Offshore photovoltaic health monitoring method and device based on dynamic perception and closed-loop control

Abstract

The invention relates to the field of photovoltaic technology, and discloses an offshore photovoltaic health monitoring method and device based on dynamic sensing and closed-loop control, and the method comprises the steps: obtaining original multi-source sensing data collected by a multi-source sensor for offshore photovoltaic equipment at a current collection moment; calculating a corresponding single health factor by using the multi-source sensing data, and calculating a health index based on the single health factor and a weight corresponding to the single health factor; and determining the health state of the offshore photovoltaic equipment based on the health index, and predicting a health index curve of the future time period and the residual life of the offshore photovoltaic equipment. According to the invention, the problems of single sensing dimension, poor model adaptability and lack of life prediction in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic technology, specifically to a method and device for monitoring the health of marine photovoltaic systems based on dynamic sensing and closed-loop control. Background Technology

[0002] With the rapid development of marine clean energy, offshore photovoltaic power generation has become an important direction for energy transformation due to its advantages such as not occupying land resources, high power generation efficiency, and strong environmental compatibility. However, the marine environment has complex characteristics such as high salt spray, strong corrosion, wind and wave impact, and biological attachment, which pose a severe challenge to the long-term stable operation of photovoltaic systems. Therefore, developing efficient and reliable marine photovoltaic health monitoring technology is crucial to ensuring its safety and economic viability throughout its entire life cycle.

[0003] Existing methods for monitoring the health of marine photovoltaic systems are mostly limited to monitoring single parameters, such as focusing only on isolated indicators like module temperature or output voltage. They lack comprehensive perception and coupled modeling of multi-source heterogeneous data such as structural stress, corrosion status, and marine organism attachment, resulting in incomplete health status assessments. At the same time, health index calculation models often use fixed weights, which cannot dynamically respond to extreme conditions such as wind and wave coupling, resulting in significant prediction errors and a general lack of effective prediction capabilities for the remaining lifespan of the equipment. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method and device for monitoring the health of marine photovoltaic systems based on dynamic sensing and closed-loop control, in order to solve the problems of single sensing dimension, poor model adaptability and lack of lifetime prediction in the prior art.

[0005] In a first aspect, embodiments of the present invention provide a method for monitoring the health of marine photovoltaic systems based on dynamic sensing and closed-loop control, the method comprising: Furthermore, acquire the raw multi-source sensing data collected by the multi-source sensors for the offshore photovoltaic equipment at the current acquisition moment; The corresponding individual health factors are calculated using the multi-source sensing data, and a health index is calculated based on the individual health factors and their corresponding weights. The health status of the offshore photovoltaic equipment is determined based on the health index, and the health index curve for future time periods and the remaining lifespan of the offshore photovoltaic equipment are predicted.

[0006] Furthermore, the acquisition of raw multi-source sensing data collected by the multi-source sensors for the offshore photovoltaic equipment at the current acquisition time includes: Obtain environmental information about the environment where the offshore photovoltaic equipment was located at the previous data collection time, and determine the weather type based on the environmental information; The collection period for each indicator at the current collection time is determined based on the weather type. The sensing data corresponding to each indicator is collected according to the collection cycle, and the multi-source sensing data is constructed based on the sensing data of each indicator.

[0007] Furthermore, the step of calculating corresponding individual health factors using the multi-source sensing data, and calculating a health index based on the individual health factors and their corresponding weights, includes: The power attenuation factor is calculated by the ratio of the maximum output power to the nominal power of the offshore photovoltaic equipment; the hot spot factor is calculated by the temperature change of the offshore photovoltaic equipment; the structural attitude factor is calculated by the cosine of the combined pitch and roll angles; the corrosion factor is calculated by the corrosion condition of the offshore photovoltaic equipment; and the adhesion factor is calculated by the adhesion-related parameters. The initial health index is obtained by weighted summation based on the individual health factors and their corresponding weights. The initial health index is corrected and smoothed to obtain the health index of the offshore photovoltaic equipment at the current data collection time.

[0008] Furthermore, the correction and smoothing of the initial health index to obtain the health index of the offshore photovoltaic equipment at the current data collection time includes: The initial health index was nonlinearly corrected using the hyperbolic tangent function to obtain the corrected health index; Obtain the health index and weighting coefficient at the previous data collection time. The weighting coefficient is obtained after calibration based on three steps: on-site data sampling, grid search, and inflection point selection. The corrected health index is smoothed using the health index from the previous data collection time and the weighting coefficient to obtain the health index of the offshore photovoltaic equipment at the current data collection time.

[0009] Furthermore, the predicted health index curve for future time periods and the remaining lifespan of the offshore photovoltaic equipment include: Acquire multidimensional time series data within a historical time period, including health index, wind speed, wave height, salinity, and load; The multidimensional time series data is analyzed using a predictive model to obtain the health index curve for future time periods; Find the future moment when the health index first falls below a preset threshold from the health index curve; Calculate the difference between the future time and the current sampling time, and use the difference as the remaining lifetime.

[0010] Furthermore, after determining the health status of the offshore photovoltaic equipment based on the health index, the method further includes: Obtain the weather type at the current data collection time; Configure the local network according to the network configuration mode corresponding to the weather type; If the weather type is normal, the cached data will be compressed and sent back in batches through the regular channel at preset intervals; if the health index is lower than the preset threshold, or if the offshore photovoltaic equipment malfunctions, the preemptive small packet mechanism will be activated, and the data will be continuously sent up through the satellite high-priority channel within a unit of time.

[0011] Furthermore, after determining the health status of the offshore photovoltaic equipment based on the health index, the method further includes: The corresponding display strategy and alarm method are determined based on the numerical range that the health index falls into; The corresponding prompts will be displayed according to the aforementioned display strategy.

[0012] Secondly, embodiments of the present invention provide a marine photovoltaic health monitoring device based on dynamic sensing and closed-loop control, the device comprising: The acquisition module is used to acquire the raw multi-source sensing data collected by the multi-source sensors from the offshore photovoltaic equipment at the current acquisition time; The calculation module is used to calculate the corresponding individual health factors using the multi-source sensing data, and to calculate the health index based on the individual health factors and the weights corresponding to the individual health factors. The prediction module is used to determine the health status of the offshore photovoltaic equipment based on the health index, and to predict the health index curve for future time periods and the remaining lifespan of the offshore photovoltaic equipment.

[0013] Thirdly, embodiments of the present invention provide a computer device, including: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method described in the first aspect or any corresponding embodiment thereof.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof.

[0015] The method provided in this application has the following beneficial effects: The method provided in this application acquires multi-source sensor data, enabling synchronous perception and deep fusion of the electrical, thermal, structural, and environmental states of offshore photovoltaic equipment. This overcomes the limitations of single-parameter monitoring and lays a reliable data foundation for comprehensively assessing the equipment's health status. By calculating a health index using multi-source data and integrating multiple health factors with adaptive weights, the assessment model can dynamically respond to complex and changing marine conditions, significantly improving the accuracy of health status assessment and robustness under severe weather conditions, effectively reducing false alarm rates. Based on the health index, the method determines the health status and predicts remaining lifespan, achieving a leap from passive alarm to proactive prediction. It can not only diagnose current faults in real time but also accurately predict equipment performance degradation trends and remaining lifespan, providing crucial decision-making basis for predictive maintenance. This significantly improves operation and maintenance efficiency and reduces the total lifespan operation and maintenance cost and power generation loss. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in 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 the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a marine photovoltaic health monitoring method based on dynamic sensing and closed-loop control according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the multi-source signal acquisition and weather pattern determination according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the workflow of a hierarchical communication strategy according to an embodiment of the present invention; Figure 4 This is a structural block diagram of a marine photovoltaic health monitoring system based on dynamic sensing and closed-loop control according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the workflow of the sensing layer in a health monitoring system according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the application layer workflow in a health monitoring system according to an embodiment of the present invention; Figure 7 This is a structural block diagram of a marine photovoltaic health monitoring device based on dynamic sensing and closed-loop control according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] According to embodiments of the present invention, a method and apparatus for monitoring the health of marine photovoltaic systems based on dynamic sensing and closed-loop control are provided. It should be noted that the steps shown in the flowcharts 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 flowcharts, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0020] This embodiment provides a method for monitoring the health of marine photovoltaic systems based on dynamic sensing and closed-loop control. Figure 1 This is a flowchart of a marine photovoltaic health monitoring method based on dynamic sensing and closed-loop control according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain the raw multi-source sensing data collected by the multi-source sensors for the offshore photovoltaic equipment at the current acquisition time.

[0021] In this embodiment, firstly, environmental information from the previous data collection time is acquired. This environmental information specifically includes the 10-minute average wind speed, significant wave height, salt spray concentration, atmospheric visibility, and lightning electric field intensity measured by an ultrasonic anemometer, radar wave height meter, ion mobility spectrometer salt spray sensor, visibility meter, and fast antenna electric field meter, respectively. Based on four preset thresholds, the current weather type is determined and marked as normal mode or severe mode. Next, according to the determined weather type, differentiated acquisition cycles are dynamically configured for sensors monitoring different indicators: for example, in normal mode, the sampling cycle of the cascade IV curve scanner and infrared thermal imager / distributed fiber optic thermometer is set to 10 minutes, and the sampling cycle of the six-degree-of-freedom MEMS and fiber optic tension meter is set to 1 minute; in severe mode, the sampling cycle of the above electrical and thermal measurements is immediately shortened to 2 minutes, and the sampling cycle of attitude and tension measurements is further shortened to 10 seconds to achieve rapid response to extreme conditions, while the laser-ultrasonic Lamb wave device and polarized light imaging device maintain their original 1-hour and 30-minute sampling cycles, respectively. Finally, all sensors synchronously collect data according to their set acquisition cycle. The sensing data covers physical quantities in multiple dimensions such as electrical, thermal, corrosion, adhesion, attitude, tension and environment. To ensure the spatiotemporal consistency of the data, all raw data are synchronized with the PPS second pulse of the Beidou satellite with high precision after acquisition via the NTP protocol, and are organized and temporarily cached to construct a complete and time-aligned raw multi-source sensing dataset for health assessment.

[0022] As an example, such as Figure 2 As shown, the multi-source signal acquisition and weather pattern judgment process can include: First, the multi-source sensors perform acquisition operations to obtain environmental information (average wind speed, significant wave height, atmospheric visibility, lightning electric field, etc.); determine whether the environmental information meets preset conditions (wind speed ≥ 20 m / s, or significant wave height ≥ 3 m, or lightning electric field ≥ 2 kV / m, or visibility < 1 km). If yes, it is determined to be severe weather, and the electrical signal sampling period is set to 2 min, and the attitude tension sampling period is set to 10 s; if no, it is determined to be ordinary weather, and the electrical signal sampling period is set to 10 min, and the attitude tension sampling period is set to 1 min; based on the sampling period, electrical, thermal, corrosion, adhesion, attitude, and tension information are obtained.

[0023] Step S102: Calculate the corresponding individual health factors using multi-source sensing data, and calculate the health index based on the individual health factors and their corresponding weights.

[0024] In this embodiment, firstly, five core individual health factors are calculated using multi-source sensing data: the power attenuation factor is calculated by the ratio of the maximum output power of the marine photovoltaic string to its factory nominal power at the current acquisition time, directly reflecting the degree of electrical performance attenuation; the hot spot factor is obtained by using two-dimensional temperature field data of the component surface obtained by fusing infrared thermal imager and distributed fiber optic temperature measurement to identify and calculate the temperature difference ratio between hot spot area and normal area, in order to quantify the hot spot effect; the structural attitude factor is obtained by calculating the cosine value of the resultant angle of the pitch and roll angles measured by a six-degree-of-freedom MEMS sensor to evaluate the tilt stability of the floating platform; the corrosion factor is obtained by using the A0 mode wave group velocity excited by the laser-ultrasonic Lamb wave device, combined with the calibration coefficient obtained by calibration test, to inversely calculate the corrosion thickness of the support and characterize the degree of structural corrosion; the adhesion factor is obtained by comparing the degree of polarization calculated by polarized light imaging with the calibration database and outputting an adhesion index of 0 to 5 levels to quantify the impact of marine organism adhesion on light intensity shading.

[0025] Next, the five individual health factors are weighted and fused to obtain an initial health index. The weights of each factor are pre-determined and fixed in the edge nodes using a Bayesian optimization algorithm. The weights are weather-adaptive. For example, in severe weather conditions, the weights of the structural attitude factor are dynamically increased and the weights of the power attenuation factor are correspondingly decreased to enhance the sensitivity to structural safety risks.

[0026] Finally, the initial health index is refined and post-processed to obtain the final health index: the hyperbolic tangent function is used to perform nonlinear correction on the initial health index to map and compress the index value into a more stable range that conforms to engineering experience; the health index at the previous acquisition time and a weight coefficient predetermined by a three-step calibration method (on-site data sampling, grid search, and inflection point selection) are obtained; and a first-order exponential smoothing algorithm is used to perform a weighted average of the corrected health index and the historical health index, thereby effectively filtering out data noise and outputting a final health index at the current acquisition time that can smoothly reflect the trend of the equipment's health status.

[0027] Step S103: Determine the health status of the offshore photovoltaic equipment based on the health index, and predict the health index curve and the remaining lifespan of the offshore photovoltaic equipment for future time periods.

[0028] In this embodiment, firstly, the current health index is matched with a preset health status grading standard to determine the real-time health status of the equipment. Then, a remaining lifespan prediction process is executed: historical multidimensional time series data from the past 168 hours is retrieved from the edge cache. This dataset includes not only the health index itself but also its key influencing factors, such as wind speed, wave height, salt spray concentration, and structural load. This time series is input into an LSTM prediction model incorporating an attention mechanism. This model captures long-term dependencies and key features in the data, outputting a health index prediction curve for a specific future time period. The length of the future time period is dynamically determined by the current weather type (30 days for normal weather and 72 hours for severe weather). The prediction curve is used to find the future point where the health index first falls below the fault threshold of 0.7, and the time difference between this predicted fault point and the current sampling point is calculated. This difference is ultimately determined as the remaining lifespan of the offshore photovoltaic equipment. Furthermore, after determining the health status, this step also links with subsequent communication and display modules to form a closed loop: for example, configuring the communication parameters of the LoRa MESH local network according to the current weather type, and deciding on the communication strategy based on the health status; at the same time, according to the numerical range of the health index, driving the ground digital twin platform to visualize it with four color layers of green, yellow, orange, and red, triggering corresponding sound alarms, and pushing alarm cards of the corresponding colors on the WeChat mini program and generating traceable closed-loop operation and maintenance work orders, thereby achieving accurate judgment of equipment status, advanced prediction of future risks, and accurate delivery of operation and maintenance instructions.

[0029] In this embodiment of the application, acquiring the raw multi-source sensing data collected by the multi-source sensors for the marine photovoltaic equipment at the current acquisition time includes the following steps A1-A3: Step A1: Obtain environmental information of the marine photovoltaic equipment at the previous data collection time, and determine the weather type based on the environmental information.

[0030] Specifically, the system first acquires marine environmental information from the environmental sensor array deployed on the offshore photovoltaic platform at the time of the previous data collection (i.e., at the end of the most recent complete sampling cycle). This information includes: the 10-minute average wind speed provided by the ultrasonic anemometer, the significant wave height measured by the radar wave height meter, the salt spray concentration detected by the ion mobility spectrometer salt spray sensor, the atmospheric visibility obtained by the visibility meter, and the lightning electric field intensity monitored by the fast antenna electric field meter. Next, the environmental parameters are compared with four preset key thresholds, and the current weather type is determined according to classification rules: when wind speed <20m / s, significant wave height <3m, lightning electric field <2kV / m, and visibility ≥1km are simultaneously met, it is classified as "normal mode"; while when any parameter reaches or exceeds its threshold, i.e., wind speed ≥20m / s, or significant wave height ≥3m, or lightning electric field ≥2kV / m, or visibility <1km, it is immediately classified as "severe mode".

[0031] Step A2: Determine the collection period for each indicator at the current collection time based on the weather type.

[0032] Specifically, the system dynamically assigns differentiated acquisition cycles to different monitoring indicators based on weather type (normal mode or severe mode). The system internally has a pre-defined acquisition cycle configuration table tied to weather type: for electrical measurements (such as maximum power point, open-circuit voltage, short-circuit current, and insulation resistance obtained from string-level IV curve scanners) and thermal measurements (such as component surface temperature field and hotspot temperature difference obtained from the fusion of infrared thermal imager and distributed fiber optic temperature measurement), the acquisition cycle is set to 10 minutes in normal mode; while in severe mode, the acquisition cycle for key operating parameters is immediately and dynamically shortened to 2 minutes to achieve rapid capture and monitoring of equipment electrical performance and thermal status. For key indicators reflecting structural safety, namely platform attitude and mooring tension measured by a six-degree-of-freedom MEMS and fiber optic tension meter, the acquisition cycle is 1 minute in normal mode and significantly increased to 10 seconds in severe mode to monitor the instantaneous impact of severe environments such as wind and waves on structural stability at high frequency. For indicators that change relatively slowly, such as the corrosion thickness of the support structure retrieved using a laser-ultrasonic Lamb wave device, and the marine organism attachment level calculated and compared using polarized light imaging, their fixed 1-hour and 30-minute acquisition cycles are maintained, respectively, and do not change with weather patterns. This weather-type-based, fine-grained dynamic adjustment strategy for acquisition cycles is automatically triggered by severe weather indicator variables without manual intervention. The ultimate goal is to optimize the resource allocation and power consumption of edge computing nodes while ensuring high-frequency, real-time monitoring of key parameters, achieving the best balance between real-time monitoring and system energy consumption.

[0033] Step A3: Collect the sensing data corresponding to each indicator according to the collection cycle, and construct multi-source sensing data based on the sensing data of each indicator.

[0034] Specifically, based on the acquisition cycle of various indicators matched with the weather type, acquisition commands are sent to the corresponding sensors to control them to perform synchronous data acquisition. The cascade-level IV curve scanner acquires electrical indicators, including maximum power point, open-circuit voltage, short-circuit current, and insulation resistance, at a set cycle (10 minutes in normal mode, 2 minutes in severe mode). The infrared thermal imager and distributed fiber optic temperature measurement device synchronously acquire the two-dimensional temperature field of the component surface and the backplane line temperature, and obtain the hot spot temperature difference after spatiotemporal registration. The six-degree-of-freedom MEMS and fiber optic tension meter acquire platform attitude and mooring tension at a set high-frequency cycle (1 minute in normal mode, 10 seconds in severe mode). The laser-ultrasonic Lamb wave device and polarized light imaging device acquire corrosion thickness and marine organism attachment levels at their inherent 1-hour and 30-minute cycles, respectively. All raw data acquired by the sensors are synchronized with the PPS second pulse of the BeiDou satellite using the NTP protocol with high precision, ensuring that the data has a unified Unix timestamp. Subsequently, using "matrix ID-string ID-pixel row and column-Unix timestamp" as the primary key, all time-aligned sensing data covering multiple dimensions such as electrical, thermal, corrosion, adhesion, attitude, tension, and environment are written into the SQLite cache database deployed on the edge computing node, thereby constructing a complete, spatiotemporally consistent, original multi-source sensing data set that can be used for health assessment.

[0035] By acquiring environmental information to determine the weather type, the system can perceive and adapt to complex changes in the marine environment. The collection cycle of each indicator is dynamically adjusted according to the weather type, achieving a balance between high-frequency monitoring of key parameters under severe weather conditions and resource conservation under normal weather conditions. Multi-source sensing data is constructed according to the collection cycle, ensuring the timeliness and completeness of data collection and providing a high-quality, multi-dimensional input data foundation for subsequent health assessments.

[0036] In this embodiment of the application, corresponding individual health factors are calculated using multi-source sensing data, and a health index is calculated based on the individual health factors and their corresponding weights, including the following steps B1-B3: Step B1 involves calculating the power attenuation factor by the ratio of the maximum output power to the nominal power of the offshore photovoltaic equipment, the hot spot factor by the temperature change of the offshore photovoltaic equipment, the structural attitude factor by the cosine of the combined pitch and roll angles, the corrosion factor by the corrosion condition of the offshore photovoltaic equipment, and the adhesion factor by the adhesion-related parameters.

[0037] Specifically, using multi-source sensing data, five key individual health factors are calculated. First, the power attenuation factor... Maximum output power acquired in real time by a cascade IV curve scanner With the factory-rated power of this series The ratio is calculated as follows: This directly quantifies the degree of degradation in current power generation performance relative to the ideal state. Secondly, the hot spot factor... The temperature difference between hot spots and normal areas on the component surface was obtained by fusing data from infrared thermal imagers and distributed fiber optic temperature measurements. The calculation is performed using the following formula: This exponential function can sensitively amplify the impact of abnormal temperature rises and effectively characterize hot spot risk. Third, structural attitude factor. The resultant angle is calculated from the pitch and roll angles measured by a six-degree-of-freedom MEMS sensor. And take the cosine value, that is: The platform's tilt angle is mapped to a stability index within the (0,1) interval. Fourth, corrosion factor. Corrosion thickness of the support obtained by laser-ultrasonic Lamb wave device The calculation is performed using the following formula: The exponential decay characteristic is used to characterize the principle that deeper corrosion has a greater negative impact on structural health. Finally, the adhesion factor... Marine organism attachment index of 0-5 output by polarized light imaging device The calculation is performed using the following formula: This achieves a linear inverse mapping between adhesion level and health factors. These five factors together form the basis for a comprehensive health assessment, encompassing electrical performance, thermal safety, structural stability, and environmental tolerance.

[0038] Step B2: The initial health index is obtained by weighted summation based on individual health factors and their corresponding weights.

[0039] Specifically, the five individual health factors calculated are: power attenuation factor... Hot spot factor Structural attitude factor Corrosive agents and adhesion factor Perform linear weighted fusion to calculate the initial health index. The calculation formula is as follows:

[0040] in, The weights corresponding to each individual health factor satisfy the constraint that the sum is 1. ).

[0041] These weights are not fixed; their baseline values ​​are pre-determined through a Bayesian optimization algorithm, using a combination of optimal parameters obtained from massive amounts of historical data and typical operating conditions. These optimal combinations are then permanently stored in edge computing nodes to ensure continuous and stable application. Crucially, this weighting strategy is weather-adaptive: it monitors weather types in real time, and when in adverse weather conditions, it automatically and dynamically increases the weight of the structural attitude factor, which reflects structural safety. At the same time, the weight of the power decay factor, which is more sensitive to short-term fluctuations, is reduced accordingly. This dynamic weighting adjustment mechanism aims to prioritize the perception and assessment of structural safety risks under adverse operating conditions, effectively suppressing false alarms caused by normal fluctuations in power generation, thereby ensuring a stable initial health index. It can reflect the overall health status of offshore photovoltaic equipment in complex marine environments.

[0042] Step B3 involves correcting and smoothing the initial health index to obtain the health index of the offshore photovoltaic equipment at the current data collection time.

[0043] By calculating the power attenuation factor, hot spot factor, structural attitude factor, corrosion factor, and adhesion factor, a comprehensive quantitative assessment of the equipment's electrical performance, thermal safety, structural stability, and environmental tolerance is achieved. An initial health index is obtained by weighted summation based on weights, enabling the assessment results to comprehensively reflect the influence of factors in various dimensions. The initial health index is corrected and smoothed, effectively improving the index's anti-interference ability and trend stability, thus obtaining a final health index that more accurately reflects the true health status of the equipment.

[0044] In this embodiment of the application, the initial health index is corrected and smoothed to obtain the health index of the offshore photovoltaic equipment at the current data collection time, including the following steps B31-B33: Step B31: The initial health index is nonlinearly corrected using the hyperbolic tangent function to obtain the corrected health index.

[0045] Specifically, the initial health index As input, the hyperbolic tangent function is used The nonlinear transformation is performed on it, and the specific calculation formula is as follows: In this formula, the constant coefficient 1.5 serves as a preset scaling factor, its main function being to adjust the range of the function's input values ​​so that the initial health index... After transformation, its output value It can be more effectively compressed and mapped to a relatively stable numerical range that conforms to engineering experience (e.g., more concentrated in the range of 0 to 1). The S-shaped curve characteristic of the hyperbolic tangent function itself makes it more sensitive to changes in the input value in the middle range, while producing a saturation effect on excessively high or low extreme input values. This characteristic makes the corrected health index... It can reflect the normal fluctuations in the health status of the equipment, and suppress abnormal peaks caused by instantaneous data jumps or calculations under extreme operating conditions.

[0046] Step B32: Obtain the health index and weight coefficient of the previous data collection time. The weight coefficient is obtained after calibration based on three steps: on-site data sampling, grid search, and inflection point selection.

[0047] Specifically, the health index from the previous data collection time is read from the edge SQLite cache database. At the same time, obtain the preset weight coefficients. The weighting coefficient These are key parameters that have been pre-determined and permanently embedded in the local firmware using a three-step calibration method: First, on-site data sampling is performed, continuously collecting health status time-series data under various operating conditions on actual photovoltaic platforms in typical marine areas; then, a grid search is executed, traversing different parameters within the (0,1) interval with a specific step size (e.g., 0.05). Candidate values ​​were selected, and the smoothed health index sequence was calculated for each candidate value. Finally, inflection points were selected by analyzing different... The smoothing curve corresponding to the value is selected based on the inflection point that best balances response speed and noise reduction capability. The value is used as the final calibration result. This optimized calibration... The value will serve as a key parameter in the first-order exponential smoothing algorithm, used to balance the contribution ratio of the latest correction results and historical health status information in the final health index.

[0048] Step B33: Use the health index and weighting coefficient from the previous data collection time to smooth the corrected health index, and obtain the health index of the offshore photovoltaic equipment at the current data collection time.

[0049] Specifically, a first-order exponential smoothing algorithm is used to smooth the time series of the health index after nonlinear correction. The corrected health index is then... Health index compared to the previous collection time and the weighting coefficients determined by the three-step calibration method. Substitute into the smoothing formula for calculation:

[0050] Among them, the weighting coefficient As a smoothing factor, its value determines the current correction value. Compared with historical values Weighting in the final result: When When the value is close to 1, the smoothing result depends more on the correction value at the current moment, responding quickly to state changes but potentially introducing more fluctuations; when... When the output is close to zero, smoothing tends to preserve historical data, resulting in a more stable output but a slower response. This can be improved by using optimized calibration. This smoothing process filters out random fluctuations in the health index caused by instantaneous sensor noise or brief environmental disturbances, while preserving the true trend of changes in the equipment's health status. The output is a smooth and stable reflection of the health index at the current acquisition moment, reflecting the evolution of the equipment's health status. This provides a basis for assessing health status and predicting remaining life expectancy.

[0051] Using the hyperbolic tangent function for nonlinear correction can map the initial health index to a more stable range, enhancing its robustness to outliers. Obtaining weight coefficients optimized by a three-step calibration method ensures the scientific validity and optimality of the smoothing algorithm parameters. Smoothing using historical health indices and weight coefficients effectively filters out data noise, enabling the final output health index to smoothly and continuously reflect the evolution trend of equipment health status.

[0052] In this embodiment of the application, predicting the health index curve for a future time period and the remaining lifespan of offshore photovoltaic equipment includes the following steps C1-C4: Step C1: Obtain multidimensional time series data within the historical time period, including health index, wind speed, wave height, salinity, and load.

[0053] Specifically, multidimensional time-series data, arranged chronologically over the past 168 hours (7 days), is extracted from the SQLite cache database of the edge computing nodes and used as the input dataset for the remaining lifetime prediction model. This dataset is a multivariate, time-aligned sequence, specifically containing the following five key indicators: a health index reflecting the overall condition of the equipment. ; 10-minute average wind speed measured by an ultrasonic anemometer Characterizing the wind load environment; wave height measured by a radar wave gauge. Characterizing the wave load environment; salt spray concentration detected by an ion mobility spectrometer salt spray sensor. (Salinity-related), characterizing the intensity of corrosive environments; and mooring tension measured by a fiber Bragg grating tension meter. (Characteristic load) directly reflects the mechanical load on the structure. These five dimensions of time-series data together constitute a data foundation that can comprehensively describe the health status of the equipment and its dynamic coupling with the marine environment, providing the necessary multi-source, heterogeneous and spatiotemporally synchronized input information for the LSTM-Attention model to perform accurate temporal pattern learning and remaining lifetime prediction.

[0054] Step C2 involves using a predictive model to analyze multidimensional time series data and obtain the health index curve for future time periods.

[0055] Specifically, multidimensional time-series data is input into a prediction model based on LSTM-Attention for analysis. The model first performs deep feature extraction on the input time-series data through its LSTM layer. LSTM, with its internal gating mechanism (input gate, forget gate, output gate), can effectively capture and remember long-term dependencies in historical data, such as the cumulative impact of persistent severe weather on health indices. Subsequently, the attention layer in the model is activated. This layer automatically identifies and focuses on the most critical time steps in the input sequence for predicting future health indices (e.g., the moment corresponding to a severe storm surge event) by calculating attention weights, and assigns higher weights to the features of these critical time steps. Finally, based on the learned time-series patterns and attention weights, the model outputs a health index prediction curve for a specific future time period. The length of this future time period is dynamically determined by the current weather type: for ordinary weather, the curve predicts the next 30 days; for severe weather, the curve predicts the next 72 hours, adapting to the differentiated needs of the prediction period under different weather conditions.

[0056] Before deployment, the LSTM-Attention prediction model requires offline supervised training. The training process is as follows: First, a large-scale training dataset is prepared, containing numerous historical multi-dimensional time-series samples covering various marine conditions and their corresponding subsequent changes in actual health indices. Then, the historical sequences are used as input, and the observed health index sequences over a subsequent period are used as training labels (i.e., the expected output). The model is iteratively trained using the backpropagation algorithm and the Adam optimizer to minimize the difference between its predicted output and the true label (typically using loss functions such as mean squared error). During training, all model parameters are adjusted, including the weights and biases of the LSTM units and the weight matrix in the attention mechanism, until the model's prediction accuracy on the validation set stabilizes and meets preset requirements. The final model parameters are then fixed and deployed to run independently on edge computing nodes.

[0057] Step C3: Find the future moment when the health index first falls below a preset threshold from the health index curve.

[0058] Specifically, the future health index prediction curve output by the LSTM-Attention model is traversed and analyzed. This curve consists of a series of discrete prediction points arranged in future time sequence, each point containing a future timestamp and its corresponding predicted health index value. The predicted health index value at each time point on this curve is compared one by one with a preset threshold (this threshold is fixed at 0.7, corresponding to the defined fault state threshold). Starting from the future point closest to the current time, the curve is scanned sequentially along the time axis in the future direction, identifying and recording the future time when the first predicted health index value is lower than the 0.7 threshold, and marking this specific time as [the next time point is missing from the original text]. .this This refers to the critical time point at which the equipment's health status, predicted by the model, enters the failure level, and is a key parameter for calculating remaining lifetime. This process is fully automated, requiring no manual interpretation, thus ensuring the objectivity and consistency of remaining lifetime assessment.

[0059] Step C4: Calculate the difference between the future time and the current sampling time, and use the difference as the remaining lifetime.

[0060] Specifically, at the determined critical moment of the predicted fault Based on this, perform a simple arithmetic operation. First, obtain the current sampling time. (That is, the absolute point in time that triggers this health assessment and prediction process, usually represented by a Unix timestamp). Then, calculate this current moment. and future failure moments The time difference between them, i.e. The difference obtained from this calculation This refers to the remaining lifespan of offshore photovoltaic equipment, which quantifies the time from the current moment until the equipment's health condition first drops to the failure threshold. The remaining runtime is less than 0.7. The value is a scalar with a clear physical meaning. Using it as a key predictive maintenance indicator output provides maintenance personnel with a direct and quantitative basis for decision-making when formulating maintenance plans and preparing maintenance resources.

[0061] By acquiring historical multidimensional time series data containing health indices and their key influencing factors, a comprehensive data foundation reflecting the evolution of equipment status is provided for the predictive model. By analyzing time series data and outputting future health index curves using the predictive model, advanced prediction of equipment performance degradation trends is achieved. By locating the moment when the health index first falls below a preset threshold and calculating the difference between it and the current moment as the remaining lifespan, a direct and quantitative decision-making basis for predictive maintenance is provided.

[0062] In this embodiment of the application, after determining the health status of the offshore photovoltaic equipment based on a health index, the method further includes: Step S201: Obtain the weather type at the current collection time.

[0063] In this embodiment, the latest environmental information measured by the environmental sensor array at the current acquisition time is acquired in real time, including the 10-minute average wind speed provided by the ultrasonic anemometer, the significant wave height measured by the radar wave height meter, the salt spray concentration detected by the ion mobility spectrometer salt spray sensor, the atmospheric visibility obtained by the visibility meter, and the lightning electric field intensity monitored by the fast antenna electric field meter. The real-time environmental parameters are compared instantly with four preset key thresholds (wind speed ≥ 20 m / s, significant wave height ≥ 3 m, lightning electric field ≥ 2 kV / m, visibility < 1 km). Based on the comparison results, the current weather type is determined: if all parameters do not exceed the thresholds, it is determined to be a normal weather type; if any parameter reaches or exceeds its threshold, it is immediately determined to be a severe weather type. This determination result updates the status of the severe weather indicator, providing a real-time and accurate decision-making basis for dynamically configuring local communication network parameters. This weather type determination process is executed in each acquisition cycle to ensure that the communication strategy can respond promptly to the latest environmental changes and achieve optimal allocation of communication resources.

[0064] Step S202: Configure the local network according to the network configuration mode corresponding to the weather type.

[0065] In this embodiment, the LoRaMESH ad hoc network deployed on the offshore photovoltaic platform is automatically selected and configured with the corresponding communication mode based on the current weather type. If the weather type is normal, a predefined normal mode configuration is used, which sets the LoRa communication spreading factor to SF7 and the transmit power to 14dBm. This configuration strikes a balance between communication rate and power consumption and is suitable for daily operation under stable channel conditions. If the weather type is severe, the system immediately switches to severe mode configuration, increasing the spreading factor to SF12 and the transmit power to 20dBm. Increasing the spreading factor significantly enhances the signal's anti-interference capability and receiver sensitivity, while increasing the transmit power helps overcome the increased path loss that may occur under severe weather conditions. The combination of these two features greatly improves the robustness and reliability of the local wireless communication link in harsh environments such as strong winds, high waves, and high salt spray. The switching between these two network configuration modes is automatically triggered and controlled by the updated BAD_WEATHER_FLAG flag, ensuring that the local communication network can adapt to environmental changes and provide a stable and reliable underlying link guarantee for data backhaul.

[0066] In step S203, if the weather type is normal, the cached data is compressed and sent back in batches through the regular channel at preset intervals; if the health index is lower than the preset threshold, or if the offshore photovoltaic equipment malfunctions, the preemptive small packet mechanism is activated, and the data is continuously sent up through the satellite high-priority channel within a unit of time.

[0067] In this embodiment, a differentiated satellite backhaul strategy is implemented based on the current weather type and equipment health status. When the weather type is normal, the regular channel is activated. Every 6 hours, the edge gateway efficiently compresses the structured data from past periods in the SQLite cache using the LZMA compression algorithm (with a measured compression rate of up to 92%), and then transmits it back in batches through the regular communication channel of the Starlink satellite network. This strategy is suitable for periodic data reporting when the system is in a healthy or sub-healthy state, and can effectively control satellite communication traffic costs. When the health index is detected to be below the preset threshold of 0.7, or when abnormal equipment operation is diagnosed, such as insulation fault (insulation resistance <1MΩ) or structural anomaly (composite angle >15° or mooring tension >40kN), the regular channel will be immediately interrupted and the emergency channel will be activated, regardless of the current weather conditions. In this emergency channel, the edge gateway initiates a preemptive small packet mechanism, transmitting small data packets containing critical alarms and status information at a low rate of 1kbps, once per minute, via the high-priority channel of the satellite network. The measured end-to-end latency can be controlled within 3 seconds, ensuring extremely low-latency delivery of critical alarms. Furthermore, a redundancy mechanism is provided. The 24-hour ring buffer built into the edge node can persist data during a momentary interruption of the satellite link, automatically retransmitting data after the link is restored. This achieves seamless hot-switching between the LoRa MESH local network and the satellite backhaul link, ensuring the integrity of data transmission and the robustness of system communication.

[0068] By acquiring real-time weather types and configuring local network parameters accordingly, the communication link becomes environmentally adaptive, ensuring the reliability of local data transmission. In normal weather conditions, batch backhaul via conventional channels optimizes the utilization efficiency of satellite communication resources. In cases of health anomalies, a preemptive small packet mechanism is activated for emergency uplink, ensuring extremely low latency delivery of critical alarm information, thus forming an efficient communication strategy of on-demand allocation and separation of priority information.

[0069] As an example, such as Figure 3As shown, the hierarchical communication strategy process includes: First, weather pattern judgment is performed. If the BAD_WEATHER_FLAG flag is 1 (i.e., severe weather), the LoRa MESH self-organizing network is switched to an SF12 spreading factor and 20dBm transmit power configuration; otherwise, the normal mode configuration of SF7 and 14dBm is used. Subsequently, data is aggregated to the edge gateway / edge node. The node then determines whether emergency conditions such as health index <0.7, insulation failure, or structural abnormality are met. If not, the regular backhaul strategy is executed, i.e., data is compressed and uploaded in batches through the Starlink regular channel every 6 hours; if any emergency condition is met, the emergency backhaul strategy is immediately activated, continuously uplinking through the Starlink emergency channel at a rate of 1kbps per minute. Throughout the process, the 24-hour ring buffer built into the edge node continues to operate. If a satellite link interruption is detected, data is automatically temporarily stored and retransmitted after the link is restored. Finally, all data is delivered to the ground control center.

[0070] In this embodiment of the application, after determining the health status of the offshore photovoltaic equipment based on a health index, the method further includes: Step S301: Determine the corresponding display strategy and alarm method based on the numerical range that the health index falls into.

[0071] In this embodiment, the ground-based digital twin platform and the WeChat mini-program receive health indices and specific device statuses in real time, and trigger a set of display strategies and alarm methods based on preset numerical ranges. A four-color layer strategy is used for visualization: when the health index... When the interface elements are displayed in green, indicating a healthy state, this corresponds to the routine inspection display strategy and does not trigger an audible alarm; when When the indicator is yellow, it signifies a sub-healthy state and suggests the need for preventative maintenance; when a specific equipment malfunctions, i.e., insulation resistance... (Insulation fault), or attitude angle or mooring tension When a structural abnormality occurs, the indicator turns orange and triggers a short beep and a corresponding voice alarm (such as insulation fault, excessive tilt angle, or excessive tension). This alarm method is intended to draw attention and prompt the need to remotely shut down backup equipment; when the health index... When the indicator turns red, it signifies a fault and triggers a long beep and a voice alarm indicating a health index below 0.7, prompting the need to initiate emergency hardening or evacuation. This multimedia alarm mechanism, combining color and sound, is strictly linked to the health status value range and specific fault types, ensuring intuitive, accurate, and efficient communication of status information.

[0072] Step S302: Display the corresponding prompts according to the display strategy.

[0073] In this embodiment, based on display strategies and alarm methods, specific visualization and interactive operations are performed on the ground digital twin platform and the WeChat mini-program. For the ground digital twin platform, firstly, in its 1:1 constructed 3D WebGL model, the corresponding four-color layers (green, yellow, orange, and red) are dynamically rendered according to the health status, and the bound sound alarms (such as short beeps, long beeps, and specific voice prompts) are triggered simultaneously. At the same time, the platform provides a one-click simulation function, which can perform twin simulations of extreme working conditions such as typhoons, wave loads, or fires, automatically generate emergency plans, and convert them into specific Modbus commands. For the WeChat mini-program, the four-color alarm cards corresponding to the status are pushed to the user in real time. The user can click on the card to jump to the 3D scene to view the details, and an automatically generated traceable closed-loop work order containing the entire process of "reporting repair - dispatching - processing - feedback" is generated, supporting maintenance personnel to transmit on-site information via voice, photos, and videos. All remote control commands (such as driving the mooring winch, shutting off the serial circuit breaker, etc.) received and executed by the SCADA system and their results are written back to the digital twin platform via the Modbus-TCP protocol, thereby completing the entire closed-loop operation and maintenance from status awareness and visual alarms to control execution.

[0074] The display strategy and alarm method are determined based on the range of health index values, so that the device status can be perceived intuitively and quickly through multiple modes such as color and sound. The corresponding prompts are executed according to the display strategy, and the status judgment results are transformed into specific operation and maintenance instructions and interaction processes. This realizes closed-loop management from status perception to operation and maintenance execution, which significantly improves the efficiency and accuracy of operation and maintenance response.

[0075] This embodiment provides a marine photovoltaic health monitoring system based on dynamic sensing and closed-loop control, such as... Figure 4 As shown, it includes: a perception layer, an edge computing layer, a communication layer, and an application layer; The perception layer is used to acquire the raw multi-source perception data collected by the multi-source sensors for the offshore photovoltaic equipment at the current acquisition time, including acquiring environmental information to determine the weather type, determining the acquisition cycle according to the weather type, and collecting the perception data corresponding to each indicator according to the acquisition cycle. The edge computing layer is used to calculate the corresponding individual health factors using multi-source sensing data, and to calculate the health index based on the individual health factors and their corresponding weights. This includes calculating the power attenuation factor, hot spot factor, structural posture factor, corrosion factor, and adhesion factor, performing weighted summation to obtain the initial health index, and then correcting and smoothing the initial health index to obtain the final health index. The communication layer is used to configure the local network according to the network configuration mode corresponding to the weather type, and to transmit data back in batches through the regular channel when the weather type is normal. When the health index is lower than the preset threshold or the equipment is abnormal, the preemptive small packet mechanism is activated to continuously upload through the satellite high-priority channel. The application layer is used to determine the health status of offshore photovoltaic equipment based on the health index, determine the corresponding display strategy and alarm method according to the numerical range of the health index, display the corresponding prompt operation according to the display strategy, and write the execution result back to the perception layer.

[0076] As an example, the workflow of the perception layer, such as Figure 5 As shown, the process includes: First, data alignment and caching are performed, and the raw data collected by multiple sensors are synchronized in time and stored in a structured manner; then, various health factors are calculated, including power attenuation factor, hot spot factor, structural attitude factor, corrosion factor, and adhesion factor; then, an initial health index is obtained through weighted fusion, and the weight allocation is dynamically adjusted under severe weather conditions, increasing the structural weight while decreasing the power weight; the initial health index is nonlinearly corrected and smoothed to output the final health index; the health status of the equipment is judged based on the health index: when the health index is ≥0.85, it is in a healthy state; when the health index is 0.7≤health index<0.85, it is in a sub-healthy state; when the health index is <0.7, it is in a fault state. At the same time, insulation faults with insulation resistance <1MΩ and structural anomalies with attitude angle >15° or tension >40kN are monitored; finally, the remaining lifespan of the equipment is predicted using an LSTM-Attention model, and the health index curves for the next 72 hours (severe weather) or 30 days (normal weather) are output to the communication layer.

[0077] As an example, the application layer workflow, such as Figure 6As shown, the process includes: the edge gateway transmits the processed data to the ground digital twin platform, which maps the array status in real time and updates the four-color layer; the system judges the health status of the equipment based on the health index, corresponding to four states: green (health index ≥ 0.85), yellow (0.7 ≤ health index < 0.85), orange (insulation fault or structural abnormality), and red (health index < 0.7); the WeChat mini program simultaneously pushes four-color alarm cards, and users can click to view alarm details and enter the closed-loop work order system to realize the full-process management of repair reporting, dispatching, processing, and feedback; corresponding operations are performed for different states: routine inspection is performed in the green state, preventive maintenance is performed in the yellow state, a short buzzer and voice broadcast are triggered in the orange state, and a long buzzer and voice broadcast are triggered in the red state; on-site maintenance personnel can transmit on-site information via text / voice, and the system automatically generates emergency plans and updates the work order status; the SCADA system remotely executes equipment operation commands (including disconnecting the array, tightening the winch, starting the backup power supply, etc.) and writes all execution results back to the digital twin platform to complete the operation and maintenance closed loop.

[0078] This embodiment also provides a marine photovoltaic health monitoring device based on dynamic sensing and closed-loop control. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0079] This embodiment provides a marine photovoltaic health monitoring device based on dynamic sensing and closed-loop control, such as... Figure 7 As shown, it includes: The acquisition module 71 is used to acquire the raw multi-source sensing data collected by the multi-source sensors for the marine photovoltaic equipment at the current acquisition time; The calculation module 72 is used to calculate the corresponding individual health factors using multi-source sensing data, and to calculate the health index based on the individual health factors and their corresponding weights. The prediction module 73 is used to determine the health status of offshore photovoltaic equipment based on the health index, and to predict the health index curve and the remaining lifespan of the offshore photovoltaic equipment for future time periods.

[0080] Furthermore, the acquisition module 71 is used to acquire environmental information of the marine photovoltaic equipment at the previous acquisition time, and determine the weather type based on the environmental information; determine the acquisition cycle of each indicator at the current acquisition time based on the weather type; acquire the sensing data corresponding to each indicator according to the acquisition cycle, and construct multi-source sensing data based on the sensing data of each indicator.

[0081] Furthermore, the calculation module 72 is used to calculate the power attenuation factor by the ratio of the maximum output power to the nominal power of the marine photovoltaic equipment, the hot spot factor by the temperature change of the marine photovoltaic equipment, the structural attitude factor by the cosine of the combined pitch and roll angles, the corrosion factor by the corrosion of the marine photovoltaic equipment, and the adhesion factor by the adhesion-related parameters; to obtain the initial health index by weighted summation based on individual health factors and their corresponding weights; and to correct and smooth the initial health index to obtain the health index of the marine photovoltaic equipment at the current acquisition time.

[0082] Furthermore, the calculation module 72 is used to perform nonlinear correction on the initial health index using the hyperbolic tangent function to obtain the corrected health index; to obtain the health index and weight coefficients at the previous acquisition time, wherein the weight coefficients are obtained after calibration based on three steps: on-site data sampling, grid search, and inflection point selection; and to smooth the corrected health index using the health index and weight coefficients at the previous acquisition time to obtain the health index of the offshore photovoltaic equipment at the current acquisition time.

[0083] Furthermore, the prediction module 73 is used to acquire multidimensional time series data within a historical time period, including health index, wind speed, wave height, salinity, and load; analyze the multidimensional time series data using a prediction model to obtain the health index curve for the future time period; find the future moment when the health index first falls below a preset threshold from the health index curve; calculate the difference between the future moment and the current sampling moment, and use the difference as the remaining lifetime.

[0084] Furthermore, the device also includes: a backhaul module for obtaining the weather type at the current collection time; configuring the local network according to the network configuration mode corresponding to the weather type; if the weather type is normal, compressing the cached data and backhauling it in batches through the regular channel at preset intervals; if the health index is lower than the preset threshold, or if the offshore photovoltaic equipment malfunctions, a preemptive small packet mechanism is activated to continuously uplink via the satellite high-priority channel within a unit of time.

[0085] Furthermore, the device also includes: a prompting module, used to determine the corresponding display strategy and alarm method based on the numerical range in which the health index falls; and to display the corresponding prompt operation according to the display strategy.

[0086] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 8As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other 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).

[0087] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 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.

[0088] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0089] The memory 20 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 as shown by a landing page for an app. Furthermore, the memory 20 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 20 may optionally include memory remotely located relative to the processor 10, which can 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.

[0090] The memory 20 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 20 may also include a combination of the above types of memory.

[0091] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0092] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then 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, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0093] Although embodiments of the invention 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 the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for monitoring the health of marine photovoltaic systems based on dynamic sensing and closed-loop control, characterized in that, The method includes: Acquire raw multi-source sensing data collected by multi-source sensors at the current acquisition time for the offshore photovoltaic equipment; The corresponding individual health factors are calculated using the multi-source sensing data, and a health index is calculated based on the individual health factors and their corresponding weights. The health status of the offshore photovoltaic equipment is determined based on the health index, and the health index curve for future time periods and the remaining lifespan of the offshore photovoltaic equipment are predicted.

2. The method according to claim 1, characterized in that, The acquisition of raw multi-source sensing data collected by multi-source sensors for the offshore photovoltaic equipment at the current acquisition time includes: Obtain environmental information about the environment where the offshore photovoltaic equipment was located at the previous data collection time, and determine the weather type based on the environmental information; The collection period for each indicator at the current collection time is determined based on the weather type. The sensing data corresponding to each indicator is collected according to the collection cycle, and the multi-source sensing data is constructed based on the sensing data of each indicator.

3. The method according to claim 1, characterized in that, The step of calculating corresponding individual health factors using the multi-source sensing data, and calculating a health index based on the individual health factors and their corresponding weights, includes: The power attenuation factor is calculated by the ratio of the maximum output power to the nominal power of the offshore photovoltaic equipment; the hot spot factor is calculated by the temperature change of the offshore photovoltaic equipment; the structural attitude factor is calculated by the cosine of the combined pitch and roll angles; the corrosion factor is calculated by the corrosion condition of the offshore photovoltaic equipment; and the adhesion factor is calculated by the adhesion-related parameters. The initial health index is obtained by weighted summation based on the individual health factors and their corresponding weights. The initial health index is corrected and smoothed to obtain the health index of the offshore photovoltaic equipment at the current data collection time.

4. The method according to claim 3, characterized in that, The process of correcting and smoothing the initial health index to obtain the health index of the offshore photovoltaic equipment at the current data collection time includes: The initial health index was nonlinearly corrected using the hyperbolic tangent function to obtain the corrected health index; Obtain the health index and weighting coefficient at the previous data collection time. The weighting coefficient is obtained after calibration based on three steps: on-site data sampling, grid search, and inflection point selection. The corrected health index is smoothed using the health index from the previous data collection time and the weighting coefficient to obtain the health index of the offshore photovoltaic equipment at the current data collection time.

5. The method according to claim 1, characterized in that, The health index curve predicting future time periods and the remaining lifespan of the offshore photovoltaic equipment include: Acquire multidimensional time series data within a historical time period, including health index, wind speed, wave height, salinity, and load; The multidimensional time series data is analyzed using a predictive model to obtain the health index curve for future time periods; Find the future moment when the health index first falls below a preset threshold from the health index curve; Calculate the difference between the future time and the current sampling time, and use the difference as the remaining lifetime.

6. The method according to claim 1, characterized in that, After determining the health status of the offshore photovoltaic equipment based on the health index, the method further includes: Obtain the weather type at the current data collection time; Configure the local network according to the network configuration mode corresponding to the weather type; If the weather type is normal, the cached data will be compressed and sent back in batches through the regular channel at preset intervals; if the health index is lower than the preset threshold, or if the offshore photovoltaic equipment malfunctions, the preemptive small packet mechanism will be activated, and the data will be continuously sent up via the satellite high-priority channel within a unit of time.

7. The method according to claim 1, characterized in that, After determining the health status of the offshore photovoltaic equipment based on the health index, the method further includes: The corresponding display strategy and alarm method are determined based on the numerical range that the health index falls into; The corresponding prompts will be displayed according to the aforementioned display strategy.

8. A marine photovoltaic health monitoring device based on dynamic sensing and closed-loop control, characterized in that, The device includes: The acquisition module is used to acquire the raw multi-source sensing data collected by the multi-source sensors from the offshore photovoltaic equipment at the current acquisition time; The calculation module is used to calculate the corresponding individual health factors using the multi-source sensing data, and to calculate the health index based on the individual health factors and the weights corresponding to the individual health factors. The prediction module is used to determine the health status of the offshore photovoltaic equipment based on the health index, and to predict the health index curve for future time periods and the remaining lifespan of the offshore photovoltaic equipment.

9. A computer device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images