Floating type offshore photovoltaic array attitude monitoring and overturning early warning method and system

Abstract

The invention discloses a floating type offshore photovoltaic array attitude monitoring and overturning early warning method and system, belongs to the technical field of ocean engineering monitoring, and aims to solve the technical problems of inaccurate attitude monitoring and delayed overturning risk early warning caused by the action of waves and gust of a floating type photovoltaic array in a dynamic ocean environment. Comprising the following steps: acquiring floating body dip angle attitude data, a wave direction spectrum and gust wind speed marine environment data of the whole array in real time, dynamically determining a current inherent rolling period by continuously monitoring free oscillation response under non-extreme excitation and performing spectral analysis, and constructing an overturning risk index; a first-level early warning threshold value and a second-level early warning threshold value are dynamically set according to the current inherent rolling period, the threshold values are lowered when the wave period is close to the inherent period, and a first-level early warning signal or a second-level early warning signal is generated by comparing the real-time risk index with the threshold values. The method is mainly used for real-time safety monitoring and overturning risk early warning of the floating type offshore photovoltaic array, stable operation of a photovoltaic power station is guaranteed, and major accidents are prevented.

Description

Technical Field

[0001] This invention belongs to the field of marine engineering monitoring technology, specifically relating to a method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays. Background Technology

[0002] Floating marine photovoltaic arrays, as an emerging form of marine renewable energy utilization, operate in a complex and harsh environment, constantly subjected to the coupled effects of dynamic marine environmental loads such as wind, waves, and currents. Against this backdrop, effectively monitoring the array's attitude and issuing timely capsizing warnings are crucial technical aspects for ensuring its safe and stable operation. However, existing monitoring and early warning methods still face several pressing problems and challenges in practice.

[0003] First, regarding the reliability of attitude monitoring, there is a significant problem of data source simplification. Currently, the system heavily relies on a single type of real-time monitoring device, such as tilt sensors, to acquire the array's float tilt data. The high salinity, high humidity, and continuous dynamic loads of the marine environment can easily lead to momentary sensor failures, data drift, or complete malfunction. Once such anomalies occur in the current float tilt data, the system will lose its ability to perceive the array's critical attitude state. Although there are ideas for incorporating historical data for estimation, ensuring the timeliness and representativeness of historical data is a challenge. The array's motion response has strong nonlinear characteristics, and its historical attitude database cannot cover all possible combinations of environmental conditions, leading to potentially large deviations between the estimated values ​​and the actual state. Furthermore, after enabling the estimation, there is a lack of an effective mechanism to verify its reliability. Directly using potentially risky estimated data for early warning judgments may introduce new erroneous decision points.

[0004] Secondly, regarding the adaptability of warning thresholds, existing methods typically employ static or semi-static threshold setting strategies. This strategy struggles to accurately reflect the dynamically changing mechanical characteristics of floating photovoltaic arrays in actual sea conditions. The array's inherent roll period is a key dynamic characteristic parameter for resisting capsizing risks, but it is not static. Ballast water distribution, minute changes in buoyancy due to prolonged immersion, wear of structural components, or biofouling can all alter its overall mass and moment of inertia, causing the inherent roll period to drift. Continuing to use fixed or outdated period values ​​to set warning thresholds will lead to a mismatch between the threshold and the array's current actual state. When the actual period has lengthened while the threshold setting remains based on a shorter period, the system may react slowly to potential resonance risks, resulting in delayed warnings; conversely, it may lead to frequent false alarms, reducing the reliability of the warning system.

[0005] Furthermore, existing technologies for handling data anomalies suffer from insufficient real-time performance and verification conflicts. When the main sensor data becomes abnormal and the system switches to estimates based on historical data, an independent and reliable data source is needed for cross-validation. Accelerometers deployed at different locations in the array can infer the overall motion trend of the array, but there is a conversion relationship between this trend and the specific physical quantity of the float's tilt angle, which may introduce calculation errors. Setting a reasonable tolerance value to judge the consistency between the estimated value and the motion trend data is itself an engineering challenge requiring extensive experimental verification. Too strict a tolerance may lead to frequent data source switching and oscillations when data fluctuations are reasonable; too wide a tolerance may fail to effectively eliminate unreliable estimates. More importantly, when verification reveals deviations exceeding limits, the system faces a decision-making difficulty: should it prioritize the estimate derived from historical environment-attitude relationships, or prioritize the motion trend calculated from real-time acceleration signals? The reliability of these two data sources is not absolute under different sea conditions, making the final decision lack a solid and reliable basis.

[0006] In summary, the challenges to the reliability of monitoring data caused by the harshness of the marine environment, the difficulties in adapting early warning thresholds due to the time-varying dynamic characteristics of the array, and the contradiction between real-time performance and accuracy when conducting multi-source verification and decision-making under abnormal data conditions, all jointly restrict the further improvement of the effectiveness of existing floating marine photovoltaic array attitude monitoring and capsizing early warning methods. Summary of the Invention

[0007] One object of the embodiments of the present invention is to solve at least the above-mentioned problems and / or defects, and to provide at least the advantages described below.

[0008] Another objective of this invention is to provide a method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays.

[0009] This study addresses the challenges of inaccurate attitude monitoring, delayed capsizing risk warnings, and susceptibility to environmental interference in floating marine photovoltaic arrays operating in dynamic marine environments due to the coupling effects of waves and gusts. Traditional methods struggle to accurately assess the dynamic response of arrays under complex loads in real time. Furthermore, the nonlinear and time-varying nature of array motion renders static monitoring strategies inadequate for adapting to changing sea conditions, leading to inappropriate warning thresholds, false alarms, or missed alarms, thus impacting system reliability.

[0010] This study aims to address the issue of reduced reliability of monitoring systems for floating body tilt angle data in harsh marine environments due to sensor malfunctions, data drift, or transient failures. High salinity, high humidity, and continuous dynamic loads can easily damage monitoring equipment. Once data anomalies occur, the system cannot obtain reliable attitude information, and historical data estimates may be inaccurate due to incomplete environmental condition coverage. Therefore, a data backup and verification mechanism is urgently needed.

[0011] To achieve the above-mentioned objectives, the present invention employs the following technical solution: A method for attitude monitoring and capsizing early warning of a floating marine photovoltaic array includes the following steps: S1. Real-time acquisition of attitude monitoring data and marine environmental monitoring data of the target photovoltaic array. The attitude monitoring data includes the overall float tilt angle of the array, and the marine environmental monitoring data includes wave direction spectrum and gust wind speed. S2. Dynamically determine the current inherent roll period of the photovoltaic array by continuously monitoring the array's free oscillation response under no extreme environmental excitation and performing spectral analysis. S3. Construct a capsizing risk index F, calculated as: F = α ×|θ| + β × S × G, where θ is the real-time floating body tilt angle after filtering, S is the energy integral value of the wave direction spectrum in the current tilt direction of the photovoltaic array, G is the current gust wind speed, and α and β are weighting coefficients set according to the array structure characteristics; S4, dynamically set the first-level warning threshold and the second-level warning threshold according to the current inherent roll period; when the monitored wave period is close to the current inherent roll period, the threshold is lowered; S5, compare the calculated real-time overturning risk index F with the dynamically set first-level warning threshold and the second-level warning threshold. When F exceeds the first-level warning threshold, a first-level warning signal is generated and the risk period is marked; when F exceeds the second-level warning threshold which is higher than the first level, a second-level alarm signal containing an emergency avoidance command is generated.

[0012] Preferably, in step S1, after acquiring the float tilt angle data in real time, a data reliability verification step is also included: if the current float tilt angle data continues to be abnormal, then a tilt angle estimation value based on historical data is enabled. The tilt angle estimation value is obtained by querying a pre-established historical attitude database that is associated with the current wave direction spectrum and gust wind speed.

[0013] Preferably, after enabling the tilt angle estimation based on historical data, the method further includes the step of: comparing and verifying the tilt angle estimation with the overall motion trend of the array calculated by inverting accelerometer signals deployed at different locations of the photovoltaic array; if the deviation exceeds the tolerance, the motion trend data is preferentially used to participate in the calculation of the overturning risk index F.

[0014] Preferably, in step S1, when acquiring the wave direction spectrum, monitoring data from the physical wave radar and wave estimation data calculated based on the inversion of accelerometer signals deployed on the floating body are integrated.

[0015] Preferably, in step S5, while generating the early warning or alarm signal, the dominant factor that causes the risk index to rise is also output. The dominant factor is determined by analyzing the instantaneous contribution rates of the two components α × |θ| and β ×S × G in the overturning risk index F formula.

[0016] Preferably, after enabling the tilt angle estimation based on historical data, the method further includes the step of: using a Kalman filter algorithm to fuse the tilt angle estimation with the real-time float tilt angle data to generate a corrected float tilt angle value, which is then used for the calculation of the subsequent capsizing risk index F.

[0017] Preferably, in step S4, when dynamically setting the first-level warning threshold and the second-level warning threshold, structural health monitoring data of the photovoltaic array is also introduced, and the threshold is adjusted according to the degree of wear or structural damage of the floating body.

[0018] Preferably, when fusing the monitoring data from the physical wave radar with the wave estimation data calculated based on accelerometer signal inversion, dynamic weights are assigned to each data source according to data quality indicators, wherein the data quality indicators include data update frequency and signal-to-noise ratio.

[0019] Preferably, in step S3, when calculating the overturning risk index F, a simplified model or lookup table pre-trained based on historical data is used for rapid calculation to reduce real-time processing latency.

[0020] Preferably, the dominant factor identifier is determined based on the average contribution rate of the overturning risk index F component within a sliding time window, wherein the length of the sliding time window is dynamically adjusted according to the degree of environmental fluctuation.

[0021] A floating marine photovoltaic array attitude monitoring and capsizing early warning system, used to implement any of the methods described above, comprising: The data acquisition module is used to acquire attitude monitoring data and marine environmental monitoring data of the target photovoltaic array in real time. The attitude monitoring data includes the overall float tilt angle of the array, and the marine environmental monitoring data includes wave direction spectrum and gust wind speed. The inherent period analysis module is used to dynamically determine the current inherent yaw period of the photovoltaic array. It is achieved by continuously monitoring the free oscillation response of the array under no extreme environmental excitation and performing spectrum analysis. The risk index calculation module is used to construct the overturning risk index F, which is calculated as follows: F = α × |θ| + β × S × G, where θ is the real-time floating body tilt angle after filtering, S is the energy integral value of the wave direction spectrum in the current tilt direction of the photovoltaic array, G is the current gust wind speed, and α and β are weighting coefficients set according to the array structure characteristics. The threshold setting module is used to dynamically set the first-level warning threshold and the second-level warning threshold according to the current inherent roll cycle; when the detected wave cycle is close to the current inherent roll cycle, the threshold is lowered. The early warning and alarm module is used to compare the calculated real-time overturning risk index F with the dynamically set first-level and second-level early warning thresholds. When F exceeds the first-level early warning threshold, a first-level early warning signal is generated and the risk period is marked. When F exceeds the second-level early warning threshold, which is higher than the first level, a second-level alarm signal containing emergency avoidance instructions is generated.

[0022] Compared with the prior art, the advantages and beneficial technical effects of the present invention are: This invention effectively improves the accuracy of attitude monitoring and the timeliness of early warning by acquiring multi-source monitoring data in real time and dynamically calculating the inherent roll period and capsizing risk index, combined with dynamic threshold settings. This method can adapt to dynamic changes in the marine environment, reduce false alarms or missed alarms caused by static strategies, enhance the system's response to complex loads, thereby ensuring the safe operation of the photovoltaic array and extending equipment lifespan.

[0023] This invention improves the robustness and fault tolerance of the monitoring system by introducing a data reliability verification step and activating historical estimates when the float tilt angle data is abnormal. This reduces the risk of data interruption due to sensor failure, ensures the continuity of attitude monitoring, and enhances the representativeness of the estimates by associating them with environmental parameters from a historical database, thus avoiding decision-making errors caused by the failure of a single data source.

[0024] This invention improves the accuracy and reliability of data verification by comparing historical estimates with motion trends retrieved from accelerometers, and prioritizing motion trend data when deviations exceed limits. This cross-validation mechanism reduces errors caused by mismatched historical data, optimizes the data selection strategy, makes the overturning risk index calculation closer to the actual array state, and improves early warning accuracy.

[0025] This invention achieves complementary enhancement of multi-source wave direction spectra by fusing wave estimation data retrieved from physical wave radar and accelerometers, thereby improving the comprehensiveness and accuracy of wave state assessment. This fusion method reduces the limitations of single data sources, overcomes data loss or noise problems caused by environmental interference, makes marine environmental monitoring more reliable, and provides a more solid data foundation for risk early warning.

[0026] This invention helps operators quickly identify the source of risk by outputting a dominant factor identifier when an early warning signal is generated and analyzing the instantaneous contribution rate, thereby improving the targeting and efficiency of emergency response. This enables maintenance teams to prioritize critical factors, such as adjusting array orientation or strengthening anchoring, reducing decision-making time and minimizing processing delays or erroneous operations caused by unclear factors.

[0027] This invention employs a Kalman filter algorithm to fuse historical estimates and real-time data to generate a corrected floating body tilt angle value, optimizing data smoothing and noise suppression. This improves the stability and accuracy of the tilt angle data, reduces calculation deviations caused by data conflicts or fluctuations, makes the capsizing risk index more reliable, and enhances the decision-making quality of the overall monitoring system.

[0028] This invention dynamically adjusts the early warning threshold by incorporating structural health monitoring data, making the threshold setting more closely match the actual state of the array and improving the adaptability and sensitivity of the early warning. This enables timely reflection of the impact of structural wear or damage on dynamic characteristics, avoids early warning failure due to outdated thresholds, and enhances the system's safety monitoring capabilities during long-term operation.

[0029] This invention optimizes the evaluation effect of wave direction spectrum by dynamically allocating multi-source data fusion weights based on data quality indicators, thereby improving the intelligence and accuracy of data fusion. This ensures the priority use of high-reliability data sources, reduces the impact of low-quality data, makes wave monitoring more adaptable to environmental changes, and enhances the robustness of the overall method.

[0030] This invention rapidly calculates the capsizing risk index using a simplified model or lookup table, significantly reducing real-time processing latency and improving system response speed under high load conditions. This ensures timely generation of early warning signals, avoids decision-making delays caused by computational complexity, and meets the stringent requirements for real-time monitoring in severe sea conditions.

[0031] This invention identifies dominant factors by using the average contribution rate within a sliding time window, smoothing out the impact of instantaneous fluctuations and improving the stability and representativeness of the identifiers. This makes dominant factor analysis more closely aligned with trend changes, reduces the risk of misjudgment, enhances the credibility of early warning information, and supports more accurate maintenance and emergency decision-making.

[0032] Other advantages, objectives, and features of the embodiments of the present invention will be apparent in part from the following description, and in part will be understood by those skilled in the art through study and practice of the embodiments of the present invention. Detailed Implementation

[0033] To further illustrate the technical means and effects of this invention, the following embodiments are provided for further explanation. The specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0034] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0035] According to one embodiment of the present invention, a method for attitude monitoring and capsizing early warning of a floating marine photovoltaic array includes the following steps: S1. Real-time acquisition of attitude monitoring data and marine environmental monitoring data of the target photovoltaic array, wherein the attitude monitoring data includes the overall float tilt angle of the array, and the marine environmental monitoring data includes wave direction spectrum and gust wind speed; S2. Dynamically determine the current inherent yaw period of the photovoltaic array, which is achieved by continuously monitoring the free oscillation response of the array under no extreme environmental excitation and performing spectrum analysis; S3. Construct a capsizing risk index F, which is calculated as follows: F = α×|θ| + β× S × G, where θ is the real-time floating body tilt angle after filtering, S is the energy integral value of the wave direction spectrum in the current tilt direction of the photovoltaic array, G is the current gust wind speed, and α and β are weighting coefficients set according to the array structure characteristics. S4. Dynamically set the first-level warning threshold and the second-level warning threshold according to the current inherent roll cycle; when the detected wave cycle is close to the current inherent roll cycle, lower the threshold. S5. Compare the calculated real-time overturning risk index F with the dynamically set first-level warning threshold and second-level warning threshold. When F exceeds the first-level warning threshold, generate a first-level warning signal and mark the risk period. When F exceeds the second-level warning threshold which is higher than the first level, generate a second-level alarm signal containing emergency avoidance instructions.

[0036] One feasible implementation is as follows: the array is deployed in a near-shore area, with its floating structure anchored to the seabed via a mooring system. During implementation, attitude monitoring data is acquired in real time using an inclination sensor installed at the center of the floating structure. This sensor measures the overall tilt angle of the array relative to the horizontal plane at a sampling frequency of 10 times per second, and the obtained data is recorded as the floating structure tilt angle θ. Marine environmental monitoring data is collected by wave radar and weather stations deployed around the array. The wave radar measures the wave direction spectrum and calculates the energy integral value S in the current tilt direction of the photovoltaic array; the weather station records gust wind speed G, taking the maximum instantaneous wind speed every 30 seconds. All data is collected and sent to a central processing unit via a wireless transmission network.

[0037] In the central processing unit, the system first performs data preprocessing. The raw float tilt angle θ data is low-pass filtered with a cutoff frequency of 0.1 Hz to eliminate short-term oscillations caused by high-frequency waves, retaining the low-frequency tilt component that reflects the overall stability of the array. The wave direction spectrum data undergoes coordinate transformation, projecting its energy distribution onto the current main tilt direction of the photovoltaic array, and the wave energy integral value S in that direction is calculated, in square meters per second. The gust wind speed G is directly taken as the instantaneous maximum value after quality control, in meters per second.

[0038] The method for dynamically determining the current inherent roll period is as follows: The system continuously monitors the array's free oscillation response during periods of relatively calm wind and waves. When the gust wind speed G is below 5 meters per second and the wave height is below 0.5 meters for 5 consecutive minutes, it is determined to be a state without extreme environmental excitation. During this period, the time series data of the float's roll angle is recorded, and a fast Fourier transform spectrum analysis is performed on the series. The period corresponding to the spectral peak is identified as the current inherent roll period T, in seconds, and updated to the system parameter database. The typical inherent roll period T of a photovoltaic array ranges from 3 to 8 seconds.

[0039] The process of constructing the capsizing risk index F is as follows: First, read the filtered real-time absolute value of the floating body's tilt angle |θ| (in degrees) from the preprocessed data. Second, read the wave energy integral value S in the current tilt direction. Third, read the current gust wind speed G. Based on the structural characteristics of this specific array, pre-set the values ​​of weighting coefficients α and β, for example, α = 0.6 and β = 0.002. The real-time capsizing risk index F is calculated using the formula F = α × |θ| + β × S × G. This index is dimensionless.

[0040] When dynamically setting warning thresholds, the system reads the latest determined current inherent roll period T. The first-level warning threshold is set to 0.5 × T, and the second-level warning threshold is set to 0.8 × T. For example, if T is 5 seconds, then the first-level warning threshold is 2.5, and the second-level warning threshold is 4.0. Simultaneously, the system monitors and calculates the average period Tw of the environmental waves in real time. When the absolute value of the difference between the environmental wave period Tw and the current inherent roll period T is less than 0.5 seconds, the system automatically lowers both warning thresholds by 20%, i.e., the first-level warning threshold is adjusted to 0.4 × T, and the second-level warning threshold is adjusted to 0.64 × T, to address the risk of resonance.

[0041] Finally, the calculated real-time overturning risk index F is compared with a dynamically set threshold. When the F value first exceeds the first-level warning threshold, the system generates a first-level warning signal and marks the risk period on the operation interface. When the F value continues to rise and exceeds the higher second-level warning threshold, the system immediately generates a second-level alarm signal, which includes an emergency evacuation instruction to the maintenance personnel.

[0042] Traditional monitoring methods also collect data on the float's tilt angle θ, wave energy integral S, and gust wind speed G, and calculate a simple risk indicator. However, their warning thresholds are fixed values ​​set based on theoretical calculations during the initial deployment of the array; for example, the first-level warning threshold is fixed at 2.0, the second-level warning threshold is fixed at 3.5, and these values ​​remain unchanged throughout the entire operation. This method, when determining the inherent roll period, only uses theoretical values ​​from the design phase, such as a fixed 4 seconds, and does not update it with changes in the array's state. When assessing risk, it only performs a static comparison between the real-time risk indicator and this set of fixed thresholds.

[0043] The comparative method faces significant problems when environmental conditions cause the array's inherent roll period to drift. For example, if biofouling increases the array's mass, the inherent roll period T increases from 4 seconds to 6 seconds. When the environmental wave period Tw approaches this new, longer inherent period, the array is more prone to resonance. However, because the comparative method's thresholds are fixed at 2.0 and 3.5, and based on older period settings, its warning threshold is too high relative to the current risk, preventing the system from triggering alarms in a timely manner and resulting in a severe delay in warning of potential capsizing risks. Conversely, if the period drifts towards shortening, false alarms may occur frequently due to excessively low thresholds.

[0044] This invention overcomes the inherent limitations of fixed-threshold methods by dynamically determining the array's current inherent roll period and adjusting the warning threshold accordingly. This method enables the warning system to adapt to long-term changes in the array's dynamic characteristics, consistently maintaining a threshold that matches the array's actual state. This adaptability significantly improves the accuracy of warnings, providing timely alerts before real risks materialize and reducing false alarms in non-hazardous conditions, thereby enhancing the overall reliability of the monitoring system. The system continuously monitors free oscillations to update key parameters, ensuring its ability to perceive the array's true state and providing a basis for operational and maintenance decisions.

[0045] According to one embodiment of the present invention, preferably, the floating marine photovoltaic array attitude monitoring and capsizing early warning method, in step S1, after acquiring the tilt angle data of the floating body in real time, further includes a data reliability verification step: If the current float tilt angle data continues to be abnormal, a tilt angle estimation based on historical data will be used. The tilt angle estimation is obtained by querying a pre-established historical attitude database that is associated with the current wave direction spectrum and gust wind speed.

[0046] During the real-time acquisition of the float's tilt angle data, the system reads data from the main tilt angle sensor at a frequency of 10Hz. The system simultaneously monitors the sensor's status parameters, including signal strength and self-diagnostic codes. When a persistent anomaly in the float's tilt angle data is detected, the system initiates a data reliability verification process. The specific criteria for determining a persistent anomaly are: data exceeding the reasonable range of ±15 degrees for 30 consecutive sampling periods, or a data change rate exceeding the physical limit of 5 degrees per second, or a fault status reported by the sensor's self-test.

[0047] At this point, the system automatically activates the tilt angle estimation based on historical data. The historical attitude database is pre-established through long-term monitoring and stores typical floating body tilt angle values ​​corresponding to different wave direction spectrum characteristic parameters and gust wind speeds. The database is organized into intervals of 10 degrees along the main wave direction and intervals of 0.5m wave energy integral values. 2 The system organizes wave direction spectrum and gust wind speed into intervals of 2 m / s each. It queries the currently measured wave direction spectrum and gust wind speed, and searches the database for historical dip angle data under the best matching conditions as an estimate.

[0048] This estimated value is then sent to the data fusion module to replace the abnormal real-time floating body tilt data in subsequent calculations. The system simultaneously records this data replacement event, including the time of occurrence, the type of abnormal data, and the source of the estimated value used. The entire verification process is completed within 200ms, ensuring that the real-time performance of the monitoring system is not affected.

[0049] In this specific case, the master tilt sensor experienced internal reference voltage drift due to long-term salt spray corrosion, resulting in consistently excessively high tilt angle outputs. Traditional monitoring methods failed to identify this systematic deviation, continuing to use the abnormal data in calculating the overturning risk index. This caused the system to continuously calculate an inflated risk index even when the array's actual attitude was normal, ultimately triggering false alarms. Maintenance personnel had to travel to the site to detect the sensor malfunction, consuming significant manpower and resources.

[0050] Another common scenario is that sensors temporarily malfunction due to splashing, outputting invalid data. Traditional methods either ignore these anomalies, resulting in data loss, or continue to use outliers, causing drastic fluctuations in the risk index. Both situations interfere with the judgment of the array's true state, either masking the real risk or generating false alarms.

[0051] This invention introduces a data reliability verification mechanism, automatically switching to estimates based on historical data when data anomalies are detected, effectively ensuring the continuity and reliability of the monitoring system. This method significantly reduces the risk of system failure due to a single sensor malfunction, ensuring reasonable attitude estimation is still provided even in the event of equipment anomalies. By utilizing a pre-established historical correlation model, the system can provide a backup data source when sensor data is unreliable, maintaining the normal operation of monitoring functions. This design enhances tolerance to equipment failures, reduces unnecessary on-site maintenance, and improves overall operational efficiency.

[0052] According to one preferred embodiment of the present invention, the floating marine photovoltaic array attitude monitoring and capsizing early warning method, after enabling the tilt angle estimation based on historical data, further includes the step of: comparing and verifying the tilt angle estimation with the overall motion trend of the array calculated by inversion from accelerometer signals deployed at different locations of the photovoltaic array; if the deviation exceeds the tolerance, the motion trend data is preferentially used to participate in the calculation of the capsizing risk index F.

[0053] Optionally, after enabling tilt angle estimation based on historical data, the system simultaneously initiates the acquisition and processing of accelerometer signals. Accelerometers deployed at the four corners of the photovoltaic array acquire triaxial acceleration data at a frequency of 50Hz. After coordinate transformation, this raw data is first filtered by a high-pass filter to remove the gravitational acceleration component, and then filtered by a low-pass filter to extract the low-frequency motion characteristics of the array. By fusing the motion data from the four points, the overall roll motion trend angle of the array is retrieved.

[0054] The system compares the estimated tilt angle obtained from historical database queries with the motion trend angle calculated by accelerometer inversion in real time. A tolerance of 3 degrees is set for the comparison. This tolerance is based on statistical analysis of the array's motion characteristics under normal sea conditions. If the deviation between the two values ​​remains below 3 degrees for five consecutive sampling periods, the system confirms the reliability of the historical estimate and continues to use it for subsequent calculations. If the deviation exceeds 3 degrees, the system automatically switches the data source, prioritizing the motion trend data calculated based on accelerometer signal inversion for calculating the capsizing risk index.

[0055] The entire comparison and verification process is fully automated and completed within each processing cycle. The system records the results of each comparison, including the deviation value, whether the limit is exceeded, and the final data source used. When a data source switch occurs, the system generates corresponding log information, but does not trigger an alert signal, because this is only an internal data source switch.

[0056] This invention utilizes motion trend data retrieved from accelerometer signals for cross-validation, providing real-time reliability checks on historical estimates. This method effectively identifies discrepancies between historical data and actual conditions, allowing for timely switching to more reliable data sources. Through cross-verification of multi-sensor data, the system significantly improves the accuracy of attitude assessment in complex sea conditions. This design enhances adaptability to unconventional operating conditions, avoids judgment errors caused by a single data source or the failure of inherent models, and provides a more reliable data foundation for capsizing early warning.

[0057] According to one embodiment of the present invention, as a preferred embodiment, in the floating marine photovoltaic array attitude monitoring and capsizing early warning method, in step S1, when acquiring the wave direction spectrum, monitoring data from physical wave radar and wave estimation data calculated based on the inversion of accelerometer signals deployed on the floating body are integrated.

[0058] Optionally, the system acquires wave direction spectrum data through two independent approaches. The first approach uses an X-band marine radar installed on a nearby fixed platform. This radar scans the surrounding sea area at a frequency of 2 Hz, and the wave direction spectrum is obtained by analyzing the backscattered signal of the radar image. The second approach utilizes a triaxial accelerometer array deployed on a floating body, collecting acceleration data at a sampling frequency of 20 Hz. Wave estimation data is obtained by integrating and performing spectral analysis on the acceleration signal, combined with the array motion model inversion calculation.

[0059] The two types of data are synchronized in time at the central processing unit before entering the fusion process. Radar data provides wave field information over a wide range, but it has blind spots at close range and is susceptible to interference from rain and snow. Accelerometer inversion data can accurately reflect the actual wave excitation experienced by the array, but it is limited by the motion characteristics of the floating body itself. The system performs weighted fusion of the wave direction spectrum data from the two sources in a 30-second processing cycle. Before fusion, confidence factors for each type of data are calculated separately. The confidence factor for radar data is calculated based on the signal-to-noise ratio and rainfall intensity, while the confidence factor for accelerometer data is calculated based on the sensor self-test status and spectral clarity.

[0060] The final wave direction spectrum is obtained by weighting the two data sources according to their confidence factors. When the radar data quality is good, its weight can reach 0.7; when rainfall interference occurs, the radar data weight automatically decreases to 0.3, and the accelerometer data weight is increased accordingly. The fused wave direction spectrum is used to calculate the wave energy integral value S in the current tilt direction of the photovoltaic array. This value serves as a key input parameter in the subsequent calculation of the overturning risk index.

[0061] This invention effectively overcomes the limitations of a single data source by integrating physical wave radar and accelerometer inversion data. This method can automatically select the more reliable data source under different environmental conditions, significantly improving the robustness and accuracy of wave monitoring. When one data source is disturbed, the system can compensate and correct using the other data source, ensuring the continuous reliability of wave direction spectrum data. This complementary design enhances adaptability to complex marine environments, provides a more solid data foundation for capsizing risk assessment, and effectively avoids misjudgments caused by the limitations of a single sensor.

[0062] According to one preferred embodiment of the present invention, the floating marine photovoltaic array attitude monitoring and capsizing early warning method, while generating an early warning or alarm signal in step S5, also outputs an identifier of the dominant factor that causes the risk index to rise. The dominant factor is determined by analyzing the instantaneous contribution rates of the two components α × |θ| and β × S × G in the capsizing risk index F formula.

[0063] Optionally, while generating the warning or alarm signal in step S5, the system simultaneously calculates the instantaneous contribution rate of the two components in the capsizing risk index F. The specific calculation process is as follows: first, obtain component A by multiplying α by the absolute value of the floating body's tilt angle at the current moment, and obtain component B by multiplying β by the wave energy integral value S and then by the gust wind speed G. Then, calculate the percentage of component A in the total risk index F, and simultaneously calculate the percentage of component B. The system presets a contribution rate threshold of 60%. When the contribution rate of any component exceeds 60%, the system determines that component is the dominant factor and adds a corresponding identifier to the generated warning signal. For example, when the tilt angle component's contribution rate exceeds 60%, it is identified as tilt angle-dominated risk; when the wave wind component's contribution rate exceeds 60%, it is identified as wave wind-dominated risk. These identifiers are displayed on the operation interface along with the warning signal and recorded in the system log.

[0064] This invention enables operations and maintenance personnel to immediately identify the primary sources of risk by outputting a dominant factor identifier during early warning. This design significantly shortens analysis and decision-making time, guiding personnel to quickly focus on the most critical risk causes. When the identifier indicates tilt as the dominant factor, personnel will prioritize checking the array structure or mooring system; when the identifier indicates wave and wind as the dominant factor, personnel will focus on changes in environmental trends. This targeted guidance significantly improves emergency response efficiency, enabling limited operations and maintenance resources to be precisely allocated to the most critical aspects, effectively avoiding delays or erroneous decisions due to unclear causes.

[0065] According to one preferred embodiment of the present invention, the floating marine photovoltaic array attitude monitoring and capsizing early warning method, after enabling the tilt angle estimation based on historical data, further includes the step of: using a Kalman filter algorithm to fuse the tilt angle estimation with the real-time floating body tilt angle data to generate a corrected floating body tilt angle value, which is then used for the subsequent calculation of the capsizing risk index F.

[0066] Optionally, after enabling tilt angle estimation based on historical data, the system simultaneously initiates a Kalman filter algorithm. The state variables of this filter are set as the float tilt angle and its rate of change. The system's process noise covariance matrix is ​​set according to the typical motion characteristics of the array, while the observation noise covariance is dynamically adjusted based on the reliability of the data source. The historical tilt angle estimate serves as the first observation input, and the real-time, but potentially anomaly-prone, raw tilt angle data from the sensors serves as the second observation input. The filter operates at a frequency of 10Hz. In each filtering cycle, state prediction is performed first, followed by state updates based on the two observation inputs, ultimately outputting a corrected float tilt angle value. This corrected value integrates the trend of historical data and the detailed characteristics of real-time data, while effectively suppressing anomalous noise. This value is directly used in the subsequent calculation of the capsizing risk index F, replacing the raw or estimated tilt angle data.

[0067] This invention employs a Kalman filter algorithm to fuse multi-source data, achieving superior data processing results compared to simple selection or fixed weighted averaging. This algorithm intelligently balances the reliability of different data sources, tracking subtle changes when sensor data is reliable and relying on historical trends when sensor anomalies occur, outputting smoother and more reliable tilt angle data. This dynamic fusion mechanism effectively suppresses sudden anomalies and systematic biases from single data sources, significantly improving the quality and stability of tilt angle data used for risk calculation. Through this optimized data preprocessing, the system provides a more reliable foundation for subsequent risk assessment, improving early warning accuracy from the outset.

[0068] According to one embodiment of the present invention, as a preferred embodiment, the floating marine photovoltaic array attitude monitoring and capsizing early warning method, in step S4, when dynamically setting the first-level early warning threshold and the second-level early warning threshold, also introduces the structural health monitoring data of the photovoltaic array, and adjusts the threshold according to the degree of wear or structural damage of the floating body; In step S3, when calculating the overturning risk index F, a simplified model or lookup table pre-trained based on historical data is used to reduce real-time processing latency.

[0069] Optionally, when dynamically setting the warning threshold in step S4, the system will additionally read the structural health monitoring data of the photovoltaic array. This data includes the average corrosion depth of key parts of the float monitored by corrosion sensors (in millimeters); the average micro-strain value of the main support structure monitored by strain gauges; and the total amount of ballast water inside the float input through periodic testing (in tons). The system calculates a structural health factor from 0 to 1 based on these parameters, where 1 represents a healthy state and 0 represents severe damage. Specifically, a corrosion depth exceeding 2 mm, a micro-strain value exceeding 500, and a ballast water loss exceeding 10% will all lead to a significant decrease in the health factor. The final first-level warning threshold is adjusted to 0.5 multiplied by the inherent roll period T and then multiplied by the structural health factor; the second-level warning threshold is adjusted to 0.8 multiplied by the inherent roll period T and then multiplied by the structural health factor. Therefore, when structural health deteriorates, the warning threshold will decrease accordingly, allowing the system to issue an early warning when the array structure's pressure-bearing capacity decreases.

[0070] In step S3, instead of directly performing floating-point operations to calculate F = α × |θ| + β × S × G, the system uses a simplified polynomial model pre-trained based on a large amount of historical data. This model takes the filtered absolute value of the floating body's tilt angle |θ|, the wave energy integral value S, and the gust wind speed G as inputs. The model form is F = k0 + k1 × |θ| + k2 × S + k3 × G + k4 × |θ| × G, where k0 to k4 are coefficients predetermined through regression analysis. Alternatively, the system maintains a three-dimensional lookup table indexed by |θ|, S, and G, with dimensions of 10 intervals, 10 intervals, and 10 intervals respectively. Each interval cell stores the corresponding historical average F value. The system maps the currently input |θ|, S, and G to specific intervals in the lookup table and quickly outputs an approximate value of the risk index F through linear interpolation. Both methods reduce computation time by more than 70% compared to the direct calculation using the original formula.

[0071] This invention dynamically adjusts the early warning threshold by introducing structural health monitoring data, enabling the early warning system to accurately reflect the current structural state and actual load-bearing capacity of the array. This design quantifies abstract structural damage into specific parameters affecting safety boundaries, making threshold setting more scientific and targeted. When the array's structural health deteriorates, the system automatically lowers the early warning threshold, implementing a more conservative and safe monitoring strategy to effectively prevent accidents caused by insufficient structural strength. Conversely, when the structural condition is good, the system avoids excessive early warnings. This dynamic adaptability significantly improves the matching degree between early warning accuracy and the actual safety state of the array, providing reliable assurance for the full life-cycle safety management of in-service photovoltaic arrays. By using pre-trained simplified models or lookup tables for rapid calculation, the computational burden of the real-time processing system is significantly reduced. This method effectively ensures the timeliness and continuity of risk index calculation under high-frequency updates or high system load conditions. The system can obtain risk assessment results that meet accuracy requirements with less computational resource consumption, avoiding monitoring blind spots or early warning delays caused by processing delays, and ensuring the real-time response capability of the early warning function at critical moments.

[0072] According to one embodiment of the present invention, preferably, the floating marine photovoltaic array attitude monitoring and capsizing early warning method, when fusing monitoring data from physical wave radar and wave estimation data calculated based on accelerometer signal inversion, assigns dynamic weights to each data source according to data quality indicators, wherein the data quality indicators include data update frequency and signal-to-noise ratio.

[0073] Optionally, when fusing physical wave radar monitoring data with wave estimation data calculated based on accelerometer signal inversion, the system calculates a dynamic weight for each data source in real time. Data quality indicators include data update frequency and signal-to-noise ratio (SNR). For radar data, the update frequency is fixed at once every 30 seconds, and the SNR is calculated as the ratio of the intensity of the radar echo signal to the noise floor, in decibels (dB). For accelerometer inversion data, the update frequency is once every 10 seconds, and the SNR is calculated as the ratio of the energy of the motion signal within the wave frequency band to the energy within the high-frequency noise frequency band. The system sets a base weight of 0.5 for each data source. When the update frequency of a data source is lower than 80% of its nominal value, its weight is reduced by 0.2. When the SNR is lower than 20 dB, the weight is further reduced by 0.3. The final weights are normalized to ensure that the sum of the weights of the two data sources is 1. Each processing cycle recalculates the weights based on the latest quality indicators and fuses the two wave direction spectra using a weighted average method.

[0074] This invention dynamically allocates fusion weights based on real-time data quality indicators, enabling the system to adaptively select more reliable data sources. This method effectively improves the accuracy and robustness of wave direction spectrum fusion results. When the quality of a data source deteriorates due to environmental interference or equipment failure, the system can automatically reduce its influence, preventing low-quality data from contaminating the final results. This intelligent weight allocation mechanism ensures the continuous reliability of marine environmental perception, laying a more solid data foundation for subsequent risk assessment.

[0075] According to one embodiment of the present invention, preferably, the floating marine photovoltaic array attitude monitoring and capsizing early warning method determines the dominant factor identifier based on the average contribution rate of the capsizing risk index F component within a sliding time window, wherein the length of the sliding time window is dynamically adjusted according to the degree of environmental fluctuation. Optionally, when identifying the dominant factor, the system no longer relies on the instantaneous contribution rate, but instead analyzes based on the average contribution rate within a sliding time window. The initial length of the sliding time window is set to 300 seconds. Within this window, the system calculates the contribution rates of the two components α × |θ| and β × S × G in the overturning risk index F formula every 10 seconds and records their values. At each analysis moment, the system calculates the average contribution rate of these two components within the most recent 300-second time window. The threshold for determining the dominant factor remains 60%, but the judgment is based on whether the average contribution rate exceeds this threshold. Furthermore, the window length is dynamically adjusted according to the degree of environmental fluctuation. The degree of environmental fluctuation is quantified by calculating the standard deviation of the gust wind speed G within the window. When the standard deviation exceeds 3 m / s, it is determined that the environmental fluctuation is severe, and the window length is automatically shortened to 120 seconds to reflect the changing trend more quickly; when the standard deviation is below 1 m / s, it is determined that the environment is stable, and the window length returns to 300 seconds to maintain stability.

[0076] This invention effectively smooths out interference from instantaneous data fluctuations by using the average contribution rate within a sliding time window for judgment. This method captures the long-term trend of risk factor contributions, rather than short-term noise, making the identification results of dominant factors more stable and reliable. Dynamically adjusting the window length allows the system to adapt to both stable and rapidly changing environments, balancing the stability and sensitivity of identification. This provides maintenance personnel with more consistent and valuable decision support, avoiding misunderstandings and decision-making difficulties caused by frequent changes in identifiers.

[0077] According to one aspect of the present invention, a floating marine photovoltaic array attitude monitoring and capsizing early warning system is provided for implementing the method described in any one of the present inventions, comprising: The data acquisition module is used to acquire attitude monitoring data and marine environmental monitoring data of the target photovoltaic array in real time. The attitude monitoring data includes the overall float tilt angle of the array, and the marine environmental monitoring data includes wave direction spectrum and gust wind speed. The inherent period analysis module is used to dynamically determine the current inherent yaw period of the photovoltaic array. It is achieved by continuously monitoring the free oscillation response of the array under no extreme environmental excitation and performing spectrum analysis. The risk index calculation module is used to construct the overturning risk index F, which is calculated as follows: F = α × |θ| + β × S × G, where θ is the real-time floating body tilt angle after filtering, S is the energy integral value of the wave direction spectrum in the current tilt direction of the photovoltaic array, G is the current gust wind speed, and α and β are weighting coefficients set according to the array structure characteristics. The threshold setting module is used to dynamically set the first-level warning threshold and the second-level warning threshold according to the current inherent roll cycle; when the detected wave cycle is close to the current inherent roll cycle, the threshold is lowered. The early warning and alarm module is used to compare the calculated real-time overturning risk index F with the dynamically set first-level and second-level early warning thresholds. When F exceeds the first-level early warning threshold, a first-level early warning signal is generated and the risk period is marked. When F exceeds the second-level early warning threshold, which is higher than the first level, a second-level alarm signal containing emergency avoidance instructions is generated.

[0078] The system comprises five main functional modules. The data acquisition module consists of a tilt sensor array mounted at the center of the buoy, surrounded by a wave radar and a weather station. The tilt sensors acquire data at a frequency of 10 Hz, the wave radar detects the wave direction spectrum, and the weather station records gust wind speeds. All data is collected and transmitted to the central processor via a wireless network.

[0079] The inherent period analysis module runs within the central processing unit. This module continuously monitors environmental data, determining a state of no extreme environmental excitation when gust wind speeds are below 5 meters per second and wave heights are below 0.5 meters for five consecutive minutes. The module then initiates free oscillation response analysis, recording the time series of the floating body's roll angle. Through Fast Fourier Transform (FFT) spectral analysis, the period corresponding to the spectral peak is identified as the current inherent roll period and stored in the system database. The inherent roll period of a typical photovoltaic array ranges from 3 to 8 seconds.

[0080] The risk index calculation module receives preprocessed data from the data acquisition module. The module first performs a low-pass filter on the float's tilt angle, with a cutoff frequency set to 0.1 Hz. It then reads the energy integral value of the wave direction spectrum in the current tilt direction of the array, along with the gust wind speed, based on the absolute value of the filtered tilt angle. Weighting coefficients α and β are set to 0.6 and β to 0.002 according to the specific array structural characteristics. The real-time capsizing risk index F is then calculated using the formula.

[0081] The threshold setting module dynamically sets the warning threshold based on the current inherent roll cycle output by the inherent cycle analysis module. The first-level warning threshold is set to 0.5 times the current inherent roll cycle, and the second-level warning threshold is set to 0.8 times the current inherent roll cycle. Simultaneously, the module monitors the environmental wave cycle in real time. When the difference between the environmental wave cycle and the current inherent roll cycle is less than 0.5 seconds, both warning thresholds are simultaneously lowered by 20%.

[0082] The early warning and alarm module continuously compares the real-time overturning risk index F output by the risk index calculation module with the dynamic threshold output by the threshold setting module. When the F value exceeds the first-level early warning threshold, the module generates a first-level early warning signal and marks the risk period on the operation interface. When the F value exceeds the second-level early warning threshold, the module immediately generates a second-level alarm signal, which includes an emergency evacuation instruction issued to maintenance personnel.

[0083] Traditional monitoring systems typically consist of a data acquisition module that only includes basic sensor periodicity analysis modules, risk index calculations based on fixed periodic values ​​determined during initial deployment, threshold setting modules using static thresholds, and alarm modules that only perform simple threshold comparisons. These modules are loosely coupled and lack a collaborative working mechanism.

[0084] In a specific case, the traditional system's inherent periodicity analysis module used a fixed 4-second period value, while the actual array's period had drifted to 6 seconds due to biological attachment. When the environmental wave period approached 6 seconds, the array experienced significant swaying. Because the threshold setting module still used a static threshold based on a 4-second period, the early warning alarm module failed to trigger the alarm in time, resulting in a response delay. Another problem was that when the tilt sensor in the data acquisition module malfunctioned, the entire system chain was interrupted, and effective monitoring could not be provided.

[0085] This invention achieves end-to-end integrated optimization from data acquisition to early warning output by constructing a highly collaborative modular system. Each specialized module performs its specific function while working closely together to ensure the integrity and reliability of the monitoring process. The modular design enhances functional independence and system maintainability; upgrading or repairing one module does not affect the normal operation of other modules. This architecture significantly improves the stability and adaptability of the entire monitoring system, providing comprehensive safety assurance for floating offshore photovoltaic arrays.

[0086] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the embodiments of the present invention. Other modifications can be readily implemented by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the embodiments of the present invention are not limited to the specific details.

Claims

1. A method for attitude monitoring and capsizing early warning of a floating marine photovoltaic array, characterized in that, Includes the following steps: S1. Real-time acquisition of attitude monitoring data and marine environmental monitoring data of the target photovoltaic array, wherein the attitude monitoring data includes the overall float tilt angle of the array, and the marine environmental monitoring data includes wave direction spectrum and gust wind speed; S2. Dynamically determine the current inherent yaw period of the photovoltaic array, which is achieved by continuously monitoring the free oscillation response of the array under no extreme environmental excitation and performing spectrum analysis; S3. Construct a capsizing risk index F, which is calculated as follows: F = α×|θ| +β × S × G, where θ is the real-time floating body tilt angle after filtering, S is the energy integral value of the wave direction spectrum in the current tilt direction of the photovoltaic array, G is the current gust wind speed, and α and β are weighting coefficients set according to the array structure characteristics. S4. Dynamically set the first-level warning threshold and the second-level warning threshold according to the current inherent roll cycle; when the detected wave cycle is close to the current inherent roll cycle, lower the threshold. S5. Compare the calculated real-time overturning risk index F with the dynamically set first-level warning threshold and second-level warning threshold. When F exceeds the first-level warning threshold, generate a first-level warning signal and mark the risk period. When F exceeds the second-level warning threshold which is higher than the first level, generate a second-level alarm signal containing emergency avoidance instructions.

2. The method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays as described in claim 1, characterized in that, In step S1, after acquiring the float tilt angle data in real time, a data reliability verification step is also included: If the current float tilt angle data continues to be abnormal, a tilt angle estimation based on historical data will be used. The tilt angle estimation is obtained by querying a pre-established historical attitude database that is associated with the current wave direction spectrum and gust wind speed.

3. The method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays as described in claim 2, characterized in that, After enabling the tilt angle estimation based on historical data, the process further includes the following steps: comparing and verifying the tilt angle estimation with the overall motion trend of the array calculated by inverting accelerometer signals deployed at different locations in the photovoltaic array; if the deviation exceeds the tolerance, the motion trend data is used first in the calculation of the overturning risk index F.

4. The method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays as described in claim 1, characterized in that, In step S1, when acquiring the wave direction spectrum, monitoring data from the physical wave radar and wave estimation data calculated based on the inversion of accelerometer signals deployed on the floating body are fused together.

5. The method for attitude monitoring and capsizing early warning of a floating marine photovoltaic array as described in claim 1, characterized in that, In step S5, while generating a warning or alarm signal, the dominant factor that causes the risk index to rise is also output. The dominant factor is determined by analyzing the instantaneous contribution rate of the two components α × |θ| and β × S × G in the overturning risk index F formula.

6. The method for attitude monitoring and capsizing early warning of a floating marine photovoltaic array as described in any one of claims 2 or 3, characterized in that, After enabling the tilt angle estimation based on historical data, the process further includes the following steps: using a Kalman filter algorithm to fuse the tilt angle estimation with the real-time float tilt angle data to generate a corrected float tilt angle value, which is then used for the subsequent calculation of the capsizing risk index F.

7. The method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays as described in claim 1, characterized in that, In step S4, when dynamically setting the first-level warning threshold and the second-level warning threshold, structural health monitoring data of the photovoltaic array is also introduced, and the threshold is adjusted according to the degree of wear or structural damage of the floating body. In step S3, when calculating the overturning risk index F, a simplified model or lookup table pre-trained based on historical data is used to reduce real-time processing latency.

8. The method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays as described in claim 4, characterized in that, When fusing monitoring data from physical wave radar with wave estimation data calculated based on accelerometer signal inversion, dynamic weights are assigned to each data source according to data quality indicators, including data update frequency and signal-to-noise ratio.

9. The method for attitude monitoring and capsizing early warning of floating marine photovoltaic arrays as described in claim 5, characterized in that, The dominant factor is identified based on the average contribution rate of the overturning risk index F component within a sliding time window, wherein the length of the sliding time window is dynamically adjusted according to the degree of environmental fluctuation.

10. A floating marine photovoltaic array attitude monitoring and capsizing early warning system, used to implement the method as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to acquire attitude monitoring data and marine environmental monitoring data of the target photovoltaic array in real time. The attitude monitoring data includes the overall float tilt angle of the array, and the marine environmental monitoring data includes wave direction spectrum and gust wind speed. The inherent period analysis module is used to dynamically determine the current inherent yaw period of the photovoltaic array. It is achieved by continuously monitoring the free oscillation response of the array under no extreme environmental excitation and performing spectrum analysis. The risk index calculation module is used to construct the overturning risk index F, which is calculated as follows: F = α × |θ| + β × S × G, where θ is the real-time floating body tilt angle after filtering, S is the energy integral value of the wave direction spectrum in the current tilt direction of the photovoltaic array, G is the current gust wind speed, and α and β are weighting coefficients set according to the array structure characteristics. The threshold setting module is used to dynamically set the first-level warning threshold and the second-level warning threshold according to the current inherent roll cycle; when the detected wave cycle is close to the current inherent roll cycle, the threshold is lowered. The early warning and alarm module is used to compare the calculated real-time overturning risk index F with the dynamically set first-level and second-level early warning thresholds. When F exceeds the first-level early warning threshold, a first-level early warning signal is generated and the risk period is marked. When F exceeds the second-level early warning threshold, which is higher than the first level, a second-level alarm signal containing emergency avoidance instructions is generated.