A cloud edge fusion-based photovoltaic tracking support energy consumption optimization method

Abstract

The application discloses a kind of based on cloud edge fusion's photovoltaic tracking support energy consumption optimization method, it is related to cloud edge fusion technical field, including S1, the real-time external environmental variable data of photovoltaic tracking support and equipment mechanical characteristic data are collected in edge side, and data is uploaded to cloud end;External environmental variable data include illumination trajectory change data, equipment mechanical characteristic data include with the energy consumption and operating state data of photovoltaic tracking support driving mechanism;The based on cloud edge fusion's photovoltaic tracking support energy consumption optimization method, reduce the overall operation energy consumption of photovoltaic tracking support, enhance the adaptability, stability and operating efficiency of system in large-scale photovoltaic power station application scenario.

Description

Technical Field

[0001] This invention relates to the field of cloud-edge fusion technology, and more specifically to a method for optimizing the energy consumption of photovoltaic tracking brackets based on cloud-edge fusion. Background Technology

[0002] In the field of energy management and equipment operation optimization, reducing equipment operating energy consumption and improving the overall system operating efficiency have always been important research directions. This is especially true in new energy power generation systems, where the maturity of related technologies directly affects power generation efficiency, operating costs, and the long-term reliability of the system. As a key actuator in photovoltaic power generation systems, the operating status and energy consumption level of photovoltaic tracking brackets have a significant impact on the overall system performance.

[0003] Existing energy management methods for photovoltaic (PV) tracking systems typically rely on equipment operation status monitoring or historical energy consumption statistical analysis to assess the system's energy consumption level over a certain period. However, these methods often depend on offline analysis or static data modeling, lacking a comprehensive consideration of real-time changes in the external environment and the dynamic characteristics of equipment operation. When illumination conditions, environmental factors, or load conditions change, existing methods struggle to reflect energy consumption trends in a timely manner, leading to delayed identification of energy consumption anomalies and impacting operational optimization. In actual operation, the energy consumption performance of PV tracking systems is influenced by multiple factors, with external environmental variables and the equipment's own mechanical characteristics being two key influencing factors. On one hand, the continuous change in the illumination trajectory over time directly affects the tracking system's movement frequency and driving load, exhibiting significant dynamism and uncertainty. On the other hand, due to manufacturing errors, long-term wear and tear, and differences in maintenance conditions, different systems or different driving components of the same system may exhibit significantly different mechanical characteristics and energy consumption behaviors under the same environmental conditions. The coupling of these external environmental fluctuations and differences in equipment mechanical characteristics makes it difficult to accurately determine the root cause when energy consumption deviations occur during PV tracking system operation. For example, under the same illumination trajectory, some support structures may experience abnormally high energy consumption. Current technologies often struggle to distinguish whether this anomaly is due to normal energy consumption changes caused by environmental fluctuations or internal factors such as decreased performance of the drive mechanism or increased mechanical friction, leading to uncertainty in determining the source of energy consumption deviation. Furthermore, as photovoltaic power plants continue to expand in scale, the limitations of a single centralized processing method in terms of data transmission real-time performance, computational response speed, and system scalability are becoming increasingly apparent. How to efficiently process large amounts of operational data from distributed photovoltaic tracking supports while ensuring real-time performance, and accurately identify energy consumption anomalies and their causes under dynamic environmental conditions, remains a pressing problem to be solved in current technologies. Summary of the Invention

[0004] The purpose of this invention is to provide a method for optimizing the energy consumption of photovoltaic tracking brackets based on cloud-edge fusion, thereby solving the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic tracking bracket energy consumption optimization method based on cloud-edge fusion, comprising: S1, collecting real-time external environmental variable data and equipment mechanical characteristic data of the photovoltaic tracking bracket at the edge side, and uploading the data to the cloud; the external environmental variable data includes light trajectory change data, and the equipment mechanical characteristic data includes energy consumption and operating status data related to the photovoltaic tracking bracket drive mechanism; S2, processing the collected external environmental variable data and equipment mechanical characteristic data in the cloud, and using a random forest algorithm to analyze the data to obtain preliminary classification results of the energy consumption deviation sources; S3, based on the preliminary classification results, establishing light trajectory fluctuation data in the cloud. The system establishes a correlation between the differences in equipment mechanical characteristics and determines whether fluctuations in the illumination trajectory exceed a preset threshold. When the preset threshold is exceeded, it extracts the characteristics of environmental fluctuations to identify potential indicators of internal performance degradation. S4: In the cloud, a support vector machine algorithm is used to classify the characteristics of environmental fluctuations and potential indicators of internal performance degradation, resulting in an abnormal energy consumption pattern under dynamic environmental change conditions. S5: Based on the abnormal energy consumption pattern, the system analyzes the interaction data between differences in equipment mechanical characteristics and fluctuations in the illumination trajectory, and determines whether the interaction data indicates that differences in equipment mechanical characteristics are the dominant factor. When differences in equipment mechanical characteristics are the dominant factor, information processing is used to separate the impact of environmental fluctuations and accurately locate the source of energy consumption deviation.

[0006] Preferably, step S1 includes receiving real-time illumination trajectory change data collected by the photovoltaic tracking bracket and instantaneous voltage and current signals of the drive mechanism; calculating real-time energy consumption values ​​based on the instantaneous voltage and current signals, and associating the real-time energy consumption values ​​with the real-time illumination trajectory change data to obtain an environmental mechanical coupling data stream; if the illumination angle change rate in the environmental mechanical coupling data stream exceeds a preset threshold, then extracting the operating status data of the drive mechanism to obtain dynamic mechanical characteristic samples; encapsulating the dynamic mechanical characteristic samples and real-time illumination trajectory change data to generate an edge-side reporting data packet, and sending the edge-side reporting data packet to the cloud management platform to realize remote digital mapping of multi-dimensional data on-site of the photovoltaic tracking bracket.

[0007] Preferably, step S2 includes acquiring external environmental variable data and equipment mechanical characteristic data to generate a standardized time series dataset; constructing a multidimensional feature vector matrix based on the standardized time series dataset and inputting the multidimensional feature vector matrix into a random forest model; calculating the feature split gain using the random forest model and determining the energy consumption deviation feature weight ranking based on the feature split gain; mapping the energy consumption deviation feature weight ranking to a fault category label library and determining the preliminary classification result of the energy consumption deviation source based on the fault category label library.

[0008] Preferably, step S3 includes acquiring preliminary classification results and light trajectory fluctuation data stored in the cloud, constructing a nonlinear correspondence model between the light trajectory fluctuation data and the differences in mechanical characteristics of the equipment, and outputting the mechanical characteristic difference value; combining the mechanical characteristic difference value with the light trajectory fluctuation data to calculate the instantaneous amplitude, and generating a fluctuation over-limit signal if the instantaneous amplitude is greater than a preset threshold; extracting the environmental fluctuation influence feature vector in response to the fluctuation over-limit signal, and analyzing the environmental fluctuation influence feature vector to determine potential indicators of internal performance degradation of the equipment.

[0009] Preferably, step S4 includes acquiring historical environmental fluctuation data and device internal performance data stored in the cloud, obtaining the feature vector of environmental fluctuation impact and the set of potential indicators of internal performance degradation through time-series alignment; mapping the feature vector of environmental fluctuation impact and the set of potential indicators of internal performance degradation to generate dynamic environmental change conditions; using a support vector machine algorithm to perform kernel function mapping on the dynamic environmental change conditions to calculate the classification decision function value, determining the energy consumption anomaly label based on the classification decision function value; and aggregating data with the same energy consumption anomaly label to obtain the energy consumption anomaly pattern under the dynamic environmental change conditions.

[0010] Preferably, step S5 includes acquiring drive motor current data and light intensity sequence, constructing a multi-dimensional time series set by combining energy consumption anomaly patterns, and generating an interactive data matrix based on the multi-dimensional time series set; extracting covariance feature values ​​from the interactive data matrix, and determining that the difference in equipment mechanical characteristics is the dominant factor if the covariance feature value is lower than a threshold; filtering out environmental fluctuation components to obtain a net energy consumption residual sequence for the state where the difference in equipment mechanical characteristics is the dominant factor; mapping the net energy consumption residual sequence to a physical topology model, and locking the coordinates of mechanical components to achieve accurate location of the source of energy consumption deviation.

[0011] Preferably, the method further includes S6: based on the precise positioning result, using a random forest algorithm in the cloud to optimize the mechanical characteristic data of the equipment to obtain an internal performance degradation compensation value, and sending the compensation value to the edge side. Specifically, this includes obtaining a spatial coordinate sequence and equipment mechanical characteristic data generated based on the precise positioning result, concatenating the spatial coordinate sequence and equipment mechanical characteristic data to construct a high-dimensional feature vector, and inputting the high-dimensional feature vector into the cloud random forest model to obtain a performance status prediction value.

[0012] Preferably, step S6 further includes calculating the deviation between the predicted performance status value and the standard performance curve; if the deviation value is greater than a preset threshold, the internal performance degradation amount is determined and an internal performance degradation compensation value is generated; the internal performance degradation compensation value is sent to the edge side to achieve device optimization.

[0013] Preferably, it also includes S7, adjusting the operating parameters of the photovoltaic tracking bracket on the edge side based on the internal performance degradation compensation value, obtaining fusion data of light trajectory fluctuation and environmental fluctuation impact under dynamic environmental change conditions, and outputting the root cause identification result of the overall energy consumption anomaly of the photovoltaic tracking bracket when the environmental fluctuation impact is less than a preset threshold. Specifically, it includes receiving the internal performance degradation compensation value, generating a corrected bracket tilt angle command based on the internal performance degradation compensation value to drive the photovoltaic tracking bracket, collecting the position feedback sequence during the operation of the photovoltaic tracking bracket, and generating environmental fluctuation impact factor and system state fusion data by combining real-time environmental data and position feedback sequence.

[0014] Preferably, step S7 further includes removing noise from the system state fusion data to construct an energy consumption anomaly sequence if the environmental fluctuation impact factor is less than a preset threshold; matching the energy consumption anomaly sequence with a fault mode library and outputting the root cause identification result of the overall energy consumption anomaly of the photovoltaic tracking bracket.

[0015] As can be seen from the above technical solution, the present invention has the following beneficial effects: This cloud-edge fusion-based photovoltaic (PV) tracking bracket energy consumption optimization method establishes a collaborative processing mechanism between the edge and cloud sides. It performs layered collection and analysis of real-time external environmental variable data and equipment mechanical characteristic data of the PV tracking bracket, enabling more accurate identification of energy consumption deviation sources under dynamic environmental conditions. This invention comprehensively considers the coupled effects of light trajectory fluctuations and differences in equipment mechanical characteristics. Through analysis of energy consumption anomaly patterns, it effectively distinguishes between environmental fluctuation factors and internal performance degradation factors, avoiding the inaccurate energy consumption anomaly judgments caused by relying on static data or single-factor analysis in existing technologies. Simultaneously, by calculating internal performance degradation compensation values ​​in the cloud and distributing them to the edge side for operational parameter adjustments, the PV tracking bracket can achieve energy consumption optimization while ensuring real-time response capabilities. This effectively improves the accuracy of energy consumption anomaly location and the timeliness of operational adjustments, reduces the overall operating energy consumption of the PV tracking bracket, and enhances the system's adaptability, stability, and operational efficiency in large-scale PV power plant applications. Attached Figure Description

[0016] Figure 1 This is a flowchart of the energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to the present invention. Detailed Implementation

[0017] 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.

[0018] like Figure 1 As shown, this invention provides a technical solution: a photovoltaic tracking bracket energy consumption optimization method based on cloud-edge fusion, including S1, collecting real-time external environmental variable data and equipment mechanical characteristic data of the photovoltaic tracking bracket at the edge side, and uploading the data to the cloud; the external environmental variable data includes light trajectory change data, and the equipment mechanical characteristic data includes energy consumption and operating status data related to the photovoltaic tracking bracket drive mechanism; S2, processing the collected external environmental variable data and equipment mechanical characteristic data in the cloud, and using a random forest algorithm to analyze the data to obtain preliminary classification results of the energy consumption deviation sources; S3, based on the preliminary classification results, establishing a correspondence between light trajectory fluctuation data and equipment mechanical characteristic differences in the cloud, and determining whether the light trajectory fluctuation exceeds a preset threshold; when it exceeds the preset threshold, extracting the environmental fluctuation impact features to determine potential indicators of equipment internal performance degradation; S4, using a support vector machine algorithm in the cloud to analyze the data... S5. Classify the characteristics of environmental fluctuations and potential indicators of internal performance degradation to obtain energy consumption anomaly patterns under dynamic environmental change conditions; S6. Based on the energy consumption anomaly patterns, analyze the interaction data between differences in equipment mechanical characteristics and fluctuations in light trajectory, and determine whether the interaction data indicates that differences in equipment mechanical characteristics are the dominant factor; when differences in equipment mechanical characteristics are the dominant factor, separate the impact of environmental fluctuations through information processing to accurately locate the source of energy consumption deviation; S7. Based on the accurate location results, use the random forest algorithm in the cloud to optimize the equipment mechanical characteristic data to obtain internal performance degradation compensation values, and send the compensation values ​​to the edge side; S8. On the edge side, adjust the operating parameters of the photovoltaic tracking bracket based on the internal performance degradation compensation values, obtain the fusion data of light trajectory fluctuations and environmental fluctuations under dynamic environmental change conditions, and output the root cause identification result of the overall energy consumption anomaly of the photovoltaic tracking bracket when the impact of environmental fluctuations is less than a preset threshold.

[0019] In the above embodiment, the photovoltaic tracking bracket forms a cloud-edge fusion architecture with the cloud analysis platform through an edge-side control unit. The edge-side control unit is located on the photovoltaic tracking bracket body and is used to collect external environmental variable data and equipment mechanical characteristic data in real time. The external environmental variable data mainly reflects the changes in illumination conditions over time, including changes in solar altitude angle, azimuth angle, and fluctuations in the light trajectory caused by cloud obstruction; the equipment mechanical characteristic data is used to characterize the working status of the drive mechanism, including motor power consumption, drive current, rotational resistance, start-stop frequency, and operational stability.

[0020] After preprocessing, the collected data is uploaded to the cloud. The cloud platform first uses a random forest algorithm to perform feature importance analysis and classification on the environmental variable data and mechanical characteristic data, thereby initially identifying the possible sources of energy consumption deviations. This step helps distinguish whether abnormal energy consumption is more likely due to environmental fluctuations or changes in the internal performance of the equipment.

[0021] After obtaining the initial classification results, the cloud platform further establishes a mapping relationship between the light trajectory fluctuation data and the differences in the mechanical characteristics of the equipment, and compares the real-time light trajectory fluctuation amplitude with a preset threshold. When the light trajectory fluctuation exceeds the threshold, it indicates that environmental changes have a significant impact on the operation of the support. At this time, environmental fluctuation impact characteristics that characterize the degree of environmental disturbance are extracted from the data. At the same time, combined with the historical data of equipment operation, potential indicators that may reflect the decline in internal performance are screened out, such as decreased drive efficiency, increased response lag, or mismatch between energy consumption and rotation angle.

[0022] Subsequently, the cloud-based system employs a support vector machine algorithm to jointly classify the characteristics of environmental fluctuations and potential indicators of internal performance degradation, forming an energy consumption anomaly pattern under dynamic environmental changes. This energy consumption anomaly pattern is used to describe the impact of changes in equipment mechanical characteristics on overall energy consumption under different environmental disturbance intensities.

[0023] Based on this, the cloud analyzes the interaction data between differences in equipment mechanical characteristics and fluctuations in light trajectory. By determining the dominant factor in the interaction data, it identifies whether energy consumption anomalies are primarily caused by differences in equipment mechanical characteristics. When differences in equipment mechanical characteristics are determined to be the dominant factor, information processing methods are used to separate the impact of environmental fluctuations, thereby achieving precise location of the source of energy consumption deviations.

[0024] After achieving precise positioning, the cloud-based system further utilizes a random forest algorithm to optimize and analyze the equipment's mechanical characteristic data, calculating a compensation value reflecting the degree of performance degradation within the equipment. This compensation value is then sent to the edge-side control unit to guide the adjustment of the support's operating parameters. The edge-side unit modifies the drive control strategy based on the compensation value, ensuring the support maintains reasonable tracking action and energy consumption levels even under dynamic environmental changes. Finally, when the impact of environmental fluctuations is less than a preset threshold, the system outputs the root cause identification result for the overall abnormal energy consumption of the photovoltaic tracking support.

[0025] By employing a cloud-edge fusion architecture, collaborative analysis of environmental data and equipment status data is achieved, improving the accuracy of energy consumption analysis while ensuring real-time performance. Random forest and support vector machine algorithms are used for hierarchical identification and pattern classification of energy consumption anomalies, effectively distinguishing between environmental fluctuations and internal equipment performance degradation, thus avoiding misjudgments. By establishing a correspondence between light trajectory fluctuations and mechanical characteristic differences, the source of energy consumption deviations can be accurately located, reducing manual investigation costs. By issuing internal performance degradation compensation values ​​and adjusting parameters at the edge, the operational stability and energy efficiency of photovoltaic tracking brackets under complex environmental conditions are improved. The overall method exhibits good adaptability and scalability, applicable to photovoltaic tracking bracket systems of different sizes and structural forms.

[0026] S1 includes receiving real-time illumination trajectory change data and instantaneous voltage and current signals of the drive mechanism collected by the photovoltaic tracking bracket; calculating real-time energy consumption values ​​based on the instantaneous voltage and current signals, and associating the real-time energy consumption values ​​with the real-time illumination trajectory change data to obtain an environmental mechanical coupling data stream; if the illumination angle change rate in the environmental mechanical coupling data stream exceeds a preset threshold, extracting the operating status data of the drive mechanism to obtain dynamic mechanical characteristic samples; encapsulating the dynamic mechanical characteristic samples and real-time illumination trajectory change data to generate an edge-side reporting data packet, and sending the edge-side reporting data packet to the cloud management platform to realize remote digital mapping of multi-dimensional data on-site of the photovoltaic tracking bracket.

[0027] In this embodiment, during system operation, a data acquisition unit connected to the electrical interface of the light sensor and the drive mechanism is installed on the edge side of the photovoltaic tracking bracket. This data acquisition unit continuously receives real-time light trajectory change data collected by the photovoltaic tracking bracket during operation, as well as instantaneous voltage and current signals from the drive mechanism. The real-time light trajectory change data reflects the change of the solar incidence angle over time, and its data source can be a photosensitive array sensor, an angle sensor, or an angle encoding device coaxially mounted with the photovoltaic tracking bracket's rotating shaft. The instantaneous voltage and current signals are directly acquired from the power supply circuit of the drive mechanism, accurately reflecting the energy consumption of the drive mechanism under different operating conditions.

[0028] After data acquisition is completed, the edge-side data acquisition unit first performs synchronous processing on the instantaneous voltage and current signals. By pairing voltage and current data within the same time slice, the instantaneous energy consumption state corresponding to that time slice is obtained. Subsequently, the instantaneous energy consumption states within consecutive time slices are accumulated to obtain the real-time energy consumption value of the drive mechanism within the corresponding time period. This real-time energy consumption value does not rely on preset model inference but is directly calculated based on actual electrical signals, thus accurately reflecting the actual energy consumption level of the drive mechanism.

[0029] After obtaining the real-time energy consumption value, the edge data acquisition unit correlates this value with real-time illumination trajectory change data on the same time axis. Specifically, through timestamp alignment, each energy consumption value is bound to the corresponding moment's incident light angle, angle change direction, and angle change amplitude, thus forming an environmental-mechanical coupling data stream. This environmental-mechanical coupling data stream can simultaneously describe changes in the external illumination environment and the mechanical actions and energy consumption of the drive mechanism in response to these changes.

[0030] After the environmental mechanical coupling data stream is generated, the edge-side data acquisition unit further analyzes the illumination trajectory change data continuously, calculates the change amplitude of the illumination angle between adjacent time slices, and obtains the illumination angle change rate accordingly. The illumination angle change rate reflects the severity of changes in illumination conditions; its calculation is based on the actual measured angle change and corresponding time intervals, without involving any empirical assumptions. Subsequently, this illumination angle change rate is compared with a preset threshold. This preset threshold is determined as follows: during the initial system deployment phase, statistical analysis is performed on the operating data of the photovoltaic tracking bracket under stable sunny conditions to extract the maximum stable value of the illumination angle change rate during normal tracking. Based on this, a safety margin is considered, ensuring that when the angle change rate exceeds this threshold, it clearly indicates a significant fluctuation in the external environment.

[0031] When the rate of change of illumination angle exceeds a preset threshold, the edge-side data acquisition unit triggers a data interception mechanism. This mechanism does not simply record data at a single moment; instead, it extends a preset time window forward and backward from the trigger moment, comprehensively capturing the operational status data of the drive mechanism within that time window. This operational status data includes changes in the drive mechanism's rotational state, energy consumption trends, and response delays. This method yields a dynamic mechanical characteristic sample that comprehensively reflects the drive mechanism's operating characteristics under dynamic environmental conditions. This dynamic mechanical characteristic sample is used to describe the drive mechanism's true mechanical response capability and energy consumption characteristics under rapidly changing illumination conditions.

[0032] After acquiring the dynamic mechanical characteristic samples, the edge-side data acquisition unit organizes and structures the samples along with the real-time illumination trajectory change data within the corresponding time period. Specifically, various data types are assigned unified time, device, and data type identifiers and encapsulated according to preset data structure rules to form an edge-side reporting data packet. This data packet contains multi-dimensional information that comprehensively describes the operating status of the photovoltaic tracking bracket under specific environmental changes.

[0033] Finally, the edge-side data acquisition unit sends the edge-side reported data packets to the cloud management platform via the communication module. Upon receiving the data packets, the cloud management platform can remotely digitally map the on-site operating status of the photovoltaic tracking bracket based on the environmental and mechanical data contained within. This provides a reliable data foundation for subsequent energy consumption anomaly analysis, energy consumption deviation source identification, and operating parameter optimization. The entire process involves real-time calculation and filtering at the edge side and centralized analysis and modeling in the cloud, thereby ensuring data authenticity while improving the overall analytical efficiency and operational stability of the system.

[0034] S2 includes acquiring external environmental variable data and equipment mechanical characteristic data to generate a standardized time series dataset; constructing a multidimensional feature vector matrix based on the standardized time series dataset and inputting the multidimensional feature vector matrix into a random forest model; calculating the feature split gain using the random forest model and determining the energy consumption deviation feature weight ranking based on the feature split gain; mapping the energy consumption deviation feature weight ranking to a fault category label library and determining the preliminary classification result of the energy consumption deviation source based on the fault category label library.

[0035] In this embodiment, after receiving external environmental variable data and equipment mechanical characteristic data reported from the edge side, the cloud management platform first performs unified reception and caching processing on the data. Each data item carries a corresponding time identifier and device identifier. The cloud platform sequentially arranges the data from different acquisition channels according to the time identifier and groups them into groups at preset time intervals, thereby generating raw time-series data with a continuous temporal order. This time interval is determined based on the actual response speed and data acquisition frequency of the photovoltaic tracking bracket. It is determined during the system deployment phase through statistical analysis of the equipment's operating cycle, ensuring that it fully reflects the environmental change process without introducing redundant data.

[0036] After constructing the original time series data, the cloud platform performs standardization on the data. The purpose of standardization is to eliminate differences in numerical ranges and units between different data types, ensuring comparability of various features in subsequent analyses. Specifically, the cloud platform first performs historical statistics on each type of external environmental variable and equipment mechanical characteristic data to obtain their minimum, maximum, and typical fluctuation ranges under normal operating conditions. Then, the currently collected data is mapped to a unified numerical range, allowing data from different sources to express their trends on the same scale. This standardization process forms a standardized time series dataset, providing stable input for feature construction.

[0037] After the standardized time series dataset is generated, the cloud platform extracts multiple indicators reflecting energy consumption changes from the data within each time slice or time window according to pre-defined feature construction rules. These indicators are then combined in a fixed order to construct a multi-dimensional feature vector matrix. Each row in this multi-dimensional feature vector matrix corresponds to the comprehensive operating status of the photovoltaic tracking bracket within a certain time slice, and each column corresponds to a specific environmental variable or mechanical characteristic. The selection of feature dimensions is based on the results of long-term operational data analysis, prioritizing indicators that have a significant impact on energy consumption changes and exhibit high stability to ensure the reliability of the model analysis results.

[0038] After constructing the multidimensional feature vector matrix, the cloud platform inputs this matrix into the random forest model for analysis and processing. The random forest model consists of multiple decision trees. During model operation, the cloud platform trains the model on historical labeled data, enabling each decision tree to judge energy consumption status based on different feature combinations. During the model inference phase, each decision tree evaluates the contribution of the features involved in the partitioning when splitting nodes, recording the gain changes generated by the feature in distinguishing different energy consumption states.

[0039] After the random forest model completes its overall analysis, the cloud platform aggregates the feature contributions recorded in all decision trees. The contributions of the same feature across different decision trees are summed and averaged to obtain the overall split gain of that feature in the entire random forest model. This split gain quantifies the influence of each feature on the energy consumption deviation discrimination result; a larger value indicates a more significant role for the feature in distinguishing between abnormal and normal energy consumption states.

[0040] After obtaining the comprehensive split gain corresponding to each feature, the cloud platform sorts all features in descending order of split gain, forming a weighted ranking of energy consumption deviation features. This ranking result reflects the relative importance of different external environmental variables and equipment mechanical characteristics to the energy consumption deviation phenomenon under the current operating state, thus providing a basis for subsequent deviation source determination.

[0041] After ranking the energy consumption deviation features by weight, the cloud platform maps and matches the ranking results with a pre-built fault category label library. This fault category label library was established during the system's long-term operation by organizing and summarizing known energy consumption anomaly cases, and each fault category label corresponds to a typical feature weight distribution pattern. During the mapping process, the cloud platform compares the current feature weight ranking with the feature patterns corresponding to each fault category label, determines their similarity, and selects the fault category label with the highest similarity as the judgment result for the current energy consumption deviation source.

[0042] The method for determining the various feature patterns in the fault category tag library is as follows: During the system initialization and operation and maintenance phases, long-term collection and analysis of operational data of photovoltaic tracking brackets under different known abnormal states are conducted to extract stable feature weight distribution characteristics, and these distribution characteristics are then solidified into corresponding fault category tags. This method ensures that the tag library content has repeatability and engineering applicability.

[0043] S3 includes acquiring preliminary classification results and light trajectory fluctuation data from cloud storage, constructing a nonlinear correspondence model between the light trajectory fluctuation data and the differences in equipment mechanical characteristics, and outputting the mechanical characteristic difference values; combining the mechanical characteristic difference values ​​with the light trajectory fluctuation data to calculate the instantaneous amplitude, and generating a fluctuation over-limit signal if the instantaneous amplitude is greater than a preset threshold; extracting the environmental fluctuation impact feature vector in response to the fluctuation over-limit signal, and analyzing the environmental fluctuation impact feature vector to determine potential indicators of equipment internal performance degradation.

[0044] In this embodiment, after completing the preliminary classification of energy consumption deviation sources, the cloud management platform uses the preliminary classification results as a prerequisite for subsequent analysis. The preliminary classification results define the main influencing directions of the current analysis, providing a clear analytical boundary for the subsequent modeling process of the relationship between light trajectory fluctuations and differences in equipment mechanical characteristics, and preventing irrelevant data from interfering with the results. Simultaneously, the cloud retrieves light trajectory fluctuation data from the data storage unit within the time period corresponding to the preliminary classification results. This light trajectory fluctuation data is collected and reported from the edge side; it has undergone time alignment and standardization processing in the cloud, continuously reflecting the true trajectory of light incidence angle changes over time.

[0045] After obtaining the preliminary classification results and light trajectory fluctuation data, the cloud further retrieves equipment mechanical characteristic data within the same time window. This equipment mechanical characteristic data includes multiple parameters reflecting the operating status and energy consumption characteristics of the drive mechanism. Using the preliminary classification results as constraints, the cloud takes the light trajectory fluctuation data and equipment mechanical characteristic data as joint inputs and continuously trains the model using historical operating data to construct a nonlinear correspondence model between the light trajectory fluctuation data and the differences in equipment mechanical characteristics. The construction process of this nonlinear correspondence model is based on a large number of normal and abnormal operating samples, enabling the model to identify the changing trends of equipment mechanical characteristics deviating from normal states under different light fluctuation intensities, thereby avoiding fitting bias caused by using linear relationships.

[0046] After the nonlinear correspondence model is constructed, the cloud inputs real-time illumination trajectory fluctuation data into the model for calculation, and the model outputs the corresponding mechanical characteristic difference value. This mechanical characteristic difference value is used to quantify the degree of deviation of the current mechanical state of the equipment from its normal operating baseline state. The normal operating baseline state is determined as follows: during the long-term stable operation phase of the equipment, the cloud performs statistical analysis on the equipment's mechanical characteristic data, filters out time periods with small fluctuations and stable operating states, and uses the characteristic value range within this time period as the normal operating reference interval. When the real-time data exceeds this reference interval, the model calculates the corresponding difference value, thereby ensuring that the difference value can truly reflect changes in the equipment state, rather than short-term noise fluctuations.

[0047] After obtaining the numerical values ​​of mechanical characteristic differences, the cloud platform jointly processes these values ​​with the light trajectory fluctuation data within the corresponding time period to calculate the instantaneous amplitude. The instantaneous amplitude comprehensively characterizes the coupling effect between the intensity of light trajectory fluctuations and the degree of deviation of the equipment's mechanical characteristics. Its calculation process involves weighted integration of the intensity of light fluctuations and the degree of mechanical differences within the same time window. The length of this time window is determined during the system deployment phase, typically selecting a time range that fully covers a significant light change process to ensure that the instantaneous amplitude reflects both environmental changes and the equipment's response characteristics.

[0048] After calculating the instantaneous amplitude, the cloud compares this instantaneous amplitude with a preset threshold. The preset threshold is determined as follows: during system initialization and operational debugging, historical data of the photovoltaic tracking bracket under normal environmental conditions is statistically analyzed, and the maximum stable instantaneous amplitude under normal operating conditions is extracted. A safety margin is then introduced based on this value to ensure that when the instantaneous amplitude exceeds this threshold, it can be clearly determined that current environmental fluctuations have an abnormal impact on equipment operation. This threshold is not fixed but can be updated and adjusted according to the equipment model, installation area, and historical operating characteristics.

[0049] When the instantaneous amplitude exceeds a preset threshold, a fluctuation exceeding the limit signal is generated in the cloud. This fluctuation exceeding the limit signal indicates that the current operating state has exceeded the normal environmental adaptation range and serves as a trigger to initiate the environmental fluctuation impact feature extraction process. The generation of this signal avoids continuous in-depth analysis of all data, thereby reducing the cloud computing load.

[0050] Upon responding to an out-of-limit fluctuation signal, the cloud platform extracts an environmental fluctuation impact feature vector from multidimensional data within the corresponding time window. This feature vector consists of multiple features reflecting the impact of environmental changes on equipment operation, selected based on stability, repeatability, and sensitivity to changes in equipment performance. During feature extraction, the cloud platform analyzes data trends over continuous time periods to avoid feature distortion caused by single instantaneous anomalies.

[0051] After extracting the feature vectors reflecting the impact of environmental fluctuations, the cloud platform analyzes these feature vectors. During analysis, the cloud platform combines historical equipment operating data and preliminary classification results to identify changes in the feature vectors that reflect equipment response lag, abnormal energy consumption amplification, and decreased operational stability. Based on this, it determines potential indicators of internal equipment performance degradation. These potential indicators describe the performance degradation characteristics that gradually develop over long-term operation, rather than transient changes caused by short-term environmental disturbances.

[0052] S4 includes acquiring historical environmental fluctuation data and device internal performance data stored in the cloud, obtaining the feature vector of environmental fluctuation impact and the set of potential indicators of internal performance degradation through time-series alignment; mapping the feature vector of environmental fluctuation impact and the set of potential indicators of internal performance degradation to generate dynamic environmental change conditions; using the support vector machine algorithm to perform kernel function mapping on the dynamic environmental change conditions to calculate the classification decision function value, and determining the energy consumption anomaly label based on the classification decision function value; aggregating data with the same energy consumption anomaly label to obtain the energy consumption anomaly pattern under the dynamic environmental change conditions.

[0053] In this embodiment, after acquiring the feature vector of environmental fluctuations and the potential indicators of internal performance degradation of the equipment, the cloud management platform first retrieves historical environmental fluctuation data and internal equipment performance data from the cloud historical data storage unit. The historical environmental fluctuation data reflects the external environmental changes of the photovoltaic tracking bracket over different time periods, while the internal equipment performance data reflects the performance change trends of the drive mechanism during long-term operation. Both types of data are appended with a unified time stamp during the storage phase; therefore, when the cloud retrieves the data, it can directly perform preliminary sorting based on the time stamp.

[0054] After data retrieval, the cloud performs time-series alignment processing on historical environmental fluctuation data and internal device performance data. Specifically, the cloud uses a system-defined time granularity as a benchmark, pairing environmental fluctuation data within the same time window with internal device performance data, and removing missing or abnormal data. The time granularity is determined based on the actual operating rhythm of the photovoltaic tracking bracket and the frequency characteristics of environmental changes. It is established during system initialization through statistical analysis of historical data, ensuring that the data within each time window fully reflects both the environmental change process and the device's performance response. Through time-series alignment processing, the cloud generates a one-to-one correspondence of environmental fluctuation impact feature vectors and a set of potential indicators of internal performance degradation along the time dimension.

[0055] The environmental fluctuation impact feature vector consists of multiple features characterizing the degree of impact of environmental changes on equipment operation. These features are selected based on their significant and long-term stable impact on energy consumption changes in historical operating data. The potential internal performance degradation indicator set consists of multiple indicators reflecting the trend of internal performance degradation. These indicators are selected based on parameters exhibiting continuously changing characteristics during long-term equipment operation. The specific number and composition of each feature and indicator are determined during the system deployment phase and can be adjusted according to the equipment model and operating scenario.

[0056] After obtaining the feature vector of environmental fluctuations and the set of potential indicators of internal performance degradation, the cloud platform maps and combines these two to generate dynamic environmental change conditions characterizing the current operating state. The generation process of these dynamic environmental change conditions involves the unified organization and correlation of environmental impact characteristics and internal performance indicators within the same time window, enabling them to jointly describe the comprehensive characteristics of the equipment's operating state under a specific environmental context. In this way, dynamic environmental change conditions can simultaneously reflect the combined effects of external environmental changes and changes in the equipment's own performance on energy consumption behavior.

[0057] After the dynamic environmental change conditions are constructed, the cloud-based system uses a support vector machine (SVM) algorithm to classify these conditions. Before the model runs, the SVM model has already been trained using historical labeled data during the system deployment phase. The kernel function type used during training is selected based on the distribution of historical data in the feature space to ensure good discriminative power across different energy consumption states in the mapping space. The kernel function parameters are determined by validating multiple parameter combinations during training and selecting the set of parameters with the highest classification stability on historical data as the fixed parameters for the model.

[0058] During model inference, the cloud inputs dynamic environmental changes into the support vector machine model. The model first performs kernel function mapping on the input data, making the overlapping data in the low-dimensional space have clearer boundaries after mapping. Subsequently, the model calculates the corresponding classification decision function value based on the mapped data. The classification decision function value is used to characterize the relative proximity between the current operating state and different energy consumption anomaly categories, and its numerical change directly reflects the positional relationship of the energy consumption state on both sides of the classification boundary.

[0059] After obtaining the classification decision function value, the cloud compares this value with the pre-set classification rules in the model. These classification rules, determined during model training, map continuous decision function values ​​to discrete energy consumption anomaly labels. When the classification decision function value meets the judgment condition corresponding to a certain energy consumption anomaly category, the cloud marks the operating state with the corresponding energy consumption anomaly label, thereby completing the identification of energy consumption anomalies under a single dynamic environmental change condition.

[0060] After identifying energy consumption anomaly labels, the cloud aggregates data with the same labels. Specifically, the cloud groups data within different time windows according to the energy consumption anomaly labels, merging data exhibiting similar anomaly characteristics under similar dynamic environmental changes into the same dataset. Subsequently, the cloud performs statistical analysis on this dataset, extracting common characteristics in environmental features, equipment performance, and energy consumption, ultimately forming an energy consumption anomaly pattern under dynamic environmental changes.

[0061] The determination of energy consumption anomaly patterns is not based on data from a single time slice, but rather through comprehensive analysis of data with consistent anomaly labels across multiple time windows. This ensures the stability and engineering applicability of the resulting anomaly patterns. Through these steps, the cloud platform achieves a systematic summary of anomaly states in photovoltaic tracking brackets under complex dynamic environmental conditions, providing a reliable basis for subsequently identifying the dominant factors of anomalies and developing targeted energy consumption optimization and compensation strategies.

[0062] S5 includes acquiring drive motor current data and light intensity sequences, constructing a multi-dimensional time series set by combining energy consumption anomaly patterns, and generating an interactive data matrix based on the multi-dimensional time series set; extracting covariance eigenvalues ​​from the interactive data matrix, and determining that the difference in equipment mechanical characteristics is the dominant factor if the covariance eigenvalue is lower than the threshold; filtering out environmental fluctuation components to obtain the net energy consumption residual sequence for the state where the difference in equipment mechanical characteristics is the dominant factor; mapping the net energy consumption residual sequence to the physical topology model and locking the coordinates of mechanical components to achieve precise location of the source of energy consumption deviation.

[0063] In this embodiment, the time range corresponding to the abnormal energy consumption mode is first used as the analysis window to retrieve drive motor current data and light intensity sequence data from the cloud data storage unit. The drive motor current data is obtained from continuous acquisition of the power supply circuit of the drive mechanism at the edge, which is used to accurately reflect the load changes of the drive mechanism during operation; the light intensity sequence data is obtained from the light sensor and is used to reflect the changing trend of external light conditions over time. Both types of data are appended with a unified time stamp during the acquisition phase, so the cloud can directly and accurately align the data based on the time stamp after retrieving the data.

[0064] After data retrieval and time alignment, the cloud platform synchronously slices the drive motor current data and illumination intensity sequence using a unified time step, ensuring that each time slice contains the corresponding motor current and illumination intensity values. This time step is determined based on the dynamic response speed of the drive mechanism and the actual frequency of illumination changes. It was established during system deployment through statistical analysis of historical operating data to ensure that the time resolution reflects instantaneous changes without introducing excessive noise. Building upon this, the cloud platform combines the aforementioned time-slice data with identified energy consumption anomaly pattern tags to construct a multi-dimensional time series set. This multi-dimensional time series set is used to comprehensively describe changes in drive load, illumination conditions, and energy consumption anomalies along the same time axis.

[0065] After the multidimensional time series dataset is constructed, the cloud platform performs matrix processing on it according to preset data arrangement rules, generating an interactive data matrix. The data in each dimension of the interactive data matrix maintains strict consistency over time, allowing for direct comparison and analysis of the relationships between different variables. Through this interactive data matrix, the cloud platform can centrally characterize the temporal coordination between changes in drive motor load and changes in light intensity.

[0066] After the interactive data matrix is ​​generated, the cloud performs statistical analysis on it, focusing on extracting covariance eigenvalues ​​to measure the strength of the correlation between data in different dimensions. Specifically, the cloud comprehensively evaluates the relationship between the drive motor current sequence and the light intensity sequence within the entire analysis window, calculating their degree of coordinated change statistically to obtain the covariance eigenvalues. These covariance eigenvalues ​​quantify the coupling strength between environmental changes and mechanical responses; the lower the value, the weaker the correlation between the two.

[0067] After obtaining the covariance eigenvalue, the cloud platform compares it with a preset threshold. The threshold is determined as follows: during the initial deployment and commissioning phases of the system, the cloud platform performs long-term statistical analysis on historical data of the photovoltaic tracking bracket under normal operating conditions, extracts the stable distribution range of the covariance eigenvalue between the drive motor current and light intensity, and uses the lower limit of this range combined with a safety margin as the judgment threshold. Setting the threshold in this way ensures that when the covariance eigenvalue is below this threshold, the impact of environmental changes on energy consumption has been significantly reduced, and abnormal energy consumption is mainly caused by differences in the mechanical characteristics of the equipment itself.

[0068] When the covariance eigenvalue is determined to be below a preset threshold, the cloud determines the current operating state as one dominated by differences in equipment mechanical characteristics. In response to this state, the cloud initiates an environmental fluctuation component filtering process. Specifically, the cloud first identifies energy consumption change components in the drive motor energy consumption data that occur synchronously with changes in light intensity, based on a historical environmental fluctuation model. Then, this portion of energy consumption change is extracted from the original energy consumption sequence, thus filtering out the energy consumption components caused by environmental fluctuations. This processing yields a net energy consumption residual sequence that reflects only the equipment's own operating characteristics. This net energy consumption residual sequence is used to accurately reflect energy consumption deviations caused by factors such as increased mechanical friction, decreased transmission efficiency, or component aging after excluding environmental influences.

[0069] After obtaining the net energy consumption residual sequence, the cloud platform maps this sequence to the physical topology model of the photovoltaic tracking bracket. This physical topology model, established during the system modeling phase, describes the spatial relationships, connection structures, and power transmission paths of the various mechanical components within the photovoltaic tracking bracket. During the mapping process, the cloud platform performs matching analysis between the net energy consumption residual sequence and the energy transmission paths corresponding to each mechanical component in the physical topology model to identify locations where abnormal energy consumption is concentrated or persistently high.

[0070] By comparing the energy consumption mapping results of different mechanical components in the physical topology model, the cloud can pinpoint the coordinates of the mechanical components most relevant to changes in net energy consumption residuals, thereby achieving precise location of the source of energy consumption deviation. This precise location result not only identifies the specific mechanical component where energy consumption anomalies occur but also provides a clear physical basis for subsequent performance compensation, maintenance decisions, and operational parameter optimization. Through the above sequential steps, the cloud has achieved reliable determination and precise location of the dominant factors of energy consumption anomalies under complex environmental conditions, significantly improving the engineering practicality and analytical accuracy of the photovoltaic tracking bracket energy consumption optimization method.

[0071] S6 includes acquiring spatial coordinate sequences and equipment mechanical characteristic data generated based on precise positioning results; concatenating the spatial coordinate sequences and equipment mechanical characteristic data to construct a high-dimensional feature vector; inputting the high-dimensional feature vector into a cloud-based random forest model to obtain performance status prediction values; calculating the deviation between the performance status prediction values ​​and the standard performance curve; if the deviation value is greater than a preset threshold, determining the internal performance degradation amount and generating an internal performance degradation compensation value; and distributing the internal performance degradation compensation value to the edge side to achieve equipment optimization.

[0072] In this embodiment, a corresponding spatial coordinate sequence is first generated based on the precise positioning results. This spatial coordinate sequence originates from the physical topology model established during the system modeling phase of the photovoltaic tracking bracket. This physical topology model, based on the actual mechanical structure, provides a unified description of the spatial positional relationships of components such as the drive motor, transmission shaft, support members, and connecting nodes within the bracket. When generating the spatial coordinate sequence, the cloud platform extracts the corresponding coordinate information from the physical topology model based on the positions of the mechanical components locked in the precise positioning results, and sorts the coordinate information according to the order of the mechanical energy transfer path, thereby forming a spatial coordinate sequence that reflects the distribution of abnormal energy consumption within the physical structure.

[0073] While generating the spatial coordinate sequence, the cloud retrieves the corresponding equipment mechanical characteristic data from the data storage unit. This equipment mechanical characteristic data includes the drive motor operating current, rotational response characteristics, energy consumption trends, and state data related to mechanical friction and transmission efficiency. After retrieving the equipment mechanical characteristic data, the cloud first filters the data based on time signatures, retaining only data within abnormal time periods corresponding to the spatial coordinate sequence. Subsequently, the filtered data undergoes time alignment processing to ensure consistency between its time dimension and the operating state corresponding to the spatial coordinate sequence, thereby guaranteeing a one-to-one correspondence between spatial information and performance information in subsequent analysis.

[0074] After data preparation, the cloud platform concatenates the spatial coordinate sequence with the equipment's mechanical characteristic data to construct a high-dimensional feature vector. This concatenation process is not a simple data overlay; rather, it follows pre-defined feature organization rules, uniformly encoding spatial coordinate information as structural features and equipment mechanical characteristic data as performance features. This ensures the high-dimensional feature vector simultaneously reflects the spatial location attributes of mechanical components and their corresponding operational performance status. The dimensions and composition of the high-dimensional feature vector are determined during the system deployment phase, based on the complexity of the photovoltaic tracking bracket structure, the types of available mechanical characteristic data, and the computational capabilities of the cloud model. This ensures analytical accuracy while avoiding unnecessary data redundancy.

[0075] After the high-dimensional feature vectors are constructed, the cloud-based system inputs them into a random forest model for performance status prediction. The random forest model is a pre-trained performance status prediction model in the cloud, and its training data comes from historical data of the equipment at different operating stages, spatial locations, and performance states. During model training, through learning from a large number of samples, the random forest model can identify the intrinsic correlation between spatial coordinate features and changes in the equipment's mechanical performance. In the model inference phase, the random forest model performs a comprehensive analysis of the input high-dimensional feature vectors and outputs the corresponding performance status prediction values. These performance status prediction values ​​characterize the overall performance level of the equipment under the current spatial location and operating conditions.

[0076] After obtaining the predicted performance status value, the cloud platform compares and analyzes it against a standard performance curve. The standard performance curve is formed during the equipment's factory testing or performance calibration phase through long-term collection and processing of operating data under ideal and healthy conditions. It describes the baseline trend of equipment performance changes with operating conditions under normal conditions. During the comparison process, the cloud platform first finds the corresponding reference interval in the standard performance curve based on the current operating conditions. Then, it compares the predicted performance status value with the standard performance value within that reference interval, calculating the deviation between the two. This deviation value quantifies the degree of performance degradation of the current equipment relative to the standard state.

[0077] Subsequently, the cloud platform compares the deviation value with a preset threshold. The preset threshold is determined as follows: during system initialization and long-term operation, the cloud platform statistically analyzes historical operating data of the device under normal performance conditions, extracts the normal fluctuation range of the predicted performance value relative to the standard performance curve, and uses the maximum stable deviation within this range as the baseline value. Based on this, a safety margin is set in conjunction with the device's safe operation requirements, thus forming a threshold for judging internal performance degradation. The threshold set in this way can effectively distinguish between normal performance fluctuations and substantive performance degradation.

[0078] When the deviation value exceeds a preset threshold, the cloud determines that the current device has internal performance degradation and determines the corresponding amount of internal performance degradation based on the magnitude of the deviation. This internal performance degradation quantifies the degree of device performance decay, and its value directly reflects the extent to which the device deviates from its ideal operating state. Based on this, the cloud generates an internal performance degradation compensation value. This compensation value indicates the extent to which device operating parameters need to be adjusted to offset the performance degradation. Its generation process incorporates device operation control strategies and safety constraints to ensure the executability of the compensation result.

[0079] After generating the internal performance degradation compensation value, the cloud sends the compensation value to the edge control unit via the communication module. Upon receiving the compensation value, the edge control unit uses it as the basis for adjusting operating parameters, and makes corresponding corrections to the drive control parameters, response strategies, or operating rhythm of the photovoltaic tracking bracket. This compensates for the energy consumption deviation caused by the internal performance degradation without changing the equipment hardware structure.

[0080] S7 includes receiving internal performance degradation compensation values, generating corrected bracket tilt angle commands based on the internal performance degradation compensation values ​​to drive the photovoltaic tracking bracket, and collecting the position feedback sequence during the operation of the photovoltaic tracking bracket; combining real-time environmental data and the position feedback sequence to generate environmental fluctuation impact factors and system state fusion data; if the environmental fluctuation impact factor is less than a preset threshold, noise stripping is performed on the system state fusion data to construct an energy consumption anomaly sequence; matching the energy consumption anomaly sequence with a fault mode library, and outputting the root cause identification result of the overall energy consumption anomaly of the photovoltaic tracking bracket.

[0081] In this embodiment, after receiving the internal performance degradation compensation value from the cloud, the edge-side control unit first performs a complete analysis of the compensation value. This internal performance degradation compensation value is generated in the cloud based on performance status prediction results and is used to quantify the operational correction required due to internal performance degradation. During the analysis process, the edge-side control unit compares the compensation value with the current operating status parameters of the photovoltaic tracking bracket to confirm the target and adjustment direction of the compensation value. Subsequently, the edge-side control unit generates a corrected bracket tilt angle command based on the compensation value. This corrected bracket tilt angle command does not directly replace the original tracking command, but rather superimposes the compensation adjustment amount on the original light tracking control strategy, enabling the bracket to correct response deviations caused by internal performance degradation while tracking sunlight.

[0082] When generating the corrected tilt angle command for the photovoltaic tracking bracket, the edge-side control unit simultaneously considers the structural constraints and safe operating boundaries of the bracket, such as the maximum allowable tilt angle variation range and the maximum allowable adjustment rate, to ensure that the corrected command will not cause mechanical shock or operational risks. After the command is generated, the edge-side control unit sends the corrected tilt angle command to the drive mechanism, driving the photovoltaic tracking bracket to perform the actual tilt angle adjustment action according to the corrected control strategy.

[0083] During the tilt adjustment process, the edge-side control unit continuously collects the actual position feedback information of the support system through a position sensing device connected to the support system's pivot or support structure. This position feedback information includes the actual tilt angle value of the support system and its trajectory data over time. The edge-side control unit organizes the continuously collected position feedback information in chronological order to form a position feedback sequence. This position feedback sequence is used to accurately reflect the dynamic response characteristics of the support system during the execution of correction commands, providing fundamental data for subsequent analysis.

[0084] While acquiring the position feedback sequence, the edge control unit simultaneously obtains real-time environmental data. This real-time environmental data includes data on changes in light intensity, changes in light trajectory, and other auxiliary data reflecting environmental stability. The edge control unit performs time alignment processing on the real-time environmental data and the position feedback sequence, ensuring a strict correspondence between the two types of data on the same time axis. Subsequently, the edge control unit analyzes the real-time environmental data, extracting feature information characterizing the magnitude and frequency of environmental changes, and generates an environmental fluctuation impact factor based on this feature information. This environmental fluctuation impact factor quantifies the degree to which current environmental changes affect the operating status of the photovoltaic tracking bracket; a higher value indicates a more significant interference from environmental fluctuations on system operation.

[0085] While generating environmental fluctuation impact factors, the edge-side control unit fuses real-time environmental data with location feedback sequences to construct system status fusion data. This fusion process, based on time synchronization, unifies environmental change information with the actual operational response information of the support structure, enabling the system status fusion data to simultaneously reflect environmental conditions and equipment operating behavior. The system status fusion data describes the overall operating status of the photovoltaic tracking support structure under the current control strategy and environmental conditions.

[0086] After the system status fusion data is generated, the edge control unit compares the environmental fluctuation impact factor with a preset threshold. The preset threshold is determined as follows: during system deployment and commissioning, historical operating data of the photovoltaic tracking bracket under stable environmental conditions is statistically analyzed to extract the stable variation range of the environmental fluctuation impact factor under normal operating conditions. The upper limit of this range, combined with a safety margin, is used as the threshold. This threshold setting reliably distinguishes between states where environmental impact is negligible and states where environmental impact is significant.

[0087] When the environmental fluctuation impact factor is determined to be less than a preset threshold, the edge-side control unit considers the current environmental interference to the support operation to be at a low level, and initiates the noise stripping process. Specifically, based on historical operating data and preset filtering rules, the edge-side control unit identifies unstable components in the system state fusion data caused by sensor noise, instantaneous jitter, or communication delays, and strips these components from the system state fusion data. Through noise stripping, an energy consumption anomaly sequence is constructed. This energy consumption anomaly sequence reflects abnormal energy consumption changes that still exist during the operation of the photovoltaic tracking support under conditions of minimal environmental impact, thereby avoiding interference from environmental factors in the anomaly judgment results.

[0088] After obtaining the energy consumption anomaly sequence, the edge control unit or cloud management platform performs pattern matching processing on the sequence. Specifically, the energy consumption anomaly sequence is compared one by one with a pre-built fault mode library. This fault mode library is formed through the collation and summarization of known energy consumption anomaly cases during long-term system operation and maintenance; each fault mode corresponds to a typical energy consumption change characteristic. By comparing the similarity between the current energy consumption anomaly sequence and the modes in the fault mode library, the system determines the fault mode that best matches the current anomaly and outputs the root cause identification result of the overall energy consumption anomaly of the photovoltaic tracking bracket.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the energy consumption of photovoltaic tracking brackets based on cloud-edge fusion, characterized in that, include: S1. Collect real-time external environmental variable data and equipment mechanical characteristic data of the photovoltaic tracking bracket at the edge and upload the data to the cloud; the external environmental variable data includes data on changes in the light trajectory, and the equipment mechanical characteristic data includes energy consumption and operating status data related to the drive mechanism of the photovoltaic tracking bracket; S2. The collected external environmental variable data and equipment mechanical characteristic data are processed in the cloud, and the random forest algorithm is used to analyze the data to obtain preliminary classification results of the sources of energy consumption deviation. S3. Based on the preliminary classification results, establish the correspondence between the light trajectory fluctuation data and the differences in the mechanical characteristics of the equipment in the cloud, and determine whether the light trajectory fluctuation exceeds the preset threshold. When the preset threshold is exceeded, the characteristics of the impact of environmental fluctuations are extracted to determine potential indicators of equipment performance degradation. S4. In the cloud, the support vector machine algorithm is used to classify the characteristics of environmental fluctuations and potential indicators of internal performance degradation, so as to obtain the energy consumption anomaly pattern under dynamic environmental change conditions. S5. Based on the energy consumption anomaly pattern, analyze the interaction data between the differences in equipment mechanical characteristics and the fluctuations in the light trajectory, and determine whether the interaction data indicates that the differences in equipment mechanical characteristics are the dominant factor; when the differences in equipment mechanical characteristics are the dominant factor, separate the influence of environmental fluctuations through information processing, and accurately locate the source of energy consumption deviation.

2. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 1, characterized in that: S1 includes: Receive real-time data on changes in the light trajectory and instantaneous voltage and current signals from the drive mechanism collected by the photovoltaic tracking bracket; Real-time energy consumption values ​​are calculated based on instantaneous voltage and current signals, and the real-time energy consumption values ​​are correlated with real-time illumination trajectory change data to obtain an environmental mechanical coupling data stream; If the rate of change of illumination angle in the environmental mechanical coupling data stream exceeds a preset threshold, the dynamic mechanical characteristic sample is obtained by extracting the operating status data of the drive mechanism. The dynamic mechanical characteristic samples and real-time light trajectory change data are encapsulated to generate edge-side reporting data packets, which are then sent to the cloud management platform to realize remote digital mapping of multi-dimensional data on the photovoltaic tracking bracket site.

3. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 1, characterized in that: S2 includes: Acquire external environmental variable data and equipment mechanical characteristic data to generate a standardized time series dataset; A multidimensional feature vector matrix is ​​constructed based on a standardized time series dataset, and then input into a random forest model. The feature splitting gain is calculated using a random forest model, and the energy consumption deviation feature weight ranking is determined based on the feature splitting gain. The energy consumption deviation feature weights are sorted and mapped to the fault category label library, and the preliminary classification results of the energy consumption deviation source are determined based on the fault category label library.

4. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 1, characterized in that: S3 includes: Obtain preliminary classification results and light trajectory fluctuation data from cloud storage, construct a nonlinear correspondence model between light trajectory fluctuation data and differences in equipment mechanical characteristics, and output the numerical values ​​of mechanical characteristic differences. The instantaneous amplitude is calculated by combining the numerical difference in mechanical characteristics with the fluctuation data of the illumination trajectory. If the instantaneous amplitude is greater than the preset threshold, a fluctuation over-limit signal is generated. In response to the signal of excessive fluctuation, the feature vector of the impact of environmental fluctuation is extracted, and the feature vector of the impact of environmental fluctuation is analyzed to determine the potential indicators of the degradation of the internal performance of the equipment.

5. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 1, characterized in that: S4 includes: Acquire historical environmental fluctuation data and device internal performance data from cloud storage, and obtain the feature vector of environmental fluctuation impact and the set of potential indicators of internal performance degradation through time-series alignment; The dynamic environmental change conditions are generated by mapping the feature vectors affected by environmental fluctuations to a set of potential indicators of internal performance degradation. The support vector machine algorithm is used to perform kernel function mapping on dynamic environmental change conditions to calculate the classification decision function value, and the energy consumption anomaly label is determined based on the classification decision function value. Data with the same energy consumption anomaly label are aggregated to obtain energy consumption anomaly patterns under dynamic environmental change conditions.

6. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 1, characterized in that: S5 includes: Acquire drive motor current data and light intensity sequence, combine energy consumption anomaly patterns to construct a multi-dimensional time series set, and generate an interactive data matrix based on the multi-dimensional time series set; Extract the covariance eigenvalues ​​from the interactive data matrix. If the covariance eigenvalues ​​are below a threshold, the difference in equipment mechanical characteristics is determined to be the dominant factor. For situations where differences in equipment mechanical characteristics are the dominant factor, environmental fluctuation components are filtered out to obtain the net energy consumption residual sequence; The net energy consumption residual sequence is mapped to the physical topology model, and the coordinates of mechanical components are locked to achieve precise location of the source of energy consumption deviation.

7. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 1, characterized in that, It also includes S6, which, based on the precise positioning results, uses a random forest algorithm in the cloud to optimize the mechanical characteristic data of the equipment, obtains an internal performance degradation compensation value, and sends the compensation value to the edge side, specifically including: Obtain spatial coordinate sequences and equipment mechanical characteristic data generated based on precise positioning results, and concatenate the spatial coordinate sequences and equipment mechanical characteristic data to construct a high-dimensional feature vector; High-dimensional feature vectors are input into a cloud-based random forest model to obtain performance status predictions.

8. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 7, characterized in that: S6 further includes: Calculate the deviation between the predicted performance status value and the standard performance curve. If the deviation value is greater than a preset threshold, determine the internal performance degradation and generate an internal performance degradation compensation value. Internal performance degradation compensation values ​​are sent to the edge side to achieve device optimization.

9. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 7, characterized in that, It also includes S7, which adjusts the operating parameters of the photovoltaic tracking bracket on the edge side based on the internal performance degradation compensation value, obtains fusion data of the influence of light trajectory fluctuation and environmental fluctuation under dynamic environmental change conditions, and outputs the root cause identification result of the overall energy consumption anomaly of the photovoltaic tracking bracket when the influence of environmental fluctuation is less than a preset threshold, specifically including: Receive internal performance degradation compensation value, generate a corrected bracket tilt angle command based on the internal performance degradation compensation value to drive the photovoltaic tracking bracket, and collect the position feedback sequence of the photovoltaic tracking bracket during operation. By combining real-time environmental data with location feedback sequences, environmental fluctuation impact factors and system state fusion data are generated.

10. The energy consumption optimization method for photovoltaic tracking brackets based on cloud-edge fusion according to claim 9, characterized in that: The S7 also includes: If the environmental fluctuation impact factor is less than the preset threshold, noise stripping is performed on the system state fusion data to construct an energy consumption anomaly sequence; Based on the fault mode library matched with the energy consumption anomaly sequence, the root cause identification results of the overall energy consumption anomaly of the photovoltaic tracking bracket are output.

Images