Diagnostic methods, products, media and equipment for inefficient photovoltaic power plants

Abstract

This invention provides a method, product, medium, and equipment for diagnosing inefficiency in photovoltaic (PV) power plants, relating to the field of energy management technology. The method includes: collecting multi-source heterogeneous data from the PV power plant and preprocessing it; extracting clean data samples from the multi-source heterogeneous data based on preset meteorological conditions; inputting the preprocessed real-time operating data into a prediction model trained from the clean data samples to predict the expected power generation efficiency of the PV power plant under ideal clean conditions; screening each effective power generation period of the PV power plant based on preset irradiance conditions and obtaining the ratio of power to irradiance in each effective power generation period as the actual power generation efficiency; and obtaining the contamination diagnosis result of the PV power plant based on the actual power generation efficiency and the expected power generation efficiency. The solution of this invention, based on the preset irradiance conditions to screen the actual power generation efficiency within the effective power generation period, achieves accurate contamination diagnosis, thereby improving the accuracy of diagnosing inefficiency problems.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, and in particular to a diagnostic method, product, medium, and equipment for the inefficiency of photovoltaic power plants. Background Technology

[0002] With the large-scale application of photovoltaic power generation technology and the intelligent upgrading of power plant operation and maintenance, efficient and accurate diagnosis of low power generation efficiency in photovoltaic power plants has become a key aspect of improving power plant efficiency and reducing operation and maintenance costs. Currently, various data-driven photovoltaic power plant anomaly diagnosis solutions have emerged in the industry, mainly achieving intelligent assessment of power plant status by integrating multi-source monitoring data and combining machine learning models.

[0003] However, current diagnostic solutions often lack the construction and effective characterization of samples of the cleanliness status of photovoltaic power plants, making it difficult for models to accurately distinguish whether performance degradation is caused by dirt and grime, and thus failing to effectively assist in operation and maintenance decisions. In addition, the machine learning models used in current diagnostic solutions are mostly general time series or regression models, which have not been optimized for the dirt and grime diagnosis scenario of photovoltaic power plants. The models have insufficient interpretability and adaptability to operating conditions, resulting in low diagnostic capabilities and difficulty in accurately locating the inefficiency problems of photovoltaic power plants. Summary of the Invention

[0004] In view of the above problems, a diagnostic method, product, medium and equipment for overcoming or at least partially solving the inefficiency problem of photovoltaic power plants are proposed.

[0005] One objective of this invention is to improve the accuracy of diagnosing inefficiencies.

[0006] A further objective of this invention is to improve the ease and reliability of obtaining training data.

[0007] Another further objective of this invention is to improve the accuracy of contamination diagnosis results for photovoltaic power plants.

[0008] Specifically, this invention provides a method for diagnosing inefficiency in photovoltaic power plants. This method includes:

[0009] Collect multi-source heterogeneous data from photovoltaic power plants and preprocess the multi-source heterogeneous data, which includes historical operating data, real-time operating data and meteorological data.

[0010] Clean data samples are extracted from preprocessed multi-source heterogeneous data based on preset meteorological conditions.

[0011] The pre-processed real-time operating data is input into the pre-trained prediction model to predict the expected power generation efficiency of the photovoltaic power plant under ideal clean conditions. The prediction model is trained from clean data samples.

[0012] Based on preset irradiance conditions, each effective power generation period of the photovoltaic power station is screened, and the actual power generation efficiency of the photovoltaic power station in each effective power generation period is obtained. The actual power generation efficiency is the ratio of the power of the photovoltaic power station to the irradiance in the effective power generation period.

[0013] The pollution diagnosis results of the photovoltaic power station are obtained based on the actual power generation efficiency and the expected power generation efficiency.

[0014] The steps for obtaining the pollution diagnosis results of a photovoltaic power plant based on the actual power generation efficiency and the expected power generation efficiency include:

[0015] The confidence level is calculated based on the number of multiple effective power generation periods and the trend of actual power generation efficiency changing over time within multiple effective power generation periods;

[0016] The pollution rate of a photovoltaic power plant is obtained by calculating the ratio of the actual power generation efficiency to the predicted expected power generation efficiency during the latest effective power generation period.

[0017] Determine whether the confidence level is higher than a preset score threshold and whether the dirt rate is lower than a preset dirt threshold;

[0018] If so, then it is confirmed that there is dirt accumulation in the photovoltaic power station.

[0019] Optionally, the effective power generation period includes the effective day; and

[0020] The steps for calculating the confidence level based on the number of multiple effective power generation periods and the trend of actual power generation efficiency over time within those periods include:

[0021] Determine whether the number of valid days exceeds a preset threshold.

[0022] If it does not exceed, then the confidence level is determined to be 0;

[0023] If the actual power generation efficiency of each valid day is exceeded, the actual power generation efficiency of each valid day is arranged in chronological order to construct an analysis dataset. A linear fitting algorithm is then used to perform regression analysis on the analysis dataset to obtain the fitting slope of the actual power generation efficiency change.

[0024] The confidence level is calculated based on the number of effective days and the fitting slope of the actual power generation efficiency change.

[0025] Optionally, the preset irradiance conditions include irradiance greater than a preset irradiance threshold, and the preset irradiance threshold is set according to the irradiance standard for sunny days.

[0026] Optionally, the steps for screening each effective power generation period of a photovoltaic power station based on preset irradiance conditions include:

[0027] Based on the preprocessed multi-source heterogeneous data, all continuous time windows where the irradiance is continuously greater than the preset irradiance threshold are identified and extracted, and each continuous time window is taken as an effective power generation period.

[0028] Optionally, the prediction model includes a lightweight gradient boosting decision tree model; and the prediction model is obtained through the following steps:

[0029] Obtain the pre-built prediction model for photovoltaic power plants;

[0030] Using clean data samples as the training set, the prediction model is incrementally trained using a pre-defined gradient boosting decision tree algorithm, and an early stopping mechanism is used to prevent overfitting, so as to obtain the prediction model.

[0031] Optionally, after obtaining the pollution diagnosis results of the photovoltaic power plant, the diagnostic method for the inefficiency problem of the photovoltaic power plant may also include:

[0032] The threshold method was used sequentially to perform string disconnection detection, frequent start-stop detection, power limit detection, and power limit detection on the preprocessed multi-source heterogeneous data.

[0033] Based on the test results and the pollution diagnosis results, the causes of the inefficiency of the photovoltaic power station were diagnosed.

[0034] Generate and output diagnostic reports containing the reasons for inefficiency to assist in operational decision-making.

[0035] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the steps of the diagnostic method for the inefficiency problem of any of the above-described photovoltaic power plants.

[0036] According to another aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of the method for diagnosing the inefficiency problem of any of the above-described photovoltaic power plants.

[0037] According to another aspect of the present invention, a computer device is also provided, which includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for diagnosing the inefficiency problem of any of the above-described photovoltaic power plants.

[0038] The method for diagnosing inefficiency in photovoltaic power plants of this invention achieves comprehensive perception of the power plant's operating status and environmental factors by collecting multi-source heterogeneous data, including historical operating data, real-time operating data, and meteorological data. Simultaneously, this method extracts clean data samples based on preset meteorological conditions and trains a prediction model accordingly, achieving sample construction and effective characterization of the ideal clean state of the photovoltaic power plant. Furthermore, this method filters effective power generation periods by preset irradiance conditions, obtains the ratio of power to irradiance during each effective power generation period as the actual power generation efficiency, and compares the actual power generation efficiency with the expected power generation efficiency output by the prediction model. This achieves precise quantitative diagnosis of pollution problems in photovoltaic power plants, thereby improving the accuracy of inefficiency diagnosis and providing a reliable basis for subsequent operation and maintenance decisions.

[0039] Furthermore, the method for diagnosing photovoltaic power plant inefficiency in this invention calculates the confidence level based on the number of multiple effective power generation periods and the trend of actual power generation efficiency over time within those periods. It then calculates the contamination rate based on the actual power generation efficiency in the latest effective power generation period and the predicted expected power generation efficiency. By combining the confidence level and the contamination rate, the reliability of the contamination diagnosis results is dynamically assessed. Therefore, the method for diagnosing photovoltaic power plant inefficiency in this invention only determines that there is contamination accumulation in the photovoltaic power plant when the confidence level is higher than a preset score threshold and the contamination rate is lower than a preset contamination threshold. This effectively avoids misjudgments caused by short-term data fluctuations, abnormal weather, or insufficient data, thereby reducing the false alarm rate and false negative rate of contamination diagnosis and ensuring the accuracy of subsequent diagnostic results.

[0040] Furthermore, the method for diagnosing the inefficiency of photovoltaic power plants in this invention provides a clear technical standard for screening effective power generation periods by explicitly setting a preset irradiance condition that the irradiance is greater than a threshold set according to the irradiance standard for sunny days. This ensures that each screened effective power generation period is under a condition where the photovoltaic module has a high and stable photoelectric conversion efficiency, eliminating the influence of interference factors such as low irradiance on the diagnostic results. This improves the accuracy and interpretability of the contamination diagnosis results, and provides an intuitive and reliable reference for operation and maintenance decisions.

[0041] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0042] The following sections will describe some specific embodiments of the invention in a detailed manner by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0043] Figure 1 This is a control block diagram of a method for diagnosing the inefficiency problem of a photovoltaic power plant according to an embodiment of the present invention;

[0044] Figure 2 This is a flowchart of a control example of a method for diagnosing the inefficiency problem of a photovoltaic power plant according to an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of a computer program product according to an embodiment of the present invention;

[0046] Figure 4 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention;

[0047] Figure 5 This is a schematic block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0048] Those skilled in the art should understand that the embodiments described below are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. These partial embodiments are intended to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention. Based on the embodiments provided by the present invention, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of the present invention.

[0049] Figure 1 This is a control block diagram of a method for diagnosing inefficiencies in photovoltaic power plants according to an embodiment of the present invention. Figure 1 As shown, the method for diagnosing the inefficiency problem of photovoltaic power plants in this embodiment generally includes:

[0050] Step S102: Collect multi-source heterogeneous data from the photovoltaic power station and preprocess the multi-source heterogeneous data, which includes historical operating data, real-time operating data, and meteorological data.

[0051] Step S104: Extract clean data samples from the preprocessed multi-source heterogeneous data based on preset meteorological conditions.

[0052] Step S106: Input the pre-processed real-time operating data into the pre-trained prediction model to predict the expected power generation efficiency of the photovoltaic power station under ideal clean conditions. The prediction model is trained from clean data samples.

[0053] Step S108: Based on preset irradiance conditions, select each effective power generation period of the photovoltaic power station and obtain the actual power generation efficiency of the photovoltaic power station in each effective power generation period. The actual power generation efficiency is the ratio of the power of the photovoltaic power station to the irradiance in the effective power generation period.

[0054] Step S110: Obtain the pollution diagnosis results of the photovoltaic power station based on the actual power generation efficiency and the expected power generation efficiency.

[0055] In this embodiment, the photovoltaic power station may include photovoltaic modules, energy storage devices, inverters, strings of photovoltaic cells, and other equipment. Operational data may include equipment status data and operational data of each device, used to sense the system's operational status. Meteorological data may include historical weather data, current weather data, and future weather data for the geographical location of the photovoltaic power station. Specifically, weather data may include temperature, irradiance, wind speed, humidity, rainfall, cloud cover, etc.

[0056] In addition, the power generation efficiency in this embodiment refers to the ability of a photovoltaic power station to output electrical energy under a unit irradiance during the power generation process, which is a quantitative indicator for the diagnosis and judgment of dirt in the diagnosis of inefficiency problems of photovoltaic power stations.

[0057] Therefore, the diagnostic method of this invention, by collecting multi-source heterogeneous data including historical operating data, real-time operating data, and meteorological data, achieves a comprehensive perception of the power plant's operating status and environmental factors. Simultaneously, the diagnostic method of this invention extracts clean data samples based on preset meteorological conditions and trains a prediction model accordingly, realizing the sample construction and effective characterization of the ideal clean state of the photovoltaic power plant. Furthermore, the diagnostic method of this invention also filters effective power generation periods by preset irradiance conditions, obtains the ratio of power to irradiance of the photovoltaic power plant in each effective power generation period as the actual power generation efficiency, and compares the actual power generation efficiency with the expected power generation efficiency output by the prediction model. This achieves accurate quantitative diagnosis of the pollution problem of the photovoltaic power plant, thereby improving the accuracy of the diagnosis of inefficiency problems and providing a reliable basis for subsequent operation and maintenance decisions.

[0058] In some embodiments, the preprocessing step S102 for multi-source heterogeneous data may include data processing operations such as anomaly detection, data standardization, and data alignment. Specifically, anomaly detection may include automatically identifying and removing outliers and outliers from the collected time-series operational data (e.g., power, voltage, current) and meteorological data using a preset algorithm to ensure the overall quality and consistency of the dataset. For example, the preset algorithm may employ the 3σ principle or the Isolation Forest algorithm.

[0059] In addition, data standardization can include converting all timestamps to a standard time zone to eliminate time zone confusion caused by different data sources. Data alignment can include aligning data from different sources and with different collection frequencies using timestamps to ensure that data points correspond one-to-one during subsequent fusion analysis. For example, power data and meteorological data can be merged and aligned by timestamp. Specifically, power data and irradiance data can be merged and aligned by timestamp to obtain time-series data for power and irradiance.

[0060] Therefore, the diagnostic method of this invention, by systematically integrating key operations such as anomaly detection, data standardization and data alignment, provides a highly reliable and fusionable underlying data foundation for the entire diagnostic method, thereby ensuring the accuracy of inefficient problem diagnosis.

[0061] In some embodiments, step S104 may include the following steps: identifying meteorological events in the meteorological data where the rainfall exceeds a preset rainfall threshold; extracting historical operational data within a preset duration after the meteorological event ends to obtain a clean data sample. Specifically, the preset rainfall threshold can be set according to the rainfall standard for heavy rain. In one specific embodiment, the preset rainfall threshold can be selected from any value between 25 and 50 mm / day. For example, the preset rainfall threshold can be set to 25 mm / day. Furthermore, the preset duration can be obtained through prior experiments. In one specific embodiment, the preset duration can be selected from any value between 0.5 and 2 days. For example, the preset duration can be set to 1 day.

[0062] In other words, the diagnostic method of this invention can automatically mark data within a preset time period after heavy rain as clean data samples, so as to train a prediction model based on the clean data samples and learn a benchmark prediction model that can predict power generation parameters under ideal clean conditions.

[0063] Therefore, the diagnostic method of this invention identifies meteorological events where rainfall exceeds a preset threshold and automatically extracts historical operating data within a preset time period after the meteorological event as a clean data sample. This ensures the effectiveness of constructing a sample of the ideal clean state of a photovoltaic power station, while avoiding the subjectivity and high cost of manual annotation. This improves the convenience and reliability of obtaining training data, thereby ensuring the accuracy of subsequent diagnostic results.

[0064] In some embodiments, the preset irradiance condition in step S108 may include an irradiance greater than a preset irradiance threshold. Specifically, the preset irradiance threshold can be set according to the irradiance standard for a sunny day. Furthermore, the preset irradiance threshold can also be adjusted according to the typical light conditions of the photovoltaic power station location.

[0065] In one specific embodiment, the preset irradiance threshold can be selected from 200~1000W / m². 2 Any value within the range. For example, the preset irradiance threshold can be set to 200 W / m². 2 .

[0066] Therefore, the method for diagnosing the inefficiency of photovoltaic power plants in this embodiment of the invention provides a clear technical standard for screening effective power generation periods by clearly defining the preset irradiance condition as an irradiance greater than the threshold set according to the irradiance standard for sunny days. This ensures that each screened effective power generation period is under a condition where the photovoltaic module has a high and stable photoelectric conversion efficiency, and eliminates the influence of interference factors such as low irradiance on the diagnostic results. This improves the accuracy and interpretability of the contamination diagnosis results, and provides an intuitive and reliable reference for operation and maintenance decisions.

[0067] In some embodiments, the step of screening each effective power generation period of a photovoltaic power station based on a preset irradiance condition in step S108 may include the following steps: based on preprocessed multi-source heterogeneous data, identifying and extracting all continuous time windows where the irradiance is continuously greater than a preset irradiance threshold, and taking each continuous time window as an effective power generation period.

[0068] Specifically, the aforementioned continuous time window is not less than a preset minimum duration. The preset minimum duration can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power plant. In one specific embodiment, the preset minimum duration can be selected from any value between 2 and 25 hours. For example, the preset minimum duration can be set to 3 hours.

[0069] Therefore, the method for diagnosing the inefficiency of photovoltaic power plants in this embodiment of the invention specifically defines the identification and selection of all continuous time windows in which the irradiance is continuously greater than a preset threshold as effective power generation periods. This effectively filters out transient irradiance fluctuations caused by cloud cover, ensuring that each selected effective power generation period is in a continuous and stable working state of the photovoltaic modules, thereby further guaranteeing the reliability of the contamination diagnosis results.

[0070] In some embodiments, the quantitative indicators for determining contamination may further include power generation. Based on this, after step S104 above, the diagnostic method of the present invention may further include the following steps: obtaining the actual power generation of the photovoltaic power station based on preprocessed multi-source heterogeneous data; inputting the preprocessed real-time operating data into a pre-trained prediction model to predict the expected power generation of the photovoltaic power station under ideal clean conditions; and obtaining the contamination diagnosis result of the photovoltaic power station based on the actual power generation and the expected power generation.

[0071] In other words, the diagnostic method of this invention provides a unified and calculable basis for judgment for the core diagnostic logic by quantitatively comparing the actual performance of the photovoltaic power station with the ideal clean state benchmark predicted by the model.

[0072] Therefore, the diagnostic method of this invention obtains the actual power generation of the photovoltaic power station through preprocessed multi-source heterogeneous data, inputs the preprocessed real-time operating data into a prediction model trained on clean data, and obtains the expected power generation under ideal clean conditions, thus achieving an effective quantitative comparison of actual and ideal power generation parameters. Furthermore, the diagnostic method of this invention also obtains the contamination diagnosis results of the photovoltaic power station based on the actual and expected power generation, providing a clear technical basis for determining the contamination status, facilitating subsequent quantitative assessment of the contamination level of the photovoltaic power station, thereby improving the accuracy and interpretability of the contamination diagnosis results, and ultimately providing an intuitive and reliable reference for operation and maintenance decisions.

[0073] In some embodiments, the steps of obtaining the pollution diagnosis results of the photovoltaic power station based on the actual power generation and the expected power generation may include the following steps: calculating the difference between the actual power generation of the photovoltaic power station and the predicted expected power generation; calculating the ratio of the difference to the expected power generation to obtain the pollution loss rate of the photovoltaic power station; determining whether the pollution loss rate is greater than a preset loss rate threshold; if so, determining that there is pollution accumulation in the photovoltaic power station.

[0074] Specifically, the formula for calculating the contamination loss rate can be: Contamination Loss Rate = (1 - Actual Power Generation / Expected Power Generation) × 100%. Furthermore, the preset loss rate threshold can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power plant. In one specific embodiment, the preset loss rate threshold can be selected from any value between 1% and 10%. For example, the preset loss rate threshold can be set to 5%.

[0075] In other words, the diagnostic method of this invention quantifies the degree of pollution of a photovoltaic power station into a specific percentage of power generation loss, so that the severity of the pollution impact can be accurately measured and classified.

[0076] Therefore, the diagnostic method of this invention uses power generation as the core power generation parameter. By calculating the difference between the actual power generation and the expected power generation, and then calculating the ratio of this difference to the expected power generation, the fouling loss rate is obtained, thus transforming the degree of fouling into a quantifiable evaluation indicator. Furthermore, the diagnostic method of this invention also determines whether the fouling loss rate exceeds a preset loss rate threshold, enabling a clear assessment of the fouling accumulation state. This facilitates maintenance personnel in developing targeted component cleaning strategies based on the specific value of the fouling loss rate.

[0077] In some embodiments, step S110 may include the following steps: calculating a confidence level based on the number of multiple effective power generation periods, the duration of each effective power generation period, and the trend of actual power generation efficiency changing over time within multiple effective power generation periods; calculating the ratio of the actual power generation efficiency within the latest effective power generation period to the predicted expected power generation efficiency to obtain the pollution rate of the photovoltaic power station; determining whether the confidence level is higher than a preset score threshold and whether the pollution rate is lower than a preset pollution threshold; if so, determining that there is pollution accumulation in the photovoltaic power station.

[0078] In other words, the diagnostic method of this invention calculates the confidence level based on the number of multiple effective power generation periods and the trend of actual power generation efficiency changing over time, calculates the fouling rate based on the actual power generation efficiency in the latest effective power generation period and the predicted expected power generation efficiency, and combines the dual judgment of confidence level and fouling rate to achieve dynamic evaluation of the reliability of fouling diagnosis results.

[0079] Therefore, the diagnostic method of this invention only determines that there is dirt accumulation in the photovoltaic power station when the confidence level is higher than the preset score threshold and the dirt rate is lower than the preset dirt threshold. This effectively avoids misjudgment caused by short-term data fluctuations, abnormal weather, or insufficient data, thereby reducing the false alarm rate and false negative rate of dirt diagnosis and ensuring the accuracy of subsequent diagnostic results.

[0080] In some embodiments, effective power generation periods may include effective days with sunny weather conditions. Based on this, the step of calculating confidence based on the number of multiple effective power generation periods and the trend of actual power generation efficiency changing over time within multiple effective power generation periods may include the following steps: determining whether the number of effective days exceeds a preset number of days threshold; if not, determining the confidence level to be 0; if so, arranging the actual power generation efficiency of each effective day in chronological order to construct an analysis dataset, and performing regression analysis on the analysis dataset using a linear fitting algorithm to obtain the fitting slope of the actual power generation efficiency change; calculating the confidence level based on the number of effective days and the fitting slope of the actual power generation efficiency change.

[0081] Specifically, the confidence level is a score ranging from 0 to 1. The preset number of days threshold and the preset score threshold can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power plant. In one specific embodiment, the preset number of days threshold can be selected from any value between 2 and 5 days, and the preset score threshold can be selected from any value between 0.6 and 0.9. For example, the preset number of days threshold can be set to 3 days, and the preset score threshold can be set to 0.7.

[0082] In other words, the diagnostic method of this invention, by introducing an effective power generation period screening and a comprehensive confidence assessment mechanism, ensures that the diagnostic process is free from weather interference and improves the reliability of the assessment results, thereby maintaining the accuracy of the diagnostic results even under complex meteorological conditions.

[0083] Therefore, the diagnostic method of this invention uses power generation efficiency as the core power generation parameter, selects the effective power generation period of the photovoltaic power station based on preset irradiance conditions, and eliminates the influence of interference factors such as low irradiance on the diagnostic results. By calculating the number, duration and trend of power generation efficiency changes of the effective power generation period, the confidence level is obtained. The ratio of the actual power generation efficiency to the expected power generation efficiency of the latest effective period is calculated to obtain the contamination rate. Then, the contamination accumulation is determined by a dual threshold, which further improves the reliability of the diagnostic results and reduces the false alarm rate and false negative rate of contamination diagnosis.

[0084] In some embodiments, the prediction model in step S106 above may include a Lightweight Gradient Boosting Machine (LightGBM) model. Specifically, the prediction model can be obtained through the following steps: obtaining a prediction model pre-built for a photovoltaic power station; using clean data samples as a training set, incrementally training the prediction model using a preset gradient boosting decision tree algorithm, and employing an early stopping mechanism to prevent model overfitting, so as to train and obtain the prediction model.

[0085] Specifically, the incremental training described above can be achieved by setting the `training_days` parameter, allowing the model to be trained using data from a recent period (e.g., 180 days) to adapt to seasonal changes and equipment aging. Additionally, an early stopping mechanism can be included during model training to prevent overfitting.

[0086] Therefore, the diagnostic method of this invention uses the LightGBM model as the core prediction model, leveraging its advantages of fast training speed and low memory consumption to meet the processing needs of large-scale time-series data from photovoltaic power plants. Simultaneously, the diagnostic method of this invention incrementally trains the model using clean data samples, enabling the LightGBM model to adapt to dynamic factors such as equipment aging and seasonal changes within the photovoltaic power plant. Combined with an early shutdown mechanism, it prevents overfitting of the LightGBM model, ensuring both training efficiency and prediction accuracy, thus facilitating the batch diagnosis of inefficiencies in large-scale photovoltaic power plants.

[0087] In some embodiments, after step S110 above, the diagnostic method of the present invention may further include the following steps: sequentially using a threshold method to perform string disconnection detection, frequent start-stop detection, power limitation detection, and power limitation detection on the preprocessed multi-source heterogeneous data; based on the detection results and the contamination diagnosis results, diagnose the causes of inefficiency in the photovoltaic power station; generate and output a diagnostic report containing the causes of inefficiency to assist in operation and maintenance decisions.

[0088] Specifically, the conditions for string disconnection detection can be set as follows: string power less than a first power threshold, total power greater than the first power threshold, string voltage less than a preset voltage threshold, and continuous abnormal time greater than a preset time threshold. It should be noted that string power refers to the MPPT (Maximum Power Point Tracking) power of any string in a photovoltaic power plant. Furthermore, the first power threshold, preset voltage threshold, and preset time threshold can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power plant. In a specific embodiment, the first power threshold can be selected from any value between 0.1 and 10W, the preset voltage threshold can be selected from any value between 10 and 100V, and the preset time threshold can be selected from any value between 0.5 and 3 hours. For example, the conditions for string disconnection detection can be set as follows: string power < 0.1W, total power > 0.1W, string voltage < 50V, and continuous abnormal time ≥ 1 hour.

[0089] The condition for frequent start-stop detection can be set as follows: the number of times the inverter starts within a day exceeds a preset threshold. It should be noted that the preset threshold can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power plant. In a specific embodiment, the preset threshold can be selected from any value between 3 and 10 times. For example, the condition for frequent start-stop detection can be set as: the number of times the inverter starts within a day > 5 times.

[0090] The conditions for power limiting detection can be set as follows: the proportion of high-power periods exceeds a first preset threshold, the coefficient of variation of high-power periods is less than a preset coefficient threshold, and the proportion of peak power plateau exceeds a second preset threshold. It should be noted that the first preset threshold, the preset coefficient threshold, and the second preset threshold can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power plant. In a specific embodiment, the first preset threshold can be selected from any value between 50% and 90%, the preset coefficient threshold can be selected from any value between 1% and 5%, and the second preset threshold can be selected from any value between 20% and 40%. For example, the conditions for frequent start-stop detection can be set as follows: the proportion of high-power periods exceeds 80%, the coefficient of variation of high-power periods is <1.5%, and the proportion of peak power plateau exceeds 30%.

[0091] The conditions for power curtailment detection can be set as follows: MPPT power is less than a second power threshold, continuous zero power duration is greater than or equal to a preset duration threshold, the ratio of power before power outage to single-circuit capacity is greater than or equal to a first ratio, and the ratio of power before power outage to single-circuit capacity is greater than or equal to a second ratio within a preset duration threshold after power outage. It should be noted that the second power threshold, preset duration threshold, first ratio, and second ratio can be configured based on historical operation and maintenance experience or the operating standards of the photovoltaic power station. In a specific embodiment, the second power threshold can be selected from any value between 1 and 10W, the preset duration threshold can be selected from any value between 10 and 60 minutes, the first ratio can be selected from any value between 5% and 10%, and the second preset ratio threshold can be selected from any value between 15% and 20%. For example, the conditions for power curtailment detection can be set as follows: MPPT power ≤ 5W, continuous zero power duration ≥ 30 minutes, the ratio of power before power outage to single-circuit capacity ≥ 10%, and the ratio of power before power outage to single-circuit capacity ≥ 15% within 30 minutes after power outage.

[0092] In addition, the causes of inefficiency in photovoltaic power plants can include module contamination, equipment failure, string disconnection, frequent start-stop cycles, power curtailment, and grid connection issues. The output diagnostic report can include: power plant code, analysis date, cause of inefficiency, actual power generation, predicted expected power generation, contamination rate, and contamination loss rate, to assist in operation and maintenance decision-making.

[0093] In other words, the diagnostic method of this invention, in addition to the core contamination diagnosis, integrates a threshold detection process based on clear engineering rules, which can systematically screen various typical hardware and operational anomalies at the inverter and string level, thereby building a complete inefficient problem tracing capability from component contamination to equipment failure.

[0094] Therefore, the diagnostic method of this invention, based on the contamination diagnosis results, employs a threshold method to sequentially perform string disconnection detection, frequent start-stop detection, power limitation detection, and power outage detection, achieving a comprehensive diagnosis of multi-level inefficiency problems at the inverter, string, and module levels. Furthermore, by integrating contamination diagnosis with various anomaly detection results, the diagnostic method of this invention accurately locates the causes of inefficiency in photovoltaic power plants and generates a diagnostic report, providing a complete basis for operation and maintenance decisions. This enables precise tracing of the causes of inefficiency in photovoltaic power plants, transforming the operation and maintenance mode from passive response to proactive prevention, and reducing subsequent operation and maintenance costs.

[0095] Figure 2 This is a flowchart of a control example of a method for diagnosing inefficiencies in photovoltaic power plants according to an embodiment of the present invention. Figure 2 As shown, the process of the diagnostic method for the inefficiency problem of photovoltaic power plants in this embodiment includes:

[0096] Step S202: Collect multi-source heterogeneous data from the photovoltaic power station. It should be noted that multi-source heterogeneous data may include historical operating data, real-time operating data, and meteorological data.

[0097] Step S204 involves preprocessing the multi-source heterogeneous data. It should be noted that preprocessing may include operations such as anomaly detection, data alignment, and data standardization.

[0098] Step S206: Identify meteorological events in the meteorological data where the rainfall exceeds a preset rainfall threshold.

[0099] Step S208: Extract historical operational data within a preset time period after the end of the meteorological event to obtain a clean data sample.

[0100] Step S210: Obtain the pre-built prediction model for the photovoltaic power station. It should be noted that a corresponding prediction model is independently built for each photovoltaic power station to adapt to its geographical location, climate conditions, and equipment characteristics.

[0101] Step S212: Using clean data samples as the training set, the prediction model is incrementally trained using a preset gradient boosting decision tree algorithm, and an early stopping mechanism is used to prevent the model from overfitting, so as to obtain the prediction model.

[0102] Step S214: Input the pre-processed real-time operating data into the pre-trained prediction model to predict the expected power generation efficiency of the photovoltaic power station under ideal clean conditions.

[0103] Step S216: Based on preset irradiance conditions, select each effective power generation period of the photovoltaic power station.

[0104] Step S218: Obtain the actual power generation efficiency of the photovoltaic power station during each effective power generation period. It should be noted that the actual power generation efficiency is the ratio of the power output to the irradiance of the photovoltaic power station during the effective power generation period.

[0105] Step S220: Calculate the confidence level based on the number of multiple effective power generation periods and the trend of actual power generation efficiency changing over time within the multiple effective power generation periods.

[0106] Step S222: Calculate the ratio of the actual power generation efficiency to the predicted expected power generation efficiency during the latest effective power generation period to obtain the pollution rate of the photovoltaic power station.

[0107] Step S224: If the confidence level is higher than the preset score threshold and the dirt rate is lower than the preset dirt threshold, it is determined that there is dirt accumulation in the photovoltaic power station.

[0108] Step S226: Generate and output a diagnostic report indicating that the inefficiency is caused by dirt accumulation, to assist in operational decision-making. This concludes the current process. Alternatively, after step S226, you can return to step S202 to restart the process.

[0109] Therefore, the diagnostic method of this invention, by collecting multi-source heterogeneous data including historical operating data, real-time operating data, and meteorological data, achieves a comprehensive perception of the power plant's operating status and environmental factors. Simultaneously, the diagnostic method of this invention extracts clean data samples based on preset meteorological conditions and trains a prediction model accordingly, realizing the sample construction and effective characterization of the ideal clean state of the photovoltaic power plant. Furthermore, the diagnostic method of this invention also presets irradiance conditions to screen effective power generation periods, obtains the ratio of power to irradiance of the photovoltaic power plant in each effective power generation period as the actual power generation efficiency, and compares the actual power generation efficiency with the expected power generation efficiency output by the prediction model. This achieves accurate quantitative diagnosis of the pollution problem of the photovoltaic power plant, thereby improving the accuracy of diagnosing inefficiencies and providing a reliable basis for subsequent operation and maintenance decisions.

[0110] Furthermore, the diagnostic method of this embodiment calculates the confidence level based on the number of multiple effective power generation periods and the trend of actual power generation efficiency changing over time within those periods. It then calculates the contamination rate based on the actual power generation efficiency within the latest effective power generation period and the predicted expected power generation efficiency. By combining the confidence level and the contamination rate, the reliability of the contamination diagnosis results is dynamically assessed. Therefore, the diagnostic method of this embodiment only determines that there is contamination accumulation in the photovoltaic power station when the confidence level is higher than a preset score threshold and the contamination rate is lower than a preset contamination threshold. This effectively avoids misjudgments caused by short-term data fluctuations, abnormal weather, or insufficient data, thereby ensuring the accuracy of subsequent diagnostic results.

[0111] Furthermore, the diagnostic method of this invention provides a clear technical standard for screening effective power generation periods by explicitly setting a preset irradiance condition that the irradiance is greater than a threshold set according to the irradiance standard for sunny days. This ensures that each screened effective power generation period is under a condition where the photovoltaic module has a high and stable photoelectric conversion efficiency, eliminating the influence of interference factors such as low irradiance on the diagnostic results. This improves the accuracy and interpretability of the contamination diagnosis results, and provides an intuitive and reliable reference for operation and maintenance decisions.

[0112] This embodiment also provides a computer program product 10, a computer-readable storage medium 20, and a computer device 30. Figure 3 This is a schematic diagram of a computer program product according to an embodiment of the present invention. Figure 4 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention. Figure 5 This is a schematic block diagram of a computer device according to an embodiment of the present invention.

[0113] Computer program product 10 includes computer program 11, which, when executed by processor 32, implements the steps of the method for diagnosing photovoltaic power plant inefficiency problems described above. Computer-readable storage medium 20 stores the computer program 11 thereon, which, when executed by processor 32, implements the steps of the method for diagnosing photovoltaic power plant inefficiency problems described above. Computer device 30 may include memory 31, processor 32, and computer program 11 stored in memory 31 and running on processor 32. Processor 32 executes computer program 11 to implement the steps of the method for diagnosing photovoltaic power plant inefficiency problems described above.

[0114] The computer program 11 used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, status setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages.

[0115] Computer program 11 may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, to perform aspects of the invention, electronic circuitry including, for example, programmable logic circuitry, a Field-Programmable Gate Array (FPGA), or a Programmable Logic Array (PLA) may execute computer-readable program instructions to personalize the electronic circuitry by utilizing state information of the computer-readable program instructions.

[0116] For the purposes of this embodiment, computer program product 10 is a related product that includes computer program 11.

[0117] For the purposes of this embodiment, the computer-readable storage medium 20 is a tangible device capable of holding and storing a computer program 11. It can be any device capable of containing, storing, communicating, propagating, or transmitting the computer program 11 for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable storage medium 20 include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable optical disc read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical encoding device, and any suitable combination thereof.

[0118] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A method for diagnosing the inefficiency of photovoltaic power plants, characterized in that... include: Collect multi-source heterogeneous data from the photovoltaic power station and preprocess the multi-source heterogeneous data, wherein the multi-source heterogeneous data includes historical operating data, real-time operating data and meteorological data; Clean data samples are extracted from the preprocessed multi-source heterogeneous data based on preset meteorological conditions. The preprocessed real-time operating data is input into a pre-trained prediction model to predict the expected power generation efficiency of the photovoltaic power station under ideal clean conditions, wherein the prediction model is trained from the clean data samples; Based on preset irradiance conditions, each effective power generation period of the photovoltaic power station is screened, and the actual power generation efficiency of the photovoltaic power station in each effective power generation period is obtained. The effective power generation period includes effective days, and the actual power generation efficiency is the ratio of the power of the photovoltaic power station to the irradiance in the effective power generation period. Determine whether the number of valid days exceeds a preset number of days threshold; If it does not exceed, then the confidence level is determined to be 0; If the actual power generation efficiency of each effective day is exceeded, the actual power generation efficiency of each effective day is arranged in chronological order to construct an analysis dataset. A linear fitting algorithm is then used to perform regression analysis on the analysis dataset to obtain the fitting slope of the actual power generation efficiency change. The confidence level is calculated based on the number of effective days and the fitting slope of the actual power generation efficiency change. The pollution rate of the photovoltaic power station is obtained by calculating the ratio of the actual power generation efficiency to the predicted expected power generation efficiency during the latest effective power generation period. Determine whether the confidence level is higher than a preset score threshold and whether the dirt rate is lower than a preset dirt threshold; If the confidence level is higher than a preset score threshold and the dirt rate is lower than a preset dirt threshold, then it is determined that there is dirt accumulation in the photovoltaic power station.

2. The diagnostic method for inefficiency in photovoltaic power plants according to claim 1, characterized in that, The preset irradiance conditions include irradiance greater than a preset irradiance threshold, and the preset irradiance threshold is set according to the irradiance standard for sunny days.

3. The method for diagnosing the inefficiency of photovoltaic power plants according to claim 2, characterized in that, The step of screening each effective power generation period of the photovoltaic power station based on preset irradiance conditions includes: Based on the preprocessed multi-source heterogeneous data, all continuous time windows where the irradiance is continuously greater than the preset irradiance threshold are identified and extracted, and each continuous time window is taken as an effective power generation period.

4. The method for diagnosing the inefficiency of photovoltaic power plants according to claim 1, characterized in that, The prediction model includes a lightweight gradient boosting decision tree model; and the prediction model is obtained through the following steps: Obtain the pre-built prediction model for the photovoltaic power station; Using the clean data samples as the training set, the prediction model is incrementally trained using a preset gradient boosting decision tree algorithm, and an early stopping mechanism is used to prevent the model from overfitting, so as to obtain the prediction model.

5. The diagnostic method for inefficiency in photovoltaic power plants according to claim 1, characterized in that, Following the step of obtaining the pollution diagnosis results of the photovoltaic power plant, the method for diagnosing the inefficiency problem of the photovoltaic power plant further includes: The threshold method is used sequentially to perform string disconnection detection, frequent start-stop detection, power limit detection, and power limit detection on the preprocessed multi-source heterogeneous data. Based on the test results and the pollution diagnosis results, the cause of the inefficiency of the photovoltaic power station is diagnosed. Generate and output a diagnostic report containing the reasons for the inefficiency to assist in operation and maintenance decisions.

6. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the diagnostic method for the inefficiency problem of a photovoltaic power plant as described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that... When the computer program is executed by the processor, it implements the steps of the diagnostic method for the inefficiency problem of photovoltaic power plants as described in any one of claims 1 to 5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the diagnostic method for the inefficiency problem of a photovoltaic power plant as described in any one of claims 1 to 5.

Images