Waterborne photovoltaic power station load prediction method and device and computer equipment

Abstract

This application relates to the field of photovoltaic power generation technology, and in particular to a method, apparatus, and computer equipment for predicting the load of a floating photovoltaic power station. The method provided in this application includes: acquiring an environmental data sequence generated based on environmental data of a target floating photovoltaic power station; dividing the environmental data sequence based on a preset time scale, and calculating the local Hurst exponent for each sub-sequence in the division result according to preset calculation rules, generating a local Hurst exponent sequence; matching the local Hurst exponent sequence with a pre-acquired set of reference data sequences to determine the matching data sequence with the highest similarity to the local Hurst exponent sequence, and acquiring the corresponding historical load data; inputting the local Hurst exponent sequence, the matching data sequence, and the historical load data into a pre-trained load analysis model to obtain the load prediction result for the target floating photovoltaic power station. The method provided in this application achieves high-accuracy load prediction for floating photovoltaic power stations.

Description

Technical Field

[0001] This application relates to the field of photovoltaic power generation technology, and in particular to a method, apparatus and computer equipment for predicting the load of a floating photovoltaic power station. Background Technology

[0002] As an important direction for the development of renewable energy, floating photovoltaic (PV) systems have been rapidly promoted in lakes, reservoirs, fishponds, and other water bodies in my country in recent years. Compared with traditional onshore PV power plants, floating PV systems have multiple advantages, including saving land resources, improving power generation efficiency, and reducing water evaporation.

[0003] However, due to the lack of significant obstacles such as buildings, vegetation, and mountains, the surface of water is flat and smooth, resulting in much lower frictional resistance to airflow compared to land. This leads to significant differences in environmental factors such as wind field characteristics and temperature variations compared to land. Therefore, the safe operation and maintenance experience of photovoltaic power stations in plains or mountains cannot be directly applied to floating photovoltaic power stations. At the same time, the unique environment of floating photovoltaic power stations makes them more prone to problems such as support collapse and photovoltaic panel overturning compared to traditional onshore photovoltaic power stations, posing new requirements for load prediction and safety assessment.

[0004] Therefore, there is an urgent need for a new load prediction method for floating photovoltaic power stations in order to analyze and evaluate the safety of floating photovoltaic power stations. Summary of the Invention

[0005] This application provides a method, apparatus, and computer equipment for predicting the load of a floating photovoltaic power station, which achieves high-accuracy load prediction for floating photovoltaic power stations and provides a data foundation for the analysis and evaluation of the safety of floating photovoltaic power stations.

[0006] To achieve the above objectives, the main technical solutions adopted in this application include: In a first aspect, embodiments of this application provide a load prediction method for a floating photovoltaic power station, the method comprising: Obtain environmental data sequences generated based on environmental data from the target floating photovoltaic power station; The environmental data sequence is divided based on at least one preset time scale, and the local Hearst exponent is calculated for each subsequence in the division result according to a preset calculation rule to generate a local Hearst exponent sequence. The local Hearst exponent sequence is matched against a pre-acquired set of reference data sequences to determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence, and the historical payload data corresponding to the matching data sequence is obtained. The local Hearst exponent sequence, the matching data sequence, and the historical load data are input into a pre-trained load analysis model to obtain the load prediction results of the target floating photovoltaic power station output by the model.

[0007] The load prediction method for floating photovoltaic power stations proposed in this application introduces a local Hearst exponent to characterize the non-stationarity and long-range correlation of environmental data sequences, effectively capturing the characteristics of various environmental factors at different time scales. It also combines reference data sequences for matching, utilizing historical reference data sequences under similar operating conditions and historical loads, and analyzes the load prediction results by fusing multi-source information through a load analysis model. This overcomes the problem of insufficient applicability of traditional onshore photovoltaic power station methods to the special environment of water surfaces.

[0008] Secondly, embodiments of this application provide a load prediction device for a floating photovoltaic power station, the device comprising: The data acquisition module is used to acquire environmental data sequences generated based on environmental data from the target floating photovoltaic power station; The sequence generation module is used to divide the environmental data sequence based on at least one preset time scale, and calculate the local Hearst exponent for each subsequence in the division result according to a preset calculation rule to generate a local Hearst exponent sequence. The data matching module is used to match the local Hearst exponent sequence in a pre-acquired set of reference data sequences, determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence, and obtain the historical load data corresponding to the matching data sequence. The model prediction module is used to input the local Hearst exponent sequence, the matching data sequence, and the historical load data into a pre-trained load analysis model to obtain the load prediction results of the target floating photovoltaic power station output by the model.

[0009] Thirdly, embodiments of this application provide a computer device, including: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the load prediction method for the floating photovoltaic power station described in the first aspect. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] Figure 1 A step diagram illustrating a load prediction method for a floating photovoltaic power station provided in this application embodiment; Figure 2 A schematic diagram of a force analysis model for a floating photovoltaic system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a load prediction device for a floating photovoltaic power station provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] This application provides a method for predicting the load of a floating photovoltaic power station. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0014] This embodiment provides a load prediction method for a floating photovoltaic power station, such as... Figure 1 As shown, it includes the following steps: Step 110: Obtain the environmental data sequence generated based on the environmental data of the target floating photovoltaic power station.

[0015] The load on a floating photovoltaic power station is related to environmental factors. Therefore, in the method described in this embodiment, environmental data related to the operating environment of the target floating photovoltaic power station is first acquired. Environmental data can come from real-time monitoring data from sensors deployed at the power station site, historical records from nearby weather stations, or weather forecasts. In one embodiment, the environmental data includes meteorological parameters such as wind speed, wind direction, temperature, humidity, and air pressure.

[0016] The acquired environmental data is arranged and organized in chronological order to form a continuous environmental data sequence. This sequence will serve as the data foundation for the target photovoltaic hydroelectric power station in subsequent load analysis.

[0017] Step 120: Divide the environmental data sequence based on at least one preset time scale, and calculate the local Hearst exponent for each subsequence in the division result according to the preset calculation rules to generate a local Hearst exponent sequence.

[0018] This embodiment proposes dividing the environmental data sequence into several subsequences according to a preset time scale to reflect the fluctuation characteristics of the environment at different time scales. Specifically, the environmental data sequence can be segmented based on a preset time scale using a sliding window or a segmentation method.

[0019] The preset time scale can be determined based on the total length of the environmental data sequence. For example, hours or days can be chosen as the scale. When the total length of the environmental data sequence is long enough, months can also be used as the scale. Different time scales can reflect different environmental conditions; for example, when using months as the scale, the overall characteristics of the environment can be analyzed. Therefore, one or more time scales can be selected according to actual needs to achieve sequence feature extraction at different time scales. In one embodiment, the time scale is less than or equal to one-tenth of the total length of the environmental data sequence to ensure that the number of subsequences is sufficiently large.

[0020] This embodiment introduces the Hurst exponent for analyzing and extracting features from the environmental data of the target floating photovoltaic power station. The Hurst exponent is a statistical indicator used to measure the long-term memory or persistence of time series data. It was proposed by British hydrologist Harold Hurst during his study of Nile River reservoir water level fluctuations. Its core meaning lies in characterizing the autocorrelation characteristics of time series data.

[0021] The local Hearst exponent is an extension of the Hearst exponent in time-varying environments. For practical applications where data often changes continuously over time, the local Hearst exponent calculates the Hearst exponent within a specific time period of the entire time series, thereby capturing the characteristic changes within a local interval of the sequence.

[0022] It should be noted that there are currently several methods for calculating the Hearst exponent. For example, the most classic original method is rescaled range analysis (R / S analysis), which calculates the Hearst exponent by calculating the ratio of the range to the standard deviation at different time scales; detrended fluctuation analysis, which calculates the fluctuation function after removing local trends through polynomial fitting; wavelet analysis, which uses the multi-scale characteristics of wavelet transform to separate signal components; and other calculation methods, which will not be elaborated on here.

[0023] The different calculation methods described above make different assumptions and handle the sequence characteristics differently, resulting in numerical differences in the Hurst exponent. However, these numerical differences do not affect the core expressive function of the Hurst exponent. As long as the calculation method remains consistent, the relative relationships between different sequences or time periods remain valid. Therefore, this embodiment does not limit the specific calculation method of the Hurst exponent; a suitable calculation method can be selected based on factors such as the characteristics of the selected environmental data.

[0024] Step 130: Match the local Hearst exponent sequence with the pre-acquired set of reference data sequences, determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence, and obtain the historical payload data corresponding to the matching data sequence.

[0025] The reference data sequence set consists of data sequences generated from historical scenarios with known environmental and load correspondences. This embodiment proposes to compare and match the environmental data sequence of the target floating photovoltaic power station with historically accumulated reference data sequences to select the reference data sequence corresponding to the historical scenario with the most similar environmental conditions.

[0026] Specifically, the matching process can measure similarity based on the morphology or statistical characteristics of the sequences, and select one or more reference data sequences that are most similar. When multiple reference data sequences need to be selected, they can be sorted from high to low based on similarity, and the top-ranked reference data sequences can be selected.

[0027] After determining the matching data sequence, the actual structural load records associated with the matching data sequence are obtained to provide physically meaningful prior information for subsequent predictions.

[0028] Step 140: Input the local Hearst exponent sequence, the matching data sequence, and the historical load data into the pre-trained load analysis model to obtain the target floating photovoltaic power station load prediction results output by the model.

[0029] After obtaining the local Hurst exponent sequence characterizing the current environmental dynamics, the matching historical environmental sequence, and their corresponding real load data, these three are fed into a pre-trained load analysis model. By learning the complex mapping relationship between environmental characteristics and structural responses from a large number of historical samples, this model can integrate current trends with similar historical cases to output predicted values ​​of the key loads on the target floating photovoltaic power station in the future, thus providing a scientific basis for structural safety assessment and operation and maintenance decisions.

[0030] The method described in this embodiment, which introduces a local Hurst exponent for feature extraction from environmental data, has the following advantages: First, unlike the common approach of directly training a single model with a large amount of historical data and only inputting current meteorological data during inference, this method actively selects the historical reference data most similar to the current operating conditions as model input through sequence matching. Simultaneously, it feeds both the local Hurst exponent sequence and the matched data sequence into the model, enabling the model to simultaneously perceive the differences between the reference data and the current data. This guides the inference process to focus more on the inherent characteristics of the floating photovoltaic environment. This approach not only improves prediction accuracy but also avoids the problem of needing to retrain the model every time new data is accumulated.

[0031] Secondly, since the model has clear reference data to support it during inference, the number of model parameters required is significantly reduced, enabling it to be deployed on edge devices and realize edge computing.

[0032] Furthermore, this method is data-driven and only requires conventional meteorological data to complete load prediction, which greatly reduces the technical threshold and implementation cost.

[0033] The second embodiment of this application further specifies the load prediction method for floating photovoltaic power stations in the first embodiment in a more detailed and specific way. Some or all of the technical features in the second embodiment can be combined with or replaced by the first embodiment, either individually or in combination, to obtain more feasible load prediction methods for floating photovoltaic power stations.

[0034] The load prediction method for the floating photovoltaic power station in the second embodiment of this application is described in detail below: Optionally, for each subsequence in the segmentation result, the local Hearst exponent is calculated according to a preset calculation rule to generate a local Hearst exponent sequence, including: fitting the environmental data sequence to obtain a fitting function and calculating the root mean square of fluctuation of the fitting function; fitting each subsequence in the segmentation result to obtain a sub-fitting function and calculating the difference between the common logarithm of the sub-fitting function and the common logarithm of the root mean square of fluctuation as the sub-root mean square of fluctuation; normalizing the sub-root mean square of fluctuation based on the time scale and the total length of the environmental data sequence to obtain the local Hearst exponent sequence.

[0035] This embodiment further defines the calculation method of the local Hearst exponent to provide a Hearst exponent calculation method that is more suitable for extracting environmental data features of floating photovoltaic power stations.

[0036] Specifically, let the environmental data sequence be... Since meteorological data is a chaotic system, based on the definition of the Hearst exponent, the environmental data sequence is determined to be a random walk structure sequence. Based on the environmental data, the calculation method is as follows: in, This is the mean operator. This is a basic data sequence formed by arranging environmental data in chronological order. Represents the first in the sequence One element, This is a time series of fluctuation increments calculated based on the underlying data sequence.

[0037] Then, the environmental data sequence Perform fitting to obtain the fitting function Calculate the root mean square of the fluctuation of the fitted function, and let the root mean square of the fluctuation of the i-th element be... If the time scale is s, the calculation formula is as follows: For each subsequence in the partitioning result, a subfit function is obtained by fitting the subsequence. The difference between the common logarithm of the subfit function and the common logarithm of the root mean square of fluctuation is calculated as the sub-root mean square of fluctuation. : Finally, based on the time scale The total length of environmental data sequences Normalize the root mean square of the sub-waves to obtain the local Hearst exponent sequence. : in, The first element of the pre-calculated local Hearst exponent sequence.

[0038] Based on the above method, by calculating according to at least one or more preset time scales, one or more local Hearst exponent sequences at one or more different time scales can be obtained, thereby achieving accurate extraction of environmental data features of the target floating photovoltaic power station.

[0039] Optionally, the load analysis model includes three parallel branches; the local Hurst exponent sequence, the matching data sequence, and the historical load data are respectively input to one of the parallel branches; each parallel branch includes at least one analysis structure, which consists of a convolutional layer and a pooling layer; the outputs of the three parallel branches are flattened and spliced, and then input to at least one fully connected layer, wherein the last fully connected layer is used as the output layer to output the power plant load prediction results.

[0040] This embodiment further defines the structure of the load analysis model.

[0041] Specifically, the load analysis model employs a lightweight structure combining a one-dimensional convolutional neural network and a fully connected network. The model inputs include a local Hurst exponent sequence, a matched data sequence, and historical load data.

[0042] This embodiment proposes a three-branch parallel structure to balance the local temporal feature extraction capabilities of various sequence data with the needs of edge deployment: the first parallel branch processes the local Hurst exponent sequence, the second parallel branch processes the matched data sequence, and the third parallel branch processes the historical payload data. The three branches have identical structures, each consisting of an overlapping analysis structure including one convolutional layer and one pooling layer. The feature vectors extracted from each branch are concatenated with auxiliary meteorological features and then input into a fully connected layer to obtain the prediction result.

[0043] In one embodiment, a parallel branch is formed by the overlap of two analysis structures, namely, a first convolutional layer - a first pooling layer - a second convolutional layer - a second pooling layer structure.

[0044] In one embodiment, a parallel branch may use the following parameters: the first one-dimensional convolutional layer has 16 convolutional kernels with a kernel length of 3 and a stride of 1; followed by a max pooling layer with a pooling window of 2; the second one-dimensional convolutional layer has 32 convolutional kernels with a kernel length of 3 and a stride of 1; followed by a max pooling layer with a pooling window of 2.

[0045] In one embodiment, the output layer employs linear activation to meet the requirements of continuous value regression prediction.

[0046] In one embodiment, the outputs of the three parallel branches are flattened. The flattened three feature vectors are concatenated with auxiliary meteorological features to form a fused feature vector, which is then sequentially input into the first fully connected layer and the second fully connected layer. The first fully connected layer can be set to 64 neurons, and the second fully connected layer can be set to 32 neurons. The output layer is set to one or more nodes depending on the prediction target.

[0047] Optionally, when the output layer has one node, the power plant load prediction result is the comprehensive equivalent load of the target floating photovoltaic power plant during the target period; when the output layer has three nodes, the power plant load prediction result is the downwind load, lift load, and overturning moment of the target floating photovoltaic power plant during the target period; when the number of nodes in the output layer is greater than three, the power plant load prediction result also includes at least one of the following parameters: critical component load and photovoltaic panel edge load.

[0048] This embodiment proposes that when the output layer is set to 1 node, the comprehensive equivalent load for the target time period can be output; when the output layer is set to 3 nodes, the downwind load Fx, the lift load Fz, and the overturning moment My can be output respectively; when the output layer is set to more nodes, the loads of key components and the edge loads of photovoltaic panels can be further output, such as the axial force of key support members, the shear force of nodes, or the peak value of suction force at the edge of photovoltaic panels.

[0049] Next, the physical model corresponding to the load analysis model provided in this embodiment will be explained. A force analysis model of the floating photovoltaic system is established, treating the photovoltaic modules, support structure, floating body, and mooring constraints as a coupled system subjected to wind load, temperature load, and water surface constraints. A simplified force analysis diagram is shown below. Figure 2 As shown.

[0050] Figure 2 In this diagram, V represents the system's air intake, Fx represents the downwind force, Fz represents the normal lift, My represents the overturning moment around the support reference point or component support point, G represents the system's self-weight, B represents the buoyancy provided by the floating body, FT represents the equivalent thermal stress load caused by temperature difference, ΔT represents the temperature change, and Mooring represents the mooring or anchoring constraint reaction force. Therefore, the floating photovoltaic system can be simplified as a coupled force-bearing system of four parts: "component – ​​support – floating body – mooring." The component is the main wind-receiving surface, the support is the load transfer path, the floating body provides buoyancy and attitude recovery, and the mooring system provides boundary constraints and additional recovery capabilities.

[0051] The environmental load acting on the surface of the photovoltaic module is decomposed. When the incoming air velocity is V, the equivalent wind pressure on the module surface can be expressed as: q =1 / 2· ρ · V ^2, where ρ Let V be the air density and V be the equivalent wind speed at the component surface. Considering the stronger pulsation and local nonstationarity of the wind field in the water scene, the equivalent wind speed can be obtained by correcting the measured wind speed, regional meteorological parameters, and the local Hurst exponent.

[0052] Under wind pressure, the aerodynamic force on the component surface can be further decomposed into a downwind force Fx and a normal lift force Fz, with the following expressions: F x = q · A · C d · μ s ; F z = q · A · C l · μ z Where A is the wind-receiving area of ​​the component, Cd is the drag coefficient, and Cl is the lift coefficient. μ s and μ zThese are the gust fluctuation correction coefficients. Since the points of application of the downwind force and normal lift are usually deviated from the support connection reference point, an overturning moment My will be generated, which can be expressed as: M y =F z ·l z +F x ·l x Where lz and lx are the lever arms of the normal force and the downwind force relative to the reference point, respectively. This overturning moment is an important indicator for assessing the risk of component rollover, support instability, and connector fatigue.

[0053] Secondly, consider the balance between the system's self-weight and the buoyancy of the floating body. If we denote the total system weight as G and the total buoyancy provided by the floating body as B, then under quasi-static conditions, B≈G should be satisfied. When the external wind load increases, the system's attitude will change through pitch or roll, altering the relative position of the center of buoyancy and the center of gravity, thus generating an additional restoring moment. For simplified engineering analysis, the restoring effect of the floating body can be approximated as linear restoring stiffness, i.e. F k =K·δ Where K is the equivalent restoring stiffness, δ This refers to the structural displacement or attitude offset. For floating photovoltaic structures constrained by mooring systems, the anchor cables or mooring components also provide additional constraint reaction forces to jointly suppress structural drift and overturning.

[0054] Consider the impact of temperature changes on components and support structures. Temperature changes cause materials to expand and contract, resulting in thermal stresses in connection nodes or confined members. Thermal stresses can be expressed as: σ T =E·α·ΔT Where E is the elastic modulus of the material. α Let be the coefficient of linear expansion, and ΔT be the temperature difference relative to the reference state. Furthermore, thermal stress can be converted into an equivalent thermal load: F T =σ T ·A The rod, where rod A is the effective cross-sectional area of ​​the corresponding component. The thermal load will change the stress distribution of the component frame, support rods and connectors, and together with the wind load, affect the overall safety status of the system.

[0055] Therefore, the total equivalent load of the floating photovoltaic structure can be expressed as: F sum =F wind +F temp+F restore +F inertia in, F wind Indicates wind-induced load, F temp Indicates temperature load. F restore This indicates the restoring effect provided by the floating body and mooring system. F inertia This represents the additional inertial term caused by gusts, wave-induced disturbances, or attitude changes. In conventional engineering applications, if the inertial term is small, it can be incorporated into the dynamic wind load correction term. Based on the above theoretical analysis, the key response quantities for the target time period can be defined as the downwind force Fx, normal lift Fz, overturning moment My, nodal thermal load FT, and the combined equivalent load. F sum .

[0056] The load analysis model provided in this embodiment is based on the aforementioned physical model. The first branch analyzes the environmental fluctuation characteristics of the target floating photovoltaic power station, the second branch analyzes the environmental characteristics of matched historical samples, and the third branch analyzes the actual load of historical samples under similar environments. Convolutional layers extract data from different time windows, such as wind speed, temperature, and other environmental data, as well as features corresponding to local Hurst exponent changes. Pooling layers suppress random noise and retain dominant trends. Fully connected layers establish a nonlinear mapping relationship between "current environmental characteristics – historical environmental characteristics – historical load response," thereby enabling inference of the load for the target time period. The meanings of the different output results of the model output layer have been explained in the analysis section of the aforementioned physical model, corresponding to parameters such as wind pressure, lift, overturning moment, and thermal stress in the physical model. During model training, historical real loads or load values ​​obtained through finite element or structural analysis inversion are used as monitoring signals. During model inference, the current local Hurst exponent sequence, the closest historical local Hurst exponent sequence, and their corresponding historical load data are used to quickly approximate key parameters or key load results in the physical model, thus balancing prediction accuracy, interpretability, and edge deployment feasibility.

[0057] Optionally, the reference data sequences in the reference data sequence set are obtained in the following manner: obtaining reference environmental data sequences generated based on historical environmental data of a reference region, wherein the type of the reference region includes at least one of the following types: the target floating photovoltaic power station, other floating photovoltaic power stations with environments similar to the target floating photovoltaic power station, onshore photovoltaic power stations adjacent to the target floating photovoltaic power station, and the region where the target floating photovoltaic power station is located; dividing the reference environmental data sequences based on a preset time scale, and calculating the local Hearst exponent for each subsequence in the division result according to a preset calculation rule to generate the reference data sequence.

[0058] This embodiment limits the method for obtaining the reference data sequence.

[0059] Specifically, the reference data sequence is constructed by processing historical environmental data from multiple reference regions. Therefore, it is necessary to acquire historical environmental data from different types of reference regions and generate corresponding reference environmental data sequences accordingly.

[0060] The reference areas include, but are not limited to, the following four categories: (1) the target floating photovoltaic power station itself, whose historical operating data can provide the most direct reference; (2) other floating photovoltaic power stations with similar geographical location, water features or climate conditions to the target floating photovoltaic power station; (3) onshore photovoltaic power stations near the target power station; and (4) the entire administrative or meteorological region where the target power station is located.

[0061] Subsequently, for each reference environmental data sequence, the same calculation method as the target floating photovoltaic power station was used to divide it and calculate the local Hearst exponent, so as to obtain a reference data sequence with consistent structure and clear physical meaning.

[0062] The reference data sequence construction method provided in this embodiment provides a basis for selecting reference regions, ensures the coverage and representativeness of the reference set, ensures scale consistency and method compatibility in the matching process, and lays a data foundation for subsequent high-precision load prediction.

[0063] Optionally, matching the local Hearst exponent sequence in a pre-acquired set of reference data sequences to determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence includes: obtaining the difference sequence between the environmental data sequence and the reference data sequence; statistically analyzing the difference sequence to obtain a first similarity; obtaining a comprehensive similarity score based on the weights pre-set for each type of reference region and the weights pre-set for the first similarity, as well as the type of the reference region corresponding to the reference data sequence and the first similarity; and selecting at least one reference data sequence with the highest comprehensive similarity score as the matching data sequence; or, according to the priority pre-set for each type of reference region, sequentially matching the reference data sequences associated with the type of the reference region corresponding to each priority, selecting the reference data sequence with a first similarity greater than a preset threshold in the reference data sequence corresponding to the current priority as the matching data sequence, and continuing to match in the reference data sequence corresponding to the next priority if there is no reference data sequence with a first similarity greater than the preset threshold in the reference data sequence corresponding to the current priority.

[0064] To achieve more reasonable and efficient reference data matching, this embodiment proposes two different matching strategies, which can be selected according to actual needs.

[0065] The first strategy employs a weighted calculation mechanism: First, the difference sequence between the target local Hearst exponent sequence and each reference data sequence is calculated, and a first similarity is obtained by performing statistical analysis (such as mean squared error, mean absolute deviation, etc.) on this difference sequence. Then, the two are weighted and fused together by pre-defined weights for different reference region types (such as the target power station itself, similar floating power stations, adjacent onshore power stations, and the region itself) and a confidence weight set for the first similarity itself, generating a comprehensive similarity score. Finally, one or more reference data sequences with the highest scores are selected as the matching results. This method balances regional representativeness and numerical similarity, and is suitable for the unified evaluation of multi-source heterogeneous reference data.

[0066] The second strategy employs a priority-driven hierarchical matching mechanism: Matching priorities are pre-defined for various reference regions (e.g., target power station itself > similar hydroelectric power stations > nearby onshore power stations > regional meteorological stations). Matching is performed sequentially within the corresponding category of reference data subsets according to priority. Within the current priority level, if a reference data sequence with a first similarity exceeding a preset threshold exists, it is directly used as the matching result, terminating subsequent searches. If no matching sequence is found, the search automatically downgrades to the next priority level. This strategy emphasizes physical proximity, prioritizing the use of the most relevant historical cases to improve matching efficiency while ensuring prediction reliability. Both strategies can be flexibly selected or combined according to actual needs to balance accuracy, robustness, and computational cost.

[0067] Optionally, the input data for the load analysis model may also include environmental data of the target floating photovoltaic power station; the load analysis model may also include a fourth parallel branch for analyzing the environmental data of the target floating photovoltaic power station.

[0068] This embodiment proposes that, in order to further improve the accuracy of load prediction and real-time response capability, the input data of the load analysis model may also include environmental data of the target floating photovoltaic power station and / or environmental data corresponding to the matching data sequence (such as current or recent measured wind speed, temperature, etc.).

[0069] Accordingly, at least a fourth parallel branch needs to be added to the model structure for feature extraction and analysis of these raw environmental data. This branch works in conjunction with the other three parallel branches that process the local Hurst exponent sequence, the matching data sequence, and the historical load data, respectively. While preserving the specificity of each input information, it achieves complementarity and enhancement of multi-dimensional features through subsequent fusion mechanisms.

[0070] This embodiment, by introducing raw environmental data and its dedicated analysis path, helps to capture features such as instantaneous meteorological disturbances or high-frequency fluctuations that are not fully represented by the Hearst index, thereby enhancing the model's sensitivity to structural load changes under sudden weather events and its predictive robustness.

[0071] Optionally, the environmental data of the target floating photovoltaic power station includes wind speed data and temperature of the target floating photovoltaic power station; the environmental data is historical environmental data collected by the target floating photovoltaic power station within a preset period before the target period and / or weather forecast data corresponding to the target period.

[0072] This embodiment proposes that the environmental data for the target floating photovoltaic power station mainly includes wind speed data and temperature data. Referring to the previous embodiment regarding the physical model, these two types of parameters have a decisive influence on the wind load, thermal stress, and aerodynamic stability of the floating photovoltaic structure. Among them, wind speed is directly related to the magnitude and pulsation characteristics of wind pressure and is the core input for calculating the downwind force, lift, and overturning moment; while temperature not only affects the thermal expansion behavior of photovoltaic module materials, but also indirectly affects the dynamic response of the structure by modulating the local turbulence intensity through the water-temperature difference.

[0073] Furthermore, the environmental data for the target floating photovoltaic power station can come from two sources: first, historical environmental data actually collected by on-site meteorological sensors within a preset period (such as the past 24 hours, 72 hours, etc.) before the target prediction period, which has high accuracy and strong correlation; second, weather forecast data corresponding to the target prediction period, which usually comes from numerical weather prediction models or professional meteorological service platforms, and is used to provide information on future environmental trends.

[0074] Among these methods, predictions based on historical environmental data of the target floating solar power station are easy to perform offline analysis and are suitable for embedded implementation and edge computing. However, the corresponding model needs to additionally undertake the task of weather forecasting, that is, predicting the future load of the floating solar power station based on current meteorological data. However, predictions based on weather forecast data are affected by the accuracy of weather forecasts, and in most cases, weather forecast data is not for the floating solar power station, but for meteorological data of a region. However, in subsequent calculations, the model can analyze and predict the load data of the current day based on the "weather data of the day", and no longer needs to undertake the task of weather forecasting.

[0075] Furthermore, the two methods mentioned above can be used individually or in combination—for example, using measured data as the primary source and forecast data as a supplement for interpolation or correction, thereby improving the coverage of future operating conditions while ensuring data timeliness. This flexible data source mechanism makes this method suitable for both short-term refined forecasting and medium- to long-term risk assessment, enhancing the system's adaptability and practicality in different operating scenarios.

[0076] The third embodiment of this application also proposes a load prediction device for a floating photovoltaic power station, such as... Figure 3 As shown, the device includes: Data acquisition module 310 is used to acquire environmental data sequences generated based on environmental data of the target floating photovoltaic power station; The sequence generation module 320 is used to divide the environmental data sequence based on at least one preset time scale, and calculate the local Hearst exponent for each subsequence in the division result according to a preset calculation rule to generate a local Hearst exponent sequence. The data matching module 330 is used to match the local Hearst exponent sequence in a pre-acquired set of reference data sequences, determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence, and obtain the historical load data corresponding to the matching data sequence. The model prediction module 340 is used to input the local Hearst exponent sequence, the matching data sequence and the historical load data into a pre-trained load analysis model to obtain the load prediction result of the target floating photovoltaic power station output by the model.

[0077] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0078] In this embodiment, the load prediction device for the floating photovoltaic power station is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0079] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 4 As shown, the computer device includes one or more processors 410, memory 420, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 4 Take a processor 410 as an example.

[0080] Processor 410 may be a central processing unit, a network processor, or a combination thereof. Processor 410 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0081] The memory 420 stores instructions executable by at least one processor 410 to cause the at least one processor 410 to perform the method shown in the above embodiments.

[0082] The memory 420 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 420 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 420 may optionally include memory remotely located relative to the processor 410, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0083] The memory 420 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 420 may also include a combination of the above types of memory.

[0084] The computer device also includes a communication interface 430 for communicating with other devices or communication networks.

[0085] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0086] This application provides a computer program product including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method of any embodiment of this application.

[0087] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

[0088] The methods, apparatus, computer devices, computer-readable storage media, or computer program products described in the above embodiments can be implemented by a computer chip or entity, or by a product having a certain function. A typical implementing device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0089] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0090] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, computer devices, computer-readable storage media, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0091] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, computer devices, computer-readable storage media, or computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0092] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0093] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0094] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0095] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, computer equipment, computer-readable storage media, or computer program products are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0096] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0097] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A water-based photovoltaic power plant load prediction method, characterized in that, The method includes: Obtain environmental data sequences generated based on environmental data from the target floating photovoltaic power station; The environmental data sequence is divided based on at least one preset time scale, and the local Hearst exponent is calculated for each subsequence in the division result according to a preset calculation rule to generate a local Hearst exponent sequence. The local Hearst exponent sequence is matched against a pre-acquired set of reference data sequences to determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence, and the historical payload data corresponding to the matching data sequence is obtained. The local Hearst exponent sequence, the matching data sequence, and the historical load data are input into a pre-trained load analysis model to obtain the load prediction results of the target floating photovoltaic power station output by the model.

2. The method of claim 1, wherein, The step of calculating the local Hurst exponent for each subsequence in the partitioning result according to a preset calculation rule, and generating a local Hurst exponent sequence, includes: The environmental data sequence is fitted to obtain a fitting function, and the root mean square of the fluctuation of the fitting function is calculated. For each subsequence in the partitioning result, a subfit function is obtained by fitting the subfit function, and the difference between the common logarithm of the subfit function and the common logarithm of the root mean square of fluctuation is calculated as the sub root mean square of fluctuation. Based on the time scale and the total length of the environmental data sequence, the root mean square of the sub-fluctuation is normalized to obtain the local Hearst exponent sequence.

3. The method of claim 1, wherein, The load analysis model includes three parallel branches; The local Hearst exponent sequence, the matching data sequence, and the historical load data are respectively input to one of the parallel branches; Each of the parallel branches includes at least one analysis structure, which consists of a convolutional layer and a pooling layer; The outputs of the three parallel branches are flattened and spliced, and then input to at least one fully connected layer, where the last fully connected layer serves as the output layer for outputting the power plant load prediction results.

4. The method of claim 3, wherein, When the output layer has one node, the power station load prediction result is the comprehensive equivalent load of the target floating photovoltaic power station in the target time period; When the output layer has 3 nodes, the power station load prediction result is the downwind load, lift load and overturning moment of the target floating photovoltaic power station during the target time period; When the number of nodes in the output layer is greater than 3, the power plant load prediction result also includes at least one of the following parameters: critical component load and photovoltaic panel edge load.

5. The method of claim 1, wherein, The reference data sequences in the reference data sequence set are obtained based on the following method: A reference environmental data sequence is obtained based on historical environmental data of a reference area. The type of the reference area includes at least one of the following types: the target floating photovoltaic power station, other floating photovoltaic power stations with environments similar to the target floating photovoltaic power station, onshore photovoltaic power stations adjacent to the target floating photovoltaic power station, and the region where the target floating photovoltaic power station is located. The reference environment data sequence is divided based on the preset time scale, and the local Hearst exponent is calculated for each subsequence in the division result according to the preset calculation rules to generate the reference data sequence.

6. The method according to claim 5, characterized in that, The step of matching the local Hearst exponent sequence with a pre-acquired set of reference data sequences to determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence includes: Obtain the difference sequence between the environmental data sequence and the reference data sequence, and perform statistical analysis on the difference sequence to obtain a first similarity. Based on the weights pre-set for each type of reference region and the weights pre-set for the first similarity, and the weighted calculation of the type of the reference region corresponding to the reference data sequence and the first similarity, a comprehensive similarity score is obtained, and at least one reference data sequence with the highest comprehensive similarity score is taken as the matching data sequence; or, According to the priority set for each type of reference region, the matching is performed sequentially in the reference data sequence associated with the type of reference region corresponding to each priority. The reference data sequence with the first similarity greater than the preset threshold in the reference data sequence corresponding to the current priority is taken as the matching data sequence. If there is no reference data sequence with the first similarity greater than the preset threshold in the reference data sequence corresponding to the current priority, the matching continues in the reference data sequence corresponding to the next priority.

7. The method of claim 5, wherein, The input data of the load analysis model also includes the environmental data of the target floating photovoltaic power station and / or the environmental data corresponding to the matching data sequence; The load analysis model also includes a fourth parallel branch for analyzing environmental data.

8. The method of claim 1, wherein, The environmental data of the target floating photovoltaic power station includes the wind speed and temperature of the target floating photovoltaic power station; The environmental data refers to historical environmental data collected by the target floating photovoltaic power station within a preset period before the target time period and / or weather forecast data corresponding to the target time period.

9. An on-water photovoltaic power plant load forecasting device characterized by comprising: The device includes: The data acquisition module is used to acquire environmental data sequences generated based on environmental data from the target floating photovoltaic power station; The sequence generation module is used to divide the environmental data sequence based on at least one preset time scale, and calculate the local Hearst exponent for each subsequence in the division result according to a preset calculation rule to generate a local Hearst exponent sequence. The data matching module is used to match the local Hearst exponent sequence in a pre-acquired set of reference data sequences, determine at least one reference data sequence with the highest similarity to the local Hearst exponent sequence as the matching data sequence, and obtain the historical load data corresponding to the matching data sequence. The model prediction module is used to input the local Hearst exponent sequence, the matching data sequence, and the historical load data into a pre-trained load analysis model to obtain the load prediction results of the target floating photovoltaic power station output by the model.

10. A computer device, comprising: include: A memory and a processor, which are in communication connection with each other, the memory has stored computer instructions, and the processor executes the computer instructions to perform the water-based photovoltaic power station load prediction method according to any one of claims 1 to 8.

Images