A real-time operation state recognition device for a photovoltaic power station

Abstract

The present application relates to photovoltaic power generation and energy storage collaborative operation technical field, especially photovoltaic power station real-time operation state recognition device, technical scheme: a photovoltaic power station real-time operation state recognition device, including: cloud server and the edge end controller deployed in photovoltaic power station side, cloud server and edge end controller signal connection;The present application has adopted group intelligence optimization scheme, through the similar equipment is grouped, and utilizes the whole equipment group in the initial operation stage's mass working condition data, trains and corrects environmental response model parameter together, so that each new equipment can quickly obtain a high-precision personalized performance model verified by a large number of real data after being put into operation, greatly improves the model accuracy, convergence speed and prediction reliability under different environmental conditions.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation and energy storage coordinated operation technology, and in particular to a real-time operation status identification device for photovoltaic power plants. Background Technology

[0002] In photovoltaic and energy storage combined equipment, the output power of photovoltaic power plants is subject to drastic fluctuations due to meteorological factors such as ambient temperature and solar irradiance. Energy storage equipment needs to optimize charging and discharging strategies based on real-time photovoltaic output, battery health status, and grid dispatch instructions to achieve peak shaving and valley filling, smooth power fluctuations, and improve grid friendliness. With the continuous expansion of photovoltaic installed capacity, a large number of photovoltaic power plants are beginning to be deeply coupled with energy storage equipment, requiring accurate and rapid identification of the real-time power generation performance, attenuation level, and abnormal status of photovoltaic strings, thereby providing reliable front-end input for energy storage management.

[0003] In existing technologies, photovoltaic power plant monitoring devices typically use fixed theoretical performance models or rely on historical data that is slowly accumulated by a single device for status assessment, which has the disadvantages of low model accuracy and poor generalization ability. Summary of the Invention

[0004] In order to overcome the shortcomings of existing technologies, such as photovoltaic power plant monitoring devices that typically use fixed theoretical performance models or rely on historical data slowly accumulated by a single device for status assessment, which have low model accuracy and poor generalization ability.

[0005] The technical solution of the present invention is: a real-time operation status identification device for a photovoltaic power station, comprising: a cloud server and an edge controller deployed on the side of the photovoltaic power station, wherein the cloud server and the edge controller are connected by signals; Cloud servers are used to perform offline training and model optimization, specifically including: S1: Predefine the basic environmental response curve and the basic equipment aging curve for the photovoltaic power station. The basic environmental response curve represents the relationship between the theoretical power generation performance of the photovoltaic string under different environmental parameters, and the basic equipment aging curve represents the theoretical relationship between the equipment performance and the cumulative irradiance or cumulative operating time. S2: Based on the equipment aging baseline curve, determine the initial operating time window where the degree of equipment aging is negligible; S3: Within the initial operating time window, ignoring the effects of aging, the parameters of the basic environmental response curve are corrected based on the initial working data and corresponding environmental data uploaded by the edge controller, and the corrected environmental response curve of the device is generated. S4: Calculate and correct the equipment aging curve based on the corrected environmental response curve and the subsequently received actual equipment performance data; Edge controllers are used to perform real-time data acquisition, local computation, and control, specifically including: A1: Collect environmental data, equipment operating parameters and grid connection data in real time, and upload some data to the cloud server; A2: Receive and store the corrected environmental response curve parameters and aging parameters sent by the cloud server; A3: Based on the received parameters, real-time collected data, and user-side energy demand obtained through a lightweight inference model at the edge, the optimal state identification and control strategy is dynamically calculated. A4: Control the operating status and power output of the photovoltaic inverter according to the optimal control strategy.

[0006] Preferably, the modification of the parameters of the basic environmental response curve in step S3 specifically includes: S31: Device grouping, which divides photovoltaic strings with similar operating characteristics into the same device group in the cloud server; S32: Initial data acquisition, collecting operating data of each device in the equipment group within the initial operating time window; S33: Group parameter optimization, based on the group operating data of the equipment group, optimizes and solves the parameters of the basic environmental response curve; S34: Parameter distribution and model update. The optimized parameters are distributed to each device in the group to generate their respective corrected environmental response curves.

[0007] As a preferred option, the equipment is grouped as follows: S311: Set the grouping dimension threshold. The grouping dimension includes the climate zone of the equipment installation location (latitude and longitude), the series of the photovoltaic module model, and the power range of the module's nominal power. S312: Calculate the Euclidean distance between the attribute values ​​of the device to be grouped in each of the above dimensions and the central values ​​of the existing device groups in the cloud; S313: If the similarity exceeds the preset threshold, the device will be assigned to the existing device group; otherwise, a new device group will be created with the device as the core.

[0008] As a preferred option, the population parameter optimization specifically involves: S331: Determine the mathematical model for the basic environmental response curve, which is a function of ambient temperature and solar irradiance as independent variables and theoretical power generation as the dependent variable. The initial form of the mathematical model for the basic environmental response curve is as follows: ,in Theoretical power generation capacity, For ambient temperature, Solar irradiance, These are the model parameters to be corrected; S332: Input the pre-processed data collected from all equipment in the equipment group into the mathematical model, use the actual power generation as the observed value, and use the weighted least squares method to globally fit and optimize the model parameters; S333: The set of parameters with the highest goodness of fit is determined as the corrected environmental response parameters.

[0009] Preferably, the calculation and correction of the equipment aging curve in S4 specifically includes: S41: Long-term performance data reception, continuously receiving and storing long-term working data packets periodically uploaded by the edge after the initial runtime window; S42: Theoretical performance calculation, based on the operating condition information in the long-term working data package, calculates the theoretical power generation using the corrected environmental response curve; S43: Performance Deviation Attribution, calculate and analyze the deviation between actual power and theoretical power, and quantify it as an aging index; S44: Individual aging curve correction, based on time series data of aging indicators, corrects the basic aging curve of the equipment for that individual. S45: Optimize the aging pattern of the group, aggregate and analyze the corrected aging curves of multiple individuals within the same equipment group, optimize the aging model parameters and distribute them.

[0010] As a preferred option, the theoretical performance calculation specifically includes: S421: Extract the average ambient temperature, average solar irradiance, and operating mode identifier from the long-term monitoring data packet during the data acquisition period; S422: Select the corresponding modified environmental response curve model based on the operating mode identifier; S423: Substitute the average ambient temperature and average solar irradiance into the selected modified environmental response curve model to calculate the theoretical power generation under this environmental condition without considering aging.

[0011] As a preferred approach, the optimization of population aging patterns specifically includes: S451: Retrieve the updated individual aging parameter set of all devices in the same device group from the cloud server; S452: Calculate the statistical characteristic values ​​for each group of aging parameters, including the mean, median, and specific quantiles; S453: The parameter set regenerated by regression of statistical characteristic values ​​is determined as the optimized group aging parameter for this equipment group; S454: The optimized group aging parameters are bound to the corresponding device group identifier and sent to the edge controllers of all devices in the group for subsequent local aging status estimation and lifespan prediction.

[0012] As a preferred option, the dynamic calculation of the optimal state identification and control strategy in A3 specifically includes: A31: User-side energy consumption habit learning modeling, establishing a quantitative model of energy demand based on historical grid-connected data or user-side electricity consumption data; A32: Future power generation and energy demand forecasting, combining user habit models, real-time system status and environmental information, to predict the power generation curve and energy demand curve for a specified period in the future; A33: Multi-objective optimization solution, based on power generation prediction curve, energy demand curve and external constraints, to solve the control parameter sequence that meets the demand and is comprehensively optimal. The multi-objective optimization simultaneously considers maximizing power generation efficiency, minimizing power fluctuation and protecting equipment life. A34: Control command generation and issuance, which converts the optimized control parameter sequence into executable control commands to drive the photovoltaic inverter.

[0013] Preferably, between A2 and A3 executed by the edge controller, a local aging statistics step is also included, specifically: A311: Aging parameter reception and update, receiving, verifying and storing optimized population aging parameters issued by the cloud server; A312: Real-time aging status calculation, based on optimized group aging parameters and local real-time monitoring data, periodically calculates the current health index of the equipment; A313: The control strategy is adaptively adjusted, using the health index as a key input to dynamically adjust the equipment's operating parameters and protection strategies; A314: Aging data feedback, compresses key aging statistics from the local system and periodically uploads them to the cloud for continuous model optimization.

[0014] As a preferred option, the device forms a complete optimization closed loop: The edge controller uploads the collected initial and long-term operating data to the cloud server; The cloud server uses group data to model and correct environmental response curves and aging curves, and then distributes the parameters. The edge controller uses the distributed model parameters, local real-time data, and user-side energy demand to perform personalized and efficient control. The newly generated data is uploaded to the cloud again for continuous iterative optimization of the model.

[0015] The beneficial effects of this invention are: 1. Compared with existing technologies, photovoltaic power station monitoring devices typically use fixed theoretical performance models or rely on historical data slowly accumulated by individual devices for status assessment, resulting in low model accuracy and poor generalization ability; this invention adopts a swarm intelligence optimization scheme, which groups similar devices and uses the massive operating condition data of the entire device group during the initial operation phase to jointly train and correct the environmental response model parameters. This allows each new device to quickly obtain a high-precision personalized performance model that has been verified by a large amount of real data after being put into operation, which greatly improves the model accuracy, convergence speed, and prediction reliability under different environmental conditions. 2. Compared to existing technologies that often mix environmental changes and equipment aging when analyzing performance degradation, making accurate attribution difficult and unable to distinguish whether the decline in power generation efficiency stems from normal fluctuations in weather conditions or actual degradation issues such as equipment aging and dust accumulation, this invention employs a factor decoupling analysis scheme. First, it defines an initial time window without aging effects based on the theoretical aging curve. Data collected within this window is considered a pure reflection of environmental influences, thus first constructing a high-precision corrected environmental response curve. Subsequently, this curve is used to remove the influence of environmental variables from long-term data, clearly quantifying performance degradation purely caused by aging, providing data for subsequent health assessments and lifespan predictions. 3. Compared to existing technologies where equipment health management is mostly limited to post-fault alarms or preventative maintenance based on fixed intervals, lacking real-time perception and predictive response to the gradual decline in equipment performance, this invention adopts an adaptive control scheme that integrates a health index. The edge controller continuously calculates the current health index of the equipment based on optimized aging parameters sent from the cloud and local real-time data. This index is directly used as a key input to dynamically adjust the performance parameters and constraints of equipment operation in the optimization model. While pursuing power generation efficiency, it proactively ensures the long-term reliability and lifespan of the equipment. Attached Figure Description

[0016] Figure 1 The diagram shown is a schematic representation of the overall framework of a photovoltaic power station real-time operation status identification device according to the present invention. Detailed Implementation

[0017] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0018] Example 1 Please see Figure 1 The present invention provides an embodiment: a real-time operation status identification device for a photovoltaic power station, including a cloud server and an edge controller deployed on the side of the photovoltaic power station, wherein the cloud server and the edge controller are connected by a signal to realize cloud-edge collaborative computing; Cloud servers are used to perform offline training and model optimization, while edge controllers are used to perform real-time data acquisition, local computation, and control. In this embodiment, the photovoltaic power station includes multiple photovoltaic strings, each string is equipped with an inverter, and the edge controller is integrated into the local data acquisition unit of the power station. After the device is initially deployed, the cloud server first predefines the basic environmental response curve and the equipment aging basic curve for the photovoltaic power station. The basic environmental response curve takes the ambient temperature and solar irradiance as inputs and outputs the theoretical power generation; the equipment aging basic curve takes the cumulative irradiance as the independent variable and outputs the performance retention rate. Based on the equipment's aging baseline curve, the cloud server determines the initial operating time window where the degree of aging is negligible to be 500 kWh / m² cumulative standard irradiance. 2 Within this window, the edge controller collects data such as ambient temperature, solar irradiance, component output power, DC voltage and DC current in real time, and uploads them to the cloud after preprocessing and encryption. During the initial window period, the cloud server ignores the effects of aging and corrects the parameters of the basic environmental response curve. Specifically, the cloud divides photovoltaic strings with similar operating characteristics into the same equipment group. The grouping is based on the climate zone of the equipment installation location, the photovoltaic module model series, and the nominal power range. By calculating the Euclidean distance, all strings of the power station are assigned to the cloud's existing "temperate monsoon climate zone monocrystalline silicon string" equipment group. The cloud calls up the massive amount of data uploaded by all devices in the group during the initial window period, uses the actual power generation as the observation value, and uses the weighted least squares method to globally fit and optimize the parameters of the basic environmental response curve model to obtain the corrected environmental response curve parameters. These parameters are sent to the edge controllers of each device in the group through an encrypted communication link to complete the local model update. The edge controller receives and stores the corrected environmental response curve parameters and aging parameters sent by the cloud server. In real-time operation, the edge controller dynamically calculates the optimal state identification and control strategy based on the received parameters, real-time collected environmental data, and equipment operating parameters, and drives the photovoltaic inverter's operating status and power output through control commands.

[0019] Example 2 This embodiment is basically the same as embodiment 1, except that after receiving the parameters sent from the cloud, the edge controller also performs a local aging statistics step, thereby integrating the health index into real-time control decisions. Specifically, the edge controller receives parameter update instruction packets from the cloud server through an encrypted communication link. The instruction packets include the target device group identifier, the optimized group aging parameter set, the model version number, and the data signature. After verifying the validity of the device group identifier and the data signature, the edge controller updates the optimized group aging parameter set to the local non-volatile memory. Subsequently, the edge controller periodically calculates the current health index of the device, continuously records the device's cumulative total operating time, cumulative power generation, and recent power output performance locally, calls the locally stored aging calculation model, substitutes the current cumulative operating time and cumulative power generation, calculates the baseline health index, and then fine-tunes the baseline health index based on the short-term average deviation between the recently monitored actual power generation and the theoretical power generation predicted based on the modified environmental response curve. If the short-term average deviation is continuously negative and exceeds the threshold, the real-time health index is reduced by a certain percentage to obtain the final health index. Before performing dynamic optimization, the edge controller reads the current real-time health index and multiplies the theoretical instantaneous power generation calculated from the corrected environmental response curve by the health attenuation coefficient to obtain the actual available power generation capacity of the device in the optimization model. At the same time, based on the level of the real-time health index, the maximum allowable output power of the inverter is dynamically adjusted: when the health index is lower than the preset warning threshold, the derating protection operation strategy is automatically activated, limiting the maximum output power to 80% of the rated value. The edge controller generates a local aging statistical summary report at a preset period, which includes the health index change curve, cumulative power generation increment, and estimated actual performance degradation slope during the reporting period. After compression and encryption, the report is uploaded to the cloud server for continuous model optimization. This embodiment achieves a shift from "post-fault repair" to "pre-aging care" by integrating health index adaptive control. The health index reflects the component's degradation status in real time and is directly embedded into the control constraints. While pursuing power generation efficiency, it proactively ensures the long-term reliability and lifespan of the equipment. The aging data feedback mechanism forms a closed-loop iteration, enabling the cloud-based group aging model to continuously approach the actual degradation pattern.

[0020] Example 3 Based on Example 2, this embodiment further refines the dynamic calculation of optimal state identification and control strategy, focusing on multi-objective optimization solution and user-side energy consumption habit learning; The edge controller continuously collects and records grid-connected power data of photovoltaic power plants and electricity consumption event data on the user side, including event start time, duration and power consumption. Based on historical electricity consumption event data, a clustering algorithm is used to identify the typical electricity consumption period of users and calculate the probability distribution of electricity consumption events in each period, which is then stored as a user habit model. During operation, the edge controller obtains the typical electricity consumption period and corresponding power demand for the next 24 hours from the user habit model. At the same time, it obtains the real-time power generation of the photovoltaic string, ambient temperature and solar irradiance. It calls the power generation prediction model based on the modified environmental response curve, combines weather forecast information, and generates the power generation prediction curve for the next 24 hours. The power generation prediction curve is superimposed and analyzed with the energy demand curve to generate the basic energy balance prediction curve. The multi-objective optimization solution simultaneously considers maximizing power generation efficiency, minimizing power fluctuations, and protecting equipment lifespan. The edge controller sets the optimization objective function as a comprehensive function of power generation revenue and equipment aging costs. The optimization variables are control sequences with multiple time steps. The optimization constraints include grid-connected power fluctuations not exceeding the upper limit required by grid dispatch, inverter power not exceeding hardware limits, and equipment operating voltage and current not exceeding safety thresholds. The model predictive control algorithm is used to solve for the optimal control parameter sequence that minimizes the objective function under the premise of satisfying all constraints. The optimized control parameter sequence is transformed into executable control instructions: the time points when the inverter output power needs to be adjusted and the target power values ​​in each time period are parsed out, and the continuous adjustment mode or intermittent operation mode is decided based on the smooth or step characteristics of the power change. The decision results are sent to the execution unit of the photovoltaic inverter. Before the start of the next optimization cycle, the control parameter sequence is rolled and re-solved based on the deviation between the actual operating state and the predicted state. This embodiment achieves dynamic matching between the power generation and consumption sides by deeply exploring users' energy consumption habits. It comprehensively balances power generation revenue, power stability, and equipment lifespan through multi-objective optimization, avoiding the drawbacks of traditional solutions that unilaterally pursue power generation and accelerate equipment aging. The rolling optimization mechanism of model predictive control enables the control strategy to respond to environmental changes and load fluctuations in real time, significantly improving the robustness and economy of the equipment under complex operating conditions.

[0021] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A real-time operating status identification device for a photovoltaic power station; characterized in that: This includes: a cloud server and an edge controller deployed on the side of the photovoltaic power station, with the cloud server and the edge controller connected by a signal. Cloud servers are used to perform offline training and model optimization, specifically including: S1: Predefine the basic environmental response curve and the basic equipment aging curve for the photovoltaic power station. The basic environmental response curve represents the relationship between the theoretical power generation performance of the photovoltaic string under different environmental parameters, and the basic equipment aging curve represents the theoretical relationship between the equipment performance and the cumulative irradiance or cumulative operating time. S2: Based on the equipment aging baseline curve, determine the initial operating time window where the degree of equipment aging is negligible; S3: Within the initial operating time window, ignoring the effects of aging, the parameters of the basic environmental response curve are corrected based on the initial working data and corresponding environmental data uploaded by the edge controller, and the corrected environmental response curve of the device is generated. S4: Calculate and correct the equipment aging curve based on the corrected environmental response curve and the subsequently received actual equipment performance data; Edge controllers are used to perform real-time data acquisition, local computation, and control, specifically including: A1: Collect environmental data, equipment operating parameters and grid connection data in real time, and upload some data to the cloud server; A2: Receive and store the corrected environmental response curve parameters and aging parameters sent by the cloud server; A3: Based on the received parameters, real-time collected data, and user-side energy demand obtained through a lightweight inference model at the edge, the optimal state identification and control strategy is dynamically calculated. A4: Control the operating status and power output of the photovoltaic inverter according to the optimal control strategy.

2. The real-time operation status identification device for a photovoltaic power station according to claim 1, characterized in that: Step S3, which involves correcting the parameters of the basic environmental response curve, specifically includes: S31: Device grouping, which divides photovoltaic strings with similar operating characteristics into the same device group in the cloud server; S32: Initial data acquisition, collecting operating data of each device in the equipment group within the initial operating time window; S33: Group parameter optimization, based on the group operating data of the equipment group, optimizes and solves the parameters of the basic environmental response curve; S34: Parameter distribution and model update. The optimized parameters are distributed to each device in the group to generate their respective corrected environmental response curves.

3. The real-time operation status identification device for a photovoltaic power station according to claim 2, characterized in that: The equipment is grouped as follows: S311: Set the grouping dimension threshold. The grouping dimension includes the climate zone of the equipment installation location (latitude and longitude), the series of the photovoltaic module model, and the power range of the module's nominal power. S312: Calculate the Euclidean distance between the attribute values ​​of the device to be grouped in each of the above dimensions and the central values ​​of the existing device groups in the cloud; S313: If the similarity exceeds the preset threshold, the device will be assigned to the existing device group; otherwise, a new device group will be created with the device as the core.

4. The real-time operation status identification device for a photovoltaic power station according to claim 2, characterized in that: The specific optimization of population parameters is as follows: S331: Determine the mathematical model for the basic environmental response curve, which is a function of ambient temperature and solar irradiance as independent variables and theoretical power generation as the dependent variable. The initial form of the mathematical model for the basic environmental response curve is as follows: ,in Theoretical power generation capacity, For ambient temperature, Solar irradiance, These are the model parameters to be corrected; S332: Input the pre-processed data collected from all equipment in the equipment group into the mathematical model, use the actual power generation as the observed value, and use the weighted least squares method to globally fit and optimize the model parameters; S333: The set of parameters with the highest goodness of fit is determined as the corrected environmental response parameters.

5. The real-time operation status identification device for a photovoltaic power station according to claim 1, characterized in that: The calculation and correction of the equipment aging curve in S4 specifically includes: S41: Long-term performance data reception, continuously receiving and storing long-term working data packets periodically uploaded by the edge after the initial runtime window; S42: Theoretical performance calculation, based on the operating condition information in the long-term working data package, calculates the theoretical power generation using the corrected environmental response curve; S43: Performance Deviation Attribution, calculate and analyze the deviation between actual power and theoretical power, and quantify it as an aging index; S44: Individual aging curve correction, based on time series data of aging indicators, corrects the basic aging curve of the equipment for that individual. S45: Optimize the aging pattern of the group, aggregate and analyze the corrected aging curves of multiple individuals within the same equipment group, optimize the aging model parameters and distribute them.

6. The real-time operation status identification device for a photovoltaic power station according to claim 5, characterized in that: The theoretical performance calculation specifically includes: S421: Extract the average ambient temperature, average solar irradiance, and operating mode identifier from the long-term monitoring data packet during the data acquisition period; S422: Select the corresponding modified environmental response curve model based on the operating mode identifier; S423: Substitute the average ambient temperature and average solar irradiance into the selected modified environmental response curve model to calculate the theoretical power generation under this environmental condition without considering aging.

7. The real-time operation status identification device for a photovoltaic power station according to claim 5, characterized in that: The optimization of population aging patterns specifically includes: S451: Retrieve the updated individual aging parameter set of all devices in the same device group from the cloud server; S452: Calculate the statistical characteristic values ​​for each group of aging parameters, including the mean, median, and specific quantiles; S453: The parameter set regenerated by regression of statistical characteristic values ​​is determined as the optimized group aging parameter for this equipment group; S454: The optimized group aging parameters are bound to the corresponding device group identifier and sent to the edge controllers of all devices in the group for subsequent local aging status estimation and lifespan prediction.

8. The real-time operation status identification device for a photovoltaic power station according to claim 1, characterized in that: The dynamic calculation of optimal state identification and control strategies in A3 specifically includes: A31: User-side energy consumption habit learning modeling, establishing a quantitative model of energy demand based on historical grid-connected data or user-side electricity consumption data; A32: Future power generation and energy demand forecasting, combining user habit models, real-time system status and environmental information, to predict the power generation curve and energy demand curve for a specified period in the future; A33: Multi-objective optimization solution, based on power generation prediction curve, energy demand curve and external constraints, to solve the control parameter sequence that meets the demand and is comprehensively optimal. The multi-objective optimization simultaneously considers maximizing power generation efficiency, minimizing power fluctuation and protecting equipment life. A34: Control command generation and issuance, which converts the optimized control parameter sequence into executable control commands to drive the photovoltaic inverter.

9. The real-time operation status identification device for a photovoltaic power station according to claim 1, characterized in that: Between steps A2 and A3 executed by the edge controller, there is also a local aging statistics step, specifically: A311: Aging parameter reception and update, receiving, verifying and storing optimized population aging parameters issued by the cloud server; A312: Real-time aging status calculation, based on optimized group aging parameters and local real-time monitoring data, periodically calculates the current health index of the equipment; A313: The control strategy is adaptively adjusted, using the health index as a key input to dynamically adjust the equipment's operating parameters and protection strategies; A314: Aging data feedback, compresses key aging statistics from the local system and periodically uploads them to the cloud for continuous model optimization.

10. A real-time operation status identification device for a photovoltaic power station according to any one of claims 1-9, characterized in that: The device forms a complete optimization closed loop: The edge controller uploads the collected initial and long-term operating data to the cloud server; The cloud server uses group data to model and correct environmental response curves and aging curves, and then distributes the parameters. The edge controller uses the distributed model parameters, local real-time data, and user-side energy demand to perform personalized and efficient control. The newly generated data is uploaded to the cloud again for continuous iterative optimization of the model.

Images