An adaptive photovoltaic support performance detection system and method

Abstract

The application discloses a kind of self-adapting photovoltaic support performance detection system and method, it is related to photovoltaic equipment detection technical field, the method includes: through sensing network acquisition support state parameters and environmental data, using adaptive Kalman filter denoising and extracting characteristic vector;Combining structural mechanics model dynamic adjustment alarm threshold, realize the real-time matching of detection criterion and environmental change;Using weighted fusion algorithm to calculate comprehensive health evaluation index, determine safety level and identify latent defect;Integrating long short-term memory neural network predicts performance degradation trend and outputs operation and maintenance suggestion.The application solves the problems of low efficiency of manual inspection and false alarm caused by environmental interference through multi-source data fusion and adaptive detection threshold adjustment, realizes accurate monitoring and proactive preventive maintenance of support performance, and improves the digital management level and operation reliability of photovoltaic power station.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic equipment testing technology, and in particular to an adaptive photovoltaic support performance testing system and method. Background Technology

[0002] With the transformation of the global energy structure, photovoltaic (PV) power generation, as an important component of renewable energy, has seen its installed capacity continuously expand. As the core structure supporting PV modules, the safety and reliability of PV mounting systems directly affect the long-term stable operation of PV power plants. PV mounting systems are exposed to a complex outdoor environment for extended periods, continuously resisting various external stresses such as wind loads, snow loads, and geological changes. To ensure the structural safety of PV systems and maximize power generation benefits, accurate performance evaluation and condition monitoring of PV mounting systems have become a crucial aspect of power operation and maintenance.

[0003] The photovoltaic (PV) support structure performance testing system aims to acquire real-time state parameters such as physical deformation and mechanical response of the support structure through sensing technology and data processing methods to identify potential structural damage. However, existing PV support testing technologies generally rely on regular manual inspections, resulting in a significant lag in capturing latent defects and early minute deformations, making it difficult to achieve 24 / 7 real-time monitoring. Furthermore, traditional monitoring equipment often uses fixed threshold alarm modes, lacking the ability to adapt to environmental fluctuations and differences in equipment models, easily leading to false alarms or missed detections. In addition, existing solutions often focus on single-dimensional performance index analysis, lacking in-depth fusion and systematic evaluation of multi-source state data, and have weak predictive analysis capabilities for support structure performance degradation trends, resulting in maintenance strategies remaining at a passive response stage.

[0004] Therefore, the present invention provides an adaptive photovoltaic support performance testing system and method, which realizes continuous monitoring of the structural performance of photovoltaic support, can comprehensively reflect the actual operating status of the support, improves the testing accuracy under harsh weather conditions, enhances the overall reliability of the photovoltaic power generation system, and realizes visualized management of the status of the entire support system. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive photovoltaic support performance testing system and method to solve the problems in the prior art that focus on single-dimensional performance index analysis, lack in-depth integration and systematic evaluation of multi-source state data, and have weak predictive analysis capabilities for support performance degradation trends.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] An adaptive photovoltaic support performance testing system includes:

[0008] The data acquisition module is used to acquire physical status information and environmental meteorological parameters of the photovoltaic support through a sensor array.

[0009] The central processing unit connects to the data acquisition module and integrates a trend analysis module. It is used to perform data cleaning and feature calculation, adaptively detect threshold matching on the obtained feature vectors, and perform modeling and prediction in the trend analysis module. The matching results and early warning information are then transmitted to the communication feedback unit.

[0010] The communication feedback unit, connected to the central processing unit, is used to transmit the matching results and early warning information to the remote monitoring platform.

[0011] Preferably, the data acquisition module in the system uses a low-power wireless transmission protocol for data aggregation, the data transmission rate is greater than or equal to the preset rate, each sensor node is equipped with an independent power supply system and energy storage unit, supports independent operation for a preset time under specific weather conditions, the system enclosure protection level reaches the preset level, and it can work stably within the preset operating temperature range.

[0012] Preferably, the central processing unit has a built-in hardware acceleration chip for parallel processing of large-scale data streams from multiple photovoltaic supports. The system updates the health status map of all supports at a preset time interval. Through spatial interpolation algorithms, it estimates the status of supports without sensors, thereby realizing digital modeling and full-coverage monitoring of the entire photovoltaic power station support array.

[0013] Preferably, the communication feedback unit supports dual backup of multiple communication protocols. When a sudden environmental disaster is detected, the system automatically switches to high-frequency sampling mode to increase the sampling rate and pushes the information of the severely damaged support to the mobile terminal of the maintenance personnel within a preset response time.

[0014] An adaptive photovoltaic mounting system performance testing method includes the following specific steps:

[0015] Step 1: Collect multi-source state parameters and environmental perception data of photovoltaic support: Through the sensor network deployed at key structural nodes of photovoltaic support, multi-source state parameters including support tilt angle, structural surface strain, vibration frequency and displacement deviation are acquired in real time, and environmental perception data including instantaneous wind speed, ambient temperature, snow thickness and solar radiation intensity are acquired simultaneously through the environmental monitoring unit.

[0016] Step 2: Preprocessing and feature extraction of the collected multi-source state parameters: Adaptive Kalman filtering algorithm is used to denoise the multi-source state parameters, eliminate abnormal fluctuations caused by electromagnetic interference, and extract the feature vector of the support structure through time domain and frequency domain analysis methods. The feature vector includes structural stiffness index, modal parameters and stress concentration factor.

[0017] Step 3, adaptive detection threshold adjustment: Based on the collected environmental perception data, the theoretical response value of the photovoltaic support under the current working conditions is calculated using a preset structural mechanics model. Combined with historical operating data, the alarm thresholds of various performance indicators are dynamically adjusted to achieve real-time matching between the detection criteria and changes in the external environment.

[0018] Step 4, perform multi-dimensional comprehensive performance evaluation: input the extracted feature vector into the multi-criteria decision model, calculate the comprehensive health evaluation index of the photovoltaic support through the weighted fusion algorithm, determine the current structural safety level of the support, and identify whether there are hidden defects such as loose bolts, material fatigue or foundation settlement.

[0019] Step 5: Predict performance degradation trends and output maintenance decisions: Use a long short-term memory neural network to perform time-series analysis on historical health evaluation indices, predict the performance degradation trend of the stent within a preset time period in the future, and automatically generate targeted maintenance suggestions and early warning instructions based on the prediction results.

[0020] Preferably, in step 1, the sensor network includes an inclination sensor installed at the connection between the photovoltaic support beam and the column. The inclination sensor has a measurement accuracy better than a preset accuracy threshold, and the sampling frequency is set to a preset frequency range. The surface strain of the structure is obtained by a fiber Bragg grating sensor attached to the stress concentration area of ​​the support. Its wavelength offset is linearly proportional to the strain value. The strain measurement has a preset measurement range and the resolution meets the preset resolution requirements. In the environmental monitoring unit, the wind speed sensor's start-up wind speed is less than or equal to a preset start-up threshold, and the measurement range covers the preset wind speed range, ensuring that accurate data support can still be provided under extreme wind load conditions.

[0021] Preferably, in step 2, the adaptive Kalman filter algorithm achieves optimal estimation of sensor random error by dynamically adjusting the process noise covariance and measurement noise covariance. In the feature extraction stage, the fast Fourier transform is used to obtain the preset order natural frequency of the photovoltaic support for the vibration signal. The overall stiffness degradation of the support structure is identified by the change in the natural frequency. For the stress signal, the rainflow counting method is used to extract the stress amplitude and mean under cyclic load to evaluate the cumulative fatigue damage of the material.

[0022] Preferably, the logic for adaptive detection threshold adjustment in step 3 includes: first, establishing a finite element analysis model of the photovoltaic support, inputting boundary conditions under different wind levels, snow pressures, and temperature changes, and calculating the static displacement threshold and dynamic response range of the support under multiple typical working conditions; second, using real-time monitored ambient temperature to correct the elastic modulus of the material, using wind speed data to calculate instantaneous wind load pressure, dynamically adjusting the alarm threshold according to changes in wind speed, and automatically compensating for zero-point drift according to changes in ambient temperature, thereby eliminating false alarms caused by seasonal changes or diurnal temperature differences.

[0023] Preferably, in step 4, the multi-dimensional comprehensive performance evaluation uses the analytic hierarchy process (AHP) to determine the weights of each indicator. Among them, the structural surface strain, natural frequency change rate, displacement deviation, and tilt angle deviation are each assigned a corresponding preset weight. The comprehensive health evaluation index is distributed within a preset value range. When the value is in the first preset range, it is defined as a healthy state. When the value is in the second preset range, it is defined as a sub-healthy state. When the value is in the third preset range, a first-level warning is triggered. When the value is in the fourth preset range, a second-level alarm is triggered. The system locates the specific location of the loose bolt by identifying the abnormal components in the feature vector.

[0024] Preferably, in step 5, the long short-term memory neural network includes a preset number of input layers, hidden layers, and output layers. The number of neurons in the hidden layers is set to a preset number based on the model complexity. The input features include a comprehensive health evaluation index sequence, an average wind speed sequence, and an environmental humidity sequence within a preset historical time period. The prediction time for performance degradation trend prediction is set to a preset time period. The prediction error is controlled within a preset error range. The generated maintenance suggestions include: reinforcing specific numbered support nodes, adjusting the installation tilt angle of photovoltaic modules, cleaning water accumulation at the bottom of the support, and replacing fasteners that have reached their fatigue limit.

[0025] Preferably, the method further includes: periodically and automatically performing online zero-point calibration and sensitivity self-check of the sensor; when the sensor failure rate exceeds a preset threshold, the system automatically initiates redundant data completion logic, uses the correlation characteristics of adjacent supports to repair the data, and ensures the continuity of the monitoring process.

[0026] Preferably, the present invention is applied to photovoltaic power station systems and can be adapted to various types of brackets. For tracking brackets, the system additionally introduces the rotation angle feedback of the drive system in the adaptive detection threshold calculation to eliminate the acceleration interference caused by the active rotation of the bracket.

[0027] Compared with the prior art, the beneficial technical effects of the present invention are as follows:

[0028] This invention enables continuous monitoring of the performance of photovoltaic support structures by deploying a sensor network, effectively solving the problems of low frequency and narrow coverage of manual inspections. It significantly improves the accuracy of identifying early minor deformations and hidden damages, and can promptly capture subtle hidden dangers.

[0029] This invention introduces environmental perception data and structural mechanics models to construct a dynamic adaptive detection threshold system. The system can adjust the judgment criteria in real time according to different load and temperature conditions, which improves the detection accuracy under harsh weather conditions and significantly reduces false alarms caused by environmental fluctuations.

[0030] This invention calculates a comprehensive health evaluation index by deeply integrating multi-source data, which can fully reflect the actual operating status of the support structure and provide quantitative data support for power plant operation and maintenance.

[0031] This invention integrates deep learning algorithms to predict performance degradation trends and anticipate the performance evolution path within a predetermined period, enabling the operation and maintenance team to plan maintenance in advance and improving the overall reliability of the photovoltaic power generation system.

[0032] This invention achieves visualized management of the status of the entire station's support structure through digital modeling and spatial interpolation technology. Its high protection level and low power consumption design ensure long-term stable operation under extreme climate conditions, effectively reducing operation and maintenance costs. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall technical solution architecture of an adaptive photovoltaic support performance testing system and method provided by the present invention;

[0034] Figure 2 This is a schematic diagram of the multi-level interaction relationships and data flow in this invention;

[0035] Figure 3 This is a schematic diagram of the core principle framework for adaptive detection threshold adjustment in this invention;

[0036] Figure 4 This is a logical flowchart of the multi-dimensional comprehensive performance evaluation performed in this invention;

[0037] Figure 5 This is a logical flowchart of the performance degradation trend prediction and operation and maintenance decision output in this invention. Detailed Implementation

[0038] 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Example 1

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0041] In an adaptive photovoltaic support performance testing method, step 1 collects multi-source state parameters and environmental perception data of the photovoltaic support. Specifically, through a sensor network deployed at key structural nodes of the photovoltaic support, multi-source state parameters including support tilt angle, structural surface strain, vibration frequency and displacement deviation are acquired in real time, and environmental perception data including instantaneous wind speed, ambient temperature, snow thickness and solar radiation intensity are acquired simultaneously through an environmental monitoring unit.

[0042] The sensor network includes tilt sensors installed at the connection between the photovoltaic support beam and the column. The tilt sensors have a measurement accuracy better than 0.01 degrees and a sampling frequency set in the range of 10 Hz to 50 Hz. In actual deployment, the tilt sensors use triaxial microelectromechanical system accelerometers to monitor the tilt offset of the support in four dimensions (north, south, east, and west) in real time by calculating the components of the gravity vector. The surface strain of the structure is obtained by fiber Bragg grating sensors attached to the stress concentration area of ​​the support. The wavelength offset is linearly proportional to the strain value. The strain measurement has a measurement range of -2000 microstrain to +2000 microstrain and a resolution of 1 microstrain. The fiber Bragg grating sensors are fixed to the base of the support column and the middle of the beam with weather-resistant adhesive, and their signal cables are connected to a data acquisition controller with multi-channel demodulation capability.

[0043] In the environmental monitoring unit, the wind speed sensor has a starting wind speed of less than or equal to 0.5 m / s and a measurement range covering wind speeds from 0 m / s to 60 m / s, ensuring accurate data support even under extreme wind load conditions. The ambient temperature sensor uses a high-precision platinum resistance thermometer, with its measurement accuracy controlled within ±0.1 degrees Celsius. The snow thickness sensor uses the ultrasonic ranging principle, is installed at a preset height above the bracket, and calculates the snow depth by transmitting pulses and receiving the time difference of reflected waves. The solar radiation intensity sensor uses a total radiation meter to provide real-time feedback on the light intensity received by the photovoltaic module surface, which helps determine the theoretical output status of the current photovoltaic power station.

[0044] In step 2, preprocessing and feature extraction are performed on the multi-source state parameters. Specifically, an adaptive Kalman filter algorithm is used to reduce noise in the multi-source state parameters, eliminating abnormal fluctuations caused by electromagnetic interference. The adaptive Kalman filter algorithm achieves optimal estimation of sensor random errors by dynamically adjusting the process noise covariance and measurement noise covariance. The specific calculation process is as follows:

[0045]

[0046] in, This represents the optimal state estimate at time k. This represents the prior state value predicted based on the previous time step. Here is the Kalman gain matrix. For real-time observations from the sensor, Using the observation matrix, this recursive filtering method enables the system to effectively separate high-frequency noise in the signal from the low-frequency response characteristics of the structure itself.

[0047] In the feature extraction stage, the first to fifth natural frequencies of the photovoltaic support are obtained by fast Fourier transform for vibration signals. The overall stiffness degradation of the support structure is identified by the change in natural frequency. For example, when the decrease in the first natural frequency exceeds 15% of the initial calibration value, the system determines that there may be loosening at the support connection or a decrease in the foundation bearing capacity. For stress signals, the stress amplitude and mean under cyclic loading are extracted by rainflow counting method to assess the cumulative fatigue damage of the material. The extracted feature vectors include structural stiffness indices, modal parameters and stress concentration factors, which constitute a high-dimensional description space of the health status of the support.

[0048] In step 3, adaptive detection threshold adjustment is performed. Based on the collected environmental perception data, the theoretical response value of the photovoltaic support under the current working conditions is calculated using a preset structural mechanics model. Combined with historical operating data, the alarm thresholds of various performance indicators are dynamically adjusted. The logic of adaptive detection threshold calculation and adjustment includes: first, establishing a finite element analysis model of the photovoltaic support, inputting boundary conditions under different wind levels, snow pressures and temperature changes, and calculating the static displacement threshold and dynamic response range of the support under multiple typical working conditions. The finite element model covers the material modulus, geometric dimensions, constraint methods and component load distribution of the support.

[0049] Secondly, the real-time monitored ambient temperature is used to correct the elastic modulus of the material. According to the temperature-induced effect of metallic materials, the material modulus decreases accordingly when the ambient temperature rises. The system corrects the theoretical judgment limits of strain and frequency through a built-in compensation function. Wind speed data is used to calculate instantaneous wind load pressure. The alarm threshold is dynamically adjusted according to the change of wind speed. Under strong wind conditions, the judgment threshold of vibration amplitude is automatically relaxed, and zero-point drift is automatically compensated according to the change of ambient temperature, thereby eliminating false alarms caused by seasonal changes or diurnal temperature differences. In addition, for tracking supports, the system introduces the rotation angle feedback of the drive system in the adaptive detection threshold calculation to eliminate the acceleration interference caused by the active rotation of the support.

[0050] In step 4, a multi-dimensional comprehensive performance evaluation is performed. The extracted feature vectors are input into a multi-criteria decision model, and a weighted fusion algorithm is used to calculate the comprehensive health evaluation index of the photovoltaic support, determining the current structural safety level of the support. The multi-dimensional comprehensive performance evaluation uses the analytic hierarchy process (AHP) to determine the weights of each indicator. Specifically, the weights for structural surface strain, natural frequency change rate, displacement deviation, and tilt angle deviation are set to 0.2 and 0.2 respectively. Calculated using the following formula:

[0051]

[0052] in, For the first The weight coefficients of the features, This is the actual observed value extracted for this feature. The corresponding adaptive dynamic detection threshold is used. The comprehensive health evaluation index is distributed in the range of 0 to 1. When the value is in the first preset range of 0 to 0.2, it is defined as a healthy state; when the value is in the second preset range of 0.2 to 0.4, it is defined as a sub-healthy state; when the value is in the third preset range of 0.4 to 0.7, a first-level warning is triggered; when the value is in the fourth preset range of 0.7 to 1, a second-level alarm is triggered. The system locates the specific location of the loose bolt by identifying abnormal components in the feature vector. For example, when the displacement deviation of the sensor at the top of the column is extremely large while the strain of the base is normal, the system determines through logical reasoning that the upper fastener has failed.

[0053] In step 5, the performance degradation trend prediction and operation and maintenance decision output are performed. The long short-term memory neural network is used to perform time series analysis on the historical health evaluation index to predict the performance degradation trend of the stent in the future preset time period. The long short-term memory neural network contains 128 input layer nodes, 64 hidden layer neurons and 1 output layer. The input features include the comprehensive health evaluation index sequence, average wind speed sequence and environmental humidity sequence in the past 30 consecutive sampling periods.

[0054] The prediction period for performance degradation trend is set to the next 3 months, and the prediction error is controlled within 5%. Based on the prediction results, the central processing unit automatically generates targeted operation and maintenance suggestions and early warning instructions. The generated operation and maintenance suggestions include: reinforcing the support nodes with specific numbers, adjusting the installation tilt angle of photovoltaic modules to reduce wind resistance, cleaning the water accumulated at the bottom of the support to prevent the foundation from weakening, and replacing fasteners that have reached their fatigue limit.

[0055] An adaptive photovoltaic support performance testing system includes: a data acquisition module for acquiring physical state information and environmental meteorological parameters of the photovoltaic support through a sensor array; the data acquisition module uses a low-power ZigBee wireless transmission protocol for data aggregation, with a data transmission rate greater than or equal to 250kbps; each sensor node is equipped with an independent solar power supply system and a lithium thionyl chloride battery energy storage unit, supporting independent operation for more than 720 hours under no-sunlight conditions; the system enclosure protection level reaches IP67, and it can operate stably within a working temperature range of -40 degrees Celsius to +85 degrees Celsius.

[0056] The central processing unit connects to the data acquisition module and is used to perform data cleaning, feature calculation, and adaptive detection threshold matching. The central processing unit has a built-in hardware acceleration chip for parallel processing of large-scale data streams from more than 1,000 photovoltaic supports in the entire power station. The system updates the health status map of all supports every 60 minutes. Through spatial interpolation algorithms, it estimates the status of supports without sensors, realizing digital modeling and full-coverage monitoring of the entire photovoltaic power station support array.

[0057] The trend analysis module is integrated into the central processing unit and is used to model and predict the long-term performance of the stent based on deep learning algorithms. The communication feedback unit is used to transmit the matching results and early warning information to the remote monitoring platform. The communication feedback unit supports dual backup communication protocols of 4G / 5G cellular network and Beidou satellite short message. When a sudden typhoon or heavy snow disaster is detected, the system automatically switches to high-frequency sampling mode, increases the sampling rate to 200 Hz, and pushes information such as the location and degree of damage of the severely damaged stent to the mobile terminal of the operation and maintenance personnel in real time within a 5-second preset response time.

[0058] In addition, the method also includes: periodically and automatically performing online zero-point calibration and sensitivity self-check of the sensor; when the sensor failure rate exceeds 5%, the system automatically starts the redundant data completion logic and uses the correlation characteristics of three adjacent supports to perform Kriging space interpolation repair to ensure the continuity of the monitoring process.

[0059] Example 2

[0060] Based on Example 1, this example elaborates on the specific application of the adaptive photovoltaic support performance testing method in large-scale mountain photovoltaic power stations. Due to the large topographic relief, extremely complex wind field environment, and geological subsidence risk in mountainous areas, this example focuses on describing the technical details of foundation subsidence identification and wind load correction for complex terrain.

[0061] In step 1, considering the special characteristics of mountain power stations, a high-precision hydrostatic level and a multi-point displacement meter are added to the sensor network to monitor the vertical displacement of the support foundation. The hydrostatic level is installed at the beginning and end of the support array and measures the settlement value of the column relative to the benchmark point in real time through the principle of connecting pipes. The measurement accuracy reaches 0.1 mm. Among the environmental sensing data collected simultaneously, the solar radiation intensity data is used to calculate the structural thermal deformation caused by uneven heating of the support, thereby eliminating the interference caused by thermal expansion and contraction in the foundation settlement judgment.

[0062] In step 2, the feature extraction stage further incorporates structural transfer function analysis. By analyzing the correlation between the vibration response at the bottom and top of the support column, the coherence function of the structure is calculated. If the coherence function shows a significant decrease within a specific frequency band, it is determined that the connection node of the support has abnormal damping characteristics. For sudden strong gusts in mountainous areas, the process noise covariance matrix Q of the adaptive Kalman filter in the preprocessing stage is dynamically scaled according to the rate of change of wind speed. When the instantaneous change of wind speed exceeds 10 meters per second, the Q matrix increases by 10 times to improve the filter's ability to track the rapid transient response of the structure.

[0063] In step 3, the adaptive detection threshold adjustment process not only relies on the basic finite element model, but also introduces the terrain correction coefficient. By establishing a micro-digital elevation model of the power station, the local slope and altitude of each support node are calculated. The basic wind speed collected by the wind speed sensor is converted through the local terrain roughness function to calculate the actual wind load pressure acting on the specific support. The dynamic adjustment of the alarm threshold not only considers the wind speed, but also combines the current rotation phase of the component relative to the flat single-axis tracking support. Using the wind tunnel test regression curve, the wind resistance stability threshold under different rotation angles is loaded into the memory lookup table.

[0064] In step 4, when performing a multi-dimensional comprehensive performance evaluation, the weight of the foundation settlement index was increased for mountainous environments. For supports located in landslide hazard zones, the weight of settlement displacement was increased to 0.45. The multi-criteria decision model introduced Markov chain analysis. By calculating the probability matrix of the support transitioning from a "healthy" state to a "sub-healthy" state, the system could probabilistically identify early hidden defects. When uneven foundation settlement exceeding 20 mm was detected, the system would not only trigger a second-level alarm, but also call surrounding sensor data through the central processing unit to conduct a regional geological disaster risk assessment.

[0065] In step 5, the performance degradation trend prediction is driven by the fusion of long short-term memory neural network and physical modeling. The physical model calculates the service life loss of the support based on the material fatigue SN curve, while the neural network captures the nonlinear degradation trend based on time series data. The output operation and maintenance suggestions include: adjusting the support counterweight in a targeted manner, grouting and reinforcing the columns that have settled, or forcibly locking the tracking support at a wind-avoiding angle before the strong wind season.

[0066] In terms of system architecture, the data acquisition module adopts a multi-level relay wireless mesh networking method in the mountain power station. Due to the obstruction of the mountain, a single node may not be able to directly connect to the gateway. Through the relay algorithm, each support detection node can act as a router to forward data from adjacent nodes. At the same time, the communication feedback unit automatically starts the ultra-narrowband modulation mode in areas with weak signals, sacrificing data rate to increase transmission distance, ensuring that critical early warning information can still be sent through the satellite link when extreme weather causes public network base stations to fail.

[0067] Example 3

[0068] This embodiment focuses on describing the implementation details of the present invention in coastal photovoltaic power stations or floating power stations. Coastal environments face severe salt spray corrosion and tidal erosion, which places extremely high demands on the monitoring of material performance degradation of the support structure.

[0069] In step 1, the multi-source state parameter acquisition module integrates a metal corrosion sensor. This sensor uses resistance analysis to measure the resistance change of the sacrificial anode or reference electrode installed at typical locations on the support surface in real time, in order to characterize the corrosion rate of the support metal coating. The environmental monitoring unit includes a salt spray concentration monitor and a humidity sensor to acquire the electrolyte concentration and saturation of the atmosphere in real time.

[0070] In step 2, a corrosion damage coefficient was added to the feature vector. This coefficient was obtained by calculating the monotonic decay slope of the structure's natural frequency over time, thus eliminating fluctuations caused by environmental loads. In the feature extraction algorithm, wavelet packet transform was introduced to decompose the strain signal of the fiber Bragg grating sensor into multiple scales. By extracting the energy distribution of high-frequency components, it was found that there were microcracks initiating inside the support material due to corrosion.

[0071] In step 3, the adaptive detection threshold adjustment process introduces a corrosion-load coupling model. As the service time of the support increases, the cross-sectional area of ​​the structure decreases due to corrosion, resulting in a decrease in load-bearing capacity. The system estimates the remaining cross-sectional modulus of the support through the real-time corrosion rate and automatically lowers the static displacement threshold and dynamic stress alarm limit accordingly. Ambient temperature monitoring is used to correct the salt spray electrochemical reaction rate, thereby dynamically updating the parameters of the corrosion prediction model.

[0072] In step 4, the multi-dimensional comprehensive performance evaluation adopts a method combining expert system and fuzzy logic. Since sensors are prone to drift in coastal environments, the fuzzy logic layer will evaluate the confidence level of sensor data. If the corrosion sensor data shows abnormal jumps, the system will perform fuzzification processing based on the service life of the support and the average level of surrounding monitoring points, and calculate a comprehensive corrected corrosion level.

[0073] In step 5, tidal cycle data and salinity data were added to the input feature layer of the long short-term memory neural network. The prediction model can predict the trend of mooring cable tension change of the floating support under extreme water levels based on future tidal forecasts. The output operation and maintenance suggestions include: applying secondary anti-corrosion coating to the support in specific areas, checking the stress points of the mooring system, and cleaning the attached organisms on the support surface to reduce its weight.

[0074] The system's communication feedback unit uses BeiDou short message service and underwater ultrasonic communication or low-orbit satellite link in the water environment. To cope with salt spray corrosion, the circuit boards of all sensor nodes are coated with a three-proof coating, namely, salt spray resistance, moisture resistance, and mold resistance. The system shell is made of special corrosion-resistant alloy material. In the digital modeling executed by the central processing unit, a hydrodynamic model is added to simulate the cyclic impact effect of wave loads on the floating support and update the fatigue life file of the support in real time.

[0075] Example 4

[0076] This embodiment details the specific implementation of this system in a large-scale desert and Gobi photovoltaic power station, a scenario characterized by extreme diurnal temperature differences, intense sandstorms, and high levels of ultraviolet radiation.

[0077] In step 1, the environmental monitoring unit is equipped with a dust concentration sensor and an ultraviolet intensity meter. The dust concentration sensor uses the principle of laser scattering to monitor the concentration of PM10 and PM100 in the air to assess the physical impact of sandstorms on the support structure. Since the temperature difference in desert areas can reach more than 60 degrees Celsius, both the tilt sensor and the strain sensor are equipped with active temperature control heating elements to ensure that the sensors can still maintain the preset accuracy range in extremely low temperature environments.

[0078] In step 2, the data preprocessing stage focuses on temperature drift compensation. In the adaptive Kalman filter algorithm, the measurement noise covariance R is dynamically modeled based on the ultraviolet intensity and temperature. Because ultraviolet aging affects the elasticity of the sensor packaging material, it introduces a slow nonlinear drift. The system extracts the temperature drift features and uses a polynomial fitting method to subtract the quasi-static thermally induced strain component from the original signal, retaining only the dynamic feature vector caused by external mechanical loads.

[0079] In step 3, an adaptive detection threshold adjustment process is carried out to address the risk of foundation loosening caused by desertification. A dynamic identification module for foundation reaction coefficient is established. By monitoring the small vibration response of the support under small wind loads, the foundation support stiffness is inferred. When the thickness of the sand and dust increases or the sand and soil are lost, the system automatically updates the earth pressure boundary conditions in the finite element model and recalculates the critical wind pressure threshold for instability of the support in real time.

[0080] In step 4, the comprehensive performance evaluation model introduces a sand and dust wear evaluation factor. This factor calculates the effective load-bearing life decay of the support structure based on the historical cumulative sand and dust intensity and the coating damage model on the support surface. In terms of weight allocation, due to the large thermal stress caused by temperature difference, the weight of the stress concentration factor is increased to 0.4 to prevent brittle fracture of the support nodes under cyclic thermal load.

[0081] In step 5, the long short-term memory neural network was optimized for the sudden characteristics of desert climate. The input feature sequence was enhanced with the hierarchical parameters of sandstorm warning signals. The system can predict whether the support structure is at risk of overturning due to sand erosion during the next sandstorm forecast cycle. The output maintenance suggestions include: sand stabilization treatment of the foundation, cleaning the sand accumulation on the surface of the components to prevent uneven stress, and lubrication maintenance of the mechanical rotating parts of the support structure.

[0082] In terms of system hardware implementation, the central processing unit adopts an industrial-grade wide-temperature chip and has an efficient passive heat dissipation structure. Due to the sparse distribution of communication base stations in desert areas, the data acquisition module supports multi-hop long-distance wireless backhaul. Each acquisition node has an independent watchdog circuit built in, which can achieve millisecond-level automatic reset when encountering extreme strong electromagnetic sandstorm interference that causes a system crash.

[0083] Through the detailed description of the above embodiments, the specific implementation path of the present invention in various complex environments has been fully explained. The system realizes intelligent monitoring and precise operation and maintenance of the entire life cycle of photovoltaic brackets through the fusion perception of multi-source parameters, the dynamic evolution of adaptive detection thresholds, and trend prediction based on deep learning, which significantly improves the safe operation level and maintenance efficiency of photovoltaic systems.

[0084] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An adaptive photovoltaic support performance testing system, characterized in that, include: Data acquisition module: Acquires physical status information and environmental meteorological parameters of photovoltaic support through sensor array and transmits the acquired data to the central processing unit; Central processing unit: Communicatively connected to the data acquisition module, and internally integrates a trend analysis module; used to clean and process the received data and extract feature vectors, the feature vectors being used for adaptive detection threshold matching and modeling and prediction in the trend analysis module, and transmitting the obtained matching results and early warning information to the communication processing unit; Communication feedback unit: It is connected to the central processing unit and is used to receive the matching results and early warning information from the central processing unit and transmit them to the remote monitoring platform.

2. A method for testing the performance of an adaptive photovoltaic support system, applied to the adaptive photovoltaic support system as described in claim 1, characterized in that, Includes the following steps: Step 1: Collect multi-source state parameters and environmental sensing data of the photovoltaic support: Multi-source state parameters are acquired in real time through a sensor network deployed at key structural nodes of the photovoltaic support, and environmental sensing data is acquired through an environmental monitoring unit; The multi-source state parameters include the support tilt angle, structural surface strain, vibration frequency and displacement deviation; The environmental sensing data includes instantaneous wind speed, ambient temperature, snow thickness and solar radiation intensity. Step 2: Preprocessing and feature extraction of the collected multi-source state parameters: Adaptive Kalman filtering algorithm is used to denoise the multi-source state parameters, and the feature vector of the support structure is extracted by time domain and frequency domain analysis methods. Step 3, adaptive detection threshold adjustment: Based on the real-time acquired environmental perception data, combined with the preset photovoltaic support structure mechanical model, calculate the theoretical response value of the support under the current working conditions, and combine with historical operating data to adaptively adjust the alarm thresholds of various performance indicators. Step 4, perform multi-dimensional comprehensive performance evaluation: input the feature vector of the support structure into the multi-criteria decision model, calculate the comprehensive health evaluation index through weighted fusion algorithm, determine the current structural safety level of the support, and identify bolt loosening, material fatigue or foundation settlement defects in the support. Step 5: Predict performance degradation trends and output maintenance decisions: Use a long short-term memory neural network to perform time series analysis on historical health evaluation index sequences, average wind speed sequences, and environmental humidity sequences to predict the performance degradation trend of the support in the future preset time period, and automatically output targeted maintenance decisions based on the prediction results.

3. The method for testing the performance of an adaptive photovoltaic support according to claim 2, characterized in that, Step 2, which involves denoising the multi-source state parameters and extracting the feature vector of the support structure, includes: using an adaptive Kalman filter algorithm to dynamically adjust the process covariance and measurement noise covariance, and making optimal estimates of the sensor random errors to remove noise; using fast Fourier transform to extract vibration mode parameters from the denoised data, using rainflow counting to extract stress cycle features, and constructing a feature vector for health evaluation.

4. The method for testing the performance of an adaptive photovoltaic support according to claim 2, characterized in that, The adaptive detection threshold adjustment in step 3 includes: calculating the response value of the support under the current working condition based on the collected environmental perception data and the structural mechanics model, and dynamically adjusting the alarm thresholds of various performance indicators based on historical operating data.

5. The method for testing the performance of an adaptive photovoltaic support according to claim 2, characterized in that, Step 4, calculating the comprehensive health evaluation index, includes: using the analytic hierarchy process (AHP) to determine the weight coefficients of various indicators such as structural surface strain, natural frequency change rate, displacement deviation, and tilt angle deviation; and using a weighted fusion algorithm to calculate the comprehensive health evaluation index based on the weight coefficients of each indicator, the actual observed values ​​of the features, and the adaptively adjusted detection threshold.

6. The method for testing the performance of an adaptive photovoltaic support according to claim 5, characterized in that, The comprehensive health evaluation index is divided into different preset value ranges, corresponding to healthy status, sub-health status, first-level warning status, and second-level alarm status, respectively.

7. The method for testing the performance of an adaptive photovoltaic support according to claim 2, characterized in that, The performance degradation trend prediction and operation and maintenance decision output in step 5 includes: configuring the input layer nodes, hidden layer neurons and output layer of the long short-term memory neural network, outputting the performance evolution path within a preset time period through the long short-term memory neural network, and controlling the prediction error within a preset error range; the generated operation and maintenance suggestions include: reinforcing the support nodes with specific numbers, adjusting the installation tilt angle of photovoltaic modules to reduce wind resistance, cleaning the water accumulation at the bottom of the support, and replacing fasteners that have reached their fatigue limit.

8. The method for testing the performance of an adaptive photovoltaic support according to claim 2, characterized in that, The method also includes sensor self-maintenance and data completion steps: periodically and automatically performing online zero-point calibration and sensitivity self-check of the sensor; real-time monitoring of the sensor failure status, and when the sensor failure rate exceeds the preset failure threshold, initiating redundant data completion logic; utilizing the correlation characteristics of adjacent supports, employing the Kriging spatial interpolation algorithm to repair data and maintain the continuity of the monitoring process.

9. The method for testing the performance of an adaptive photovoltaic support according to claim 2, characterized in that, For complex terrain environments, the method also includes the following steps: using static levels installed at the beginning and end of the support array to measure the vertical displacement of the columns relative to the reference point using the principle of connecting pipes; introducing a terrain correction coefficient during the adaptive detection threshold adjustment process, calculating the local slope and altitude of the support nodes using a micro-digital elevation model of the power station, and converting the foundation wind speed into the actual wind load pressure acting on a specific support using a local terrain roughness function; increasing the weight of the foundation settlement index when performing multi-dimensional comprehensive performance evaluation, and using Markov chain analysis to calculate the probability matrix of the support transitioning from a healthy state to a sub-healthy state, thereby achieving probabilistic identification of early latent defects.

10. The method for testing the performance of an adaptive photovoltaic mounting system according to claim 2, characterized in that, For coastal corrosive environments, the method further includes the following steps: using a metal corrosion sensor integrated in a multi-source state parameter acquisition module, the resistance change on the support surface is measured by resistance analysis to characterize the corrosion rate of the support's metal coating; a corrosion damage coefficient is introduced into the feature vector, which is obtained by calculating the monotonic decay slope of the structure's natural frequency over time; a coupling model of corrosion and load is introduced during the adaptive detection threshold adjustment process, the remaining cross-sectional modulus of the support is estimated based on the real-time corrosion rate, and the static displacement threshold and dynamic stress alarm limit are adjusted accordingly.

Images