Power generation anomaly early warning system and method for photovoltaic power generation device

Abstract

The application relates to the technical field of photovoltaic power generation, and discloses a power generation anomaly early warning system and method for a photovoltaic power generation device, which comprises the following steps: acquiring real-time operation parameters and real-time environmental parameters, combining historical data to construct a dynamic reference power model to output a theoretical reference power, calculating a power generation residual according to the real-time output power and the theoretical reference power, comparing the power generation residual with a dynamic residual threshold to mark a preliminary abnormal device, performing space-time correlation verification on the preliminary abnormal device to confirm or cancel an abnormal early warning and output an early warning signal, and feeding back the early warning signal and residual data to the dynamic reference power model for online updating. The system comprises a collection module, a preprocessing module, a historical database, a model construction module, a residual calculation module, a threshold determination module, an abnormality identification module, a correlation verification module, an early warning module and an updating module. The application can effectively distinguish environmental fluctuations from device failures, and reduce the false alarm rate and the missed alarm rate.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and in particular to a power generation anomaly early warning system and method for photovoltaic power generation equipment. Background Technology

[0002] Photovoltaic power generation, as a clean and renewable energy source, plays an increasingly important role in the global energy structure transformation. With the continuous expansion of photovoltaic power plant installed capacity and the widespread deployment of distributed photovoltaic systems, the operational safety and power generation efficiency of photovoltaic power generation systems have received increasing attention. During long-term operation, photovoltaic power generation equipment is inevitably affected by various factors such as component aging, hot spot effects, loose connectors, and inverter failures, leading to abnormal conditions such as decreased power generation. If these abnormalities are not detected and addressed in a timely manner, they can not only cause power generation losses but also potentially lead to more serious equipment damage or even safety accidents. Therefore, providing early warning of power generation anomalies for photovoltaic power generation equipment is of great significance for ensuring the safe and stable operation of photovoltaic power plants and improving power generation efficiency.

[0003] Currently, anomaly warning methods for photovoltaic (PV) power generation equipment mainly include methods based on fixed thresholds and methods based on predictive model comparisons. Methods based on fixed thresholds set upper and lower limits for current, voltage, or power; when monitored data exceeds these limits, an anomaly is identified. However, the output power of PV modules is significantly affected by meteorological factors such as solar radiation intensity, ambient temperature, and wind speed, resulting in substantial differences in output power under different environmental conditions. Fixed threshold methods struggle to adapt to complex and variable actual operating environments, easily leading to false alarms or missed alarms. Methods based on predictive model comparisons construct PV power generation prediction models, compare predicted power generation with actual power generation, and identify abnormal modules based on the deviation. While this method overcomes the limitation of fixed thresholds in ignoring environmental influences to some extent, the accuracy of its prediction models highly depends on the quality and coverage of the training data. In extreme weather or drastic changes in environmental conditions, prediction deviations are significant, and it is also difficult to accurately distinguish between normal power generation fluctuations caused by environmental factors and power generation anomalies caused by equipment malfunctions. In addition, existing studies have attempted to distinguish between real and false faults by constructing power fluctuation maps and performing cluster analysis. However, this method mainly relies on power fluctuation characteristics for judgment, and the quantitative analysis of the combined effects of multiple environmental factors is still insufficient. It is still difficult to guarantee the accuracy of early warning in complex scenarios such as rapid changes in illumination and cloudy weather.

[0004] Therefore, how to effectively distinguish between normal power generation fluctuations caused by environmental factors and power generation anomalies caused by equipment malfunctions, and improve the accuracy and reliability of abnormal early warning for photovoltaic power generation equipment, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a power generation anomaly early warning system and method for photovoltaic power generation equipment, which solves the problem of low early warning accuracy and high false alarm rate in the prior art due to the difficulty in accurately distinguishing between normal power generation fluctuations caused by environmental factors and power generation anomalies caused by equipment malfunctions.

[0006] To achieve the above objectives, the present invention provides a method for early warning of abnormal power generation in photovoltaic power generation equipment, comprising the following steps: The real-time operating parameters and real-time environmental parameters of the photovoltaic power generation equipment are obtained. The real-time operating parameters include output power, and the real-time environmental parameters include solar irradiance. Obtain historical operating parameters and historical environmental parameters of photovoltaic power generation equipment under normal operating conditions; Based on the mapping relationship between the historical environmental parameters and the historical output power in the historical operating parameters, a dynamic reference power model is constructed. The dynamic reference power model is used to output the corresponding theoretical reference power based on the input real-time environmental parameters. Calculate the power generation residual based on the real-time output power in the real-time operating parameters and the theoretical reference power; A dynamic residual threshold is determined based on the distribution of power generation residuals under historical normal operating conditions. The power generation residuals are compared with the dynamic residual threshold. When the power generation residuals exceed the dynamic residual threshold, the corresponding photovoltaic power generation equipment is marked as a preliminary abnormal equipment. Spatiotemporal correlation verification is performed on the preliminary abnormal equipment. The spatiotemporal correlation verification includes at least one of spatial correlation verification and temporal correlation verification. The spatial correlation verification is used to determine the source of interference based on the synchronicity of the power generation residuals of multiple photovoltaic power generation equipment in the same area. The temporal correlation verification is used to eliminate instantaneous interference based on the persistence of the power generation residuals of a single photovoltaic power generation equipment within a preset time period. Based on the results of the spatiotemporal correlation verification, the abnormal warning for the preliminary abnormal device is confirmed or cancelled, and a warning signal is output.

[0007] Specifically, a dynamic reference power model is constructed based on the mapping relationship between the historical environmental parameters and the historical output power in the historical operating parameters, including: The dynamic reference power model introduces a temperature correction coefficient and a wind speed correction coefficient to perform nonlinear compensation on the theoretical reference power. The temperature correction coefficient is used to characterize the degree of influence of the ambient temperature deviating from the standard temperature on the power generation efficiency of the photovoltaic module, and the wind speed correction coefficient is used to characterize the degree of influence of wind speed on the heat dissipation effect and power generation efficiency of the photovoltaic module.

[0008] The calculation of power generation residuals specifically includes: The power generation residual includes absolute power residual and relative power residual.

[0009] Among them, determining the dynamic residual threshold based on the distribution of power generation residuals under historical normal operating conditions specifically includes: Historical environmental parameters are divided into multiple subsets according to irradiance intervals, and the dynamic residual thresholds corresponding to each irradiance interval are determined based on the statistical distribution of power generation residuals within each subset.

[0010] The specific steps of the spatial association verification are as follows: Obtain the real-time residual power generation of other photovoltaic power generation devices located in the same photovoltaic power generation area as the initially abnormal device. If multiple photovoltaic power generation devices in the same area show an abnormal trend in residual power generation at the same time period, it is determined to be a normal fluctuation caused by environmental factors and the abnormal warning is cancelled. If only the initially abnormal device shows an abnormal residual power generation while other photovoltaic power generation devices in the same area are normal, it is determined to be a fault of the device itself and the abnormal warning is confirmed.

[0011] The specific steps of the time correlation verification are as follows: The trend of the power generation residual change of the preliminarily abnormal equipment over a preset time period is obtained. If the power generation residual abnormality is sudden and persistent, it is judged as equipment failure and an abnormality warning is confirmed. If the power generation residual abnormality fluctuates briefly and then returns to the normal range, it is judged as instantaneous interference from environmental factors and the abnormality warning is canceled.

[0012] Specifically, the process of confirming or canceling the anomaly warning for the preliminary abnormal device based on the spatiotemporal correlation verification result, and outputting a warning signal, includes: When the spatiotemporal correlation verification confirms a device malfunction, a Level 1 warning is output; when the spatiotemporal correlation verification cannot confirm a device malfunction and the possibility of a device malfunction cannot be ruled out, a Level 2 warning is output.

[0013] After confirming or canceling the anomaly warning for the preliminary abnormal device based on the result of the spatiotemporal correlation verification and outputting the warning signal, the process further includes: The early warning signal and the corresponding power generation residual data are fed back to the dynamic reference power model, and the dynamic reference power model is updated online.

[0014] The power generation anomaly early warning system for photovoltaic power generation equipment includes a data acquisition module, a preprocessing module, a historical database, a model building module, a residual calculation module, a threshold determination module, an anomaly identification module, a correlation verification module, an early warning module, and an update module. The acquisition module is connected to the preprocessing module, the preprocessing module is connected to the model building module and the residual calculation module, the historical database is connected to the model building module, the model building module is connected to the residual calculation module, the residual calculation module is connected to the anomaly identification module and the update module, the threshold determination module is connected to the anomaly identification module, the anomaly identification module is connected to the correlation verification module, the correlation verification module is connected to the early warning module, the early warning module is connected to the update module, and the update module is connected to the model building module. The acquisition module is used to acquire real-time operating parameters and real-time environmental parameters of the photovoltaic power generation equipment; The preprocessing module is used to perform time alignment and outlier removal on the collected data; The historical database is used to store historical operating parameters and historical environmental parameters of photovoltaic power generation equipment under normal historical operating conditions; The model building module is used to build a dynamic reference power model based on the mapping relationship between historical environmental parameters and historical output power. The dynamic reference power model is used to output the corresponding theoretical reference power based on the input real-time environmental parameters. The residual calculation module is used to calculate the power generation residual based on the real-time output power and the theoretical reference power. The threshold determination module is used to determine the dynamic residual threshold based on the distribution of power generation residuals under historical normal operating conditions. The anomaly identification module is used to compare the power generation residual with the dynamic residual threshold. When the power generation residual exceeds the dynamic residual threshold, the corresponding photovoltaic power generation equipment is marked as a preliminary abnormal equipment. The correlation verification module is used to perform spatiotemporal correlation verification on the preliminary abnormal equipment to confirm or cancel the abnormal warning; The early warning module is used to output an early warning signal based on the verification result of the associated verification module; The update module is used to update the dynamic reference power model online based on the early warning signal output by the early warning module and the corresponding power generation residual data.

[0015] This invention discloses a power generation anomaly early warning system and method for photovoltaic power generation equipment. By acquiring real-time operating parameters and real-time environmental parameters of the photovoltaic power generation equipment and combining them with historical normal operation data, a dynamic reference power model integrating environmental parameter mapping relationships is constructed to output a theoretical reference power. The power generation residual is calculated based on the real-time output power and the theoretical reference power. This residual is compared with a dynamic residual threshold determined based on historical residual distribution to mark initially abnormal equipment. Spatial and temporal correlation verifications are then performed on these initially abnormal equipment to eliminate normal fluctuations or transient interference caused by environmental factors and confirm equipment failure. Finally, a graded early warning signal is output based on the verification results. Simultaneously, the early warning signal and power generation residual data are fed back to the dynamic reference power model for online updates. This invention can effectively distinguish between normal power generation fluctuations caused by environmental factors and power generation anomalies caused by equipment malfunctions, significantly reducing false alarm and false negative rates, and improving the accuracy and reliability of photovoltaic power generation equipment anomaly early warning. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0017] Figure 1 This is a flowchart of the steps of the power generation anomaly early warning method for photovoltaic power generation equipment according to the first embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the power generation anomaly early warning system for photovoltaic power generation equipment according to the second embodiment of the present invention.

[0019] In the diagram: 201-Acquisition module, 202-Preprocessing module, 203-Historical database, 204-Model building module, 205-Residual calculation module, 206-Threshold determination module, 207-Anomaly identification module, 208-Correlation verification module, 209-Early warning module, 210-Update module. Detailed Implementation

[0020] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0021] First embodiment: Please refer to Figure 1 This invention provides a method for early warning of abnormal power generation in photovoltaic power generation equipment, comprising the following steps: S101: Obtain real-time operating parameters and real-time environmental parameters of the photovoltaic power generation equipment, wherein the real-time operating parameters include output power and the real-time environmental parameters include solar irradiance; Specifically, the output power of photovoltaic (PV) power generation equipment is affected by various factors during operation. Data acquisition devices deployed in PV power plants are used to acquire real-time operating and environmental parameters of the PV power generation equipment. Operating parameters include at least the output power of the PV strings, and may further include output voltage and output current. Environmental parameters include at least solar irradiance, and may further include ambient temperature and wind speed. Solar irradiance is the most significant factor affecting the output power of PV modules; the short-circuit current of PV modules is essentially proportional to irradiance. Therefore, accurate irradiance data acquisition is crucial for subsequent power baseline calculations.

[0022] In practical implementation, solar irradiance data can be collected using an irradiance sensor, ambient temperature data using a temperature sensor, and wind speed data using a wind speed sensor. Operating parameters can be obtained through the monitoring unit inside the photovoltaic inverter or combiner box, or through the power acquisition module 201 deployed in each photovoltaic string. The acquisition frequency can be set according to actual needs, for example, once every 5 minutes, which can be increased to once every 1 minute during periods of drastic weather changes. The acquired data must be timestamped for subsequent time alignment processing. The selection of the aforementioned acquisition equipment and the setting of the acquisition frequency can be adjusted according to the actual scale of the photovoltaic power station and monitoring requirements.

[0023] S102: Obtain historical operating parameters and historical environmental parameters of photovoltaic power generation equipment under normal operating conditions; Specifically, to establish a benchmark model that reflects the normal power generation characteristics of the equipment, it is necessary to obtain operational data and corresponding environmental data of the photovoltaic power generation equipment under historical normal operating conditions. Historical normal operating conditions refer to the period during which the equipment has not experienced any faults and its power generation performance is normal. In practice, historical normal operating periods can be determined through manual screening or automatic identification. For example, operational data confirmed by maintenance personnel to be fault-free within the first full year after the photovoltaic power station is put into operation can be selected as historical normal data, or the remaining data can be used as historical normal data after identifying and eliminating abnormal data points through statistical methods.

[0024] Historical data should cover a variety of typical weather conditions, including sunny, cloudy, overcast weather, and periods of rapid change in irradiance in the morning and evening, to ensure that the subsequently established dynamic baseline power model has good adaptability under various environmental conditions. The parameter types for historical data acquisition should be consistent with those for real-time data, i.e., at least historical output power and historical environmental parameters should be included. The storage period for historical data can be determined based on actual conditions, but it is generally recommended to be no less than 3 months to cover a sufficient number of weather types and operating conditions. The acquired historical data is stored in the historical database 203 for use in subsequent model construction.

[0025] S103: Based on the mapping relationship between the historical environmental parameters and the historical output power in the historical operating parameters, a dynamic reference power model is constructed. The dynamic reference power model is used to output the corresponding theoretical reference power based on the input real-time environmental parameters. Specifically, the dynamic reference power model function outputs the corresponding theoretical reference power based on the input real-time environmental parameters, that is, the normal power value that the photovoltaic power generation equipment should output under the given environmental conditions. This model is built upon historical normal operation data and reflects the inherent mapping relationship between environmental parameters and output power when the equipment is in a healthy state.

[0026] In practical implementation, the dynamic reference power model can be implemented using various modeling methods. For example, a multiple linear regression method can be used, with historical solar irradiance, historical ambient temperature, historical wind speed, and other environmental parameters as input variables and historical output power as the dependent variable, to fit a linear regression equation between environmental parameters and output power. However, the output power of photovoltaic modules is not strictly linearly related to environmental parameters. Under low irradiance conditions, the power generation efficiency of photovoltaic modules is often lower than that under high irradiance conditions; increased ambient temperature leads to increased operating temperature of photovoltaic modules, which in turn reduces open-circuit voltage and output power; wind speed indirectly affects power generation efficiency by influencing the heat dissipation effect of the modules. Therefore, preferably, the dynamic reference power model of this invention introduces temperature correction coefficients and wind speed correction coefficients to perform nonlinear compensation on the reference power obtained based on irradiance.

[0027] The temperature correction factor is introduced as follows: A standard temperature is set, and the power generation efficiency of the photovoltaic module changes when the ambient temperature deviates from the standard temperature. The temperature correction factor is used to quantify the degree of this change, and its value can be determined through historical data analysis or module factory parameters. Generally, for every certain increase in ambient temperature, the output power of the crystalline silicon photovoltaic module will decrease accordingly. In actual modeling, the temperature correction factor can be set as a function of the ambient temperature. The wind speed correction factor is introduced similarly. Wind speed affects the convective heat dissipation on the module surface, changing the module's operating temperature and thus affecting the output power. Under low wind speed conditions, the module's heat dissipation is poor, the operating temperature is high, and the power generation efficiency is reduced; under higher wind speed conditions, heat dissipation improves, and the power generation efficiency is improved. The wind speed correction factor can be determined by fitting the relationship between wind speed and power generation efficiency from historical data.

[0028] Taking the multiple regression method as an example, the dynamic baseline power model can be expressed as follows: the theoretical baseline power is determined by a base power value based on irradiance, a temperature correction factor, and a wind speed correction factor. The base power value is directly proportional to irradiance, the temperature correction factor is a function of ambient temperature, and the wind speed correction factor is a function of wind speed. The specific forms of these functions can be determined through regression analysis or curve fitting of historical data.

[0029] In another embodiment, the dynamic baseline power model can also be implemented using a neural network model. Historical environmental parameters are used as input layer features, and historical output power is used as output layer labels, resulting in a multi-layer neural network model through training. The number of hidden layers and nodes in the neural network can be determined based on the amount of data and the required modeling accuracy. Unlike pure black-box models, the neural network model of this invention can introduce a physical constraint layer after the input layer, embedding the physical laws of temperature correction coefficients and wind speed correction coefficients into the model structure, giving the model both data-driven fitting capabilities and physical interpretability.

[0030] Regardless of the modeling method used, the dynamic reference power model can be put into online use once it is built. When real-time environmental parameters are input into the model, the model outputs the theoretical reference power corresponding to the current environmental conditions.

[0031] S104: Calculate the power generation residual based on the real-time output power in the real-time operating parameters and the theoretical reference power; Specifically, the power generation residual is a key indicator reflecting the difference between the actual power generation state and the theoretical normal state of the equipment. After obtaining the theoretical reference power output from the dynamic reference power model, the actual output power collected in real time is compared with the theoretical reference power to calculate the power generation residual.

[0032] Power generation residuals include at least two forms: absolute power residuals and relative power residuals. Absolute power residuals reflect the absolute difference between actual power and the theoretical reference power. Relative power residuals reflect the ratio of the absolute residual to the theoretical reference power. These two types of residuals characterize the degree of power generation anomalies from different dimensions. Absolute residuals are suitable for judging whether the absolute magnitude of power decline reaches a warning level, while relative residuals are suitable for uniformly measuring the degree of anomalies under different power levels. For example, under low irradiance conditions, the absolute residual may be small, but the relative residual may be large. In this case, the relative residual can more sensitively reflect the relative decline in the equipment's power generation efficiency.

[0033] In practical implementation, the rate of change of power residuals can be further calculated, i.e., the trend of residual change over time. The rate of change of residuals helps to determine whether an anomaly is sudden or gradual. Sudden anomalies cause the residuals to increase sharply in a short period of time, while gradual anomalies are characterized by a slow increase in residuals over time. Multidimensional residual feature vectors provide richer discriminative information for subsequent anomaly identification.

[0034] The residual calculation is performed at a preset time period, such as once every 5 minutes. The calculated residual data is stored together with the corresponding timestamp for use in subsequent steps.

[0035] S105: Determine the dynamic residual threshold based on the distribution of power generation residual under historical normal operating conditions, compare the power generation residual with the dynamic residual threshold, and when the power generation residual exceeds the dynamic residual threshold, mark the corresponding photovoltaic power generation equipment as a preliminary abnormal equipment. Specifically, after obtaining the real-time power generation residual, it is necessary to determine whether the residual has reached an abnormal level. This invention uses a dynamic residual threshold instead of a fixed threshold for judgment to overcome the poor adaptability of the fixed threshold method under different environmental conditions.

[0036] The method for determining the dynamic residual threshold is as follows: First, calculate the historical power generation residual using environmental parameters and output power data under historical normal operating conditions to obtain the residual distribution under historical normal operating conditions. Then, divide the historical environmental parameters into multiple subsets according to irradiance intervals. The division method of irradiance intervals can be determined according to actual conditions, for example, dividing irradiance from low to high into multiple intervals at certain intervals. For each irradiance interval, calculate the statistical distribution parameters of the historical power generation residual within that interval, including the mean and standard deviation. Based on the statistical distribution parameters, determine the dynamic residual threshold corresponding to that irradiance interval. For example, the threshold can be set as the mean plus or minus a preset multiple multiplied by the standard deviation, where the specific value of the preset multiple can be adjusted according to the requirements for false alarm rate and false negative rate.

[0037] Through the above method, different residual thresholds correspond to different irradiance ranges. Under high irradiance conditions, due to the large power base and the large residual fluctuation range, the threshold is correspondingly relaxed; under low irradiance conditions, the threshold is correspondingly tightened. This dynamic threshold mechanism enables anomaly detection to adaptively adjust with changes in environmental conditions, effectively avoiding the problems of fixed thresholds being prone to missed detections under low irradiance conditions and prone to false alarms under high irradiance conditions.

[0038] During real-time assessment, the irradiance range to which the current real-time environmental parameters belong is first determined. Then, the dynamic residual threshold corresponding to this range is obtained, and the real-time power generation residual is compared with this threshold. When the power generation residual exceeds the corresponding dynamic residual threshold, the corresponding photovoltaic power generation equipment is marked as a preliminary abnormal device. It should be noted that this preliminary marking means that the marking still needs to undergo subsequent spatiotemporal correlation verification to finally confirm whether it is a real equipment failure, in order to avoid misjudging normal fluctuations caused by environmental factors as equipment abnormalities.

[0039] S106: Perform spatiotemporal correlation verification on the preliminary abnormal equipment. The spatiotemporal correlation verification includes at least one of spatial correlation verification and temporal correlation verification. The spatial correlation verification is used to determine the source of interference based on the synchronicity of the power generation residuals of multiple photovoltaic power generation equipment in the same area. The temporal correlation verification is used to eliminate instantaneous interference based on the persistence of the power generation residuals of a single photovoltaic power generation equipment within a preset time period. Specifically, equipment marked as initially abnormal may fall into two categories: first, the equipment itself has indeed malfunctioned, leading to an abnormal decrease in power generation; second, environmental factors have caused a temporary decrease in power, but the equipment itself is not faulty. Spatiotemporal correlation verification is used to distinguish between these two scenarios, which includes at least one of spatial correlation verification and temporal correlation verification.

[0040] The specific implementation method of spatial correlation verification is as follows: obtain the real-time power generation residuals of other photovoltaic power generation devices located in the same photovoltaic power generation area as the initially abnormal device. The same photovoltaic power generation area refers to a set of photovoltaic devices that are geographically adjacent and affected by the same environmental conditions. If multiple photovoltaic power generation devices in the same area show an abnormal trend in power generation residuals at the same time period, it indicates that the anomaly is likely a normal fluctuation caused by regional environmental factors, and the anomaly warning should be cancelled. Conversely, if only the initially abnormal device shows abnormal power generation residuals, while other photovoltaic power generation devices in the same area are normal, it indicates that the anomaly is caused by a fault in the device itself, and the anomaly warning should be confirmed.

[0041] In practice, spatial correlation verification can be achieved by calculating the correlation coefficient between the residual sequences of the initially abnormal equipment and surrounding equipment. If the correlation coefficient is higher than a preset threshold, it is judged as a synchronous change, belonging to environmental factors; if the correlation coefficient is lower than the preset threshold, it is judged as an independent change, belonging to equipment failure. The number of other photovoltaic power generation equipment involved in spatial correlation verification can be determined according to the actual situation.

[0042] The specific implementation method of time-related verification is as follows: Obtain the trend of power generation residual changes of the initially abnormal equipment over a preset time period. The preset time period can be set according to actual conditions. If the abnormal power generation residual occurs suddenly and persists, it is determined that the equipment has experienced a continuous fault, and an abnormality warning should be confirmed. Conversely, if the abnormal power generation residual fluctuates briefly and then quickly returns to the normal range, it is determined to be a momentary disturbance from environmental factors, and the abnormality warning should be cancelled.

[0043] In practice, time-related verification can be achieved by analyzing the duration and rate of change of the residual time series. If the duration of the residual exceeding the threshold exceeds a preset duration and the rate of change of the residual remains negative, it is judged as a persistent fault; if the duration of the residual exceeding the threshold is short and the rate of change of the residual subsequently turns positive, it is judged as a transient disturbance.

[0044] Spatial correlation verification and temporal correlation verification can be used individually or in combination. Preferably, the two verification methods are used in combination to more comprehensively eliminate the interference of environmental factors. Spatial correlation verification eliminates regional environmental factors, while temporal correlation verification eliminates instantaneous environmental factors. The two complement each other and further improve the accuracy of early warning.

[0045] S107: Based on the results of the spatiotemporal correlation verification, confirm or cancel the abnormal warning for the preliminary abnormal device, and output a warning signal.

[0046] Specifically, based on the spatiotemporal correlation verification results of step S106, a final judgment is made on the preliminary abnormal equipment and a corresponding early warning signal is output.

[0047] When spatiotemporal correlation verification confirms a device malfunction, a Level 1 warning is issued. A Level 1 warning indicates a confirmed device malfunction, requiring immediate intervention from operations and maintenance personnel. When spatiotemporal correlation verification cannot confirm a device malfunction and the possibility of a malfunction cannot be ruled out, a Level 2 warning is issued. A Level 2 warning indicates a suspected device malfunction, and manual verification by operations and maintenance personnel is recommended. When spatiotemporal correlation verification determines that the malfunction is due to normal fluctuations caused by environmental factors, no warning is issued, but the relevant data can be recorded for subsequent model optimization.

[0048] Early warning signals can be output in various ways, such as displaying warning information on the monitoring system interface of the photovoltaic power station, notifying operation and maintenance personnel via SMS or email, or issuing alarms via audible and visual alarm devices. Warning information should at least include the number or location of the abnormal equipment, the time of the abnormality, the type of abnormality, and related residual data.

[0049] To further improve the accuracy and adaptability of the dynamic reference power model in long-term operation, while outputting early warning signals, the early warning signals and corresponding power generation residual data are fed back to the dynamic reference power model as new training samples to update the model online.

[0050] Specifically, for data whose fluctuations are confirmed to be caused by environmental factors through spatiotemporal correlation verification, these data are added to the historical database 203 as new normal operation samples for incremental training of the model, enabling the model to output a more accurate theoretical baseline power when facing similar environmental conditions in the future. For data confirmed to be equipment failures, these data are marked as anomalous samples and are not used for model training to avoid contaminating the model with failure data.

[0051] Online updates can be achieved using incremental learning, which fine-tunes the model parameters using new samples without retraining the entire model. Incremental learning offers advantages such as low computational cost and fast update speed, making it suitable for real-time operation scenarios in photovoltaic power plants. The frequency of online updates can be set according to actual conditions. Through online updates, the dynamic reference power model can continuously adapt to long-term trends such as photovoltaic module aging and environmental changes, maintaining the long-term effectiveness of the early warning system.

[0052] Second embodiment: Please refer to Figure 2This invention provides a power generation anomaly early warning system for photovoltaic power generation equipment, including a data acquisition module 201, a preprocessing module 202, a historical database 203, a model building module 204, a residual calculation module 205, a threshold determination module 206, an anomaly identification module 207, a correlation verification module 208, an early warning module 209, and an update module 210. The acquisition module 201 is connected to the preprocessing module 202. The preprocessing module 202 is connected to the model building module 204 and the residual calculation module 205. The historical database 203 is connected to the model building module 204. The model building module 204 is connected to the residual calculation module 205. The residual calculation module 205 is connected to the anomaly identification module 207 and the update module 210. The threshold determination module 206 is connected to the anomaly identification module 207. The anomaly identification module 207 is connected to the association verification module 208. The association verification module 208 is connected to the early warning module 209. The early warning module 209 is connected to the update module 210. The update module 210 is connected to the model building module 204.

[0053] Specifically, the acquisition module 201 acquires the operating parameters and environmental parameters of the photovoltaic power generation equipment in real time and sends the acquired data to the preprocessing module 202. The preprocessing module 202 performs time alignment and outlier removal on the received data, and sends the processed real-time data to the model building module 204 and the residual calculation module 205, respectively. The historical database 203 stores the operating parameters and environmental parameters of the photovoltaic power generation equipment under historical normal operating conditions. The model building module 204 reads historical data from the historical database 203 and constructs a dynamic reference power model based on the mapping relationship between historical environmental parameters and historical output power. This model can output the corresponding theoretical reference power based on the input real-time environmental parameters. The residual calculation module 205 receives the real-time output power sent by the preprocessing module 202 and the theoretical reference power sent by the model building module 204, and calculates the power generation residual. The threshold determination module 206 determines the dynamic residual threshold corresponding to each irradiance interval based on the historical data in the historical database 203. The anomaly identification module 207 receives the power generation residual sent by the residual calculation module 205 and the dynamic residual threshold sent by the threshold determination module 206. It compares the power generation residual with the dynamic residual threshold. When the power generation residual exceeds the dynamic residual threshold, the corresponding photovoltaic power generation equipment is marked as a preliminary anomaly, and the marking result is sent to the association verification module 208. The association verification module 208 performs spatiotemporal association verification on the preliminary anomaly equipment to confirm or cancel the anomaly warning, and sends the verification result to the warning module 209. The warning module 209 outputs a corresponding warning signal based on the verification result, and simultaneously sends the warning signal and the corresponding power generation residual data to the update module 210. The update module 210 updates the dynamic reference power model online based on the received warning signal and power generation residual data, and feeds back the updated model parameters to the model construction module 204.

[0054] The modules of the above system can be deployed as software modules on the monitoring server of the photovoltaic power station, or they can be implemented as dedicated early warning devices using a combination of hardware and software. Data communication between modules can be achieved through wired or wireless networks. The above module division is only for descriptive convenience. In actual implementation, the functions of multiple modules can be integrated into one unit, or the functions of one module can be further divided into multiple sub-modules.

[0055] This invention, by constructing a dynamic benchmark power model and combining it with multidimensional residual analysis and spatiotemporal correlation dual verification, can effectively distinguish between normal power generation fluctuations caused by environmental factors and power generation anomalies caused by equipment malfunctions. This significantly reduces the false alarm rate and missed alarm rate of the early warning system, and improves the accuracy and reliability of abnormal early warning for photovoltaic power generation equipment.

[0056] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A method for early warning of abnormal power generation in photovoltaic power generation equipment, characterized in that, Includes the following steps: The real-time operating parameters and real-time environmental parameters of the photovoltaic power generation equipment are obtained. The real-time operating parameters include output power, and the real-time environmental parameters include solar irradiance. Obtain historical operating parameters and historical environmental parameters of photovoltaic power generation equipment under normal operating conditions; Based on the mapping relationship between the historical environmental parameters and the historical output power in the historical operating parameters, a dynamic reference power model is constructed. The dynamic reference power model is used to output the corresponding theoretical reference power based on the input real-time environmental parameters. Calculate the power generation residual based on the real-time output power in the real-time operating parameters and the theoretical reference power; A dynamic residual threshold is determined based on the distribution of power generation residuals under historical normal operating conditions. The power generation residuals are compared with the dynamic residual threshold. When the power generation residuals exceed the dynamic residual threshold, the corresponding photovoltaic power generation equipment is marked as a preliminary abnormal equipment. Spatiotemporal correlation verification is performed on the preliminary abnormal equipment. The spatiotemporal correlation verification includes at least one of spatial correlation verification and temporal correlation verification. The spatial correlation verification is used to determine the source of interference based on the synchronicity of the power generation residuals of multiple photovoltaic power generation equipment in the same area. The temporal correlation verification is used to eliminate instantaneous interference based on the persistence of the power generation residuals of a single photovoltaic power generation equipment within a preset time period. Based on the results of the spatiotemporal correlation verification, the abnormal warning for the preliminary abnormal device is confirmed or cancelled, and a warning signal is output.

2. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, Based on the mapping relationship between the historical environmental parameters and the historical output power in the historical operating parameters, a dynamic reference power model is constructed, specifically including: The dynamic reference power model introduces a temperature correction coefficient and a wind speed correction coefficient to perform nonlinear compensation on the theoretical reference power. The temperature correction coefficient is used to characterize the degree of influence of the ambient temperature deviating from the standard temperature on the power generation efficiency of the photovoltaic module, and the wind speed correction coefficient is used to characterize the degree of influence of wind speed on the heat dissipation effect and power generation efficiency of the photovoltaic module.

3. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, The calculation of power generation residuals specifically includes: The power generation residual includes absolute power residual and relative power residual.

4. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, The dynamic residual threshold is determined based on the distribution of power generation residuals under historical normal operating conditions, specifically including: Historical environmental parameters are divided into multiple subsets according to irradiance intervals, and the dynamic residual thresholds corresponding to each irradiance interval are determined based on the statistical distribution of power generation residuals within each subset.

5. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, The specific steps for spatial association verification are as follows: Obtain the real-time residual power generation of other photovoltaic power generation devices located in the same photovoltaic power generation area as the initially abnormal device. If multiple photovoltaic power generation devices in the same area show an abnormal trend in residual power generation at the same time period, it is determined to be a normal fluctuation caused by environmental factors and the abnormal warning is cancelled. If only the initially abnormal device shows an abnormal residual power generation while other photovoltaic power generation devices in the same area are normal, it is determined to be a fault of the device itself and the abnormal warning is confirmed.

6. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, The specific steps for the time correlation verification are as follows: The trend of the power generation residual change of the preliminarily abnormal equipment over a preset time period is obtained. If the power generation residual abnormality is sudden and persistent, it is judged as equipment failure and an abnormality warning is confirmed. If the power generation residual abnormality fluctuates briefly and then returns to the normal range, it is judged as instantaneous interference from environmental factors and the abnormality warning is canceled.

7. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, Based on the results of the spatiotemporal correlation verification, the abnormal warning for the preliminary abnormal device is confirmed or canceled, and a warning signal is output, specifically including: When the spatiotemporal correlation verification confirms a device malfunction, a Level 1 warning is output; when the spatiotemporal correlation verification cannot confirm a device malfunction and the possibility of a device malfunction cannot be ruled out, a Level 2 warning is output.

8. The method for early warning of abnormal power generation in photovoltaic power generation equipment as described in claim 1, characterized in that, After confirming or canceling the anomaly warning for the preliminary abnormal device based on the result of the spatiotemporal correlation verification and outputting the warning signal, the method further includes: The early warning signal and the corresponding power generation residual data are fed back to the dynamic reference power model, and the dynamic reference power model is updated online.

9. A power generation anomaly early warning system for photovoltaic power generation equipment, used to implement the power generation anomaly early warning method for photovoltaic power generation equipment as described in claim 1, characterized in that, It includes a data acquisition module, a preprocessing module, a historical database, a model building module, a residual calculation module, a threshold determination module, an anomaly identification module, a correlation verification module, an early warning module, and an update module; The acquisition module is connected to the preprocessing module, the preprocessing module is connected to the model building module and the residual calculation module, the historical database is connected to the model building module, the model building module is connected to the residual calculation module, the residual calculation module is connected to the anomaly identification module and the update module, the threshold determination module is connected to the anomaly identification module, the anomaly identification module is connected to the correlation verification module, the correlation verification module is connected to the early warning module, the early warning module is connected to the update module, and the update module is connected to the model building module. The acquisition module is used to acquire real-time operating parameters and real-time environmental parameters of the photovoltaic power generation equipment; The preprocessing module is used to perform time alignment and outlier removal on the collected data; The historical database is used to store historical operating parameters and historical environmental parameters of photovoltaic power generation equipment under normal historical operating conditions; The model building module is used to build a dynamic reference power model based on the mapping relationship between historical environmental parameters and historical output power. The dynamic reference power model is used to output the corresponding theoretical reference power based on the input real-time environmental parameters. The residual calculation module is used to calculate the power generation residual based on the real-time output power and the theoretical reference power. The threshold determination module is used to determine the dynamic residual threshold based on the distribution of power generation residuals under historical normal operating conditions. The anomaly identification module is used to compare the power generation residual with the dynamic residual threshold. When the power generation residual exceeds the dynamic residual threshold, the corresponding photovoltaic power generation equipment is marked as a preliminary abnormal equipment. The correlation verification module is used to perform spatiotemporal correlation verification on the preliminary abnormal equipment to confirm or cancel the abnormal warning; The early warning module is used to output an early warning signal based on the verification result of the associated verification module; The update module is used to update the dynamic reference power model online based on the early warning signal output by the early warning module and the corresponding power generation residual data.

Images