Cleaning cycle recommendation method and system for photovoltaic power station

Abstract

The invention relates to the technical field of photovoltaic power station maintenance, and discloses a cleaning cycle recommendation method and system for a photovoltaic power station, and the method comprises the steps: recognizing various orientations of an inverter string based on preprocessed historical current data through employing a preset clustering algorithm, and carrying out the anomaly detection to delete an abnormal string, and obtaining the current data of normal strings in different orientations; calculating a photovoltaic string dispersion rate, and calculating a cleaning index based on the string dispersion rate and the operation state parameters of the inverter; and further based on the cleaning index and the weather data, screening the date as the initial date of the preliminarily recommended cleaning period, dividing the cleaning period in combination with the power station capacity, and finally determining the recommended cleaning period based on the weather data and the historical cleaning record. According to the method, the photovoltaic cleaning opportunity and period are accurately recommended, so that the rationality and economy of the cleaning decision are improved, and powerful support is provided for efficient operation and maintenance and long-term income guarantee of a photovoltaic power station.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power station maintenance, and particularly to a method and system for recommending cleaning cycles of a photovoltaic power station. Background Art

[0002] In recent years, photovoltaic power generation, as a clean energy source, has received extensive attention. Not only are large-scale ground photovoltaic power stations continuously expanding, but the proportion of distributed photovoltaic power stations is also increasing year by year. However, during long-term operation, photovoltaic modules are vulnerable to environmental factors such as dust, bird droppings, and haze, resulting in dust accumulation and a significant decline in power generation efficiency, directly affecting the economic benefits of photovoltaic power stations. Currently, the operation and maintenance difficulties of the inability to quantify the impact of dust accumulation and the difficulty in determining the cleaning cycle have become urgent problems to be solved in the photovoltaic industry.

[0003] Traditional expert experience guidance and fixed-cycle cleaning methods often ignore the natural environment, meteorological conditions, and actual on-site situations, resulting in unsatisfactory economic benefits. Moreover, for large-scale photovoltaic power stations, manual visual inspection has a large workload and is greatly affected by subjective factors. Image processing methods have high requirements for algorithm accuracy and equipment investment, and it is difficult to accurately reflect the actual on-site situation only relying on operation data analysis. Summary of the Invention

[0004] In view of this, the present invention provides a method and system for recommending cleaning cycles of a photovoltaic power station, which solves the problem of how to accurately recommend the cleaning timing and cycle of a photovoltaic power station, thereby improving the rationality and economy of cleaning decisions, and providing strong support for the efficient operation and maintenance and long-term revenue guarantee of a photovoltaic power station.

[0005] In a first aspect, the present invention provides a method for recommending cleaning cycles of a photovoltaic power station, the method comprising:

[0006] Obtaining historical current data of each inverter in the photovoltaic power station and performing preprocessing;

[0007] Based on the preprocessed historical current data, using a preset clustering algorithm to identify multiple orientations of inverter strings;

[0008] Deleting abnormal strings by performing anomaly detection on the identified multiple orientation strings respectively, and obtaining current data of normal strings with different orientations;

[0009] Calculating the string dispersion rate of the photovoltaic based on the current data of the normal strings with different orientations, and calculating a cleaning index based on the string dispersion rate and the operating state parameters of the inverter;

[0010] Obtaining weather data of the geographical location where the photovoltaic power station is located and performing preprocessing;

[0011] Filter dates based on the cleaning index and pre-processed weather data as the starting date of the initially recommended cleaning cycle, divide the cleaning cycle in combination with the power station capacity, and finally determine the recommended cleaning cycle based on the weather data and historical cleaning records.

[0012] The cleaning cycle recommendation method for a photovoltaic power station provided by the embodiment of the present invention can identify different orientations of inverter strings through a clustering algorithm, and perform anomaly detection on the strings of each orientation to delete abnormal strings, which can ensure that the obtained normal string current data truly reflects the power generation performance of the strings of each orientation. On this basis, it is beneficial to accurately calculate the dispersion rate of the inverter strings, and calculate the cleaning index of the photovoltaic power station based on the dispersion rate of the inverter strings, which can scientifically quantify the impact of dust accumulation on the power generation performance of photovoltaic modules, overcome the limitation that it is difficult to evaluate the degree of component dust accumulation in traditional methods. Finally, in combination with the cleaning index and weather data, dynamically screen the starting date of the cleaning cycle, and consider the power station capacity and historical cleaning records to accurately recommend the cleaning cycle, ensure that the cleaning plan is adapted to the actual environmental conditions, provide effective guidance for the operation and maintenance party for cleaning activities, reduce the operation and maintenance cost while improving the power generation efficiency of photovoltaic modules.

[0013] In an optional implementation manner, the obtaining of the historical current data of each inverter of the photovoltaic power station and pre-processing thereof includes:

[0014] Obtain the daily available hours of the inverter within a preset time period;

[0015] Take the number of days greater than the preset available hours threshold as the effective days of the daily available hours of the inverter, and obtain the corresponding historical current data of each inverter within the effective days;

[0016] Resample the obtained historical current data based on a preset time granularity, use linear interpolation to fill in the missing values, and correct the negative current values.

[0017] The embodiment of the present invention obtains the daily available hours of the inverter within a preset time period, and takes the number of days greater than the preset available hours threshold as the effective days, which can delete those days when the inverter runs for too short a time and the data is unreliable due to equipment failures, extreme weather, etc., and can ensure that the data used for subsequent analysis is obtained when the inverter is operating normally or basically normally, improving the effectiveness and representativeness of the data, avoiding interference of invalid data on the analysis results, and providing a more reliable basis for subsequent analysis and decision-making based on current data.

[0018] In an optional implementation manner, the preset available hours threshold is calculated by the following formula:

[0019]

[0020] where θ represents the available hours threshold, H annual represents the annual available hours, and α represents the coefficient for dynamically correcting the power generation potential according to the geographical location of the distributed power station;

[0021] The effective days of the available hours of the inverter per day are obtained by the following formula:

[0022]

[0023] where S represents the set of effective days greater than θ within the time period D d represents the d-th day. Sort S in descending order according to the available hours, and take the dates of the first P days as the effective days.

[0024] In the embodiment of the present invention, a coefficient for dynamically correcting the power generation potential according to the geographical location of the distributed power station is introduced, which can reflect the influence of natural conditions such as light and climate in different geographical locations on the power generation capacity of the photovoltaic power station, making the available hours threshold more in line with the local actual power generation potential, and making the subsequent judgment of the effective operation days of the inverter more in line with the actual situation of the power station. It is beneficial to screen out the most representative and reliable data from a large amount of data and provide a high-quality data basis for subsequent analysis.

[0025] In an optional embodiment, based on the preprocessed historical current data, using a preset clustering algorithm to identify multiple orientations of the inverter strings, including:

[0026] Extract the average current value and current standard deviation of each string at a preset moment as the current feature vector, and construct a feature vector matrix of the strings;

[0027] Based on the feature vector matrix, use the DBSCAN algorithm for orientation grouping, and divide multiple strings into multiple orientation categories according to the clustering results.

[0028] ​In the embodiments of the present invention, the average current and the current standard deviation of each string at a preset moment are extracted as the current feature vectors. These two indicators can effectively summarize the key characteristics of the string current. The average current reflects the average power generation capacity of the string at a specific moment, while the current standard deviation reflects the fluctuation of the current around the average value. For strings with different orientations, due to factors such as the intensity, angle, and duration of received sunlight, their power generation capabilities and current stabilities will vary. Through these two features, these differences can be well captured, providing representative feature information for subsequent clustering analysis and accurately distinguishing strings with different orientations; in an actual photovoltaic power station, the distribution of strings with different orientations may not be a simple regular shape but complex and diverse. The DBSCAN algorithm can well adapt to this complex distribution situation, automatically determine the appropriate number of orientation categories according to the distribution of the current feature vectors of the strings, making the clustering results more objectively and accurately reflect the actual situation, and accurately dividing the strings with different orientations into different categories.

[0029] In an alternative embodiment, the deletion of abnormal strings by performing anomaly detection on the identified multi-class orientation strings respectively includes:

[0030] Calculate the average current of the strings during the large power generation period of the photovoltaic system, and mark the strings with a daily average current less than a preset threshold as dropped strings;

[0031] Perform data cleaning and normalization on the current data for the obtained valid days, and convert the current value of each string into a ratio relative to the total current of all strings;

[0032] Based on the ratio of the current value of each string converted to the total current of all strings, calculate the normalized average current during the high-irradiance period and the non-high-irradiance period respectively;

[0033] Take the ratio of the normalized average current during the non-high-irradiance period to the normalized average current during the high-irradiance period as the string aging judgment index, and take the string aging judgment index greater than the preset index threshold as the aging condition;

[0034] Obtain the average current of each string during the large power generation period of the photovoltaic system within a preset time period, and perform a descending order sorting of the average current. If there are strings that are all within a preset range at the end of the ranking within a preset number of days and meet the aging condition continuously for multiple days, the corresponding strings can be confirmed as aging strings;

[0035] Delete the dropped strings and aging strings from the recognition results of the strings with different orientations.

[0036] In the embodiments of the present invention, the average current of the string is calculated during the large power generation period of the photovoltaic, and the string with abnormal current is marked by the daily average current being less than the preset threshold, which can accurately locate the string whose power generation ability is significantly lower than the normal level. By calculating the normalized current averages during the high-irradiance and non-high-irradiance periods respectively, and using the ratio of the two as the aging judgment index of the string, a quantitative basis is provided for aging detection. The average current during the large power generation period of the photovoltaic within the preset time period is obtained and sorted. Combining the aging conditions and the ranking of the strings can more reliably confirm the aging strings. Aging is a gradual process. By synthesizing multi-day data and ranking, misjudgment caused by occasional abnormal fluctuations can be avoided, and the accuracy of identifying aging strings can be improved. By accurately identifying and deleting abnormal strings, a reliable basis is provided for subsequent cleaning and operation and maintenance decisions.

[0037] In an alternative embodiment, the cleaning index is calculated by the following formula:

[0038]

[0039] where CI i represents the cleaning index of the inverter SN i , γ represents the difference correction coefficient of the component tilt orientation, P represents the effective number of days for calculation, δ p represents the discretization rate of the inverter SN i on the pth day, represents the available hours of the inverter SN i on the pth day, represents the available hours of the reference inverter on the pth day;

[0040] The average value of the cleaning indices of each inverter is taken as the cleaning index of the photovoltaic power station.

[0041] The calculation method of the cleaning index in the embodiments of the present invention takes into account the differences in component tilt orientation, effective days, the discretization rate of the inverter, the available hours of the inverter, and the available hours of the reference inverter on the pth day at the same time, organically combines multiple data dimensions, makes full use of various data information during the operation of the inverter, makes the calculation of the cleaning index more scientific and comprehensive, can more accurately discover the impact of cleaning factors on the operation of the photovoltaic power station, provides a unified quantitative standard for the cleaning status of the photovoltaic power station. Taking the average value of the cleaning indices of each inverter as the cleaning index of the photovoltaic power station can reflect the cleaning degree of the photovoltaic power station as a whole, avoid the influence of the special situation of a single inverter on the overall evaluation, and represent the cleaning level of the entire power station with a comprehensive value. This quantitative index is convenient for operation and maintenance personnel to compare the cleaning status between photovoltaic power stations in different time periods and different regions, and helps to formulate reasonable cleaning and maintenance plans and resource allocation schemes.

[0042] In an alternative embodiment, obtaining the weather data of the geographical location where the photovoltaic power station is located and performing preprocessing includes:

[0043] Obtaining the weather data of a historical preset number of days and the predicted weather data of a future preset number of days, where the weather data includes: rainfall, air quality index, and weather description;

[0044] Taking the weather data of the historical preset number of days and the predicted weather data of the future preset number of days as a weather data set.

[0045] In the embodiment of the present invention, the weather data of the historical preset number of days and the predicted weather data of the future preset number of days are obtained simultaneously, providing a comprehensive weather information perspective for the operation and maintenance of the photovoltaic power station. Incorporating multiple data dimensions such as rainfall, air quality index, and weather description, it is possible to comprehensively understand the impact of weather on the photovoltaic power station.

[0046] In an alternative embodiment, screening the date based on the cleaning index and the preprocessed weather data as the starting date of the initially recommended cleaning cycle includes:

[0047] Screening out the dates with a cleaning index lower than the cleaning threshold from the multi-day cleaning index data as potential starting dates of the cleaning cycle;

[0048] For the potential cleaning dates, if within the time window {W d-c ,W d+c}, the rainfall does not exceed the set threshold and the air quality does not meet the preset excellent quality condition, then determine the recommended starting date D clean(Wd) of the cleaning cycle. If within the time window {W d-c ,W d+c}, the rainfall exceeds the set threshold and the air quality meets the preset excellent quality condition, then postpone the cleaning date by c days and re-check the weather data within the new time window {W d ,W d+2c} until the rainfall within the time window does not exceed the set threshold and the air quality does not meet the preset excellent quality condition, and finally determine the recommended starting date D clean of the cleaning cycle.

[0049] In the embodiment of the present invention, by screening the date as the potential starting date of the cleaning cycle, it is possible to accurately identify the time point when the cleaning degree of the photovoltaic power station components has affected the power generation efficiency. For the potential cleaning dates, adjusting the cleaning time according to the weather data within the time window helps to reasonably plan the cleaning equipment, materials, and personnel arrangements, reducing the operation and maintenance costs.

[0050] In an alternative embodiment, dividing the cleaning cycle in combination with the power station capacity and finally determining the recommended cleaning cycle based on the weather data and historical cleaning records includes:

[0051] Determine the recommended cleaning period time range of the photovoltaic power station according to the corresponding relationship between the preset installed capacity and the recommended cleaning period;

[0052] When there are cloudy or overcast days during the recommended cleaning period time range, then use the cloudy or overcast day D cloudy As the recommended date;

[0053] Obtain the historical cleaning records, and use the date that does not conflict with the historical cleaning records as the cleaning date. If the recommended date does not conflict with the historical cleaning records, then readjust the time window and determine the recommended start date.

[0054] In the embodiment of the present invention, the recommended cleaning period time range is determined according to the corresponding relationship between the preset installed capacity and the recommended cleaning period, taking into account the differences of power stations of different scales, which can ensure that the setting of the cleaning period conforms to the characteristics of the power station itself, improve the pertinence and effectiveness of the cleaning work. When there are preset adverse weather conditions during the recommended cleaning period time range, avoid these adverse weather conditions as the recommended date, which guarantees the safety and effectiveness of the cleaning work. Selecting cloudy or overcast days as the recommended date makes full use of the favorable factors of such weather for the cleaning work. Obtain the historical cleaning records, and use the date that does not conflict with the historical cleaning records as the cleaning date, avoiding unnecessary repeated cleaning. This method of determining the cleaning period comprehensively considers various factors such as the power station capacity, weather data, and historical cleaning records, provides a comprehensive and scientific basis for the operation and maintenance decision-making, can ensure that the photovoltaic modules maintain a good working state, reduce the decline in power generation efficiency and the risk of equipment failure caused by dirt accumulation, and thus guarantee the stable operation of the photovoltaic power station and improve the power generation benefit. At the same time, an orderly cleaning plan also helps to extend the service life of the photovoltaic modules and related equipment and reduce the long-term operation and maintenance costs.

[0055] In a second aspect, the present invention provides a recommended cleaning period system for a photovoltaic power station, including:

[0056] A historical data acquisition module, configured to acquire the historical current data of each inverter of the photovoltaic power station and perform preprocessing;

[0057] An orientation recognition module, configured to recognize multiple types of orientations of the inverter strings based on the preprocessed historical current data by using a preset clustering algorithm;

[0058] An anomaly detection module, configured to delete the abnormal strings by performing anomaly detection on the multiple types of orientation strings obtained by recognition respectively, and acquire the current data of the normal strings with different orientations;

[0059] A cleaning index calculation module, configured to calculate the string discreteness rate of the photovoltaic based on the current data of the normal strings in different orientations, and calculate the cleaning index based on the string discreteness rate and the operating state parameters of the inverter;

[0060] A weather data acquisition module, configured to acquire the weather data of the geographical location where the photovoltaic power station is located and perform preprocessing;

[0061] A cleaning cycle recommendation module, configured to screen dates based on the cleaning index and the preprocessed weather data as the starting date of the initially recommended cleaning cycle, divide the cleaning cycle in combination with the power station capacity, and finally determine the recommended cleaning cycle based on the weather data and historical cleaning records.

[0062] In a third aspect, the present invention provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the cleaning cycle recommendation method for a photovoltaic power station according to the first aspect or any corresponding embodiment thereof.

[0063] In a fourth aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute the cleaning cycle recommendation method for a photovoltaic power station according to the first aspect or any corresponding embodiment thereof.

[0064] In a fifth aspect, the present invention provides a computer program product, including computer instructions, and the computer instructions are used to cause a computer to execute the cleaning cycle recommendation method for a photovoltaic power station according to the first aspect or any corresponding embodiment thereof. Description of the Drawings

[0065] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0066] Figure 1 is a schematic flowchart of a cleaning cycle recommendation method for a photovoltaic power station according to an embodiment of the present invention;

[0067] Figure 2 is a schematic flowchart of acquiring and preprocessing the historical current data of each inverter of a photovoltaic power station according to an embodiment of the present invention;

[0068] Figure 3 is a schematic flowchart of identifying multiple orientations of inverter strings according to an embodiment of the present invention;

[0069] Figure 4 It is a schematic flow chart for performing anomaly detection on multiple types of orientation strings according to an embodiment of the present invention;

[0070] Figure 5 It is a schematic flow chart for processing weather data of a photovoltaic power station according to an embodiment of the present invention;

[0071] Figure 6 It is a schematic flow chart for determining a recommended cleaning period according to an embodiment of the present invention;

[0072] Figure 7 It is a schematic flow chart for determining a recommended start date of a cleaning period according to an embodiment of the present invention;

[0073] Figure 8 It is a structural block diagram of a cleaning period recommendation system for a photovoltaic power station according to an embodiment of the present invention;

[0074] Figure 9 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed implementation manners

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0076] To solve the problems that the influence of dust accumulation during the operation and maintenance of a distributed photovoltaic power station cannot be quantified and the cleaning period is difficult to determine, a cleaning period recommendation method for a photovoltaic power station is provided in this embodiment, which can be applied to various types of photovoltaic power stations. Figure 1 It is a flow chart of a cleaning period recommendation method for a photovoltaic power station according to an embodiment of the present invention. As Figure 1 shown, this flow includes the following steps:

[0077] S101, obtain the historical current data of each inverter in the photovoltaic power station and perform preprocessing.

[0078] In the embodiments of the present invention, the historical current data corresponding to each inverter within the valid days is obtained by removing the interference of invalid data, so as to focus on the valid historical current data, making the obtained data more targeted and valuable. In the actually collected current data, there may be missing values or abnormal values in the original current data collected due to reasons such as sensor failures and signal interferences. The preprocessing of the historical current data in the embodiments of the present invention helps to accurately analyze key indicators such as the working characteristics and power generation efficiency of the inverter, providing more accurate data support for the subsequent cleaning and operation and maintenance of the photovoltaic power station.

[0079] S102. Based on the preprocessed historical current data, use a preset clustering algorithm to identify multiple orientations of the inverter strings.

[0080] Specifically, extract features from the obtained historical current data, and use the DBSCAN algorithm to group the orientations of the inverter strings based on the extracted features, so as to effectively distinguish the strings with different orientations, ensure the accuracy of the calculation of the dispersion rate, and provide a reliable basis for subsequent evaluation and cleaning decisions.

[0081] S103. Perform anomaly detection on the identified multiple-orientation strings respectively to delete the abnormal strings, and obtain the current data of the normal strings with different orientations.

[0082] Since abnormal strings (including string dropout and aging) may cause significant deviations in the current data, thus affecting the accuracy of the dispersion rate calculation and even causing deviations in subsequent analysis and cleaning decisions. The embodiments of the present invention introduce an anomaly detection mechanism, accurately identify and delete abnormal strings in combination with the historical current characteristics, ensure that the dispersion rate calculation truly reflects the operating state of the strings, and provide reliable data support for the calculation of the subsequent cleaning index.

[0083] S104. Calculate the string dispersion rate of the photovoltaic based on the current data of the normal strings with different orientations, and calculate the cleaning index based on the string dispersion rate and the operating state parameters of the inverter.

[0084] Specifically, the current data of the strings with different orientations will vary due to factors such as the illumination angle, etc., but there is a certain pattern under normal circumstances. By calculating the string dispersion rate, this difference can be quantified, and the dispersion degree of the current data of each orientation string relative to the average value can be sensitively captured, thereby judging the performance differences between each component. Combining the string dispersion rate and the operating state parameters of the inverter to calculate the cleaning index can more comprehensively reflect the actual operating conditions of the photovoltaic system.

[0085] S105. Obtain the weather data of the geographical location where the photovoltaic power station is located and perform preprocessing.

[0086] Specifically, different weather conditions will have different impacts on photovoltaic devices. Accurate weather data is an important reference for evaluating the performance of photovoltaic power plants. Through the analysis and preprocessing of weather data, preventive maintenance and precise maintenance can be achieved, avoiding unnecessary maintenance work and resource waste, thereby effectively reducing the operation and maintenance costs.

[0087] S106. Screen dates based on the cleanliness index and the preprocessed weather data as the starting dates of the initially recommended cleaning cycles, divide the cleaning cycles in combination with the power station capacity, and finally determine the recommended cleaning cycles based on the weather data and historical cleaning records.

[0088] For photovoltaic power plants with different capacities, the difficulty, cost, and time required for the cleaning work are different. In the embodiments of the present invention, the cleaning cycles are divided in combination with the power station capacity, which can reasonably allocate the cleaning resources. For large-capacity power stations, longer cleaning cycles can be arranged and the cleaning can be carried out in stages to avoid the tension of manpower and material resources caused by centralized cleaning; for small-capacity power stations, the cleaning cycles can be appropriately shortened according to the actual situation to improve the cleaning efficiency and ensure the stable power generation efficiency. At the same time, by comprehensively considering the weather data and historical cleaning records, the dirt rules of the components under different seasons and weather conditions can be summarized, the cleaning personnel, equipment, and materials can be reasonably arranged in advance, the utilization efficiency of the cleaning resources can be improved, the cleaning cost can be reduced, the cleaning cycles can be accurately determined, the photovoltaic components can maintain good power generation performance, the power generation loss caused by the dirt of the components can be reduced, the overall power generation efficiency of the power station can be improved, and the power generation income can be increased. A reasonable cleaning cycle can not only avoid increasing costs due to overly frequent cleaning but also prevent the impact on power generation caused by untimely cleaning, achieving the maximization of economic benefits.

[0089] In the embodiments of the present invention, different orientations of the inverter strings are identified through a clustering algorithm, and abnormal detection is performed on the strings of each orientation to delete the abnormal strings, which can ensure that the obtained normal string current data truly reflects the power generation performance of the strings of each orientation. On this basis, it is beneficial to accurately calculate the discrete rate of the inverter strings, and calculate the cleanliness index of the photovoltaic power station based on the discrete rate of the inverter strings, which can scientifically quantify the impact of dust accumulation on the power generation performance of the photovoltaic components, overcome the limitation that it is difficult to evaluate the degree of dust accumulation of the components by traditional methods. Finally, in combination with the cleanliness index and weather data, the starting dates of the cleaning cycles are dynamically screened, and the cleaning cycles are accurately recommended considering the power station capacity and historical cleaning records, ensuring that the cleaning plan is adapted to the actual environmental conditions, providing effective guidance for the operation and maintenance party for cleaning activities, reducing the operation and maintenance costs while improving the power generation efficiency of the photovoltaic components.

[0090] Specifically, as Figure 2 shown, step S101 in the embodiments of the present invention includes the following steps:

[0091] Step S1011, obtain the daily available hours of the inverter within a preset time period.

[0092] In one embodiment, within a preset time period D, calculate the daily available hours h of the inverter SN i of the inverter SN a , and the calculation method is as follows:

[0093]

[0094] where h a is the daily available hours of SN i , E inv is the daily actual power generation of SN i , P max is the rated power of SN i , and the set h a is obtained.

[0095] Step S1012: Take the number of days greater than the preset available hours threshold as the effective days of the daily available hours of the inverter, and obtain the corresponding historical current data of each inverter within the effective days.

[0096] Based on the available hours obtained above and combined with the preset available hours threshold θ, this embodiment of the present invention screens out the P days with the highest power generation efficiency of SN i , obtains the corresponding historical current data, and performs preprocessing and normalization operations on it. Among them, the available hours threshold θ is dynamically adjusted according to the geographical location of the power station and the annual available hours, and the specific formula is as follows:

[0097]

[0098] where θ represents the available hours threshold, H annual represents the annual available hours, and α represents the coefficient for dynamically correcting the power generation potential according to the geographical location of the distributed power station.

[0099] This embodiment of the present invention introduces a coefficient for dynamically correcting the power generation potential according to the geographical location of the distributed power station, which can reflect the influence of natural conditions such as sunlight and climate in different geographical locations on the power generation capacity of the photovoltaic power station. The solar radiation intensity, sunshine duration, and climate conditions (such as dust, rain, etc.) in different regions are different, which directly affect the available hours of the photovoltaic power station. Through this dynamic coefficient, the available hours threshold can be more in line with the local actual power generation potential, making the subsequent judgment of the effective operation days of the inverter more in line with the actual situation of the power station, helping to more accurately evaluate whether the inverter is in an effective operation state under local conditions, and avoiding overestimating or underestimating the operation efficiency and performance of power stations in some regions due to the adoption of a unified fixed standard.

[0100] Based on the calculated available hours threshold θ, this embodiment of the present invention performs calculations on H aThe valid days in are filtered, and the process is as follows:

[0101]

[0102] Where S represents the time period D The set of valid days greater than θ, d represents the dth day. Further, the set of valid days S is divided into the number of available hours. Arrange in descending order and take the date index of the previous P days. The process is as follows:

[0103]

[0104] Among them, P d =(d1,d2,…,d P ) represents the set of P days with the highest number of available hours, recorded as the effective days. d The valid days in the date correspond to the SN i The historical current data of the present invention is expressed as follows: I = (I1, I2, ..., I N ),in And N and T represent SN respectively. i The number of strings and the number of sampling time points.

[0105] The embodiment of the present invention preferentially selects days with a higher number of available hours as valid data, and the corresponding current data can better represent the performance of the inverter during normal operation. The most representative and reliable data can be screened out from a large amount of data, providing a high-quality data basis for subsequent analysis.

[0106] Step S1013, resample the acquired historical current data based on a preset time granularity, use linear interpolation to fill in missing values, and correct negative currents.

[0107] Specifically, the embodiment of the present invention resamples the historical current data I with a 5-minute granularity, fills the missing values ​​with linear interpolation, and corrects the negative current, for example, changing the negative current to 0. Linear interpolation can not only fill the missing values, but also maintain the continuity of the data, so that the change of the data in the time series is smoother, which conforms to the continuity characteristics of the actual physical process. In the actual collected current data, negative current may appear due to sensor failure, signal interference and other reasons. In the physical sense, the current of the photovoltaic power station is usually positive. Correcting the negative current can correct these obvious data errors, make the data conform to the actual physical laws, and improve the quality and credibility of the data.

[0108] like Figure 3 As shown, step S102 in the embodiment of the present invention specifically includes the following steps:

[0109] S1021, extract the average current and current standard deviation of each string at a preset moment as the current feature vector, and construct the feature vector matrix of the strings;

[0110] Specifically, for the purpose of orientation grouping, the embodiments of the present invention extract the current feature vector F of each string i as the input of the DBSCAN algorithm. The specific formula is as follows:

[0111]

[0112] where represents the average current of each string at time point t, reflecting the overall power generation level, represents the current standard deviation of each string at time point t, reflecting the stability of the string current. Thus, the feature vector matrix of the strings F = (F1, F2,..., F N ) is constructed.

[0113] The embodiments of the present invention extract the average current and current standard deviation of each string at a preset moment as the current feature vector. These two indicators can effectively summarize the key characteristics of the string current. The average current reflects the average power generation ability of the string at a specific moment, while the current standard deviation reflects the fluctuation of the current around the average value. For strings with different orientations, due to factors such as the intensity, angle, and duration of received sunlight, their power generation ability and current stability will vary. Through these two features, these differences can be well captured, thereby providing representative feature information for subsequent clustering analysis and accurately distinguishing strings with different orientations.

[0114] S1022, based on the feature vector matrix, use the DBSCAN algorithm for orientation grouping, and divide multiple strings into multiple orientation categories according to the clustering results.

[0115] Specifically, the DBSCAN algorithm is a density-based clustering algorithm that can effectively identify clusters of any shape and can identify noise points as separate categories. In an actual photovoltaic power station, the distribution of strings with different orientations may not be a simple regular shape but complex and diverse. The embodiments of the present invention can well adapt to this complex distribution situation by using the DBSCAN algorithm, and can automatically determine the appropriate number of orientation categories according to the distribution of the current feature vectors of the strings, making the clustering results more objectively and accurately reflect the actual situation and accurately dividing strings with different orientations into different categories.

[0116] Furthermore, the embodiments of the present invention input the feature vector matrix F into the DBSCAN algorithm for orientation grouping. The specific formula is as follows:

[0117] N ∈ (Fi ) = {F j ∈ F | distance(F i , F j ) ≤ ∈}

[0118] By calculating the Euclidean distance between feature vectors, the ∈-neighborhood of each string is defined. If |N ∈ (F i ) ≥ MinPts|, then F i is a core point, where MinPts represents the minimum number of neighborhood points required to define a point as a core point. If the feature vector F j of the string is located within the ∈-neighborhood of the core point F i , then F j is density-reachable from F i , proving that F j ∈ N ∈ (F i ), that is, F i and F j belong to the same cluster. Starting from the core point, expand the density-reachable points to generate the cluster G k = {F i ∈ F | F i is density-reachable from the core point F h}. If a certain feature vector F i is neither a core point nor belongs to any cluster, it is marked as a noise point.

[0119] According to the clustering results of the DBSCAN algorithm, the N strings are divided into K categories, and the obtained string orientation set can be expressed as G = (G1, G2,..., G K ). Corresponding it with I, we get IG = (IG1, IG2,..., IG K ), where IG k = (I1, I2,..., I Z ), and K and Z represent the number of orientations and the number of strings for each orientation respectively.

[0120] As Figure 4 shown, step S103 in the embodiment of the present invention specifically includes the following steps:

[0121] S1031, calculate the average current of the strings during the large power generation period of the photovoltaic, and mark the strings with an average daily current less than the preset threshold as dropped strings. Specifically, the dropped string detection first calculates the average current of the strings during the large power generation period of the photovoltaic, and the calculation formula is as follows:

[0122]

[0123] Among them, represents the average current value of the i-th string during the photovoltaic high-generation period on the p-th day, and T represents the total number of sampling points during the photovoltaic high-generation period. represents the current value of the i-th string at the t-th sampling point on the p-th day.

[0124] It should be noted that the photovoltaic high-generation period usually refers to the period when the photovoltaic power generation system has a relatively high power generation and a large power generation volume in a day, which is generally closely related to the sunlight intensity and angle of the sun. Generally, it is from 9 am to 3 pm. The actual photovoltaic high-generation period will also be affected by factors such as geographical location, seasonal changes, weather conditions, installation angle and orientation of photovoltaic modules. In the embodiment of the present invention, by calculating the average current of the string during the photovoltaic high-generation period and marking the open circuit by the daily average current being less than the preset threshold, the string with significantly lower power generation ability than the normal level can be accurately located.

[0125] Furthermore, by setting the open-circuit current threshold μ, the daily average current less than μ of the string is marked as an open circuit, and the calculation process is as follows:

[0126]

[0127] where, I drop represents the set of strings with open circuit.

[0128] Furthermore, the set of open-circuit dates P f is determined according to the open-circuit strings, and the string data corresponding to the open-circuit dates is deleted from the valid days. Deleting this part of the data can avoid the deviation of the overall analysis result caused by abnormal data and improve the accuracy of subsequent analysis.

[0129] S1032, perform data cleaning and normalization processing on the current data of the obtained valid days, and convert the current value of each string into a ratio relative to the total current of all strings.

[0130] Specifically, in the embodiment of the present invention, the current value of each string I i is converted into a ratio relative to the total current of all strings, and the calculation process is as follows:

[0131]

[0132] where, I norm,i (t) represents the normalized current value of the i-th string at time t, and I i (t) is the original current value of the i-th string at time t.

[0133] In the embodiments of the present invention, the current data of the remaining valid days is cleaned and normalized, and the current value of each string is converted into a ratio relative to the total current of all strings, enhancing the comparability of the data. Due to factors such as orientation and installation location, the original current values of different strings may vary greatly. After normalization, this difference can be eliminated, enabling the strings in different orientations to be compared and analyzed under a unified standard, and more accurate string performance characteristics can be mined.

[0134] S1033. Based on the conversion of the current value of each string into a ratio relative to the total current of all strings, calculate the normalized current average values for the high-irradiance time period and the non-high-irradiance time period respectively.

[0135] Specifically, calculate the normalized current average values for the high-irradiance time period and the non-high-irradiance time period respectively, and the calculation process is as follows:

[0136]

[0137] Among them, A1 and A0 are the normalized current average values for the high-irradiance and non-high-irradiance time periods respectively, and T1 and T0 are the time sets for the high-irradiance and non-high-irradiance time periods respectively.

[0138] S1034. Use the ratio of the normalized current average value in the non-high-irradiance time period to the normalized current average value in the high-irradiance time period as the string aging judgment index, and take the string aging judgment index being greater than the preset index threshold as the aging condition.

[0139] In the embodiments of the present invention, calculate the string aging judgment index according to the normalized current average value If β i is greater than the aging judgment threshold, it is initially considered that the string has an aging phenomenon. This index comprehensively considers the normalized current average values in the high-irradiance time period and the non-high-irradiance time period. Under different light intensities, the aging performance of photovoltaic modules may be different. Some modules may have more obvious aging problems under high irradiance, while some may also experience performance degradation under low irradiance. By calculating the ratio by comprehensively considering the current average values of these two time periods, the overall performance changes of the string in various light environments can be more comprehensively reflected.

[0140] S1035. Obtain the current average values of each string during the photovoltaic large-generation time period within the preset time period, and perform a descending order sorting of the current average values. If there are strings that are all within the preset range at the back in terms of ranking within the preset number of days, and meet the aging condition continuously for multiple days, then the corresponding strings can be confirmed as aging strings.

[0141] In one example, for instance, when detecting aging strings in a PV power station with 10 strings, the preset time period is one month (30 days), the peak PV power generation time period is from 10 am to 2 pm every day, the preset number of days is 15 days, the preset range for the bottom rankings is the last 3, and the aging condition is that the ratio of the average normalized current in the high-irradiance time period to the average normalized current in the non-high-irradiance time period is greater than 1.5 (the preset index threshold). Within this one month, the current values of the 10 strings during the peak PV power generation time period (10:00 - 14:00) are recorded every day, and then the average current value of each string per day is calculated. For example, on the first day, the average current values of string 1 - string 10 are respectively: 10A, 9A, 8A, 7A, 6A, 5A, 4A, 3A, 2A, 1A.

[0142] After one month, calculate the average current value of each string during the peak PV power generation time period within these 30 days, and sort them in descending order. For example, string 8, string 9, and string 10 are always among the last 3 with the bottom rankings. Calculate the average normalized current values of each string during the high-irradiance time period (such as 11:00 - 13:00 on sunny days) and the non-high-irradiance time period (such as cloudy days or evening periods), and calculate their ratio. Suppose after calculation, the ratio of string 8 is greater than 1.5 for 16 days within these 30 days, the ratio of string 9 is greater than 1.5 for 13 days, and the ratio of string 10 is greater than 1.5 for 14 days, all of which meet the aging condition of being continuously satisfied multiple times (set here as greater than or equal to 10 times) within multiple days. Based on the above data, it can be confirmed that string 8, string 9, and string 10 are aging strings. In this way, through the analysis and judgment of multi-dimensional data, aging strings can be identified more accurately.

[0143] S1036, delete the dropped strings and aging strings from the recognition results of strings with different orientations.

[0144] Specifically, after identifying the dropped strings and aging strings, delete them from the list of strings with different orientations. By cleaning the abnormal detection results of strings, ensure the integrity and data quality of the strings in each orientation.

[0145] In the embodiment of the present invention, for the set of strings IG = (IG1, IG2,..., IG K ) with different orientations in step S104, calculate the orientation of the strings respectively, and the calculation formula for the dispersion rate is as follows:

[0146]

[0147] Where is the string dispersion rate of the kth orientation of the distributed PV at time t, i is the identifier of the string, N is the number of strings, is the current data of the ith string at time t, is the standard deviation of the current data at time t, is the average current of all strings at time t.

[0148] After obtaining the discrete rates at multiple time points in the embodiments of the present invention, the average value of the discrete rates within the time period is taken to obtain the daily discrete rate of the string in this orientation. The calculation process is as follows:

[0149]

[0150] where, represents the discrete rate of the k-th orientation on the p-th day, and T p is the set of valid time points on the p-th day. After calculating the discrete rates of multiple groups of strings in this orientation, the weights are comprehensively calculated according to the normalized current of each group, and then the comprehensive weighted discrete rate of SN i is calculated. The calculation process is as follows:

[0151]

[0152] where, w k represents the weight of the k-th group, and δ p represents the discrete rate of SN i on the p-th day.

[0153] Furthermore, using the string discrete rate and the operating state parameters of SN i , the cleaning index is calculated through the following formula:

[0154]

[0155] where, CI i represents the cleaning index of the inverter SN i , γ represents the difference correction coefficient of the component inclination orientation, P represents the number of valid days for calculation, δ p represents the discrete rate of the inverter SN i on the p-th day, represents the available hours of the inverter SN i on the p-th day, represents the available hours of the reference inverter on the p-th day; where the reference inverter is selected as the benchmark by evaluating and comparing the performance of the inverters in neighboring power stations and choosing the inverter with the best power generation performance and stable operation.

[0156] The calculation method of the cleaning index in the embodiments of the present invention takes into account the differences in the inclination directions of components, the number of effective days, the discrete rate of the inverter, the available hours of the inverter, and the available hours of the reference inverter at the same time. It organically combines multiple data dimensions, makes full use of various data information during the operation of the inverter, makes the calculation of the cleaning index more scientific and comprehensive, can more accurately explore the impact of cleaning factors on the operation of the photovoltaic power station, and provides a unified quantitative standard for the cleaning status of the photovoltaic power station.

[0157] Further, the average value of the cleaning indexes of each inverter is taken as the cleaning index of the photovoltaic power station. For example, the set of cleaning indexes CI of M inverters in the power station is CI=(CI1, CI2, …, CI M ), and its average value is taken as the cleaning index of the distributed photovoltaic. Taking the average value of the cleaning indexes of each inverter as the cleaning index of the photovoltaic power station can reflect the cleaning degree of the photovoltaic power station as a whole, avoid the influence of the special situation of a single inverter on the overall evaluation, and use a comprehensive value to represent the cleaning level of the entire power station. This quantitative index is convenient for operation and maintenance personnel to compare the cleaning status between photovoltaic power stations in different time periods and different regions, and helps to formulate reasonable cleaning and maintenance plans and resource allocation schemes.

[0158] As Figure 5 shown, step S105 in the embodiments of the present invention specifically includes the following steps:

[0159] S1051, obtain the weather data of the historical preset number of days and the predicted weather data of the future preset number of days. The weather data includes: rainfall, air quality index, and weather description;

[0160] S1052, use the weather data of the historical preset number of days and the predicted weather data of the future preset number of days as the weather data set.

[0161] Embodiments of the present invention simultaneously obtain weather data for a preset number of historical days and predicted weather data for a preset number of future days, providing a comprehensive perspective on weather information for the operation and maintenance of photovoltaic power plants. Historical weather data can reflect the long-term patterns and regularities of weather changes in the region, such as rainfall frequency and air quality fluctuations over a certain period in the past, helping operation and maintenance personnel understand the past operating performance of the photovoltaic power plant under different weather conditions. The future predicted weather data enables the operation and maintenance team to plan in advance and make forward-looking decisions for upcoming weather conditions. The granularity of the original weather data is one hour. Embodiments of the present invention group the data by date, calculate the total daily rainfall, and count the number of various weather descriptions. The final weather description for each day is determined through ratio analysis. By incorporating multiple data dimensions such as rainfall, air quality indicators, and weather descriptions, a comprehensive understanding of the impact of weather on the photovoltaic power plant can be obtained. Rainfall affects the cleanliness of the component surface. Excessive rainfall may cause soil erosion and affect the power plant foundation, while too little may cause dust accumulation and affect power generation efficiency. Air quality indicators reflect the content of pollutants in the air and are directly related to the degree of contamination of photovoltaic components. Weather descriptions cover various weather conditions such as sunny, cloudy, and overcast. Different weather conditions have different light intensities, which affect the power generation power of the power plant.

[0162] As Figure 6 shown, in the embodiments of the present invention, step S106 specifically includes the following steps:

[0163] S1061, Screen out the dates with a cleaning index lower than the cleaning threshold from the multi-day cleaning index data as potential starting dates for the cleaning cycle.

[0164] Embodiments of the present invention can accurately identify the time points when the cleaning degree of the photovoltaic power plant components has affected the power generation efficiency by screening out the dates with a cleaning index lower than the preset threshold as potential starting dates for the cleaning cycle. The cleaning index comprehensively reflects the impact of the cleaning condition of the components on the power generation performance. Based on this, it can ensure that when the dirtiness of the components is sufficient to reduce the power generation efficiency, cleaning is arranged in a timely manner, minimizing the power generation loss caused by dirtiness and improving the overall power generation efficiency of the photovoltaic power plant.

[0165] S1062, For potential cleaning dates, obtain the weather data within the time window {W d-c ,W d+c}. Let d represent the current day, d - c and d + c represent the historical c days and the future c days respectively. If there is no more than a set threshold for rainfall and the air quality does not meet the preset excellent quality condition within the time window {W d-c ,W d+c}, then determine the recommended starting date D clean(Wd) for the cleaning cycle. If within the time window {W d-c ,W d+cIf the rainfall within {W} exceeds the set threshold and the air quality reaches the preset excellent quality condition, the cleaning date will be postponed by c days, and the weather data within the new time window {W} will be rechecked until the rainfall within the time window does not exceed the set threshold and the air quality does not reach the preset excellent quality condition, and finally the recommended start date D of the cleaning cycle is determined. d ,W d+2c} until the rainfall within the time window does not exceed the set threshold and the air quality does not reach the preset excellent quality condition, and finally the recommended start date D of the cleaning cycle is determined. clean 。

[0166] Furthermore, as Figure 7 shown, the cleaning date is postponed by c days, the cleaning index for W days is calculated, and the weather data within the new time window {W} is checked. If the cleaning index within the time window is lower than the threshold, and the rainfall within the time window does not exceed the set threshold and the air quality reaches the preset excellent quality condition, the recommended start date D of the cleaning cycle is finally determined. d+c If the cleaning index within the time window is higher than the threshold, or the rainfall within the time window does not exceed the set threshold, the cleaning date is postponed by c days again. d ,W d+2c} until the rainfall within the time window does not exceed the set threshold and the air quality does not reach the preset excellent quality condition, and finally the recommended start date D of the cleaning cycle is determined. clean If the cleaning index within the time window is higher than the threshold, or the rainfall within the time window does not exceed the set threshold, the cleaning date is postponed by c days again.

[0167] In the embodiment of the present invention, the rainfall and weather description for the next 7 days are obtained through weather forecast information, and the weather data for the historical 7 days are obtained, that is, c is 7. For the potential cleaning date, the cleaning time is adjusted according to the weather data within the time window. If there is rainfall exceeding the set threshold and the air quality reaches the preset excellent quality condition within the time window (for example, when the air quality is at mild pollution or below, it does not reach the excellent quality condition), the cleaning date is postponed because the rainfall at this time may naturally clean the dirt on the surface of the components, and good air quality means that new pollutants will not accumulate quickly on the components in the short term. Postponing the cleaning can avoid unnecessary manual cleaning costs and achieve a cleaning effect using natural rainfall, further optimizing the power generation efficiency.

[0168] S1063. Determine the recommended cleaning cycle time period of the photovoltaic power station according to the corresponding relationship between the preset installed capacity and the recommended cleaning cycle.

[0169] In one embodiment, the cleaning cycle time period is divided according to the power station capacity P station , and the division standard is as follows:

[0170]

[0171] In the embodiment of the present invention, the recommended cleaning cycle time period is determined according to the corresponding relationship between the preset installed capacity and the recommended cleaning cycle, taking into account the differences of power stations of different scales. For large-scale power stations, due to their large installed capacity and numerous components, the cleaning work is more complex and has a greater impact on power generation efficiency. Therefore, a more frequent or reasonably planned cleaning cycle may be required; while small-scale power stations are relatively flexible. This capacity-based division method can ensure that the setting of the cleaning cycle conforms to the characteristics of the power station itself, improving the pertinence and effectiveness of the cleaning work.

[0172] S1064, when there is cloudy or overcast weather during the recommended cleaning cycle time period, then take the cloudy or overcast weather D cloudy as the recommended date (i.e., D clean+1 ).

[0173] In the embodiment of the present invention, when there is preset adverse weather (such as heavy rain, sand and dust, etc. weather that may affect the cleaning effect or cause damage to equipment) during the recommended cleaning cycle time period, avoid these adverse weather as the recommended date, ensuring the safety and effectiveness of the cleaning work. Conducting cleaning under adverse weather may not only fail to achieve the expected cleaning effect, but also may damage the photovoltaic modules and cleaning equipment, increasing the operation and maintenance costs and safety risks. By reasonably avoiding, the quality and efficiency of the cleaning work are improved. Selecting cloudy or overcast weather as the recommended date makes full use of the favorable factors of this kind of weather for the cleaning work, so as to avoid cleaning on sunny days with high photovoltaic power generation efficiency, thereby reducing the economic loss of power generation caused by cleaning downtime.

[0174] S1065, obtain the historical cleaning records, and take the date that does not conflict with the historical cleaning records as the cleaning date. If the recommended date does not conflict with the historical cleaning records, then readjust the time window and determine the recommended start date.

[0175] In the embodiment of the present invention, by referring to the historical records, the operation and maintenance personnel can understand the past cleaning situation, prevent cleaning the same area or components multiple times in a short period, saving manpower, material resources and time resources. At the same time, avoiding conflicts with the historical cleaning records helps to maintain the coherence and planning of the cleaning work, making the entire operation and maintenance process more orderly. If the recommended date conflicts with the historical cleaning records, then readjust the time window and determine the recommended start date. This dynamic adjustment mechanism can continuously optimize the cleaning strategy according to the actual situation. By combining historical experience and current various factors, the determined cleaning cycle is made more scientific and reasonable, better adapting to the actual operation requirements of the photovoltaic power station and improving the refined level of operation and maintenance management.

[0176] The method provided by the embodiment of the present invention has achieved remarkable effects when applied to a distributed photovoltaic power station: the DBSCAN algorithm is used to detect the string orientation of the inverter, effectively reducing the problem of large dispersion rate caused by orientation differences, and improving the calculation accuracy by 10%-15% compared with the prior art. A string anomaly detection mechanism is introduced to accurately identify abnormal strings, and the identification accuracy reaches more than 95%, avoiding the problem of distorted dispersion rate calculation caused by abnormal current data.

[0177] In this embodiment, a cleaning cycle recommendation system for a photovoltaic power station is also provided. This system is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that realizes a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0178] This embodiment provides a cleaning cycle recommendation system for a photovoltaic power station, as Figure 8 shown, including:

[0179] A historical data acquisition module 801, configured to acquire historical current data of each inverter of the photovoltaic power station and perform preprocessing;

[0180] An orientation identification module 802, configured to identify multiple types of orientations of the inverter strings based on the preprocessed historical current data by using a preset clustering algorithm;

[0181] An anomaly detection module 803, configured to delete abnormal strings by performing anomaly detection on the multiple types of orientation strings obtained by identification, and acquire the current data of normal strings with different orientations;

[0182] A cleaning index calculation module 804, configured to calculate the string dispersion rate of the photovoltaic based on the current data of the normal strings with different orientations, and calculate the cleaning index based on the string dispersion rate and the operating state parameters of the inverter;

[0183] A weather data acquisition module 805, configured to acquire weather data of the geographical location where the photovoltaic power station is located and perform preprocessing;

[0184] A cleaning cycle recommendation module 806, configured to screen dates based on the cleaning index and the preprocessed weather data as the starting date of the initially recommended cleaning cycle, divide the cleaning cycle in combination with the power station capacity, and finally determine the recommended cleaning cycle based on the weather data and historical cleaning records.

[0185] In some alternative implementation manners, the historical data acquisition module 801 includes:

[0186] Daily available hours acquisition unit, configured to acquire the daily available hours of the inverter within a preset time period;

[0187] Historical current acquisition unit within valid days, configured to use the number of days greater than the preset available hours threshold as the valid days of the daily available hours of the inverter, and acquire the corresponding historical current data of each inverter within the valid days;

[0188] Data preprocessing unit, configured to resample the acquired historical current data based on a preset time granularity, use linear interpolation to fill in missing values, and perform correction processing on negative current values.

[0189] In some optional embodiments, the preset available hours threshold is calculated by the following formula:

[0190]

[0191] where θ represents the available hours threshold, H annual represents the annual available hours, and α represents a coefficient for dynamically correcting the power generation potential according to the geographical location of the distributed power station;

[0192] The valid days of the daily available hours of the inverter are obtained by the following formula:

[0193]

[0194] where S represents the set of valid days within the time period D greater than θ, d represents the d-th day, and S is sorted in descending order according to the available hours The dates of the first P days are used as the valid days.

[0195] In some optional embodiments, the orientation recognition module 802 includes:

[0196] Current feature vector extraction unit, configured to extract the average current and current standard deviation at a preset moment of each string as the current feature vector, and construct the feature vector matrix of the string;

[0197] Orientation category recognition unit, configured to perform orientation grouping on the basis of the feature vector matrix using the DBSCAN algorithm, and divide multiple strings into multiple orientation categories according to the clustering results.

[0198] In some optional embodiments, the anomaly detection module 803 includes:

[0199] String dropout recognition unit, configured to calculate the average current of the strings during the large power generation period of the photovoltaic system, and mark the strings with a daily average current less than the preset threshold as dropped strings;

[0200] The aging string identification unit is used to clean and normalize the current data with the obtained valid days, and convert the current value of each string into a proportion relative to the total current of all strings;

[0201] Based on the proportion of the current value of each string converted to the total current of all strings, calculate the normalized current average values in the high-irradiance time period and the non-high-irradiance time period respectively;

[0202] Taking the ratio of the normalized current average value in the high-irradiance time period to the normalized current average value in the non-high-irradiance time period as the string aging judgment index, and taking the string aging judgment index being greater than the preset index threshold as the aging condition;

[0203] Obtain the current average value of each string in the photovoltaic peak power generation time period within the preset time period, and sort the current average values in descending order. If there are strings that are all within the preset range at the end of the ranking within the preset number of days, and the aging condition is continuously met multiple times within multiple days, then the corresponding strings can be confirmed as aging strings;

[0204] The abnormal string deletion unit is used to delete the missing strings and aging strings from the string identification results in different orientations.

[0205] In some alternative embodiments, the cleanliness index is calculated by the following formula:

[0206]

[0207] where CI i represents the cleanliness index of the inverter SN i , γ represents the difference correction coefficient of the component tilt orientation, P represents the valid days for calculation, δ p represents the discretization rate of the inverter SN i on the p-th day, represents the available hours of the inverter SN i on the p-th day, represents the available hours of the reference inverter on the p-th day;

[0208] Taking the average value of the cleanliness indexes of each inverter as the cleanliness index of the photovoltaic power station.

[0209] In some alternative embodiments, the weather data acquisition module 805 includes:

[0210] The weather data acquisition unit is used to acquire the weather data of the historical preset days and the predicted weather data of the future preset days. The weather data includes: rainfall, air quality index, and weather description;

[0211] A weather data set generation unit, which is used to use the weather data of a preset number of historical days and the predicted weather data of a preset number of future days as a weather data set.

[0212] In some optional embodiments, the cleaning cycle recommendation module 806 includes:

[0213] An initial cleaning cycle start date generation unit, which is used to screen out the dates lower than the cleaning threshold from the multi-day cleaning index data as potential cleaning cycle start dates;

[0214] A recommended start date generation unit for the cleaning cycle, which is used to obtain the weather data within the time window {W d-c ,W d+c} for potential cleaning dates. d represents the current day, d-c and d+c respectively represent the historical c days and the future c days. If there is rainfall within the time window {W d-c ,W d+c} that does not exceed the set threshold and the air quality does not meet the preset excellent quality condition, then determine the recommended start date D clean(Wd) of the cleaning cycle. If there is rainfall within the time window {W d-c ,W d+c} that exceeds the set threshold and the air quality meets the preset excellent quality condition, then postpone the cleaning date by c days and re-check the weather data within the new time window {W d ,W d+2c} until the rainfall within the time window does not exceed the set threshold and the air quality does not meet the preset excellent quality condition, and finally determine the recommended start date D clean .

[0215] A recommended cleaning cycle time period generation unit, which is used to determine the recommended cleaning cycle time period of the photovoltaic power station according to the corresponding relationship between the preset installed capacity and the recommended cleaning cycle;

[0216] A recommended date generation unit, which is used to use the cloudy or overcast weather D cloudy as the recommended date when there is cloudy or overcast weather within the recommended cleaning cycle time period;

[0217] A recommended date adjustment unit, which is used to obtain the historical cleaning records and use the date that does not conflict with the historical cleaning records as the cleaning date. If the recommended date does not conflict with the historical cleaning records, then re-adjust the time window and determine the recommended start date.

[0218] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be elaborated here.

[0219] The cleaning cycle recommendation system of the photovoltaic power station in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0220] An embodiment of the present invention further provides a computer device having the above Figure 8 shown cleaning cycle recommendation system of the photovoltaic power station.

[0221] Please refer to Figure 9 , Figure 9 which is a schematic structural diagram of the computer device provided by an optional embodiment of the present invention. As Figure 9 shown, the computer device includes: one or more processors 10, a memory 20, and an interface for connecting each component, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (such as an array of servers, a set of blade servers, or a multi-processor system). Figure 9 In

[0222] Figure 16, one processor 10 is taken as an example.

[0223] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0224] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the computer device and the like. In addition, the memory 20 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0225] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, a hard disk, or a solid-state drive; the memory 20 may further include a combination of the above types of memory.

[0226] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0227] The embodiments of the present invention further provide a computer-readable storage medium. The methods according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and to be stored in a local storage medium, so that the methods described herein can be stored in such software processes on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above types of memory. It can be understood that a computer, a processor, a microprocessor central control system, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods shown in the above embodiments are implemented.

[0228] A part of the present invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can call or provide the methods and / or technical solutions according to the present invention through the operations of the computer. Those skilled in the art should understand that the forms in which computer program instructions exist in a computer-readable medium include but are not limited to source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include but are not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible by the computer.

[0229] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for recommending a cleaning cycle of a photovoltaic power station, characterized in that: include: Obtain historical current data of each inverter in the photovoltaic power station and perform preprocessing; Based on the pre-processed historical current data, a preset clustering algorithm is used to identify multiple directions of inverter strings; By performing abnormal detection on the identified strings of multiple orientations, abnormal strings are deleted, and current data of normal strings in different orientations are obtained; Calculating a photovoltaic string discrete rate based on the current data of the normal strings in different directions, and calculating a cleaning index based on the string discrete rate and an operating state parameter of an inverter; Obtain weather data for the geographical location of the photovoltaic power station and perform preprocessing; The date is screened based on the cleaning index and the pre-processed weather data as the preliminary recommended start date of the cleaning cycle, and the cleaning cycle is divided in combination with the power station capacity, and the recommended cleaning cycle is finally determined based on the weather data and historical cleaning records.

2. The method according to claim 1, characterized in that The obtaining and preprocessing of historical current data of each inverter of the photovoltaic power station includes: Get the daily available hours of the inverter within a preset time period; The days that are greater than the preset available hours threshold are used as the effective days of the inverter's daily available hours, and the corresponding historical current data of each inverter within the effective days are obtained; The acquired historical current data is resampled based on the preset time granularity, and linear interpolation is used to fill in missing values, and negative currents are corrected.

3. The method according to claim 2, characterized in that The preset hour threshold can be calculated using the following formula: Among them, θ represents the threshold of available hours, H annual represents the annual available hours, and α represents the coefficient for dynamic correction of power generation potential according to the geographical location of the distributed power station; The effective number of days that the inverter can be used every day is obtained by the following formula: Where S represents the time period D The set of valid days greater than θ, d represents the dth day, and S is calculated according to the number of available hours. Arrange in descending order, taking the dates of the previous P days as valid days.

4. The method according to claim 1, characterized in that: The method of identifying multiple directions of inverter strings based on the preprocessed historical current data using a preset clustering algorithm includes: Extract the current average value and current standard deviation of each string at a preset time as the current feature vector, and construct the feature vector matrix of the string; The DBSCAN algorithm is used to perform orientation grouping based on the eigenvector matrix, and the plurality of group strings are divided into a plurality of orientation categories according to the clustering result.

5. The method according to claim 1, characterized in that The method of performing abnormal detection on the identified multiple types of orientation strings to delete abnormal strings includes: Calculate the average current of the photovoltaic strings during the peak photovoltaic power generation period, and mark the strings with daily average current less than the preset threshold as dropped strings; The current data of valid days are cleaned and normalized, and the current value of each string is converted into a ratio relative to the total current of all strings; Based on the current value of each string converted into a ratio relative to the total current of all strings, the normalized current average values ​​of the high irradiation time period and the non-high irradiation time period are calculated respectively; The ratio of the normalized current average value in the non-high irradiation time period to the normalized current average value in the high irradiation time period is used as a string aging judgment index, and the string aging judgment index being greater than a preset index threshold is used as an aging condition; The current average value of each string in the photovoltaic peak time period within the preset time period is obtained, and the current average values ​​are sorted in descending order. If there are strings that are in the preset range of the lower ranking within the preset number of days, and the aging condition is met for multiple consecutive days, the corresponding string can be confirmed as an aging string; The dropped strings and aged strings are deleted from the string identification results of different orientations.

6. The method according to claim 1 or 3, characterized in that: The cleaning index is calculated by the following formula: Among them, CI i Indicates inverter SN i The cleaning index, γ represents the difference correction factor of the component inclination direction, P represents the effective days used for calculation, δ p Indicates the inverter SN on day p i The discrete rate, Indicates the inverter SN on day p i The number of hours available, Indicates the benchmark inverter on day p The number of hours available; The average value of the cleanliness index of each inverter is taken as the cleanliness index of the photovoltaic power station.

7. The method according to claim 1, characterized in that The obtaining of weather data of the geographical location of the photovoltaic power station and preprocessing the data include: Obtain weather data for a preset number of historical days and forecast weather data for a preset number of future days. Weather data includes: rainfall, air quality index and weather description; The weather data of the historical preset number of days and the predicted weather data of the future preset number of days are taken as the weather data set.

8. The method according to claim 7, characterized in that The screening date based on the cleaning index and the pre-processed weather data as the preliminary recommended cleaning cycle start date includes: Filter the dates below the cleaning threshold from the multi-day cleaning index data as potential cleaning cycle start dates; For potential cleaning dates, get the time window {W d-c ,W d+c }, d represents the current day, dc and d+c represent the historical c days and the future c days respectively. If the time window {W d-c ,W d+c If the rainfall does not exceed the set threshold and the air quality does not reach the preset good quality condition, the recommended start date of the cleaning cycle is determined. clean(Wd) , if the time window {W d-c ,W d+c If the rainfall exceeds the set threshold and the air quality reaches the preset good quality condition, the cleaning date will be postponed for c days and the new time window will be rechecked. d ,W d+2c } until the rainfall in the time window does not exceed the set threshold and the air quality does not reach the preset good quality condition, and finally determine the recommended start date D of the cleaning cycle clean .

9. The method according to claim 8, characterized in that The cleaning cycle is divided according to the power plant capacity, and the recommended cleaning cycle is finally determined based on weather data and historical cleaning records, including: Determine the recommended cleaning cycle time period of the photovoltaic power station according to the correspondence between the preset installed capacity and the recommended cleaning cycle; When there is cloudy or overcast weather during the recommended cleaning cycle time period, the cloudy or overcast weather D cloudy As a recommended date; Obtain historical cleaning records, and use the recommended date as the cleaning date if it does not conflict with the historical cleaning records. If the recommended date does not conflict with the historical cleaning records, readjust the time window and determine the recommended start date.

10. A cleaning cycle recommendation system for a photovoltaic power station, characterized in that: include: A historical data acquisition module is used to acquire historical current data of each inverter of a photovoltaic power station and perform preprocessing; An orientation recognition module, used to identify multiple orientations of inverter strings using a preset clustering algorithm based on preprocessed historical current data; An abnormality detection module is used to delete abnormal strings by performing abnormality detection on the identified multiple types of orientation strings respectively, and obtain current data of normal strings in different orientations; A clean index calculation module, used to calculate the photovoltaic string discrete rate based on the current data of the normal strings in different directions, and calculate the clean index based on the string discrete rate and the operating state parameters of the inverter; The weather data acquisition module is used to acquire the weather data of the geographical location of the photovoltaic power station and perform preprocessing; The cleaning cycle recommendation module is used to filter the date based on the cleaning index and the pre-processed weather data as the preliminary recommended cleaning cycle start date, divide the cleaning cycle in combination with the power station capacity, and finally determine the recommended cleaning cycle based on the weather data and historical cleaning records.

Images