternative families of Copula functions can be selected to perform trial calculations and fitting of the extreme marginal distributions of wind speed and water level.

[0258] The alternative families include the Gumbel family, the Clayton family, and the Frank family.

[0259] Next, the Akaike Information Criterion values ​​for each alternative fitting result are calculated. The Copula family of functions with the smallest Akaike Information Criterion value is selected as the optimal preset correlation connection function.

[0260] The above uses this function to mathematically assemble the separated extreme marginal distributions of wind speed and water level, ultimately generating the joint probability distribution of time-series environmental driving variables input into the Monte Carlo sampling pool.

[0261] According to one aspect of this application, a reliability design method for a surface photovoltaic base based on extreme environmental loads is provided, the method comprising the following steps:

[0262] Collect basic data on extreme environmental loads in the target water area, including historical extreme wind speed, wind direction, wave height, wave period, flow velocity, flow direction and water level changes, as well as statistical parameters of extreme conditions.

[0263] A parameterized model of extreme environmental loads is constructed, and wind, wave, and current loads are transformed into mathematical expressions with wind speed, wave height, and current velocity as independent variables, respectively, and a coupling relationship model between loads is established.

[0264] Multi-condition load simulation is carried out. Based on the parametric model, the load distribution under single extreme load and coupled extreme load is simulated respectively, and the pressure and shear force data of the base are output.

[0265] The simulated load is input into the structural mechanics software to calculate the stress and deformation of the base, and the Monte Carlo simulation method is used to calculate the failure probability of the base under extreme loads.

[0266] If the failure probability exceeds the standard, adjust the base structure parameters, such as size and material, and resubmit them into the load simulation and reliability calculation until the failure probability meets the design requirements.

[0267] Where, d _bar The dimensionless water depth obtained based on local water depth transformation has the following transformation relationship:

[0268] d _bar =g×d _local / (U _A ) 2 ;

[0269] Among them, F _bar The dimensionless wind zone length is obtained based on the transformation of the dynamic effective wind zone length, and its transformation relationship is as follows:

[0270] F _bar =g×F _eff / (U _A ) 2 ;

[0271] In other embodiments, based on a unified physical dimensionality reduction logic framework, the spectral peak period is calculated synchronously, and the calculation process of the spectral peak period satisfies the following formula:

[0272] T _p =(U _A / g)×7.54×tanh(0.833×(d _bar ) 0.375 )×tanh(0.0379×(F _bar ) 0.333 / tanh(0.833×(d _bar ) 0.375 ));

[0273] Among them, T_p 7.54, 0.833 and 0.0379 are dimensionless constant coefficients, and 0.375 and 0.333 are dimensionless exponential coefficients.

[0274] In some other embodiments, the process of calculating the equivalent wind zone length using the duration of wind speed can also be as follows:

[0275] F _eq =((U _A ) 2 / g)×(g×t _actual / (U _A ×62.38)) 1.5 ;

[0276] Among them, F _eq 62.38 is the equivalent wind zone length, 1.5 is a dimensionless exponential coefficient, and 62.38 is a constant coefficient.

[0277] In some embodiments, under catenary operation, the horizontal tension of the anchor cable and the suspension configuration are linked by a residual function established through the following geometric constraints:

[0278] f(T _current )=(T _current / w)×(cosh(w×L h / T _current )-1)-h _v ;

[0279] Where w corresponds to the underwater weight per unit length of the anchor cable, cosh is the hyperbolic cosine function operator, and L h The corresponding horizontal projection length of the anchor cable from the suspension end to the bottom contact point in the horizontal direction, h _v Corresponding vertical span.

[0280] The first derivative of the residual function with respect to horizontal tension is obtained by analytically differentiating the above equation.

[0281] In other embodiments, the parameter sensitivity is specifically calculated using a finite difference approximation method based on a converged Kriging surrogate model.

[0282] Accordingly, within the neighborhood of the current design parameter, a small perturbation increment is applied to each physical design parameter, and the limit state response value before and after the perturbation is predicted using the Kriging surrogate model. The ratio of the difference in response value to the perturbation increment is calculated as the sensitivity estimate of the physical design parameter.

[0283] In this invention, a dynamic surface wind zone linked to water level is extracted, and a force analysis mechanism is established that covers the fluid diffraction effect of large-size flat structures and the nonlinear transition of anchor cables. This mechanism is used to capture the energy suppression effect of wind zone shortening at low water levels and the cascade feedback of abrupt changes in anchor stiffness at extreme water levels, thus avoiding misjudgments of structural safety caused by the traditional fixed water depth assumption under complex hydrological evolution.

[0284] Furthermore, a non-stationary extreme value model with a time trend term was introduced to mitigate the risk of climate deterioration, and a multi-output agent prediction framework was constructed by combining active learning evaluation criteria. While covering multiple systemic safety risks such as stress, overturning, and displacement, the number of full-scale physical simulation calls was reduced to the order of hundreds of agent optimizations, freeing up computing resources and making closed-loop iterative optimization of engineering entities based on parameter sensitivity possible.

[0285] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A reliability design method for a photovoltaic base on a water surface under extreme environmental loads, characterized in that, include: Acquire time-series data of environmental driving variables for the target water area. The time-series data of environmental driving variables should include at least wind speed time-series data and operating water level time-series data. Based on wind speed time series data and operating water level time series data, wave characteristic parameters are deterministically generated using a shallow water wave forecasting physical model. Based on wave characteristic parameters, wind speed time series data, and operating water level time series data, a dynamic response model of the water surface photovoltaic base is constructed to obtain the multi-degree-of-freedom mechanical response of the base and mooring system. Based on the joint probability distribution of time-series data of environmental driving variables and the multi-degree-of-freedom mechanical response of the base and mooring system, Monte Carlo simulation accelerated by surrogate model is performed to obtain the reliability index of the water surface photovoltaic base.

2. The method according to claim 1, characterized in that, Before deterministically generating wave characteristic parameters based on shallow water wave forecasting physical models using wind speed and operating water level time series data, the process also includes: Obtain the digital elevation model of the target water area; Based on the time series data of operating water level and digital elevation model, the dynamic water surface boundary corresponding to the current operating water level is extracted; Based on the dynamic water surface boundary and the pre-configured base installation position, the distance is calculated along the wind direction ray to obtain the length of the dynamic effective wind zone.

3. The method according to claim 2, characterized in that, Based on wind speed time-series data and operating water level time-series data, wave characteristic parameters are deterministically generated using a shallow water wave forecasting physical model, specifically including: Based on the time series data of the operating water level and the digital elevation model, the local water depth at the base installation location was determined. Convert wind speed time series data into corresponding wind stress factors; The wind stress factor, local water depth, and dynamic effective wind zone length are input into the shallow water wave forecast physical model, and nonlinear physical coupling calculations are performed to obtain wave characteristic parameters. The wave characteristic parameters include at least the meaningful wave height and spectral peak period.

4. The method according to claim 3, characterized in that, The process of inputting wind stress factor, local water depth, and dynamic effective wind zone length into the shallow water wave forecast physical model for nonlinear physical coupling calculation to obtain wave characteristic parameters also includes a physical boundary verification mechanism: Obtain the wind speed duration corresponding to the wind speed time series data; Based on the wind stress factor and the dynamic effective wind zone length, the shortest development time required for waves to reach steady state is determined. When the duration of wind speed is less than the shortest development time, the equivalent wind zone length is calculated using the duration of wind speed, and the equivalent wind zone length is used to replace the dynamic effective wind zone length in the shallow water wave forecast physical model. After obtaining the wave characteristic parameters, the shallow water wave breaking limit is determined based on the local water depth; The wave characteristic parameters are truncated and verified using the shallow water wave breaking limit.

5. The method according to claim 4, characterized in that, The wave characteristic parameters are truncated and verified using the shallow water wave breaking limit, specifically including: Based on the meaningful wave height and pre-configured statistical exceedance probability in the wave characteristic parameters, the design wave height under extreme conditions is extracted; The shallow water wave breaking limit is set as a physical cutoff threshold that is strictly positively correlated with the local water depth. When the design wave height is greater than the physical cutoff threshold, the design wave height is updated to the physical cutoff threshold.

6. The method according to claim 1, characterized in that, The process of constructing a dynamic response model for the water-surface photovoltaic base and obtaining the multi-degree-of-freedom mechanical response of the base and mooring system also includes cascade mechanical analysis of the mooring components: Calculate the vertical span between the anchor cable suspension end and the bottom anchoring end based on the operating water level time series data; Based on the geometric relationship between the vertical span and the total length of the pre-configured anchor cable, it is determined whether the anchor cable is in a catenary state with a bottom contact section or in a tensioned state without a contact section. Based on the determined working state, the restoring force of the anchor cable is calculated using the corresponding nonlinear force mechanism; By inputting the restoring force as a displacement constraint boundary condition into the dynamic response model, the multi-degree-of-freedom mechanical response of the base and mooring system is obtained.

7. The method according to claim 1, characterized in that, Monte Carlo simulations accelerated by performing a proxy model were used to obtain reliability indices for the water-surface photovoltaic base, specifically including: A candidate sample pool is generated based on the joint probability distribution of time series data of environmental driving variables. Select a subset of sample points from the candidate sample pool, and use the obtained multi-degree-of-freedom mechanical responses of the base and mooring system to calculate the corresponding limit state response values ​​and construct an initial training set. Based on this, a Kriging surrogate model is constructed to predict the remaining sample points in the candidate sample pool and obtain the predicted response value and prediction uncertainty. Based on this, an active learning evaluation criterion is constructed, which selects the target sample point with the highest uncertainty from the candidate sample pool, adds the target sample point to the initial training set, and iteratively updates the model until the convergence condition is met. The converged model is used to evaluate the candidate sample pool, and the proportion of failed samples is statistically analyzed to obtain the reliability index of the water surface photovoltaic base.

8. The method according to claim 1, characterized in that, The joint probability distribution of time series data for environment-driven variables was obtained through the following offline construction steps: Extract extreme environment sample points from historical environmental data sequences; A non-stationary generalized extreme value distribution model incorporating a time trend term is used to statistically fit extreme environment sample points. The location and scale parameters of the non-stationary generalized extreme value distribution model are set to vary over time to obtain the extreme value marginal distribution during the target design life. By using a correlation link function, the extreme value marginal distributions of the corresponding wind speed time series data and the operating water level time series data are coupled and correlated to obtain the joint probability distribution of the environmental driving variable time series data.

9. The method according to claim 1, characterized in that, After obtaining the reliability indicators of the water-surface photovoltaic base, the method is also used to guide the dynamic optimization of structural parameters, specifically including: Compare the reliability indicators with the target safety threshold; If the reliability index does not meet the target safety threshold requirement, the physical design parameters of the water surface photovoltaic base shall be adjusted according to the parameter sensitivity. The physical design parameters shall include at least the geometric dimensions or material properties. Based on the adjusted physical design parameters, the Monte Carlo simulation of the acquisition of multi-degree-of-freedom mechanical response and the acceleration of the surrogate model are re-executed until the updated reliability index meets the target safety threshold requirements.

10. An apparatus, characterized in that, It includes at least a water-surface photovoltaic base, the design parameters of which are determined using the method described in any one of claims 1-9.

Images