An automatic operation and maintenance control method for a photovoltaic power station drawer type alternating current power distribution cabinet

Abstract

The application provides an automatic operation and maintenance control method for a photovoltaic power station drawer type AC power distribution cabinet, relates to the technical field of automatic control of electric power equipment, and has the beneficial effects that: (1) the standardized coordinate library and the automatic action sequence are constructed through an algorithm, replacing manual operation processes of manual operation of personnel entering a relay protection room; and (2) a multilevel algorithm interlocking mechanism is used, replacing a single protection mode that only relies on personnel to follow safety specifications.

Description

Technical Field

[0001] This invention relates to the field of power equipment automation control technology, and in particular to an automated operation and maintenance control method for a drawer-type AC distribution cabinet in a photovoltaic power station. Background Technology

[0002] In the AC power system of a photovoltaic power station, the drawer-type AC distribution cabinet is the core control unit, integrating all system control switches. Its stable operation directly determines the power station's power generation efficiency and safety. According to operation and maintenance specifications, AC distribution cabinets need to undergo regular operations such as switch activation / deactivation, tripping maintenance, insulation testing, and terminal tightening. The maintenance process must strictly adhere to safety procedures such as power outage testing, sectional isolation, and insulation protection. With the expansion of photovoltaic power station scale and the increase in distributed energy access, the number of distribution cabinets has surged, and the operation and maintenance environment has become more complex. The traditional manual operation and maintenance model can no longer meet the demands for efficiency and safety. Current operation and maintenance requires manual operation in the relay protection room, equipped with tools such as infrared thermometers and insulation resistance testers, and manual recording of equipment status. However, photovoltaic power stations generally face problems such as a shortage of professional operation and maintenance personnel and significant on-site management difficulties.

[0003] However, the current operation and maintenance of AC distribution cabinets mainly relies on manual labor. Maintenance requires manual power outages, installation of isolation devices in sections, operation with insulated equipment, and monitoring of capacitor discharge by a dedicated person. These processes lack standardized positioning and action logic, and a single operation takes far more than 30 seconds. Furthermore, manual judgment of drawer positions based on experience is prone to errors due to positioning deviations. Simply following safety regulations to avoid risks lacks real-time risk assessment algorithms, making it impossible to identify potential electric shock hazards in short-circuit / grounding scenarios in advance. There are also no operational interlocking mechanisms. If human fatigue or misjudgment occurs, it can easily lead to equipment damage or personal injury. Although there are technologies for monitoring the status of photovoltaic equipment and predicting the insulation performance of combiner boxes, allowing for maintenance to be scheduled 15 days in advance, these do not cover the internal temperature and operating conditions of AC distribution cabinets. The current distribution cabinet requires manual on-site inspection of indicator lights and temperature measurement by touch. It lacks real-time data processing and anomaly prediction algorithms, resulting in delayed fault detection and inability to prevent potential faults such as overheating and discharge. During inspections, the terminal status and fuse status are manually recorded, but the "on / off" status of the drawers lacks an algorithmic recording and verification mechanism. Manual records are prone to omissions and tampering, leading to incomplete traceability of drawer on / off status. Subsequent maintenance is prone to incorrect on / off states, and it is impossible to trace historical status. All of these factors contribute to the current situation of low maintenance efficiency, high safety risks, and weak fault prevention capabilities of AC distribution cabinets in photovoltaic power plants. Therefore, an integrated technical solution is needed to achieve standardization of maintenance processes, real-time monitoring, traceability of records, and safe interlocking of operations. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, the technical problem to be solved by this invention is to provide an automated operation and maintenance control method for drawer-type AC distribution cabinets in photovoltaic power plants. This method integrates intelligent multi-axis collaborative control, modular safety machinery, remote intelligent inspection, and a dedicated safety system. Through coordinate encapsulation algorithms, insulating gripper design, multi-sensor fusion, and 4G communication, it achieves automated operation and deactivation of the drawer cabinet, real-time status monitoring, fault early warning, and safe operation, thereby solving problems such as low efficiency of manual operation, significant safety hazards, and lack of status monitoring.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: The present invention provides an automated operation and maintenance control method for a drawer-type AC distribution cabinet in a photovoltaic power station, comprising the following steps: S1. Obtain the drawer module parameters of the target drawer-type AC distribution cabinet, construct a three-dimensional coordinate model of the distribution cabinet through coordinate mapping and error correction algorithms, convert the drawer module parameters into digital coordinate data that the control system can recognize, and encapsulate the digital coordinate data into a standardized coordinate library. ; S2. Based on the standardized coordinate library Generate a preset inspection path, control the sensor module to move along the path, collect status data of the power distribution cabinet's operating status, and form a status dataset. S3. Compare the preset parameters with the status dataset, generate the comparison result, determine the status abnormality based on the comparison result, generate the abnormality alarm information based on the status abnormality, report the abnormality alarm information to the remote control center, and receive the operation instructions fed back by the remote control center. S4. Perform security verification on abnormal alarm information and status data according to the operation instructions, obtain the verification results, and output "Allow operation" or "Lock operation" instructions according to the verification results. S5. If the "Allow Operation" command is output, then according to the standardized coordinate library... Calculate the motion parameters of the actuator, generate the action sequence and execute the drawer opening and closing operation. During the operation, monitor the deviation in real time and trigger the safety control to generate the operation execution result. S6. Record the final status of the drawer module parameters based on the operation execution results, update the historical records and synchronize them to the remote control center to form a traceable operation and maintenance ledger.

[0006] In the preferred embodiment, the specific steps of step S1 are as follows: S11. Let the total number of drawer modules be... , No. The physical locations of the parameters for each drawer module are as follows: Horizontal physical coordinates Vertical physical coordinates Depth direction physical coordinates Physical threshold for switch rotation angle ; S12. Convert the physical parameters into digital coordinates using a mapping formula, the formula of which is: ; ; ; ; in, X-axis transformation factor Y-axis transformation factor Z-axis transformation factor This is the rotation angle conversion factor; For the first The horizontal digital coordinates of each drawer module; For the first The vertical digital coordinates of each drawer module; For the first Digital coordinates of the depth direction of each drawer module; For the first Digital threshold for the opening and closing rotation angle of each drawer module; , , , The first The error correction values ​​for each drawer module in the X, Y, Z axes and rotation angle directions were calculated using the least squares method after multiple calibration experiments. The formula is as follows: ; in, For the first The measured value of the digital X-axis coordinates after the second calibration. For the number of calibrations ( ); S13. Verify the positioning accuracy corresponding to the digitized coordinates using the accuracy verification formula, which is: ; in, , The shaft accuracy verification formula is similar; if it meets the requirements, then... Encapsulated as a standardized coordinate library ;in, For the first A standardized set of coordinates for each drawer module.

[0007] In the preferred embodiment, the specific steps of step S2 are as follows: S21, Let the first The historical operation frequency of each drawer module is Then its weight ,in, For the first The operation frequency weight of each drawer module is assigned, with higher weights indicating higher priority. S22, from the standardized coordinate library Extract the vertical physical coordinates of all modules. The system is divided into K regions based on numerical sorting. The formula for dividing the module set in the k-th region is as follows: ; Wherein, the vertical boundary of the k-th region is , ( =), S23, Calculate the region weight ,in, The weight of the k-th region is calculated as follows: Traverse the regions from largest to smallest, and within each region, based on... Horizontal physical coordinates in Sort to generate subpaths Overall path ,in, This is a sub-path of the k-th region; S24. Acquire the HSV value of the drawer switch indicator image using the HSV color space feature extraction algorithm. ,in, For hue, For saturation, For brightness, the HSV threshold range of the red indicator light is defined as follows: The HSV threshold range of the green indicator light is By comparing the HSV threshold ranges of the red and green indicator lights, the color status is output: ;in, For the first The indicator light status of each module: 1 = red, 0 = green; S25. Collect the temperature inside the cabinet via an infrared temperature measurement module. ,in, For the first Each module in Temperature at any moment; For collection timestamps; The sliding window algorithm is used to calculate the rate of temperature change, with a window size of [missing value]. The formula is: ,in, For the first Each module in The rate of temperature change over time; like The inspection interval of this module will be dynamically adjusted from the default interval. shortened to ; in, For the first The default inspection interval for each module; For the first The inspection interval after adjustment of each module; The threshold for the rate of temperature change is set to 2℃ / s; S26, Collection of the first The closed status signal of each drawer ,in, For the first The closed state of each drawer is represented by 1 = closed and 0 = open, forming a state dataset. ,in, It is a multi-source state dataset.

[0008] In the preferred embodiment, step S3 consists of the following steps: S31. Define the drawer target position signal ;in, For the first The target position signal for each drawer is 1 = should be closed, 0 = should be pulled out. Anomalies are determined using the position deviation formula. ,in, For the first The positional deviation of each drawer, if Then, the anomaly type is further determined by the hot standby / cold standby status verification formula: Hot standby status verification: If the operation command is "close", ;in, For the first The operation commands for each drawer are 1 = closed, 0 = withdrawn. Therefore, the hot standby anomaly detection formula is: ,in, For the first The hot standby abnormality judgment result for each drawer is 1 = abnormal, 0 = normal; Cold standby status verification: If the operation command is "extract", The formula for determining cold standby anomalies is: ;in, For the first The cold standby abnormality judgment result of each drawer: 1 = abnormal, 0 = normal; like or Then, an abnormal location alarm message will be generated. ;in, For the first Alarm message indicating abnormal position of drawer; S32, Set temperature safety threshold ,in, The safe temperature threshold is set at 40℃. Anomalies are determined using the temperature deviation formula. ,in, For the first The temperature deviation value of each module, if Then a temperature anomaly alarm message will be generated. ;in, For the first Temperature anomaly alarm information for each module; S33. Compare the indicator light status using the status similarity formula: ,in, For the first The indicator light status deviation value of each module, if Then an indicator light malfunction alarm message will be generated. ;in, For the first Alarm information for abnormal indicator lights in each module; S34. Define data priority weights and abnormal alarm signal weights. Weight of status record data Real-time video data weights Bandwidth is allocated through a data transmission scheduling formula: ; in, Allocate bandwidth for the corresponding data type; This refers to the total bandwidth of the 4G communication module. Weights for the corresponding data types; Collection of abnormal alarm information The data is transmitted to the remote control center according to the bandwidth allocation strategy described above. The remote control center then processes the abnormal alarm information based on the set of data. Generate operation instructions And feedback, among which, It is a set of operation instructions.

[0009] In the preferred embodiment, step S4 consists of the following steps: S41, Collection of the first Short circuit / ground signal strength of each drawer ,in, For the first Short-circuit / ground signal strength for each drawer; setting signal thresholds. The following is determined through the verification formula: ; in, For the first The short circuit / grounding test result for each drawer is 1 = qualified, 0 = unqualified; S42. Call the standardized coordinate library The Middle Safety travel boundaries of each module , , ,in, , For the first Minimum / maximum safe travel of each module's X-axis. For the first Minimum / maximum safe travel of each module's Y-axis. For the first Minimum / maximum Z-axis safe travel of each module; Calculate the current operation point Distance from the boundary: ; in, , , For the first The coordinates of the current operation point of each module; For the first The minimum distance between the current point on the X-axis of each module and the boundary. , The calculation is similar; Define safety margin The verification formula is used to determine: ,in, For the first The safety travel verification result for each module is 1 = qualified, 0 = unqualified. S43, Insulation resistance of the acquisition actuator Define insulation threshold The verification formula is used to determine: ; in, For the first The insulation status verification result of each module is 1 = qualified and 0 = unqualified. S44. Generate a definite result using logical AND formulas: ,in, For the first The overall security assessment results of each module; like Then output the "Allow operation" command. ; like Then the "lock operation" command will be output. The cause of the risk was reported via the 4G module. ; For the first Risk reasons information for each module.

[0010] In the preferred embodiment, step S5 is as follows: S51. According to the standardized coordinate library Extract the first The target operation coordinates of each module are: Calculate the displacement: ; Current coordinates are ;in, For the first The module performs X-axis displacement. , The calculation is similar; S52. Let the basic step angle of the stepper motor be... Subdivision multiple is Then the actual step angle ; Calculate the required number of steps using the step size formula: ;in, For the first Number of steps required for the X-axis stepper motor of each module; Let be the radius of the motor drive mechanism. , The calculation is similar; S53. Generate an action sequence based on the drawer switch type using a sequence generation formula: Rotary switch: ; in, For the first The action sequence of each module; Push-pull switch: ; in, Add displacement to push and pull; S54, Level 1 Threshold (Emergency Stop Trigger): Based on the generated action sequence in step S52, monitor operational deviations in real time: ;in, For the first Operational deviation values ​​for each module; , For the first The real-time measured X-axis coordinates of each module, if This will trigger an emergency stop command. Cut off operation command; Level 2 Threshold (Travel Protection): According to the safety travel verification result of step S42, if Then the motor speed is adjusted using the deceleration formula: ; in, For the first The motor speed after adjusting each module; This is the motor's default speed. , The same principle applies to shaft speed adjustment; Level 3 Threshold (Electric Shock Risk Prediction): According to the short-circuit / grounding verification result of step S41, if Then a locking command is triggered. Action sequences are prohibited from execution.

[0011] S55. Generate an execution result set based on the execution result of step S54: .

[0012] In the preferred embodiment, step S6 is as follows: S61. Based on the execution result set E , and Completion state, encoding the final state : ; S62. Extract from the execution result set E , , ,from Extract module identifier j and target coordinates. ,generate ; S63, Update the formula using historical records Write to the local database and synchronize to the remote center: ,in, For the new historical record set, the ledger order is maintained using a timestamp sorting formula: ,in, This is a sorted historical record set; ensuring that the operation and maintenance status is traceable and preventing accidental insertion of drawers.

[0013] In the preferred embodiment, in step S12 The range of values ​​is System errors are reduced through multiple calibrations; the control unit uses an Arduino Mega 2560 chip, and the digital conversion coefficients are... , , The value can be: ; ; ; in, , , These represent the maximum physical travel of the distribution cabinet on the X, Y, and Z axes, respectively.

[0014] In the preferred embodiment, the HSV threshold range of the OpenMV machine vision module in step S24 is determined by the following algorithm: S241. Collect 100 sets of sample images of red indicator lights and 100 sets of sample images of green indicator lights, and extract the HSV value of each sample. ( This refers to the HSV value of a single sample. Representative sample; S242. Calculate the mean H channel value of red samples using a clustering algorithm. Standard deviation The threshold for the red H channel is then... The threshold calculations for the S and V channels are similar, with the mean value of the red S channel being... The standard deviation is The mean value of the red V channel is The standard deviation is ; S243. The HSV threshold for green samples was determined using the same clustering algorithm, and the mean value of the green H channel was [value missing]. The standard deviation is The average value of the green S channel is The standard deviation is The mean value of the green V channel is The standard deviation is To ensure color recognition accuracy .

[0015] In the preferred embodiment, the stepper motor microstepping factor in step S52 Dynamically adjust using the following algorithm: S521, if the displacement This indicates a large displacement. For the first If a module performs X-axis displacement, then set the subdivision factor. Ensure movement speed; S522, if If it is a small displacement, then set the subdivision factor. Using the step angle formula ,in This refers to the actual step angle of the stepper motor. This is the basic step angle.

[0016] This invention provides an automated operation and maintenance control method for a drawer-type AC distribution cabinet in a photovoltaic power station. Through the coordination of the above-mentioned structures, compared with existing methods, it has the following advantages: First, by constructing a standardized coordinate library and automated action sequences through algorithms, the manual operation process of entering the relay protection room is replaced. This eliminates the need for human experience to determine drawer positions and perform operations such as switch activation / deactivation and tripping maintenance, fundamentally reducing manual intervention and avoiding time-consuming repetitive operations caused by human experience bias. Second, by integrating a multi-level algorithmic interlocking mechanism for short-circuit / grounding prediction, over-limit travel protection, and insulation status verification, the single protection method relying solely on manual adherence to safety regulations is replaced. This allows for early identification of electric shock risks and operational hazards. Furthermore, the algorithmic interlocking of operating permissions and equipment status prevents equipment damage and personal injury accidents caused by human fatigue or misjudgment. Third, through real-time acquisition and algorithmic analysis of multi-source sensor data, the lagging monitoring methods of manually checking indicator lights and contact temperature measurement are transformed into real-time perception and anomaly prediction of cabinet temperature and switch status. This enables timely detection of potential fault signs such as overheating and discharge, avoiding maintenance delays due to untimely fault detection and ensuring the continuous and stable operation of the distribution cabinet. Attached Figure Description

[0017] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a main view structural diagram of the process of this invention; Figure 2 This is a graph showing the training accuracy of the model of this invention; Figure 3 This is a diagram illustrating the effect of coordinate error correction in this invention; Figure 4 This is a parameter rendering of the stepper motor of the present invention; Figure 5 This is a diagram illustrating the temperature compensation effect of the present invention; Figure 6 This is a graph showing the relationship between dust concentration and recognition rate in this invention; Figure 7 This is a comparison chart of the stuttering rate of the actuator of this invention; Figure 8 This is the high humidity insulation monitoring diagram of the present invention; Figure 9 This is a graph showing the relationship between humidity and insulation resistance in this invention. Detailed Implementation

[0018] To better understand the purpose, technical solution, and function of this invention, the embodiments and technical features in these embodiments can be combined with each other without conflict. The exemplary embodiments of this invention will be described below with reference to the accompanying drawings, including specific technical details of the embodiments to aid understanding. These details should be considered illustrative only. Therefore, those skilled in the art should recognize that various improvements and adjustments can be made to the embodiments described herein without departing from the scope of protection and core concept of this invention. Similarly, for clarity and conciseness, detailed descriptions of well-known photovoltaic power station distribution cabinet structures, conventional manual operation and maintenance procedures (such as power outage testing procedures), and the working principles of general sensors are omitted in the following description.

[0019] In the field of photovoltaic power plant operation and maintenance technology, the drawer-type AC distribution cabinet, as the core control unit of the AC power system, directly determines the power generation efficiency and operational safety of the power plant through its operation and maintenance quality. As photovoltaic power plants expand to larger scales, higher altitudes, and more windy and sandy environments, the traditional manual operation and maintenance model is no longer sufficient to meet the demands. Manual intervention is required in the relay protection room to perform switch activation / deactivation and tripping maintenance, using tools such as infrared thermometers and insulation resistance testers to collect status data and manually record operation and maintenance data. To improve efficiency and safety, the industry is gradually exploring partial automation technologies, such as combiner box insulation performance prediction models and semi-automatic operation devices for distribution cabinets, to reduce manual intervention and lower the intensity of operation and maintenance.

[0020] With the development of intelligent photovoltaic operation and maintenance technology, automated operation and maintenance is gradually evolving from "partial function replacement" to "full-process closed-loop control," requiring the integration of drawer cabinet status perception, automatic operation, safety protection, and remote monitoring. An ideal intelligent operation and maintenance solution should be able to integrate multi-dimensional data through algorithm-driven approaches to achieve precise location operation, real-time risk prediction, and operation and maintenance status traceability. However, in existing technologies, most related automation solutions are limited to single functions or only achieve temperature monitoring, or only support semi-automatic operation. They have not formed a full-process algorithmic closed loop of "coordinate positioning - data acquisition - anomaly judgment - safety execution - status recording," and lack adaptation design for the modular structure of drawer-type distribution cabinets.

[0021] The relevant technologies have significant limitations in achieving automated operation and maintenance of distribution cabinets: Firstly, in terms of operation and control, they rely heavily on the physical positioning of mechanical structures, failing to achieve ±0.5mm-level precise positioning through a standardized coordinate library built using algorithms, making them susceptible to operational deviations due to environmental changes. Secondly, in terms of safety protection, basic protection is achieved only through hardware emergency stop buttons, lacking a multi-level algorithmic interlocking mechanism of "short circuit / grounding prediction - over-travel protection - insulation status verification." Furthermore, in terms of status management, maintenance records are mostly manually entered, lacking algorithmic status verification and traceability logic, making them prone to record omissions or incorrect operation risks. Therefore, these technologies are ill-suited to the "full-process intelligent, high-safety, and strong environmental adaptability" operation and maintenance requirements of photovoltaic power plant drawer-type AC distribution cabinets, and cannot completely replace manual labor for closed-loop management.

[0022] To make the purpose, technical solution, and effects of this solution clearer, the following five examples, combined with typical application scenarios of substations in Qinghai Province (high altitude and low temperature, high dust in the Gobi Desert, and high temperature and high humidity in the eastern region), illustrate the specific implementation process of this solution.

[0023] Example 1 like Figures 1-3 As shown, this embodiment provides a "pre-trained model that can be directly called," focusing on training coordinate mapping and error correction, OpenMV color recognition, and temperature anomaly prediction to ensure that the model's accuracy and stability meet actual operation and maintenance needs. Figure 1 As shown, the training process strictly follows the machine learning workflow of "data acquisition → preprocessing → model building → validation". The output of each step serves as the input for the next step. The specific steps are as follows: S1. Coordinate Error Correction Model Training: The coordinate error correction model is the foundation for achieving "precise positioning" of the drawer module. If there is a deviation between the physical coordinates and the digital coordinates, it will directly lead to the deviation of the subsequent inspection path and the operation of the actuator exceeding the tolerance. Therefore, it is necessary to establish a high-precision mapping relationship through training first.

[0024] S11. Training Data Acquisition: A commonly used 10kV drawer-type AC distribution cabinet, model GGD212, was selected from a photovoltaic power station in Qinghai. This model has a penetration rate of over 60% in photovoltaic power stations in Qinghai and is representative. Ten typical drawer modules, numbered J1J10, were extracted from the cabinet, covering different current ratings: 100A, 200A, and 400A, to ensure data universality. Eight calibration points were selected for each module, specifically: the upper left edge, upper right edge, lower left edge, and lower right edge (four edge points) to ensure overall module position calibration, the center of the switch operating handle, the drawer closing limit point, the drawer opening limit point, and the center of the grounding bolt inside the cabinet.

[0025] Data acquisition is conducted simultaneously in two dimensions: physical coordinate measurement and digital coordinate reading. A high-precision laser rangefinder is used to collect the physical coordinates of each calibration point. , , , Module number, For the calibration point numbering, a total of [number] samples were collected. The system collects physical coordinate data; simultaneously, it controls the actuators via Arduino Mega2560 to move to each calibration point and reads the measured digital coordinate values ​​output by the control system. , , To reduce random errors, data was collected five times at each calibration point, ultimately yielding... Group of digital coordinate data.

[0026] S12. Data Preprocessing: After data acquisition, outliers need to be removed. On-site issues such as laser rangefinder glare and momentary actuator malfunctions may cause some data to deviate from the normal range. The 3σ criterion, a commonly used statistical outlier removal method, is employed to effectively filter out random errors. The mean of each data set is calculated. with standard deviation Remove Six sets of invalid data were removed, mainly due to glare interference from edge points of the J6 module. The remaining 394 sets of valid data were grouped by module, and the mean physical coordinates of each group were calculated. , average of digital coordinates This reduces the impact of random errors on model training.

[0027] S13. The model objective is to establish an unbiased mapping between "physical coordinates and digital coordinates," therefore a linear regression model is constructed, which conforms to the mapping formula. ,in For conversion factors, This is the error correction value, used to compensate for system errors such as module installation deviations and actuator clearances.

[0028] The model parameters are trained using gradient descent: the loss function is set to mean squared error. The number of training iterations was set to 1000. Preliminary experiments verified that the loss function converged to a stable value after 1000 iterations. The learning rate was set to 0.001 to balance the convergence speed and avoid oscillations.

[0029] Key parameters obtained after training: X-axis conversion factor Due to the maximum physical travel of the X-axis of the power distribution cabinet The GGD212 cabinet has standard dimensions. The Arduino Mega2560's I / O ports have a 10-bit resolution and a maximum digitization value of 1023, calculated according to the "full-scale mapping" principle. Error correction values ​​for each module —such as module J1 (Due to the proximity of module J1 to the cabinet's cooling fan, there was a slight misalignment during installation.) Module J2 (The installation position is centered with minimal deviation), and the difference in correction values ​​reflects the actual installation status of different modules.

[0030] S14. Model Validation: To ensure the model's generalization ability, three untrained drawer modules, J11 and J13, with current ratings of 100A, 200A, and 400A respectively, were selected, covering the same specifications as the training modules. 24 sets of validation data were collected, with 8 calibration points for each module, collected once. The data were then substituted into the trained model to calculate digital coordinates, and verified using the accuracy formula. The positioning accuracy was verified. The results showed that the accuracy of all verification points was ≤0.3mm, which is better than the specified positioning accuracy requirement of ±0.5mm, proving that the model can be used as a positioning reference for subsequent automated operations.

[0031] In this embodiment, the color status of the drawer switch indicator light (red = standby / fault, green = running) is the core basis for judging the status of the power distribution cabinet. If the recognition is incorrect, it will lead to abnormal misjudgment or missed judgment. Therefore, it is necessary to train a highly robust color recognition model to adapt to the complex lighting environment of the Qinghai substation, where natural light and LED lighting inside the cabinet are superimposed, resulting in large fluctuations in illuminance.

[0032] In this embodiment, sample collection was conducted at a substation in Qinghai Province, simulating actual lighting conditions during operation and maintenance: 9:00-11:00 AM (primarily natural light, illuminance 100-300 lux), 2:00-4:00 PM (direct sunlight, illuminance 300-500 lux), and nighttime (LED lighting inside the cabinet, illuminance 100-200 lux). The samples were collected from drawer switch indicator lights, specifically red LEDs with wavelengths of 620-630 nm and green LEDs with wavelengths of 520-530 nm, both industry standard models. 1000 samples were collected for each color, covering different angles (0°~45°, simulating different sensor shooting angles during inspection) and different dust coverage levels: light (dust coverage <10%), moderate (10-30%), and heavy (30-50%), simulating sensor contamination after long-term operation and maintenance. The sample resolution was set to 640×480 pixels (the default resolution of OpenMV4H7, balancing recognition accuracy and data processing speed).

[0033] In this embodiment, the HSV color space is used to extract features. The HSV space is more robust to changes in light intensity than the RGB space, thus reducing the impact of ambient light fluctuations. For each sample image, the HSV value of each pixel is extracted using the OpenMV image processing library. ,in, Hue determines the color category; Saturation determines the vibrancy of a color; To determine the brightness, manually label the value: red = 1, green = 0.

[0034] The color threshold range is determined using the K-means clustering algorithm (an unsupervised learning algorithm that can automatically identify color cluster centers). Red indicator light: , (The red hue is concentrated around 0°); , (Excluding low-saturation gray interference); , (Excluding low-brightness dark-state interference); Green indicator light: , (The green hue is concentrated between 60° and 80°); , ; , ; This threshold range conforms to the HSV threshold setting, effectively distinguishing between red and green colors while filtering out light and pollution interference.

[0035] In this embodiment, a classification model is constructed using Support Vector Machine (SVM). SVM exhibits excellent classification performance on small sample sizes and high-dimensional data, making it suitable for indicator light color recognition. The feature dimension is HSV (Hypervector) 3D, and the sample size is 2000 sets. 80% of the samples (1600 sets) are used as the training set, and 20% of the samples (400 sets) are used as the test set. A linear kernel is selected as the training kernel function (HSV features and color labels are linearly separable, and linear kernels are computationally fast).

[0036] The trained model achieved a recognition accuracy of 99.2% on the test set, meeting the requirement of ≥99% accuracy. The three misidentified samples were all red indicator lights heavily covered in dust (dust caused the saturation to drop to 35%, below the set threshold). The subsequent optimization can be achieved through the "Inspection and Cleaning Reminder" function in step S2. When the recognition accuracy is below 95% for three consecutive times, a sensor lens cleaning reminder will be triggered.

[0037] In this embodiment, abnormal internal temperature (such as overheating or discharge) is the main cause of distribution cabinet failures in photovoltaic power plants. It is necessary to predict these failures in advance using a model to prevent them from escalating. The model is trained based on historical temperature data from the Qinghai substation to ensure it is adapted to the local climate and equipment operating characteristics.

[0038] Temperature data inside a photovoltaic power station in Qinghai Province was collected over one year (from January 2023 to December 2023). The data collection frequency was once per second (to ensure the capture of instantaneous temperature changes). A total of [data missing] were obtained. The data was divided into sets. In addition, abnormal samples were marked in conjunction with the operation and maintenance records: when the temperature was >40℃ (the upper limit of the normal operating temperature of the distribution cabinet) or the temperature rose by >2℃ within 1 minute (sudden temperature rise is usually accompanied by faults), it was marked as an abnormal sample, totaling 1200 sets (mainly concentrated in the high temperature period in summer and the peak load period in winter); the rest of the data was marked as normal samples.

[0039] In this embodiment, a combined model of "sliding window and logistic regression" is adopted: the sliding window is used to calculate the rate of temperature change, reflect the temperature trend, and avoid misjudgment based on a single temperature value; the logistic regression is used to output the probability of anomalies, a binary classification task, which is suitable for fault prediction.

[0040] Sliding window size set This refers to a 1-minute window, a commonly used temperature trend analysis window in the industry, which balances real-time performance and stability. The formula for calculating the rate of temperature change is... The numerator is the temperature difference before and after the window (reflecting the magnitude of temperature change), and the denominator is the average temperature within the window, avoiding deviations in the rate of change under different reference temperatures.

[0041] Model input is (rate of temperature change) and (Current temperature), output as the probability of an anomaly. ( If the result is deemed abnormal (P < 0.8 indicates normal), the weight parameters of the logistic regression are optimized using gradient descent, with cross-entropy loss selected as the loss function.

[0042] In this embodiment, 100,000 sets of new data from January 2024 were selected as the validation set. The model's accuracy in predicting temperature anomalies was 98.5%, with 1,164 out of 1,182 abnormal samples correctly identified. The average warning lead time was 8.2 seconds—meaning the model could issue a warning 8.2 seconds before the temperature reached the anomaly threshold, providing sufficient emergency processing time for the remote control center. This effectively supports the temperature anomaly judgment process and prevents equipment damage caused by temperature anomalies.

[0043] Example 2 like Figure 4 , 5As shown, a high-altitude, low-temperature scenario was selected in Yushu, Qinghai Province, specifically at an altitude of 3500m, a temperature of 15℃~5℃, humidity ≤25%, and wind speed ≤5m / s. This example is located at a photovoltaic power station in Yushu Tibetan Autonomous Prefecture, Qinghai Province, where the lowest winter temperature can reach 15℃. Furthermore, the high altitude results in atmospheric pressure being only 65% ​​of standard atmospheric pressure, approximately 65kPa. Low temperatures lead to increased resistance in the stator windings of the stepper motor and weakened magnetism in the rotor magnets, resulting in a decrease in output torque. At 10℃, the torque is approximately 60% of that at room temperature, making it prone to "step loss." Simultaneously, low temperatures reduce the sensitivity of the infrared temperature measurement module's detector, increasing the measurement error from ±0.5℃ at room temperature to ±1.5℃. High altitude and low air pressure also reduce the insulation strength of the air, meaning the air breakdown voltage decreases with decreasing air pressure. If the room temperature insulation resistance threshold is still used, situations may arise where "insulation is qualified but misjudged as unqualified," affecting maintenance efficiency.

[0044] In this embodiment, the algorithm parameters are set as follows: Coordinate library parameter adjustment: The Q235 steel used in the cabinet will expand and contract at low temperatures, with a contraction coefficient of approximately 1.2 × 10⁻⁶. -5 ℃ - ¹, and the vertical shrinkage has the greatest impact on the drawer's closing accuracy, meaning the drawer needs to be aligned with the guide rails vertically. Therefore, the physical coordinates in the vertical direction in the coordinate library need to be corrected, as shown in the formula. ,in The ambient temperature at the site. The values ​​represent the vertical physical coordinates at room temperature, and the correction compensates for coordinate shifts caused by cabinet shrinkage. Furthermore, low temperatures increase the clearance of the actuator's slide rails, therefore the error correction value is adjusted accordingly. Adding 0.5 digits, where 1 digit corresponds to 0.156 mm, and 0.5 digits is approximately 0.078 mm, can cover gap variations.

[0045] Data acquisition parameters: The infrared temperature measurement module selected is MLX90614. To compensate for low-temperature measurement errors, a compensation coefficient is introduced. Based on experimental data: the measured temperature of the module at 10℃ is 1.5℃ lower than the actual temperature. This is compensated using the formula... , For actual measured temperature, To correct for the temperature, the temperature measurement error can be controlled within ±0.5℃; The default speed of the stepper motor is reduced from 800 steps / s at room temperature to 500 steps / s. Since the motor torque is only 60% of that at room temperature when the temperature is 10℃, reducing the speed can increase the output torque of the motor (the torque of the stepper motor is inversely proportional to the speed; a 37.5% reduction in speed can increase the torque by about 60%), thus avoiding step loss.

[0046] Safety judgment parameters: The insulation resistance threshold is reduced from 100MΩ at room temperature to 80MΩ. At high altitudes and low air pressure, the air insulation strength decreases. If the 100MΩ threshold is still used, the actual insulation will be mistakenly judged as unqualified (experimental data shows that at an altitude of 3500m, the measured insulation resistance is about 20% lower than at room temperature). Reducing the threshold can reduce misjudgments. The short-circuit / ground signal threshold was adjusted from 40dB at room temperature to 38dB. High altitude causes radio wave propagation attenuation to increase (about 2dB). If the original threshold is used, normal signals will be misjudged as "short-circuit / ground signals". Increasing the threshold can compensate for signal attenuation.

[0047] Execution parameters: The stepper motor microstepping factor is increased from 8 times at room temperature to 16 times. The increased microstepping factor reduces the minimum step distance of the motor (from 0.225° to 0.1125°), improves positioning accuracy, and offsets mechanical deformation caused by low temperature (the drawer guide rail deforms by about 0.1mm at low temperature), ensuring that the operating accuracy meets the requirements.

[0048] In this embodiment, the automated operation and maintenance process is demonstrated in detail using drawer module J5 (200A, current state is "working position closed", target state is "cold standby withdrawal", that is, the drawer is completely withdrawn from the cabinet and the main circuit and control circuit are disconnected). The input of each step comes from the output of the previous step: Input the standardized coordinate library trained in Example 1 Physical coordinates of J5 module at room temperature , , Ambient temperature at the site The current drawer closure status is collected in real time by a temperature sensor inside the cabinet. The status is collected by an infrared status sensor, and 1 represents a closed state.

[0049] S1 coordinate library correction: First, correct the vertical physical coordinates: (The following text appears to be unrelated and possibly a separate excerpt: "At room temperature...") Substituting into the contraction formula, we can calculate... —This value represents the actual vertical physical coordinates of the J5 module at 10℃, ensuring that subsequent path planning is based on the actual cabinet dimensions; Next, calculate the vertical digitized coordinates: based on the results obtained in Example 1. and the corrected error correction value ,Original Add a 0.5-digit compensation gap and substitute it into the formula. ,have to This digital coordinate will serve as the vertical positioning reference for the actuator.

[0050] S2 Data Acquisition: S21. First, obtain the historical operation frequency of the J5 module. The weight of the operation and maintenance records for the past 6 months is calculated based on the number of times per month. , The total operation frequency of the 10 modules; then from the coordinate library Extract the vertical physical coordinates of all modules, sort them by value from smallest to largest, and divide them into K=3 regions, each region covering approximately 34 modules. (J5 module...) If it belongs to region 2, calculate the weight of this region. , The module set for region 2, containing J5, J6, and J7; traversed from largest to smallest region weight, region 1. Area 2 Area 3 In the second region, according to horizontal physical coordinates Sort by size from smallest to largest and generate sub-paths Overall inspection route This path planning can prioritize the inspection of modules with high operation frequency, reducing unnecessary movement; S22. Control the infrared temperature measurement module to move to the temperature measurement point of module J5, using the preset coordinates in the coordinate library. , Measured temperature Substitute into the compensation formula: The corrected temperature is close to the actual operating temperature of the module. S23, Infrared status sensor acquires the closing signal of J5 module. The trigger voltage is 0.9V, which is greater than the threshold of 0.5V; the OpenMV module acquires the indicator light image, and the HSV value is... Compare the indicator light status with the red threshold range. The color red represents the standby state, which aligns with the expected state of "working position closed"; this ultimately forms the state dataset. .

[0051] S3. First, set the target position signal for the J5 module. In the cold standby state, it should be disconnected, and the position deviation should be calculated. The location was determined to be abnormal; then a cold standby status verification was performed: because the target status was cold standby, the operation command was initially determined to be "extract" (…). Substitute into the cold standby anomaly determination formula ,current The device is in a closed state, which does not meet the requirements for cold standby, and an abnormal location alarm message is generated. ; Temperature and indicator light abnormality judgment: Temperature Temperature deviation Indicator light status With historical records Consistency, deviation , so there are no abnormalities in temperature and indicator lights; Data reporting and instruction reception: The total bandwidth of the 4G communication module , and the bandwidth utilization rate is monitored in real time (not exceeding 0.8, sufficient bandwidth), allocate bandwidth according to priority: abnormal signal bandwidth , and aggregate the abnormal alarms and encrypt and transmit them to the remote control center; Based on the alarm information, the remote control center confirms that the J5 module needs to perform the "cold standby extraction" operation and feedbacks the operation instruction .

[0052] S4. Safety determination: S41. The short - circuit / grounding sensor collects the signal strength of the J5 module , and compares it with the adjusted threshold value , , and determines that there is no short - circuit / grounding abnormality, ; S42. Extract the X - axis safety travel boundary of the J5 module from the coordinate library , the current operation point position , calculate the minimum distance (safety margin), similarly , , , and determines that the travel is safe, ; S43. The insulation resistance tester collects the insulation resistance of the actuator , and compares it with the adjusted threshold value of 80 MΩ, , and determines that the insulation is qualified, ; Through logical AND operation , output the "permission to operate" instruction , and record the risk cause at the same time (no risk).

[0053] S5. Execute the operation: S�1. Extract the cold standby target coordinates of the J5 module from the coordinate library (the depth coordinate when fully extracted), the current depth coordinate , calculate the displacement (the negative sign represents moving along the negative Z - axis direction, that is, extraction); S52. The basic step angle of the stepper motor , the subdivision multiple , the actual step angle ; The radius of the motor drive mechanism , substitute it into the step number formula , and this step number can ensure that the drawer moves precisely by 50 mm; S53. The drawer of the J5 module is a push - pull switch, so generate an action sequence , First, align the actuator with the drawer handle using X and Y axis movements, then pull it out along the Z axis; S54. During operation, monitor operational deviations in real time. (Level 1 threshold), no emergency stop; current (Secondary threshold), no deceleration required; (Level 3 threshold), no locking; finally complete the drawer extraction operation and generate the execution result set. .

[0054] S6. Status Recording and Tracing: Based on the execution result set , (No emergency stop) (Unlocked) Complete, and the drawer is fully extended (detected by infrared sensor). ), encoding the final state (Cold backup); Record operation timestamps , combined , , Generate operation and maintenance ledger ;Will Write to the local database (storage on an SD card) and synchronize to the remote control center; the updated historical record set. Sort by timestamp in descending order. This facilitates subsequent queries and anomaly tracing.

[0055] Example 3 like Figure 6 , 7 As shown in the example, this embodiment selects a Gobi photovoltaic power station in Haixi Mongolian and Tibetan Autonomous Prefecture, Qinghai Province, at an altitude of 2800m, with temperatures ranging from 5℃ to 25℃ and a dust concentration of 1520mg / m³. Sandstorms occur 23 times per month, with no surrounding obstructions and strong winds, resulting in a persistently high dust concentration of 1520mg / m³ and sandstorms occurring 23 times per month, with dust concentrations reaching over 50mg / m³. This high-dust environment presents two major problems: First, dust covers the sensor lenses (OpenMV, infrared temperature measurement module), causing image blurring and reduced temperature measurement accuracy. For example, after dust covers the OpenMV lens, the color recognition accuracy drops from 99% to 92%, and the infrared temperature measurement module's temperature measurement error increases from ±0.5℃ to ±2℃ due to dust obstruction. Second, dust enters the actuator's slide rail and motor interior, increasing mechanical resistance and causing "jamming" (jamming rate of approximately 5%) during stepper motor start-stop, which can severely damage the motor windings. Therefore, it is necessary to optimize the algorithm parameters to address "dust interference" and ensure the accuracy of sensor data acquisition and the reliability of the actuator.

[0056] In this embodiment, the algorithm parameters are set as follows: Data acquisition parameters: OpenMV image preprocessing optimization. Dust can cause particle noise in images, i.e., dark or bright spots, and reduce color contrast. Therefore, a preprocessing process of "median filtering and grayscale stretching" is added. The median filtering uses a 3×3 window, which can effectively remove isolated dust particle noise without excessively blurring the image edges. Grayscale stretching expands the image grayscale range from [50,200] to [0,255] through linear transformation, enhancing the distinction between the indicator light and the background. At the same time, in order to detect lens dust coverage in a timely manner, cleaning is required when the recognition accuracy is below 95% for three consecutive times. The inspection interval is shortened from 60s at room temperature to 30s to ensure that lens contamination can be detected quickly. Infrared status sensor threshold adjustment: Dust covering the sensor's transmitter / receiver end will cause the reflected signal strength to decrease. At room temperature, the trigger threshold is 0.5V. After dust covering, the reflected signal may drop to 0.4V, which is still in a closed state and is easily misjudged as "not closed". Therefore, the trigger threshold is increased to 0.8V to ensure that it is only judged as closed when the signal strength is sufficient, thereby reducing false judgments. Infrared temperature measurement module protection: A polytetrafluoroethylene dust cover is installed in front of the module lens, and the number of temperature samplings is increased from 1 to 3, and the average value is taken as the final temperature to offset the temperature fluctuations caused by dust.

[0057] Anomaly detection parameters: Dust may cause misidentification in a single image acquisition. For example, if a red indicator light is partially obscured by dust, it may be mistakenly identified as green. Therefore, the rule of "acquiring 3 consecutive frames of images, and only if more than 2 frames are abnormal" is set to avoid single misidentification. The acquisition time for 3 frames is about 0.3 seconds, which does not affect the overall process. Temperature anomaly threshold adjustment: Dust covering the infrared temperature measurement module will cause poor heat dissipation and increase its own temperature (about 2°C). If the 40°C anomaly threshold is still used, there will be a situation where "the actual temperature is 38°C but it is mistakenly judged as 40°C". Therefore, the temperature anomaly threshold is increased to 42°C to offset the effect of the module's own temperature rise.

[0058] Execution parameters: Dust increases the resistance of the slide rail. If the acceleration and deceleration time is still 100ms, the sudden torque change when the motor starts and stops is likely to cause jamming. Therefore, the acceleration and deceleration time is extended to 200ms. By slowly increasing / decreasing the speed, that is, the time from 0 to rated speed is changed from 100ms to 200ms, the torque impact is reduced and the risk of jamming is reduced. Increased additional displacement for push-pull: Dust entering the drawer guide rails can increase the friction between the drawer and the rails. If the original additional displacement for push-pull is 1mm, there may be a situation where "the drawer appears to be fully engaged but there is actually poor contact". Therefore, the additional displacement for push-pull is increased from 1mm to 2mm to ensure that the drawer can be fully pushed in or pulled out, and to avoid overheating caused by poor contact. A "count of executions" function was added to the algorithm. When the same actuator operates 50 times consecutively, a cleaning reminder is automatically triggered (reported to the remote control center via a 4G module) to prevent mechanical failures caused by long-term dust accumulation.

[0059] In this embodiment, taking drawer module J7 (i.e., 400A) as an example, with its current state being "test position closed" and its target state being "hot standby closing," meaning the drawer is pushed into the working position, the control circuit is energized, the main circuit is disconnected, and it is ready to close at any time, the automated operation and maintenance process in this scenario is demonstrated, with input and output correlation maintained between steps: Input data: Standardized coordinate library trained in Example 1 That is, the physical coordinates of the J7 module , , The on-site dust concentration is 18 mg / m³, collected in real time by a dust sensor; the OpenMV lens is currently covered by light dust; current drawer status. (Test position closed), the indicator light initially identifies as (Green, suspected misjudgment).

[0060] S2, Data Acquisition: S21 and J7 modules are 400A high-capacity modules, historical operation frequency. Next / Month, Weight The total operation frequency of the 10 modules is 90; from the coordinate library Extract the vertical physical coordinates and divide the region into K=3 regions according to the values. (J7 module) It belongs to region 3, with regional weight. Traverse the regions from largest to smallest according to their weights, starting with the first region. Area 2 Area 3 In the third region, according to horizontal physical coordinates Sort by size from smallest to largest and generate sub-paths Overall path ; S22, Control the OpenMV module to move to the position of the indicator light on the J7 module ( , The original image was captured; due to dust coverage, the original image's HSV value was [value missing]. The red indicator light is present, but dust causes a decrease in saturation and brightness. First, a 3×3 median filter is performed to remove dust particle noise from the image, resulting in a lower HSV value. Next, grayscale stretching was performed, increasing the brightness from 42% to 45% and the saturation from 62% to 65%, which is the HSV value after stretching. Finally, compare the red threshold range. Determine the status of the indicator light The color red indicates a correction of the initial misjudgment. S23, The infrared temperature measurement module collects the temperature of module J7, and samples it 3 times. , , ,average value —Due to the addition of a dust cover, the temperature measurement error is approximately ±0.8℃, requiring no additional compensation; calculate the temperature change rate. The temperature rose gradually from 38℃ to 38.5℃ in about one minute; S24. The infrared status sensor collects the reflected signal from module J7. If the trigger voltage is 0.9V ≥ 0.8V, the closed state is determined. ; Forming a state dataset .

[0061] S3. Anomaly Analysis and Command Generation: The target state of module J7 is hot standby closing, i.e., the test position is closed and the control circuit is energized. Therefore... (Should be closed), positional deviation No abnormal location; temperature (Adjusted threshold), temperature deviation No temperature anomaly; three frames were continuously acquired, the first frame... Frame 2 Frame 3 If two or more frames are consistent, it is determined that there is no abnormality in the indicator light; based on the analysis result that there is no abnormality, the remote control center will issue a "hot standby closing" operation command. Perform the closing operation to ensure the control circuit is energized.

[0062] S4. Safety Determination: S41. Signal Strength Acquired by Short Circuit / Ground Sensor The scene is at an altitude of 2800m with high air pressure, so there is no need to adjust the signal threshold, and it is determined that there is no short circuit / grounding. S42 and J7 module X-axis safety boundaries Current position , Similarly , , S43, Insulation Resistance , ; Comprehensive judgment Output (Permission granted).

[0063] S5 Operation: S51 and J7 modules are rotary switches (90° rotation is required for closing), target rotation angle. (Taken from coordinate library) ); Basic step angle of stepper motor Subdivision multiples Actual step angle ; Rotating mechanism radius , number of steps S52, the action sequence of the rotary switch is as follows: First, the rotating jaws are aligned with the switch handle via X, Y, and Z axis movements, then the rotary closing mechanism is executed; S53, the stepper motor acceleration / deceleration time is set to 200ms, gradually increasing from 0 to 500step / s (rated speed) to avoid jamming; because the switch is rotary, no additional push-pull displacement is required, but an axial clamping displacement of 1mm is added after rotation (to compensate for gaps caused by dust) to ensure good switch contact; S54, during operation, deviations... No emergency stop; sufficient travel distance, no need to decelerate; normal short circuit / grounding signal, no lockout; after execution, an execution result set is generated. .

[0064] S6. Status Recording and Tracing: , , Completed, and the control circuit is energized (voltage detection is 220V), coded. (Hot standby); The actuator corresponding to the J7 module has been operated 48 times, and it is 2 times away from 50 cleaning reminders. Update the history. .

[0065] Example 4 like Figure 8 , 9 As shown, this embodiment selects a photovoltaic power station located in Haidong City, Qinghai Province, at an altitude of 2200m, with a temperature of 25℃~35℃ and humidity of 55%~70%. It is rainy in summer and is located in the Huangshui Valley. In summer, it is affected by the monsoon and has a lot of precipitation, which causes the air humidity to remain at 55%~70% for a long time, far exceeding the 30% in dry areas. The temperature can reach 25℃~35℃, forming a "high temperature and high humidity" environment. The core impacts of this environment are: First, high temperature and humidity cause condensation inside the cabinet. When the temperature drops sharply or the humidity exceeds 70%, water vapor in the air condenses on the metal surface of the cabinet, reducing electrical insulation performance. For example, the insulation resistance of the actuator drops from 150MΩ at room temperature to below 90MΩ. If the insulation threshold of 100MΩ is still used, it will be mistakenly judged as "insulation unqualified." Second, high temperature accelerates the aging of electrical components and increases the risk of abnormal temperature inside the cabinet. For example, oxidation of contactor contacts causes heating, and the temperature rise rate increases from 0.5℃ / min to 1℃ / min. Third, condensation adheres to the surface of the OpenMV lens, forming water mist, which reduces the color recognition accuracy (from 99% to 93%). Therefore, it is necessary to optimize the algorithm parameters for "high temperature and humidity condensation" to ensure accurate insulation judgment, timely prediction of temperature anomalies, and reliable sensor data.

[0066] In this embodiment, the algorithm parameters are set as follows: Data acquisition parameters: A humidity-linked heating function has been added. When the humidity sensor detects humidity > 65%, the lens anti-fog coating heating is automatically activated at 5W and 40℃ to prevent lens condensation. The anti-fog coating is made of polytetrafluoroethylene, which can quickly evaporate moisture on the lens surface after heating, restoring image clarity. At high temperatures, the temperature change rate accelerates. If a 60-second sliding window is still used, it may miss abnormal temperature rises (such as a 3℃ increase within 1 minute). Therefore, the sampling period has been adjusted. The time limit has been shortened from 60 seconds to 30 seconds, improving the real-time performance of temperature change rate calculation and ensuring timely detection of anomalies. Humidity data has been incorporated into the status dataset, and when humidity is greater than 65%, it is automatically marked as "high humidity risk". The insulation status will be the focus of subsequent safety assessments to avoid insulation problems caused by condensation.

[0067] Safety judgment parameters: High temperature and high humidity condensation will reduce the insulation resistance of the actuator by 10%~15%. Experimental data: When the humidity is 70%, the insulation resistance is about 85% of that when it is dry. Therefore, the insulation resistance threshold is reduced from 100MΩ to 90MΩ to avoid misjudging as "insulation unqualified". High temperatures can cause thermal expansion of the metal components of the cabinet and actuator. If a safety margin of 5mm is still used, there is a possibility of "exceeding the limit after expansion." Therefore, the safety margin should be increased. The thickness was increased from 5mm to 6mm to allow for sufficient expansion space.

[0068] In this embodiment, the motor winding temperature is prone to rise under high temperature. When the ambient temperature is 35°C, the motor temperature can reach 60°C. Long-term operation will accelerate insulation aging. Therefore, a cooling fan is added, namely model 4010, with a speed of 5000r / min and an air volume of 1.2CFM. It automatically turns on when the motor casing temperature is >40°C to reduce the motor temperature, which can reduce the motor temperature to below 45°C. At high temperatures, the mechanical stability of the actuator decreases, such as the viscosity of the slide rail grease decreases and the clearance increases slightly. To ensure operational accuracy, the emergency stop trigger deviation is reduced from 1mm to 0.8mm to improve the sensitivity of safety control and avoid equipment collisions caused by exceeding tolerances.

[0069] In this embodiment, taking drawer module J3 (i.e., 100A) as an example, its current state is "cold standby withdrawn" and its target state is "operating closed," meaning the drawer is pushed into the operating position and both the main circuit and the control circuit are energized, the process in this scenario is demonstrated. The inputs and outputs between the steps are closely related: Input the coordinate library of Example 1 J3 module , , The ambient temperature is 32℃, and the humidity is 68% > 65%, posing a high humidity risk. The current drawer status is as follows: (Cold standby extraction); Preliminary insulation resistance test .

[0070] S2, Data Acquisition: Historical Operation Frequency of S21 and J3 Modules Next / Month, Weight (Total frequency 75); Vertical physical coordinates Belongs to the first region ( (axis 250-350mm), area weight ;path Overall path When humidity is 68% > 65%, the lens heating (5W) will automatically activate. After 10 seconds, an image of the indicator light will be captured, along with the HSV value. Compare with the green threshold to determine (Green indicates normal cold standby status); S22, sampling period 30s, sampling temperature , ,average value ; Calculate the rate of temperature change (The temperature rises by 0.9℃ within 30 seconds, with a gradual change); S23, (Extracted), humidity 68%, marked "high humidity risk"; Forming a state dataset .

[0071] S3. Anomaly Analysis and Command Generation: Target Location (When the circuit is closed, it should be shut down), deviation. The location is abnormal; No, but If the temperature is determined to be "close to the threshold but without a sudden increase", no alarm will be issued temporarily; an alarm will be reported. Receive "operational closing" command .

[0072] S4. Safety Assessment: S41 , S42, , S43 ?no, Feedback risk Remotely start dehumidification (after 10 minutes) ), re-evaluate .

[0073] S5. Perform the operation: Displacement. , number of steps Motor temperature 42℃, fan turned on, action sequence ;deviation It will be executed normally.

[0074] S6, Record: (Work) The logbook indicates a humidity level of 68% and dehumidification operation.

[0075] Example 5 To comprehensively verify the application of this method in different scenarios in Qinghai Province, this embodiment selects the Golmud Substation in Qinghai as the main experimental site. Located in the Qaidam Basin, this substation exhibits typical plateau environmental characteristics and covers the scenarios described in Embodiments 2, 3, and 4. Specific information is as follows: Geographical location: 36°25'N, 94°54'E, altitude 2800m; Climate conditions: Winter minimum temperature 30℃, summer maximum temperature 35℃, annual average temperature difference 65℃; Environmental characteristics: PM 10 The peak concentration reached 351.3 μg / m³, the annual precipitation was <100 mm, and the ultraviolet radiation intensity was 810. Experimental subjects: 12 drawer-type AC distribution cabinets (model GGD2-12) in the substation were selected, with a total of 156 drawer modules. They were divided into 3 groups according to the current level: 100A (52 modules), 200A (52 modules), and 400A (52 modules), covering the needs of "different specifications of drawer modules".

[0076] Experimental period: 3 months (January 1, 2024 - March 31, 2024), implemented in three phases, covering low temperatures in winter, sandstorms in spring, and high temperatures in summer to ensure comprehensive results.

[0077] In this embodiment, the experimental equipment specifically includes: Control unit: Arduino Mega2560, 10-bit I / O resolution; Image acquisition: OpenMV4H7, 640×480 resolution, supports HSV color space processing; Temperature acquisition: MLX90614 infrared temperature measurement module, accuracy ±0.5℃, measurement range 40℃~125℃; Status acquisition: GP2Y0A21YK infrared distance sensor, detection distance 1080cm, accuracy ±1cm; Communication module: EC204G module, supports AES256 encryption, bandwidth ≥10Mbps; Actuator: 42 stepper motor, torque It features adjustable subdivision magnification (116) and a three-axis slide rail with a positioning accuracy of ±0.1mm. Software model: The pre-trained model trained in Example 1 is used, including the coordinate error correction model, the OpenMV color recognition model (accuracy 99.2%), and the temperature anomaly prediction model (accuracy 98.5%). At the same time, the scene parameters of Examples 2 to 4 are loaded to ensure that the software and hardware work together.

[0078] In this embodiment, the experiment is divided into three stages, with the output of each stage serving as the input for the next stage, forming a complete verification process: Phase 1, Data Collection Phase (Month 1: January 1, 2024 - January 31, 2024) On-site calibration was conducted at the Golmud substation: the dimensions of 156 drawer modules in 12 distribution cabinets were measured, the physical coordinates of each drawer in the horizontal / vertical / depth directions were recorded, and the physical threshold of switch rotation angle was measured. Digital transformation: Application of conversion factors: / mm; Rotation angle conversion factor: °; implement In the second calibration experiment, the error correction value was calculated using the least squares method; coordinate accuracy verification formula: Actual measurement results: maximum deviation 0.42mm, meeting accuracy requirements; Based on steps S21 to S22, establish an operation and maintenance baseline: collect 24 / 7 operational data to form a status dataset. Establish historical operation frequency statistics Set the inspection weights for each area. ; Phase 1, Multi-Performance Testing Phase (Month 2: February 1, 2024 - February 29, 2024) Test Scenario 1: Low Temperature Environment Operation Test (25℃ at Night) Input data: Ambient temperature Target drawer, cabinet 3, module 12, operation command: closing operation ( ); S1 coordinate compensation: Physical coordinates: [x=150mm, y=320mm, z=180mm]; Temperature compensation coefficient: The compensated coordinates are: [149.7mm, 319.5mm, 179.6mm]. S2 Low Temperature Inspection: Sensor warm-up time: 180 seconds; Sampling interval adjustment: 2 × default interval; Data acquisition: , (red), ; S3 Anomaly Analysis: Positional Deviation: (Anomaly exists); Temperature deviation: (Normal); Generate hot standby anomaly alarm; S4 Safety Assessment: Insulation Resistance: (Pass); Short circuit signal: (Passed); Trip safety: (Pass); Output (Permission granted); S5 Motion Control: Displacement Calculation Speed ​​reduction: Step angle correction: Perform a rotation operation: ; S6 State Record: Final State: (Hot standby); Operation time: Deviation Record: ; Test results: Operation time: 28 seconds; Positioning accuracy: 0.35 mm; Status recognition accuracy: 100%; Insulation detection accuracy: 100%; Test Scenario 2: Durability Test in Sandstorm Environment (PM) 10 =300μg / m³) Input data: Environmental conditions: wind speed Dust concentration Test content: 72-hour continuous automated inspection; Inspection frequency: full cabinet inspection every 2 hours; S1 coordinate library update: Based on the wear model: Total wear = 0.001 × 36 × 1.3 = 0.047 mm; Update coordinate library: ; S2 Enhanced Inspection: Path Redundancy: Sensor cleaning cycle: Automatic cleaning every 4 hours; Data integrity: 98.7%; S3 anomaly detection: Three modules were found to have abnormal temperatures. ); Identify two indicator lights with abnormal status; Generate 5 abnormal alarm messages; S4 Safety Protection: Dynamic Adjustment of Insulation Threshold Safety margin for the trip: Risk warning accuracy rate: 96.3%; S5 fault-tolerant operation: One operation was triggered by sand accumulation, resulting in an emergency stop; the automatic retry mechanism successfully completed the operation; motion stability was maintained at 92.5%; S6 Data Protection: Storage Redundancy: Data integrity verification: 100% passed; Communication interruption rate: 0.08%; Test results: System availability: 99.2%; False alarm rate: 2.1%; Data integrity: 100%; Operation success rate: 97.8%; Phase 3, Comparative Testing Phase (Month 3: March 1, 2024 - March 31, 2024) Objective: To compare this solution with existing technologies and verify its superiority. Two control groups were selected, performing the same operations on the same scenario and modules as this solution, and recording the same indicators. Control group 1: Traditional manual operation - operated by 3 senior operation and maintenance personnel (with more than 5 years of experience in photovoltaic power station distribution cabinet operation and maintenance) in accordance with the "Photovoltaic Power Station Operation and Maintenance Procedures" (DL / T19732019) and using conventional tools (wrench, multimeter, infrared thermometer). Control group 2: The method used in this protocol; Each control group performed 100 operations in each of the three scenarios (900 operations in total), recording the same indicators as the control group to ensure fairness in the comparison. Same conditions: same time period (9:00-17:00 daily), same modules (10 typical modules), same operations (10 times each for closing / opening / exiting); Data recording: each group recorded four indicators: "single operation time, accuracy, anomaly detection rate, and safety incidents," for a total of 3 × 10 × 3 = 90 comparison operations. Statistical validation: The t-test was used to analyze the significance of the differences (significance level p<0.05).

[0079] In this embodiment, after the experiment, all data were statistically analyzed to eliminate random errors, such as operational failures caused by sudden power outages. The final comparison results of key indicators for each group were obtained, as shown in Table 1 below: Table 1: Comparison and Analysis of Core Performance Indicators

[0080] In this embodiment, the environmental adaptability analysis is shown in Table 2 below: Table 2: Performance under extreme environments

[0081] Based on the above results, a thorough technical analysis of the advantages of this solution is conducted, and the verification results are as follows: Long-term operational stability: Continuous operation for 30 days (720 hours); System availability: 99.3% (downtime <5 hours); Number of major failures: 0; Communication interruption rate: <0.1%; Software system stability: Zero crashes, zero freezes; Extreme environmental tolerance: 100% normal start-up rate at -30℃; Wind and sand protection capability: PM 10Functional integrity rate of 96.2% under 500μg / m³ environment; UV resistance: no component damage during continuous operation under UV index level 10 environment; Temperature cycling test: performance degradation <2% after 100 cycles of temperature cycling from -30℃ to +60℃; Through 30 days of field testing, the automated operation and maintenance control method for the photovoltaic power station drawer-type AC distribution cabinet of this invention has demonstrated excellent performance in practical application at the Golmud substation in Qinghai Province: it exhibits superior performance in terms of positioning accuracy, operational efficiency, and identification accuracy. Even under harsh environments such as high altitude, low temperatures, strong winds, sandstorms, and intense ultraviolet radiation, the system maintains stable operation, and the degradation of various performance indicators is controlled within acceptable limits.

[0082] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0083] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. An automated operation and maintenance control method for a drawer-type AC distribution cabinet in a photovoltaic power station, characterized in that, Includes the following steps: S1. Obtain the drawer module parameters of the target drawer-type AC distribution cabinet, construct a three-dimensional coordinate model of the distribution cabinet through coordinate mapping and error correction algorithms, convert the drawer module parameters into digital coordinate data that the control system can recognize, and encapsulate the digital coordinate data into a standardized coordinate library. ; S2. Based on the standardized coordinate library Generate a preset inspection path, control the sensor module to move along the path, collect status data of the power distribution cabinet's operation, and form a status dataset. ; S3. Compare the preset parameters with the status dataset, generate the comparison result, determine the status abnormality based on the comparison result, generate the abnormality alarm information based on the status abnormality, report the abnormality alarm information to the remote control center, and receive the operation instructions fed back by the remote control center. S4. Perform security verification on abnormal alarm information and status data according to the operation instructions, obtain the verification results, and output "Allow operation" or "Lock operation" instructions according to the verification results; S5. If the "Allow Operation" command is output, then according to the standardized coordinate library... Calculate the motion parameters of the actuator, generate the action sequence and execute the drawer opening and closing operation. During the operation, monitor the deviation in real time and trigger the safety control to generate the operation execution result. S6. Record the final status of the drawer module parameters based on the operation execution results, update the historical records and synchronize them to the remote control center to form a traceable operation and maintenance ledger.

2. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 1, characterized in that, The specific steps of step S1 are as follows: S11. Let the total number of drawer modules be... , No. The physical locations of the drawer module parameters are as follows: horizontal physical coordinates Vertical physical coordinates Depth direction physical coordinates Physical threshold for switch rotation angle ; S12. Convert physical parameters into digital coordinates using mapping formulas. The formula is: ； ； ； ； in, X-axis transformation factor Y-axis transformation factor Z-axis transformation factor This is the rotation angle conversion factor; For the first The horizontal digital coordinates of each drawer module; For the first The vertical digital coordinates of each drawer module; For the first Digital coordinates of the depth direction of each drawer module; For the first Digital threshold for the opening and closing rotation angle of each drawer module; , , , The first Error correction values ​​for each drawer module in the X, Y, Z axes and rotation angle directions; S13. Verify the positioning accuracy corresponding to the digitized coordinates using the accuracy verification formula. Encapsulated as a standardized coordinate library ;in, For the first A standardized set of coordinates for each drawer module.

3. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 2, characterized in that, The specific steps of step S2 are as follows: S21, Let the first The historical operation frequency of each drawer module is Then its weight The higher the weight, the higher the priority; S22, from the standardized coordinate library Extract the vertical physical coordinates of all modules. Divide into K regions according to numerical sorting; S23, Calculate the region weight ,in, The weight of the k-th region is calculated as follows: Traverse the regions from largest to smallest, and within each region, based on... Horizontal physical coordinates in Sort to generate subpaths Overall path ; S24. Acquire the HSV value of the drawer switch indicator image using the HSV color space feature extraction algorithm. Define the HSV threshold range of the red indicator light as follows: The HSV threshold range of the green indicator light is By comparing the HSV threshold ranges of the red and green indicator lights, the color status is output: ;in, For the first The status of the indicator lights for each module; S25. Collect the temperature inside the cabinet via an infrared temperature measurement module. The sliding window algorithm is used to calculate the rate of temperature change, with a window size of [missing value]. ; like The inspection interval of this module will be dynamically adjusted from the default inspection interval. Shorten to the adjusted inspection interval ; in, For the first The default inspection interval for each module; For the first The inspection interval after adjustment of each module; The threshold for the rate of temperature change; S26, Collection of the first The closed status signal of each drawer To form a state dataset .

4. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 3, characterized in that, The specific steps of step S3 are as follows: S31. Set drawer target position signal Anomalies are determined using the positional deviation formula: ,like The verification is as follows: Hot standby status verification: If the operation command is "close", the first... Operating instructions for each drawer Then the first The formula for determining the hot standby anomaly of each drawer is: ; Cold standby status verification: If the operation command is "extract", the first... Operating instructions for each drawer Then the first The formula for determining the abnormality of each drawer's cold standby is: ; like or Then, an abnormal location alarm message will be generated. ;in, For the first Alarm message indicating abnormal position of drawer; S32, Set temperature safety threshold Anomalies are determined using the temperature deviation formula: ,in, For the first The temperature deviation value of each module, if Then a temperature anomaly alarm message will be generated. ; S33. Compare the indicator light status deviation values ​​using the status similarity formula: ,like Then an indicator light malfunction alarm message will be generated. ; S34. Define data priority weights and abnormal alarm signal weights. Weight of status record data Real-time video data weights Bandwidth is allocated through data transmission scheduling formulas; Collection of abnormal alarm information The bandwidth allocation strategy, as per step S34, is transmitted to the remote control center, which then processes the abnormal alarm information. Generate operation instruction set And provide feedback.

5. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to any one of claims 1 to 4, characterized in that, The specific steps of step S4 are as follows: S41, Collection of the first Short circuit / ground signal strength of each drawer Set signal threshold The first is determined by the verification formula. Short circuit / grounding test results for each drawer: ; S42. Call the standardized coordinate library The Middle Safety travel boundaries of each module , , ; Calculate the current operation point Distance from the boundary; Define safety margin The verification formula is used to determine: ; S43, Insulation resistance of the acquisition actuator Set insulation threshold The insulation condition verification result is determined using the verification formula: ; S44. Generate and determine the comprehensive security assessment result through logic and formulas: ; like Then output the "Allow operation" command. ; like If so, the "lock operation" command will be output. The cause of the risk was reported via the 4G module. .

6. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 5, characterized in that, The specific steps of step S5 are as follows: S51. According to the standardized coordinate library Extract the first The target operation coordinates of each module are: Calculate the displacement along the X, Y, and Z axes. The current coordinates are... ; S52. Let the basic step angle of the stepper motor be... Subdivision multiple is Then the actual step angle ; The required number of steps is calculated using the step size formula. , and ; S53. Generate an action sequence based on the drawer switch type using a sequence generation formula: Rotary switch: ； in, For the first The action sequence of each module; Push-pull switch: ; in, Add displacement to push and pull.

7. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 6, characterized in that, Step S5 also includes the following steps for performing a Level 3 security threshold determination: S54, Level 1 Threshold: Real-time monitoring of operational deviations based on the action sequence generated in step S52. ,like This will trigger an emergency stop command. Cut off operation command; Secondary threshold: Based on the safety travel verification result of step S42, if Then the motor speed is adjusted using a deceleration formula; Level 3 threshold: Based on the short-circuit / grounding verification result of step S41, if Then a locking command is triggered. Action sequences are prohibited from execution. S55. Generate an execution result set based on the execution result of step S54: 。 8. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 7, characterized in that, The specific steps of step S6 are as follows: S61. Based on the execution result set E , and Completion state, encoding the final state : ; S62. Extract from the execution result set E , , ,from Extract module identifier j and target coordinates. ,generate ; S63, Update the formula using historical records Write to the local database and synchronize to the remote center: ,in, For the new historical record set, the ledger order is maintained using a timestamp sorting formula: ,in, This is a sorted historical record set; ensuring that the operation and maintenance status is traceable and preventing accidental insertion of drawers.

9. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 2, characterized in that, In step S12 The range of values ​​is System errors are reduced through multiple calibrations; the control unit uses an Arduino Mega 2560 chip, and the digital conversion coefficients are... , , The value can be: ； ； ； in, , , These represent the maximum physical travel of the distribution cabinet on the X, Y, and Z axes, respectively.

10. The automated operation and maintenance control method for the drawer-type AC distribution cabinet of a photovoltaic power station according to claim 6, characterized in that, The stepper motor microstepping factor in step S52 Dynamically adjust using the following algorithm: S521, if the first The module performs X-axis displacement. This represents a large displacement; therefore, a subdivision factor should be set. ; S522, if the first The module performs X-axis displacement. For small displacements, set the subdivision factor. .

Images