A solar panel control method and system based on a photovoltaic monitoring platform

Abstract

The application discloses a solar panel control method and system based on a photovoltaic monitoring platform, relates to the intelligent control technical field of photovoltaic power generation, and comprises the following steps: step one, establishing a control mapping table; step two, calling panel-end monitoring data; step three, correcting historical records to obtain a reference curve; step four, synchronously comparing to obtain control solving characteristics; step five, inputting an improved Hammerstein-Wiener model to obtain response prediction characteristics; step six, inputting an improved Tube-MPC algorithm to obtain a control target; step seven, issuing a control instruction to obtain execution feedback data; and step eight, updating control parameters to generate a solar panel control result. Through the improved Hammerstein-Wiener model and the improved Tube-MPC algorithm embedded with a controllable response decoupling mechanism, the application realizes solar panel deviation prediction, target solving and closed-loop regulation.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for photovoltaic power generation, and in particular to a solar panel control method and system based on a photovoltaic monitoring platform. Background Technology

[0002] As the scale of photovoltaic (PV) power generation continues to expand, the number of solar panels, string relationships, support control units, and panel-side data acquisition devices in PV power plants are constantly increasing. Traditional methods relying on manual inspections and total inverter-side monitoring are no longer sufficient to reflect the operational differences of individual solar panels in a timely manner. Existing PV monitoring platforms can typically collect data on power generation, irradiance, panel surface temperature, string current, and equipment alarm information, and display, statistically analyze, and issue alarms on the power plant's operating status, providing fundamental data support for PV power plant operation and maintenance.

[0003] However, existing platforms are mostly limited to monitoring and alarm functions, lacking sufficient control over the solar panels themselves. When solar panels are affected by cloud cover, temperature increases, bracket angle deviations, or changes in panel output status, the platform often only alerts to power anomalies, failing to establish a power generation baseline based on current environmental conditions or determine which parts of the power generation deviation are worth controlling and restoring. Even systems with remote control capabilities often rely on fixed thresholds or manual commands, leading to issues such as control lag, frequent adjustments, inaccurate control targets, and inconsistent updates to feedback parameters.

[0004] Therefore, how to provide a solar panel control method and system based on a photovoltaic monitoring platform is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a solar panel control method and system based on a photovoltaic monitoring platform. This invention establishes the data acquisition and control association and control constraints between the solar panel and the data acquisition and control resources at the panel end through the photovoltaic monitoring platform, retrieves the monitoring data at the panel end according to the mapping relationship, and corrects the historical monitoring records according to the current environmental conditions to form a solar panel power generation baseline curve.

[0006] Based on this, the monitoring data at the solar panel is synchronously compared with the power generation baseline curve to generate the control solution features at the solar panel. The control solution features at the solar panel are then input into the improved Hammerstein-Wiener model with a controllable response decoupling mechanism to obtain the predicted features of the solar panel response. The perturbation bundle boundary is then constructed and the solar panel control target is solved by the improved Tube-MPC algorithm. Finally, the photovoltaic monitoring platform issues control commands and updates the platform control parameters based on the execution feedback, thereby realizing the closed-loop control of the solar panel.

[0007] A solar panel control method based on a photovoltaic monitoring platform according to an embodiment of the present invention includes the following steps: Step 1: Read the topology information and panel control configuration of the photovoltaic power station in the photovoltaic monitoring platform, establish the data acquisition and control association and control constraints between the solar panel and the panel-side data acquisition and control resources, and obtain the solar panel control mapping table; Step 2: Locate the panel-end data acquisition resources according to the solar panel control mapping table and retrieve the corresponding monitoring records to obtain panel-end monitoring data with acquisition and control association; Step 3: Based on the environmental conditions in the panel monitoring data, perform benchmark correction on the matching historical monitoring records to obtain the solar panel power generation benchmark curve; Step 4: Synchronously compare the panel monitoring data with the solar panel power generation baseline curve to obtain the panel control solution characteristics with acquisition and control correlation; Step 5: Input the board-end control solution features into the improved Hammerstein-Wiener model to obtain the board-end response prediction features with acquisition and control correlation. The improved Hammerstein-Wiener model includes an input nonlinear mapping module, a board-end dynamic response module, and an output nonlinear correction module. The output nonlinear correction module embeds a controllable response decoupling mechanism. Step 6: Input the panel end response prediction features into the improved Tube-MPC algorithm, construct the perturbation tube bundle boundary based on the panel end response prediction features, and obtain the solar panel control target with acquisition and control correlation within the control constraints by solving in the control increment method. Step 7: Generate solar panel control instructions for the corresponding panel control resources based on the data acquisition and control association in the solar panel control target, and issue the solar panel control instructions through the photovoltaic monitoring platform to obtain control execution feedback data; Step 8: Update the control parameters in the photovoltaic monitoring platform based on the control execution feedback data to generate the solar panel control results.

[0008] Optionally, step one specifically includes: Read the topology information of the photovoltaic power station in the photovoltaic monitoring platform, extract the array area number, string number and panel position number, and generate the solar panel topology identifier according to the combination of array area number, string number and panel position number; Read the board-side control configuration, extract board-side acquisition resources, board-side control resources, communication addresses, control boundaries, and resource topology binding information, and generate resource topology binding identifiers based on the resource topology binding information; Perform consistency matching between solar panel topology identifier and resource topology binding identifier, write the successfully matched panel-side acquisition resources into the acquisition association, and write the successfully matched panel-side control resources into the control association; Send a status query command to the communication address and retain the acquisition and control associations that return valid responses; When there are multiple acquisition or control associations for the same solar panel, resource ranking results are generated according to string consistency, panel location consistency and communication response time, and the acquisition and control association with the highest ranking result is selected as the target acquisition and control resource. Based on the target acquisition and control resources and control boundaries, acquisition and control associations and control constraints are generated. These associations and constraints are then written into the mapping data table of the photovoltaic monitoring platform to obtain the solar panel control mapping table.

[0009] Optionally, step two specifically includes: Read the data acquisition and control association item by item according to the solar panel control mapping table, and extract the panel-end data acquisition resources and data acquisition communication addresses from the data acquisition and control association; Based on the acquisition communication address, a monitoring record reading command is sent to the board-side acquisition resource, and the original monitoring records returned by the board-side acquisition resource within the platform acquisition cycle are read. The original monitoring records are verified. Original monitoring records whose collection timestamps fall within the platform's collection period are identified as valid monitoring records, while original monitoring records whose collection timestamps do not fall within the platform's collection period are removed. When multiple valid monitoring records are returned by the same board-end acquisition resource within the platform acquisition cycle, the time difference between the acquisition timestamp and the end time of the platform acquisition cycle is calculated, and the records are sorted in ascending order of time difference. The valid monitoring record with the highest ranking is selected as the target monitoring record. The target monitoring records are bound to the data acquisition and control association in the solar panel control mapping table, and then aggregated according to power generation records, environmental records, and controlled status records to obtain panel-end monitoring data with data acquisition and control association.

[0010] Optionally, step three specifically includes: Read the data acquisition and control association and current environmental conditions from the panel monitoring data, and retrieve historical monitoring records under the same data acquisition and control association from the historical monitoring record library of the photovoltaic monitoring platform; Perform field integrity verification and irradiation validity verification on historical monitoring records, and retain historical monitoring records that simultaneously contain power generation, historical environmental conditions, controlled status and historical irradiation intensity that meet the irradiation validity threshold. The controlled status in the board-end monitoring data is compared with the controlled status in the historical monitoring records. The historical monitoring records with the same controlled status are selected as the matching historical monitoring records. When the historical monitoring records with the same controlled status are empty, the angle difference between the current support angle and the historical support angle is calculated, and the historical monitoring record with the smallest angle difference is selected as the matching historical monitoring record. Divide the current irradiance intensity by the historical irradiance intensity to obtain the irradiance correction coefficient. Subtract the current plate temperature from the historical plate temperature to obtain the temperature deviation. Multiply the temperature deviation by the temperature attenuation coefficient, take the negative value, and input it into the exponential function to obtain the temperature correction coefficient. The corrected power generation is obtained by multiplying the power generation in the historical monitoring records by the irradiance correction factor and the temperature correction factor in turn. The power generation correction is performed in time alignment according to the platform's data collection cycle, and the power generation correction is arranged in ascending order according to the data collection timestamp to form a solar panel power generation baseline curve.

[0011] Optionally, step four specifically involves: Read the data acquisition and control association, acquisition timestamp, real-time power generation and current controlled status from the monitoring data at the board end, and extract the corresponding benchmark power generation from the solar panel power generation benchmark curve according to the acquisition timestamp; Perform validity verification on the benchmark power generation and retain the data collection points that meet the benchmark validity conditions; Subtract the real-time power generation from the benchmark power generation to obtain the power deviation value, and divide the power deviation value by the benchmark power generation to obtain the power generation deviation ratio; The number of data points that continuously meet the deviation triggering condition based on the data collection timestamp is counted, and the number of data points is divided by the length of the control observation window to obtain the deviation duration ratio. The control margin is calculated based on the current controlled state and the control constraints in the solar panel control mapping table. The control margin is then divided by the corresponding control constraint range to obtain the control margin ratio. The power generation deviation ratio, deviation duration ratio, and controlled margin ratio are spliced ​​together according to the mining control correlation to obtain the plate end control solution characteristics with mining control correlation.

[0012] Optionally, step five specifically includes: The input nonlinear mapping module reads the board-end control solution features according to the data acquisition and control association, and extracts the power generation deviation ratio, deviation duration ratio, and control margin ratio from the board-end control solution features; compares the power generation deviation ratio with the deviation trigger threshold, sets it to zero when the power generation deviation ratio does not reach the deviation trigger threshold, and subtracts the deviation trigger threshold when the power generation deviation ratio reaches the deviation trigger threshold to obtain the effective deviation amount; The deviation from the continuous ratio is input into a piecewise linear function to obtain the continuous enhancement amount; the controlled margin ratio is pruned according to the effective margin interval to obtain the margin constraint amount; the effective deviation amount, the continuous enhancement amount, and the margin constraint amount are combined to obtain the plate-end nonlinear input vector. The panel-end dynamic response module takes the nonlinear input vector at the panel end as the input state of the current acquisition cycle, reads the panel-end response state of the previous acquisition cycle from the state cache of the panel-end dynamic response module, and performs autoregressive dynamic recursion based on the input state, the panel-end response state and the response attenuation coefficient to obtain the panel-end dynamic response quantity. The panel-end dynamic response quantity is used to characterize the output recovery trend of the solar panel under the current deviation degree and controllable margin. The output nonlinear correction module executes a controllable response decoupling mechanism, multiplying the effective deviation by the continuous enhancement to obtain the stable deviation, multiplying the stable deviation by the margin constraint to obtain the controllable response, and using the non-negative difference between the effective deviation and the controllable response as the disturbance response. The controllable response and disturbance response are input into the Softmax function to obtain the controllable response weight and disturbance response weight. Based on the controllable response weight and disturbance response weight, the dynamic response at the board end is gating and corrected to obtain the board end response prediction characteristics with acquisition and control correlation.

[0013] Optionally, step six specifically includes: According to the control association in the panel response prediction features, the corresponding control constraints and current controlled state are read. The control prediction time domain is read from the control parameters in the photovoltaic monitoring platform. The control execution boundary and control increment boundary are read from the control constraints. The panel dynamic response quantity, controllable response weight and disturbance response weight are extracted from the panel response prediction features. The perturbation control boundary is constructed based on the perturbation response weights and the control execution boundary. The upper and lower boundaries of the control execution boundary are tightened according to the perturbation control boundary to obtain the time-domain tightening constraint. In the control prediction time domain, candidate control increment sequences are generated according to the control increment boundary, and the predicted controlled state corresponding to each candidate control increment sequence is recursively deduced based on the dynamic response of the board end. The power generation recovery benefit corresponding to the predicted controlled state is calculated based on the controllable response weights, the control execution cost is calculated based on the change magnitude of the candidate control increment sequence, and the bundle boundary penalty is calculated based on the deviation between the predicted controlled state and the time-domain tightening constraint. The power generation recovery benefit, control execution cost, and bundle boundary penalty are combined into a control evaluation function. The candidate control increment sequence with the highest control evaluation function value is selected, and the first control increment in the candidate control increment sequence is superimposed on the current controlled state to obtain the solar panel control target with acquisition-control correlation.

[0014] Optionally, step seven specifically includes: Read the data acquisition and control association and target control status in the solar panel control target, determine the corresponding panel-end control resources and corresponding panel-end data acquisition resources in the solar panel control mapping table according to the data acquisition and control association, and read the communication address and instruction template corresponding to the panel-end control resources; The control action quantity is obtained by performing state difference calculation between the target's controlled state and the current controlled state in the panel monitoring data, and then written into the instruction template to generate solar panel control instructions. The photovoltaic monitoring platform sends solar panel control commands to the corresponding panel control resources according to the communication address, and receives the command response records returned by the panel control resources. When the instruction response record indicates that the instruction has been executed, the monitoring record of the corresponding board-side acquisition resource after the control execution is called. The monitoring record after the control execution is used as the feedback monitoring record. The state difference operation is performed between the feedback monitoring record and the target controlled state to obtain the target execution deviation. When the instruction response record indicates that the instruction execution has failed, the current controlled state in the board monitoring data is retained, the board monitoring data is used as a feedback monitoring record, and a target execution deviation with a failure mark is generated. Control execution feedback data is generated based on instruction response records, feedback monitoring records, and target execution deviations.

[0015] Optionally, step eight specifically includes: Read the instruction response record, target execution deviation, and feedback monitoring record from the control execution feedback data, and determine whether the solar panel control instruction has been successfully executed based on the instruction response record; When the solar panel control command is executed, the difference between the feedback monitoring record and the panel monitoring data before the control execution is calculated to obtain the power generation recovery amount and the controlled state change amount. The target fit is generated based on the absolute value of the target execution deviation, and the control effectiveness score is generated based on the power generation recovery amount, the controlled state change amount and the target fit. When the solar panel control command is not executed, an invalid control score is generated based on the target execution deviation marked with a failure, and the invalid control score is used as the control effectiveness score. Compare the control effectiveness score with the score update threshold; When the solar panel control command is executed and the control effectiveness score meets the score update threshold, the power generation recovery benefit weight of the control evaluation function in the improved Tube-MPC algorithm is corrected according to the power generation recovery amount, the control execution boundary and control increment boundary are corrected according to the target execution deviation, and the response decay coefficient in the improved Hammerstein-Wiener model is corrected according to the change in the controlled state. When the solar panel control command is not executed or the control effectiveness score does not meet the score update threshold, the current control parameters in the photovoltaic monitoring platform are retained, and the control execution feedback data is written into the feedback record of the photovoltaic monitoring platform. Write the current or updated control parameters, control effectiveness score, and feedback monitoring records into the photovoltaic monitoring platform to generate solar panel control results.

[0016] Optionally, a solar panel control system based on a photovoltaic monitoring platform includes the following modules: The control mapping establishment module is used to read the topology information of the photovoltaic power station and the control configuration of the panel in the photovoltaic monitoring platform, establish the acquisition and control association and control constraints between the solar panel and the acquisition and control resources of the panel, and obtain the solar panel control mapping table. The panel-side monitoring and acquisition module is used to locate the panel-side acquisition resources according to the solar panel control mapping table and retrieve the corresponding monitoring records to obtain panel-side monitoring data with acquisition and control association. The baseline curve generation module is used to perform baseline correction on the matching historical monitoring records based on the environmental conditions in the panel monitoring data to obtain the solar panel power generation baseline curve. The control feature generation module is used to synchronously compare the panel monitoring data with the solar panel power generation baseline curve to obtain the panel control solution features with acquisition and control correlation. The response prediction module is used to input the plate-end control solution features into the improved Hammerstein-Wiener model to obtain the plate-end response prediction features with acquisition and control correlation. The control target solving module is used to input the panel end response prediction features into the improved Tube-MPC algorithm, construct the perturbation tube bundle boundary based on the panel end response prediction features, and obtain the solar panel control target with acquisition and control correlation in the control constraint by solving the control increment method. The control command issuing module is used to generate solar panel control commands for corresponding panel control resources based on the data acquisition and control association in the solar panel control target, and issue the solar panel control commands through the photovoltaic monitoring platform to obtain control execution feedback data; The feedback update module is used to update the control parameters in the photovoltaic monitoring platform based on the control execution feedback data, and generate solar panel control results.

[0017] The beneficial effects of this invention are: The improved Hammerstein-Wiener model embedded in this invention incorporates a controllable response decoupling mechanism. It performs nonlinear mapping, dynamic recursion, and gating correction on the power generation deviation, deviation persistence, and control margin in the panel control solution characteristics. This can distinguish between the high controllable response part that can be recovered by control actions and the low controllable response part that is greatly affected by environmental disturbances. This makes the panel response prediction characteristics more consistent with the actual output recovery trend of the solar panel after it is controlled, and reduces misjudgments caused by short-term irradiance fluctuations or temperature rise disturbances.

[0018] The improved Tube-MPC algorithm constructs a perturbation bundle boundary based on the panel response prediction characteristics and generates a candidate control increment sequence within the control constraints. The control increment is evaluated by the power generation recovery benefit, control execution cost, and bundle boundary penalty. This makes the solar panel control objective no longer only pursue instantaneous power increase, but also achieve a balance between power generation recovery, regulation stability, and execution safety, reducing the risk of frequent support adjustments and repeated fluctuations in panel output.

[0019] By using a solar panel control mapping table to connect the acquisition of resources, control resources, and control constraints, and combining control execution feedback data to update the control parameters in the photovoltaic monitoring platform, a closed-loop control is formed from monitoring data acquisition, response prediction, control target solving, command issuance to feedback correction, thereby improving the photovoltaic monitoring platform's ability to control solar panels more precisely and stably. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a solar panel control method based on a photovoltaic monitoring platform proposed in this invention; Figure 2 This is a schematic diagram illustrating the working principle of the improved Hammerstein-Wiener model for a solar panel control method based on a photovoltaic monitoring platform proposed in this invention. Figure 3 This is a schematic diagram illustrating the association between the photovoltaic monitoring platform and the solar panel-side data acquisition and control resources proposed in this invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0022] refer to Figures 1-3 A solar panel control method and system based on a photovoltaic monitoring platform includes the following steps: Step 1: Read the topology information and panel control configuration of the photovoltaic power station in the photovoltaic monitoring platform, establish the data acquisition and control association and control constraints between the solar panel and the panel-side data acquisition and control resources, and obtain the solar panel control mapping table; Step 2: Locate the panel-end data acquisition resources according to the solar panel control mapping table and retrieve the corresponding monitoring records to obtain panel-end monitoring data with acquisition and control association; Step 3: Based on the environmental conditions in the panel monitoring data, perform benchmark correction on the matching historical monitoring records to obtain the solar panel power generation benchmark curve; Step 4: Synchronously compare the panel-end monitoring data with the solar panel power generation baseline curve to obtain the panel-end control solution characteristics with acquisition and control correlation; Step 5: Input the board-end control solution features into the improved Hammerstein-Wiener model to obtain the board-end response prediction features with acquisition and control correlation. The improved Hammerstein-Wiener model includes an input nonlinear mapping module, a board-end dynamic response module, and an output nonlinear correction module. The output nonlinear correction module embeds a controllable response decoupling mechanism. Step 6: Input the panel end response prediction features into the improved Tube-MPC algorithm, construct the perturbation tube bundle boundary based on the panel end response prediction features, and obtain the solar panel control target with acquisition and control correlation within the control constraints by solving in the control increment method. Step 7: Generate solar panel control instructions for the corresponding panel control resources based on the data acquisition and control association in the solar panel control target, and issue the solar panel control instructions through the photovoltaic monitoring platform to obtain control execution feedback data; Step 8: Update the control parameters in the photovoltaic monitoring platform based on the control execution feedback data to generate the solar panel control results.

[0023] In this embodiment, step one specifically includes: The photovoltaic monitoring platform reads the power station topology table and the panel control configuration table. The power station topology table records the solar panel topology identifier according to the array area, string, and panel position. The panel control configuration table records the panel acquisition resources, panel control resources, communication address, and control boundary according to the same level. The solar panel topology identifier is generated using a combination coding method of "array area number - string number - panel position number". The resource topology binding identifier is generated using the same coding rule as the solar panel topology identifier. The platform compares the solar panel topology identifier and the resource topology binding identifier item by item. When they match, acquisition association and control association are generated. The photovoltaic monitoring platform sends status query commands to the communication addresses corresponding to the acquisition and control resources. If the status query command returns the online status and resource type within 3 seconds, the corresponding resource is determined to be a valid resource. Resources that do not return a valid response are not written into the solar panel control mapping table. When there are multiple acquisition or control associations for the same solar panel, the platform sorts them in order of string consistency, panel position consistency, and communication response time, and selects the acquisition and control association with the highest ranking as the target acquisition and control resource. The photovoltaic monitoring platform reads the bracket angle boundary, panel output boundary, and control increment boundary based on the target acquisition and control resources, generates control constraints, and writes the acquisition and control association and control constraints into the mapping data table to obtain the solar panel control mapping table. The solar panel control mapping table is used to uniformly determine the acquisition source of panel monitoring data, the constraint range of control target solution, and the target of solar panel control command issuance, providing the object basis for subsequent panel monitoring data acquisition and control command issuance.

[0024] In this embodiment, step two specifically involves: The photovoltaic monitoring platform reads the data acquisition and control associations item by item according to the solar panel control mapping table, and extracts the panel-end data acquisition resources and data acquisition communication addresses from the data acquisition and control associations. The panel-end data acquisition resources include the panel-end power sampling channel, the panel surface temperature sampling channel, the bracket angle sampling channel, and the environmental sensor sampling channel. The platform's data acquisition cycle is set to 60 seconds. At the beginning of each data acquisition cycle, a monitoring record reading command is sent to the data acquisition communication address. After the photovoltaic panel collects resources and returns the original monitoring records, the photovoltaic monitoring platform reads the collection timestamp, power generation, panel temperature, support angle, irradiance, and current controlled status from the original monitoring records. The platform compares the collection timestamp with the current platform collection cycle. Original monitoring records whose collection timestamps fall within the current platform collection cycle are considered valid monitoring records, while original monitoring records whose collection timestamps are earlier than the start time of the cycle or later than the end time of the cycle are directly discarded. When multiple valid monitoring records are returned by the same panel acquisition resource within the same platform acquisition cycle, the photovoltaic monitoring platform calculates the time difference between the acquisition timestamp of each valid monitoring record and the end time of the platform acquisition cycle, sorts them in ascending order of time difference, and selects the valid monitoring record with the highest ranking as the target monitoring record. The photovoltaic monitoring platform binds the target monitoring records with the data acquisition and control association in the solar panel control mapping table, and collects them according to power generation records, environmental records, and controlled status records to obtain panel-end monitoring data with data acquisition and control association. The panel-end monitoring data is used to provide a unified data foundation for the generation of solar panel power generation baseline curves, the extraction of panel-end control solution features, and the comparison of control execution feedback.

[0025] In this embodiment, step three specifically includes: The photovoltaic monitoring platform reads the current irradiance, current panel temperature, current controlled status, acquisition timestamp, and acquisition-control association from the panel monitoring data with acquisition-control association. Using the acquisition-control association as the search key, it searches the historical monitoring record database for the same solar panel. The search range is set to the daytime valid power generation records within 90 days before the current acquisition date. The photovoltaic monitoring platform performs field integrity verification and irradiance validity verification on historical monitoring records, and retains historical monitoring records that simultaneously include power generation, historical irradiance intensity, historical panel temperature, historical controlled status, and historical acquisition timestamp; historical monitoring records with historical irradiance intensity below 150W / m² are not included in the benchmark correction. The photovoltaic monitoring platform compares the current controlled state with the historical controlled state. When the current support angle range, panel output range, and access state are all consistent, the corresponding historical monitoring record is determined as the matching historical monitoring record. When there are no historical monitoring records with consistent controlled states, the historical monitoring record with the smallest support angle difference and consistent access state is selected as the matching historical monitoring record. The photovoltaic monitoring platform divides the current irradiance by the historical irradiance to obtain the irradiance correction coefficient; subtracts the historical panel temperature from the current panel temperature to obtain the temperature deviation; multiplies the temperature deviation by the temperature attenuation coefficient 0.004, takes the negative value, and inputs it into an exponential function to obtain the temperature correction coefficient; and multiplies the power generation in the historical monitoring records by the irradiance correction coefficient and the temperature correction coefficient in sequence to obtain the corrected power generation. The photovoltaic monitoring platform aligns the power generation correction time according to the platform's data acquisition cycle and arranges them in ascending order according to the data acquisition timestamp to form a solar panel power generation baseline curve. The solar panel power generation baseline curve is used to represent the reference power generation level under the current environmental conditions and the current controlled state, providing a benchmark for subsequent synchronous comparison of panel monitoring data and generation of panel control solution features.

[0026] In this embodiment, step four specifically includes: The photovoltaic monitoring platform reads the data acquisition and control association, acquisition timestamp, real-time power generation and current controlled status from the monitoring data at the panel end, and extracts the benchmark power generation corresponding to the same acquisition period from the solar panel power generation benchmark curve according to the acquisition timestamp; when there are multiple benchmark power generation in the same acquisition period, the benchmark power generation closest to the acquisition timestamp of the monitoring data at the panel end is selected as the corresponding benchmark power generation. The photovoltaic monitoring platform performs validity verification on the corresponding benchmark power generation. If the benchmark power generation is greater than zero and the corresponding collection period is within the effective daytime power generation period, the corresponding collection point is retained. The power deviation value is obtained by subtracting the real-time power generation from the benchmark power generation value, and the power deviation value is divided by the benchmark power generation value to obtain the power generation deviation ratio. The photovoltaic monitoring platform counts the number of data collection points with a continuous deviation ratio greater than or equal to 0.08 based on the data collection timestamp, and divides the number of data collection points by 5 consecutive data collection cycles to obtain the deviation duration ratio. The photovoltaic monitoring platform calculates the control margin based on the current controlled state and the control constraints in the solar panel control mapping table, and divides the control margin by the corresponding control constraint range to obtain the control margin ratio. The power generation deviation ratio, deviation duration ratio, and control margin ratio are concatenated according to the data acquisition and control correlation to obtain the panel-end control solution features with data acquisition and control correlation. The panel-end control solution features are used to characterize the degree of power generation deviation, deviation duration, and control adjustability of the solar panel, providing input basis for subsequent improvement of the Hammerstein-Wiener model to predict the panel-end response.

[0027] In this embodiment, step five specifically includes: The improved Hammerstein-Wiener model is based on the original Hammerstein-Wiener model's input nonlinearity, dynamic response, and output nonlinearity series structure, retaining the original model's basic ability to perform input mapping, dynamic response modeling, and output correction for nonlinear dynamic systems. The improvement in this implementation does not change the original model's series processing framework, but limits the input object to the panel-end control solution characteristics with acquisition-control correlation. In the dynamic response stage, the panel-end response state of the previous acquisition cycle is introduced, and in the output correction stage, a controllable response decoupling mechanism is embedded, so that the model output is transformed from ordinary response prediction values ​​into panel-end response prediction characteristics suitable for solving solar panel control objectives. The input nonlinear mapping module reads the board-end control solution characteristics according to the data acquisition and control correlation, and extracts the generation deviation ratio, deviation duration ratio, and controlled margin ratio. The generation deviation ratio is compared with the deviation trigger threshold of 0.08. If the generation deviation ratio is less than 0.08, the effective deviation is set to 0; if the generation deviation ratio is greater than or equal to 0.08, 0.08 is subtracted from the generation deviation ratio to obtain the effective deviation. The deviation duration ratio is input into a piecewise linear function. If the deviation duration ratio is less than 0.4, 0.5 times the deviation duration ratio is output; if the deviation duration ratio is greater than or equal to 0.4 and less than 0.8, the original deviation duration ratio is output; if the deviation duration ratio is greater than or equal to 0.8, 1.2 times the deviation duration ratio is output, and the output upper limit is limited to 1, resulting in the duration enhancement. The controlled margin ratio is trimmed to the effective margin range of 0 to 1 to obtain the margin constraint. The effective deviation, duration enhancement, and margin constraint are combined to obtain the board-end nonlinear input vector. The panel-end dynamic response module uses the panel-end nonlinear input vector as the input state of the current acquisition cycle and reads the panel-end response state formed in the previous acquisition cycle for the same acquisition and control association from the state buffer. When the panel-end response state of the previous acquisition cycle is not found in the state buffer, the panel-end nonlinear input vector of the current acquisition cycle is used as the initial panel-end response state. Based on the response attenuation coefficient of 0.65, an autoregressive dynamic recursion is performed on the panel-end response state of the previous acquisition cycle and the input state of the current acquisition cycle to obtain the panel-end dynamic response quantity. The panel-end dynamic response quantity represents the output recovery trend of the solar panel under the current power generation deviation and control margin. The output nonlinear correction module executes a controllable response decoupling mechanism. It multiplies the effective deviation by the continuous enhancement to obtain the stable deviation, multiplies the stable deviation by the margin constraint to obtain the controllable response, and uses the non-negative difference between the effective deviation and the controllable response as the disturbance response. The controllable response and the disturbance response are input into the Softmax function to obtain the controllable response weight and the disturbance response weight. Based on the controllable response weight, the high controllable response portion of the dynamic response at the plate end that is consistent with the control margin is retained, and based on the disturbance response weight, the low controllable response portion of the dynamic response at the plate end that corresponds to short-term irradiance fluctuations and temperature rise disturbances is suppressed, thus obtaining the plate end response prediction characteristics with acquisition-control correlation. Through the above improvements, the improved Hammerstein-Wiener model can, while retaining the original model's nonlinear dynamic modeling capabilities, unify the transformation of power generation deviation, continuous deviation, and control margin into response prediction results oriented towards control solution. The plate-end response state of the previous acquisition cycle enhances the continuity of response prediction, and the controllable response decoupling mechanism reduces the interference of short-term environmental disturbances on the prediction results, making the plate-end response prediction characteristics more reflective of the recoverable trend of the solar panel after control action adjustment. This provides a stable input basis for constructing the disturbance bundle boundary and solving the solar panel control target in the improved Tube-MPC algorithm.

[0028] In this embodiment, step six specifically includes: The improved Tube-MPC algorithm is based on the original Tube-MPC algorithm's prediction time-domain optimization, perturbation bundle construction, constraint tightening, and first control output. It retains the original algorithm's basic ability to solve constrained control under perturbation conditions. The improvement in this implementation does not change the original algorithm's processing framework of ensuring control stability through the bundle boundary. Instead, it limits the input object to the panel end response prediction characteristics with acquisition and control correlation, changes the perturbation bundle boundary from a fixed perturbation range to a dynamic construction based on the panel end response prediction characteristics, and limits the control evaluation function to simultaneously consider power generation recovery benefits, control execution costs, and bundle boundary penalties. This transforms the algorithm output from ordinary control quantities into solar panel control targets suitable for execution by the photovoltaic monitoring platform. The photovoltaic monitoring platform reads the corresponding control constraints and current controlled state according to the acquisition-control correlation in the panel response prediction characteristics. It reads the control prediction time domain from the platform control parameters, which is set to 5 acquisition cycles. It reads the control execution boundary and control increment boundary from the control constraints, and extracts the panel dynamic response quantity, controllable response weight, and disturbance response weight from the panel response prediction characteristics. The control execution boundary is used to limit the allowable range of support angle, panel output, and connection status, and the control increment boundary is used to limit the adjustment range of a single control action. The improved Tube-MPC algorithm constructs a perturbation bundle boundary based on the perturbation response weight and the control execution boundary. When the perturbation response weight increases, the perturbation bundle boundary widens to cover the response offset caused by short-term irradiation fluctuations, temperature rise perturbations, and execution errors. The upper and lower boundaries of the control execution boundary are tightened according to the perturbation bundle boundary. The upper limit of the control execution boundary is subtracted from the perturbation bundle boundary, and the lower limit of the control execution boundary is added to the perturbation bundle boundary to obtain the time-domain tightening constraint. Within the control prediction time domain, the improved Tube-MPC algorithm generates candidate control increment sequences according to the control increment boundary. Each candidate control increment sequence contains control increments within 5 acquisition cycles. The predicted controlled state corresponding to each candidate control increment sequence is recursively derived based on the dynamic response of the board end, and the predicted controlled state is restricted within the time domain tightening constraint. When the predicted controlled state deviates from the time domain tightening constraint, the deviation between the predicted controlled state and the time domain tightening constraint boundary is calculated as the basis for calculating the bundle boundary penalty. The improved Tube-MPC algorithm calculates the power recovery benefit based on the controllable response weight and the predicted power recovery magnitude corresponding to the controlled state, calculates the control execution cost based on the change magnitude of the candidate control increment sequence and the number of control operations, and calculates the bundle boundary penalty based on the deviation between the predicted controlled state and the time-domain tightening constraint. The power recovery benefit, control execution cost, and bundle boundary penalty are combined into a control evaluation function. The candidate control increment sequence with the highest control evaluation function value is selected, and the first control increment in the candidate control increment sequence is superimposed on the current controlled state to obtain the solar panel control target with acquisition-control correlation. Through the above improvements, the improved Tube-MPC algorithm can directly use the panel response prediction features output by the improved Hammerstein-Wiener model to solve the control target while retaining the original algorithm's ability to resist disturbance constraints. The dynamic disturbance bundle boundary improves the algorithm's adaptability to irradiance fluctuations and panel response uncertainties, the time-domain tightening constraint reduces the risk of the control target exceeding the execution boundary, and the control evaluation function enables the solar panel control target to take into account power generation recovery, smooth operation, and execution safety, providing an executable target basis for the photovoltaic monitoring platform to generate solar panel control commands.

[0029] In this embodiment, step seven specifically includes: The photovoltaic monitoring platform reads the data acquisition and control association and the target control status in the solar panel control target. Based on the data acquisition and control association, it determines the corresponding panel-end control resources and corresponding panel-end data acquisition resources in the solar panel control mapping table, and reads the communication address and instruction template corresponding to the panel-end control resources. The panel-end control resources include the bracket control unit, the panel-end output control unit, and the access status control unit, which are used to assign the solar panel control target to the specific execution object. The photovoltaic monitoring platform reads the current controlled state from the monitoring data at the panel end, performs state difference calculation between the target controlled state and the current controlled state to obtain the control action quantity; when the control action quantity corresponds to bracket adjustment, the angle increment is written into the instruction template of the bracket control unit; when the control action quantity corresponds to panel output adjustment, the output adjustment quantity is written into the instruction template of the panel output control unit; when the control action quantity corresponds to access status adjustment, the access switching identifier is written into the instruction template of the access status control unit, generating solar panel control instructions; The photovoltaic monitoring platform sends solar panel control commands to the corresponding panel control resources according to the communication address and receives command response records. When the command response record indicates that the command execution is completed, it calls the monitoring record of the corresponding panel resource after the control execution as the feedback monitoring record, and performs state difference operation between the controlled state in the feedback monitoring record and the target controlled state to obtain the target execution deviation. When the command response record indicates that the command execution fails, it retains the current controlled state in the panel monitoring data, uses the panel monitoring data as the feedback monitoring record, and generates a target execution deviation with a failure mark. The photovoltaic monitoring platform generates control execution feedback data based on command response records, feedback monitoring records, and target execution deviations. This enables the control execution results to reflect the command execution status, changes in the controlled state, and target deviations, providing feedback basis for subsequent updates to the control parameters in the photovoltaic monitoring platform.

[0030] In this embodiment, step eight specifically includes: The photovoltaic monitoring platform reads the command response record, target execution deviation, and feedback monitoring record from the control execution feedback data, and determines whether the solar panel control command has been executed based on the command response record. When the command is executed, the difference between the feedback monitoring record and the panel monitoring data before control execution is calculated to obtain the power generation recovery amount and the controlled state change amount. The power generation recovery amount is obtained by subtracting the real-time power generation before control execution from the feedback power generation, and the controlled state change amount is obtained by the difference between the feedback controlled state and the current controlled state before control execution. The photovoltaic monitoring platform generates target alignment based on the absolute value of the target execution deviation, and generates a control effectiveness score by combining the power generation recovery and the change in controlled state. When the instruction is not executed, an invalid control score is generated based on the target execution deviation marked with a failure, and the invalid control score is used as the control effectiveness score. The photovoltaic monitoring platform compares the control effectiveness score with the score update threshold of 0.75. When the instruction is executed and the control effectiveness score is greater than or equal to 0.75, the platform calculates the revenue correction based on the ratio of the power generation recovery to the real-time power generation before control execution, and adds the revenue correction to the power generation recovery revenue weight of the control evaluation function in the improved Tube-MPC algorithm. The platform also calculates the boundary correction based on the absolute value of the target execution deviation, and uses the boundary correction to tighten the control execution boundary and the control increment boundary. The response decay correction is calculated based on the difference between the change in the controlled state and the target controlled state, and then the response decay correction is added to the response decay coefficient in the improved Hammerstein-Wiener model. When the instruction is not executed or the control effectiveness score is less than 0.75, the current control parameters are retained and the control execution feedback data is written to the feedback record. The photovoltaic monitoring platform writes the current or updated control parameters, control effectiveness score, and feedback monitoring records into the control result table to generate solar panel control results. The solar panel control results are used to record the control execution status, power generation recovery status, target alignment status, and parameter update status, providing parameter basis for solving the control target in the next acquisition cycle.

[0031] In this embodiment, a solar panel control system based on a photovoltaic monitoring platform includes the following modules: The control mapping establishment module is used to read the topology information of the photovoltaic power station and the control configuration of the panel in the photovoltaic monitoring platform, establish the acquisition and control association and control constraints between the solar panel and the acquisition and control resources of the panel, and obtain the solar panel control mapping table. The panel-side monitoring and acquisition module is used to locate the panel-side acquisition resources according to the solar panel control mapping table and retrieve the corresponding monitoring records to obtain panel-side monitoring data with acquisition and control association. The baseline curve generation module is used to perform baseline correction on the matching historical monitoring records based on the environmental conditions in the panel monitoring data to obtain the solar panel power generation baseline curve. The control feature generation module is used to synchronously compare the panel monitoring data with the solar panel power generation baseline curve to obtain the panel control solution features with acquisition and control correlation. The response prediction module is used to input the plate-end control solution features into the improved Hammerstein-Wiener model to obtain the plate-end response prediction features with acquisition and control correlation. The control target solving module is used to input the panel end response prediction features into the improved Tube-MPC algorithm, construct the perturbation tube bundle boundary based on the panel end response prediction features, and obtain the solar panel control target with acquisition and control correlation in the control constraint by solving the control increment method. The control command issuing module is used to generate solar panel control commands for corresponding panel control resources based on the data acquisition and control association in the solar panel control target, and issue the solar panel control commands through the photovoltaic monitoring platform to obtain control execution feedback data; The feedback update module is used to update the control parameters in the photovoltaic monitoring platform based on the control execution feedback data, and generate solar panel control results.

[0032] Example 1: To verify the feasibility of the present invention in practice, it was applied to the solar panel control scenario of a rooftop distributed photovoltaic array. The test objects included 96 solar panels, 8 strings, 8 sets of bracket control units, 16 panel-end output control units, and 1 photovoltaic monitoring platform. Under conditions of midday cloud shadow movement, panel surface temperature rise, and local bracket angle deviation, the photovoltaic array is prone to single-panel power generation deviation, bracket adjustment lag, and panel-end output fluctuation. Traditional platforms only issue control commands based on real-time power thresholds, which has the problems of short-term fluctuation false triggering and unstable control targets.

[0033] During testing, the photovoltaic monitoring platform first establishes a solar panel control mapping table, binding the solar panels with panel-end acquisition resources, panel-end control resources, and control constraints. The platform retrieves panel-end monitoring data with a 60-second acquisition cycle, generates a solar panel power generation baseline curve based on historical monitoring records within 90 days, and then synchronously compares the real-time power generation with the baseline power generation to form panel-end control solution features. Subsequently, the panel-end control solution features are input into the improved Hammerstein-Wiener model to obtain panel-end response prediction features, which are then input into the improved Tube-MPC algorithm to solve the solar panel control objective. Finally, the photovoltaic monitoring platform issues solar panel control commands and records control execution feedback data.

[0034] For comparison, three control methods were set up: the first is a fixed threshold control method, which triggers the support or plate-end output control only when the power generation deviation ratio reaches a preset threshold; the second is a normal MPC control method, which solves the control objective based on real-time power generation deviation and control constraints, but does not introduce a controllable response decoupling mechanism and dynamic disturbance bundle boundary; the third is the control method of this invention, which adopts an improved Hammerstein-Wiener model with an embedded controllable response decoupling mechanism and an improved Tube-MPC algorithm. Data were continuously collected for seven effective power generation periods during the day. After removing nighttime and communication interruption records, a total of 8064 sets of plate-end monitoring records were obtained, and the results are shown in Table 1 below.

[0035] Table 1. Comparison of solar panel control effects under different control methods

[0036] As shown in Table 1, the average power generation recovery rate of the control method of the present invention reaches 10.5%, which is higher than the 4.9% of the fixed threshold control method and the 7.2% of the ordinary MPC control method; the number of invalid control is reduced to 12 times, which is 26 times less than the fixed threshold control method and 11 times less than the ordinary MPC control method; the average target execution deviation is reduced to 0.063, indicating that the solar panel control target is closer to the controlled state after execution, and the control effectiveness score is improved to 0.88, indicating that the present invention has a more stable control effect under complex light disturbance and panel end state change conditions.

[0037] Among them, the average power generation recovery rate is used to represent the recovery range of power generation after control execution relative to before control; the number of invalid control operations is used to represent the number of times control failed to produce effective power generation recovery or the target execution deviation was too large after control; the average target execution deviation is used to represent the difference between the controlled state after control execution and the solar panel control target; the average response delay is used to represent the time between the issuance of the control command and the update of the feedback monitoring record; and the control effectiveness score is used to comprehensively evaluate power generation recovery, execution deviation and response stability.

[0038] The reason why this invention achieves the above-mentioned effects is that the improved Hammerstein-Wiener model separates low controllable responses such as short-term irradiance fluctuations and panel temperature rise disturbances from high controllable responses corresponding to support adjustment and panel output adjustment through a controllable response decoupling mechanism, making the panel response prediction characteristics more closely match the recovery trend of the solar panel after control. The improved Tube-MPC algorithm constructs the disturbance bundle boundary based on the panel response prediction characteristics and solves the control increment within the control constraints, so that the solar panel control objective takes into account the power generation recovery benefits, control execution costs, and boundary stability. By updating the control parameters in the photovoltaic monitoring platform through control execution feedback data, the photovoltaic monitoring platform can form a closed loop of monitoring, prediction, control, and feedback updates, improving the engineering applicability and stability of solar panel control.

[0039] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A solar panel control method based on a photovoltaic monitoring platform, characterized in that, Includes the following steps: Step 1: Read the topology information and panel control configuration of the photovoltaic power station in the photovoltaic monitoring platform, establish the data acquisition and control association and control constraints between the solar panel and the panel-side data acquisition and control resources, and obtain the solar panel control mapping table; Step 2: Locate the panel-end data acquisition resources according to the solar panel control mapping table and retrieve the corresponding monitoring records to obtain panel-end monitoring data with acquisition and control association; Step 3: Based on the environmental conditions in the panel monitoring data, perform benchmark correction on the matching historical monitoring records to obtain the solar panel power generation benchmark curve; Step 4: Synchronously compare the panel-end monitoring data with the solar panel power generation baseline curve to obtain the panel-end control solution characteristics with acquisition and control correlation; Step 5: Input the board-end control solution features into the improved Hammerstein-Wiener model to obtain the board-end response prediction features with acquisition and control correlation. The improved Hammerstein-Wiener model includes an input nonlinear mapping module, a board-end dynamic response module, and an output nonlinear correction module. The output nonlinear correction module embeds a controllable response decoupling mechanism. Step 6: Input the panel end response prediction features into the improved Tube-MPC algorithm, construct the perturbation tube bundle boundary based on the panel end response prediction features, and obtain the solar panel control target with acquisition and control correlation within the control constraints by solving in the control increment method. Step 7: Generate solar panel control instructions for the corresponding panel control resources based on the data acquisition and control association in the solar panel control target, and issue the solar panel control instructions through the photovoltaic monitoring platform to obtain control execution feedback data; Step 8: Update the control parameters in the photovoltaic monitoring platform based on the control execution feedback data, and generate the solar panel control results; Step five specifically involves: The input nonlinear mapping module reads the board-end control solution features according to the acquisition and control association, and extracts the power generation deviation ratio, deviation duration ratio and control margin ratio from the board-end control solution features; The generation deviation ratio is compared with the deviation trigger threshold. When the generation deviation ratio does not reach the deviation trigger threshold, it is set to zero. When the generation deviation ratio reaches the deviation trigger threshold, the deviation trigger threshold is subtracted to obtain the effective deviation amount. The deviation from the continuous ratio is input into a piecewise linear function to obtain the continuous enhancement amount; The controlled margin ratio is pruned according to the effective margin range to obtain the margin constraint amount; The effective deviation, continuous enhancement, and margin constraint are combined to obtain the plate-end nonlinear input vector; The panel-end dynamic response module takes the nonlinear input vector at the panel end as the input state of the current acquisition cycle, reads the panel-end response state of the previous acquisition cycle from the state cache of the panel-end dynamic response module, and performs autoregressive dynamic recursion based on the input state, the panel-end response state and the response attenuation coefficient to obtain the panel-end dynamic response quantity. The panel-end dynamic response quantity is used to characterize the output recovery trend of the solar panel under the current deviation degree and controllable margin. The output nonlinear correction module executes a controllable response decoupling mechanism, multiplying the effective deviation by the continuous enhancement to obtain the stable deviation, multiplying the stable deviation by the margin constraint to obtain the controllable response, and using the non-negative difference between the effective deviation and the controllable response as the disturbance response. The controllable response and disturbance response are input into the Softmax function to obtain the controllable response weight and disturbance response weight; gating correction is performed on the dynamic response at the board end based on the controllable response weight and disturbance response weight to obtain the board end response prediction characteristics with acquisition and control correlation. Step six specifically involves: According to the control association in the panel response prediction characteristics, the corresponding control constraints and current controlled state are read. The control prediction time domain is read from the control parameters in the photovoltaic monitoring platform. The control execution boundary and control increment boundary are read from the control constraints. The panel dynamic response quantity, controllable response weight and disturbance response weight are extracted from the panel response prediction characteristics. The perturbation control boundary is constructed based on the perturbation response weights and the control execution boundary. The upper and lower boundaries of the control execution boundary are tightened according to the perturbation control boundary to obtain the time-domain tightening constraint. In the control prediction time domain, candidate control increment sequences are generated according to the control increment boundary, and the predicted controlled state corresponding to each candidate control increment sequence is recursively deduced based on the dynamic response of the board end. The power generation recovery benefit corresponding to the predicted controlled state is calculated based on the controllable response weights, the control execution cost is calculated based on the change magnitude of the candidate control increment sequence, and the bundle boundary penalty is calculated based on the deviation between the predicted controlled state and the time-domain tightening constraint. The power generation recovery benefit, control execution cost, and bundle boundary penalty are combined into a control evaluation function. The candidate control increment sequence with the highest control evaluation function value is selected, and the first control increment in the candidate control increment sequence is superimposed on the current controlled state to obtain the solar panel control target with acquisition-control correlation.

2. The solar panel control method based on a photovoltaic monitoring platform according to claim 1, characterized in that, Step one specifically involves: Read the topology information of the photovoltaic power station in the photovoltaic monitoring platform, extract the array area number, string number and panel position number, and generate the solar panel topology identifier according to the combination of array area number, string number and panel position number; Read the board-side control configuration, extract board-side acquisition resources, board-side control resources, communication addresses, control boundaries, and resource topology binding information, and generate resource topology binding identifiers based on the resource topology binding information; Perform consistency matching between solar panel topology identifier and resource topology binding identifier, write the successfully matched panel-side acquisition resources into the acquisition association, and write the successfully matched panel-side control resources into the control association; Send a status query command to the communication address and retain the acquisition and control associations that return valid responses; When there are multiple acquisition or control associations for the same solar panel, resource ranking results are generated according to string consistency, panel location consistency and communication response time, and the acquisition and control association with the highest ranking result is selected as the target acquisition and control resource. Based on the target acquisition and control resources and control boundaries, acquisition and control associations and control constraints are generated. These associations and constraints are then written into the mapping data table of the photovoltaic monitoring platform to obtain the solar panel control mapping table.

3. The solar panel control method based on a photovoltaic monitoring platform according to claim 1, characterized in that, Step two specifically involves: Read the data acquisition and control association item by item according to the solar panel control mapping table, and extract the panel-end data acquisition resources and data acquisition communication addresses from the data acquisition and control association; Based on the acquisition communication address, a monitoring record reading command is sent to the board-side acquisition resource, and the original monitoring records returned by the board-side acquisition resource within the platform acquisition cycle are read. The original monitoring records are verified. Original monitoring records whose collection timestamps fall within the platform's collection period are identified as valid monitoring records, while original monitoring records whose collection timestamps do not fall within the platform's collection period are removed. When multiple valid monitoring records are returned by the same board end resource within the platform's collection cycle, the time difference between the collection timestamp and the end time of the platform's collection cycle is calculated, and the records are sorted in ascending order of time difference. The valid monitoring record with the highest time difference is selected as the target monitoring record. The target monitoring records are bound to the data acquisition and control association in the solar panel control mapping table, and then aggregated according to power generation records, environmental records, and controlled status records to obtain panel-end monitoring data with data acquisition and control association.

4. The solar panel control method based on a photovoltaic monitoring platform according to claim 1, characterized in that, Step three specifically involves: Read the data acquisition and control association and current environmental conditions from the panel monitoring data, and retrieve historical monitoring records under the same data acquisition and control association from the historical monitoring record library of the photovoltaic monitoring platform; Perform field integrity verification and irradiation validity verification on historical monitoring records, and retain historical monitoring records that simultaneously contain power generation, historical environmental conditions, controlled status and historical irradiation intensity that meet the irradiation validity threshold. The controlled status in the board-end monitoring data is compared with the controlled status in the historical monitoring records. The historical monitoring records with the same controlled status are selected as the matching historical monitoring records. When the historical monitoring records with the same controlled status are empty, the angle difference between the current support angle and the historical support angle is calculated, and the historical monitoring record with the smallest angle difference is selected as the matching historical monitoring record. Divide the current irradiance intensity by the historical irradiance intensity to obtain the irradiance correction coefficient. Subtract the current plate temperature from the historical plate temperature to obtain the temperature deviation. Multiply the temperature deviation by the temperature attenuation coefficient, take the negative value, and input it into the exponential function to obtain the temperature correction coefficient. The corrected power generation is obtained by multiplying the power generation in the historical monitoring records by the irradiation correction factor and the temperature correction factor in turn. The power generation correction is performed in time alignment according to the platform's data collection cycle, and the power generation correction is arranged in ascending order according to the data collection timestamp to form a solar panel power generation baseline curve.

5. The solar panel control method based on a photovoltaic monitoring platform according to claim 1, characterized in that, Step four specifically involves: Read the data acquisition and control association, acquisition timestamp, real-time power generation and current controlled status from the monitoring data at the board end, and extract the corresponding benchmark power generation from the solar panel power generation benchmark curve according to the acquisition timestamp; Perform validity verification on the benchmark power generation and retain the data collection points that meet the benchmark validity conditions; Subtract the real-time power generation from the benchmark power generation to obtain the power deviation value, and divide the power deviation value by the benchmark power generation to obtain the power generation deviation ratio; The number of data points that continuously meet the deviation triggering condition based on the data collection timestamp is counted, and the number of data points is divided by the length of the control observation window to obtain the deviation duration ratio. The control margin is calculated based on the current controlled state and the control constraints in the solar panel control mapping table. The control margin is then divided by the corresponding control constraint range to obtain the control margin ratio. The power generation deviation ratio, deviation duration ratio, and controlled margin ratio are spliced ​​together according to the mining control correlation to obtain the plate end control solution characteristics with mining control correlation.

6. The solar panel control method based on a photovoltaic monitoring platform according to claim 1, characterized in that, Step seven specifically involves: Read the data acquisition and control association and target control status in the solar panel control target, determine the corresponding panel-end control resources and corresponding panel-end data acquisition resources in the solar panel control mapping table according to the data acquisition and control association, and read the communication address and instruction template corresponding to the panel-end control resources; The control action quantity is obtained by performing state difference calculation between the target's controlled state and the current controlled state in the panel monitoring data, and then written into the instruction template to generate solar panel control instructions. The photovoltaic monitoring platform sends solar panel control commands to the corresponding panel control resources according to the communication address, and receives the command response records returned by the panel control resources. When the instruction response record indicates that the instruction has been executed, the monitoring record of the corresponding board-side acquisition resource after the control execution is called. The monitoring record after the control execution is used as the feedback monitoring record. The state difference operation is performed between the feedback monitoring record and the target controlled state to obtain the target execution deviation. When the instruction response record indicates that the instruction execution has failed, the current controlled state in the board monitoring data is retained, the board monitoring data is used as a feedback monitoring record, and a target execution deviation with a failure mark is generated. Control execution feedback data is generated based on instruction response records, feedback monitoring records, and target execution deviations.

7. The solar panel control method based on a photovoltaic monitoring platform according to claim 1, characterized in that, Step eight specifically involves: Read the instruction response record, target execution deviation, and feedback monitoring record from the control execution feedback data, and determine whether the solar panel control instruction has been successfully executed based on the instruction response record; When the solar panel control command is executed, the difference between the feedback monitoring record and the panel monitoring data before the control execution is calculated to obtain the power generation recovery amount and the controlled state change amount. The target fit is generated based on the absolute value of the target execution deviation, and the control effectiveness score is generated based on the power generation recovery amount, the controlled state change amount and the target fit. When the solar panel control command is not executed, an invalid control score is generated based on the target execution deviation marked with a failure, and the invalid control score is used as the control effectiveness score. Compare the control effectiveness score with the score update threshold; When the solar panel control command is executed and the control effectiveness score meets the score update threshold, the power generation recovery benefit weight of the control evaluation function in the improved Tube-MPC algorithm is corrected according to the power generation recovery amount, the control execution boundary and control increment boundary are corrected according to the target execution deviation, and the response decay coefficient in the improved Hammerstein-Wiener model is corrected according to the change in the controlled state. When the solar panel control command is not executed or the control effectiveness score does not meet the score update threshold, the current control parameters in the photovoltaic monitoring platform are retained, and the control execution feedback data is written into the feedback record of the photovoltaic monitoring platform. Write the current or updated control parameters, control effectiveness score, and feedback monitoring records into the photovoltaic monitoring platform to generate solar panel control results.

8. A solar panel control system based on a photovoltaic monitoring platform, executing the solar panel control method based on a photovoltaic monitoring platform as described in any one of claims 1 to 7, characterized in that, Includes the following modules: The control mapping establishment module is used to read the topology information of the photovoltaic power station and the control configuration of the panel in the photovoltaic monitoring platform, establish the acquisition and control association and control constraints between the solar panel and the acquisition and control resources of the panel, and obtain the solar panel control mapping table. The panel-side monitoring and acquisition module is used to locate the panel-side acquisition resources according to the solar panel control mapping table and retrieve the corresponding monitoring records to obtain panel-side monitoring data with acquisition and control association. The baseline curve generation module is used to perform baseline correction on the matching historical monitoring records based on the environmental conditions in the panel monitoring data to obtain the solar panel power generation baseline curve. The control feature generation module is used to synchronously compare the panel monitoring data with the solar panel power generation baseline curve to obtain the panel control solution features with acquisition and control correlation. The response prediction module is used to input the plate-end control solution features into the improved Hammerstein-Wiener model to obtain the plate-end response prediction features with acquisition and control correlation. The control target solving module is used to input the panel end response prediction features into the improved Tube-MPC algorithm, construct the perturbation tube bundle boundary based on the panel end response prediction features, and obtain the solar panel control target with acquisition and control correlation in the control constraint by solving the control increment method. The control command issuing module is used to generate solar panel control commands for corresponding panel control resources based on the data acquisition and control association in the solar panel control target, and issue the solar panel control commands through the photovoltaic monitoring platform to obtain control execution feedback data; The feedback update module is used to update the control parameters in the photovoltaic monitoring platform based on the control execution feedback data, and generate solar panel control results.

Images