Low-cost photovoltaic low-efficiency string identification analysis and inspection cleaning operation and maintenance method and device

Abstract

The invention discloses a low-cost photovoltaic low-efficiency string identification analysis and polling cleaning operation and maintenance method and device. The method mainly comprises the steps of benchmark string selection, benchmark string generating capacity analysis, non-benchmark string abnormity identification and polling cleaning linkage, cleaning plan optimization and the like. According to the method and the device, the daily relative power generation amount is calculated, and transverse sorting and longitudinal slope analysis are combined, so that accurate identification of short-term abnormality, long-term attenuation and limited-voltage photovoltaic strings is realized; only group strings with short-term abnormity and individual component defects are subjected to linkage unmanned aerial vehicle inspection, so that invalid operation is avoided, and the operation and maintenance cost is reduced; and in combination with future electric quantity loss caused by dust coverage and weather prediction, the cleaning plan is optimized, the cleaning time is dynamically adjusted or cleaning is canceled, the operation and maintenance cost is further reduced, and the overall power generation efficiency of the power station is improved.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic operation and maintenance technology, and in particular to a low-cost method and system for identifying, analyzing, inspecting, cleaning and maintaining inefficient photovoltaic strings. Background Technology

[0002] Currently, photovoltaic power plants are developing towards large-scale and intensive development, and the power generation efficiency of the power plant strings directly determines the overall capacity of the power plant.

[0003] To improve the power generation efficiency of photovoltaic (PV) strings, large-scale PV power plants now widely use drones equipped with visible light and thermal infrared sensors combined with AI algorithms to identify defects such as cracks and hot spots in the modules, as well as surface contamination caused by dust, bird droppings, and fallen leaves. For surface contamination, semi-automatic cleaning vehicles and fully automatic cleaning robots replace manual cleaning. However, drone inspections and module cleaning frequency still face challenges such as a lack of standardization and high operation and maintenance costs. With the continuous upgrading of operation and maintenance needs, accurately identifying inefficient strings, reducing operation and maintenance costs, and avoiding ineffective inspections and cleaning operations have become core requirements in the operation and maintenance field of large-scale PV power plants. This has also driven the transformation of PV operation and maintenance technology from "fixed-cycle operation and maintenance" to "data-driven precision operation and maintenance." Summary of the Invention

[0004] This disclosure provides a low-cost method and system for identifying, analyzing, inspecting, cleaning, and maintaining inefficient photovoltaic strings in large-scale photovoltaic power plants. It addresses issues such as imperfect identification and analysis of inefficient photovoltaic strings, lack of scientific rigor in inspection and cleaning plans, and misjudgment of power-limited components in existing technologies.

[0005] The low-cost photovoltaic inefficient string identification, analysis, inspection, cleaning, and maintenance method disclosed herein mainly includes the following steps: S1, Select benchmark strings to represent the normal operating status of different areas of the entire site; perform cleaning and maintenance on the benchmark strings; S2, perform power generation analysis on the selected benchmark string, including: anomaly analysis, benchmark calibration, and relative power generation calculation; S3, based on a set threshold, identifies non-benchmark strings with abnormal power generation, and coordinates with drones for inspection or string cleaning based on the identification results; S4, Cleaning plan optimization.

[0006] Furthermore, in step S1, the specific method for selecting the benchmark string includes: Based on the power station's layout, three differentiated areas are divided into east, central and west, or south, central and north, to ensure that the number of inverters and string distribution density covered by each area are basically the same. At least one inverter string with no historical fault records, the same component installation age, and far away from the source of obstruction is randomly selected from each region and used as the benchmark string for that region.

[0007] Furthermore, in step S1, the specific method for ensuring the cleaning of the benchmark string includes: Each benchmark string is equipped with a low-cost cleaning robot adapted to the photovoltaic module type, and a high-frequency cleaning frequency is set; the cleaning time is fixed in the early morning when there is no sunlight and no power generation operation. Meanwhile, a dust sensor is installed at the end of the cleaning robot to automatically detect the dust coverage rate on the component surface after each cleaning. It is necessary to ensure that the coverage rate is ≤ the set threshold and that there are no stubborn stains. This ensures that the benchmark string is always in a standardized clean operating state, and its power generation serves as a reliable benchmark value for comparing the efficiency of all strings in the field.

[0008] Furthermore, in step S2, the specific steps of the power generation analysis include: The power plant's SCADA system collects the actual power generation of the benchmark string and the actual power generation of all other strings in the plant, Pi, i=1,2,...n, where n is the total number of strings in the plant minus the number of benchmark strings. It also collects the daily weather data and maintenance record data. The system determines whether the benchmark strings are abnormal by verifying the cleanliness status, checking the baseline deviation of power generation, and cross-verifying multiple benchmarks. If two or more benchmark strings in different areas are abnormal, the system retrieves power grid monitoring data, SCADA system communication logs, and images of the benchmark string areas to prioritize identifying common interference factors. If common interference factors are identified, data is re-collected after the interference is eliminated. If no common interference is identified, the "historical benchmark value replacement scheme" is activated, which uses the average power generation of benchmark strings under the same weather conditions over a past period as a temporary benchmark value. At the same time, an alert is triggered on the operation and maintenance terminal, prompting manual on-site verification of the benchmark string status. Benchmark string baseline calibration: If all benchmark strings are normal, the average of the actual power generation of each benchmark string on the day is taken as the baseline value P0; if only one benchmark string in a region is abnormal, the average of the actual power generation of the remaining normal benchmark strings on the day is taken as the baseline value P0; if the historical baseline replacement scheme is used, the calculated historical average power generation is taken as the temporary baseline value P0_temp, and the baseline value is recalibrated after the benchmark strings return to normal. Relative power generation calculation: Calculate the daily relative power generation of each non-benchmark string: R_i = Pi / P0; if a temporary base value P0_temp is used, then R_i = Pi / P0_temp.

[0009] Furthermore, step S3 specifically includes: (1) Horizontal comparison Sort all non-benchmark strings in descending order of R_i, filter out strings with R_i < set threshold, and record them as "strings with low power generation". When the number of selected strings is small, they are marked as "inefficient candidate strings for the day"; if the number of selected results is large, longitudinal analysis is performed. (2) Vertical comparison For each non-benchmark string, R_i data was continuously collected for 30 days to construct a "daily relative power generation - time" variation curve; Outliers in the curve are removed using the “3σ principle”: calculate the mean μ and standard deviation σ of R_i over 30 days, and remove values ​​that are <μ-3σ or >μ+3σ. For the curve after removing outliers, the downward slope k is calculated by linear regression: k = ΔR / Δt, where ΔR is the difference between R_i on two adjacent days, and Δt is the time interval in days; Sort all strings in descending order of the absolute value of k, filter out strings whose |k|> set threshold, and mark them as "long-term decay candidate strings"; (3) Power rationing identification From the "strings with low power generation", select the strings whose power generation has decreased due to grid curtailment, and exclude them from the list of abnormal candidate strings to avoid wasting unnecessary operation and maintenance costs: If a string R_i < 0.9 and satisfies "there is a power grid rationing instruction on the day" and "there is no maintenance record for the string", then the string is determined to be a "power rationing string". (4) Judgment of abnormal string type If a string is only marked as a "single-day inefficient candidate string" and is not a power-limited string, it is determined to be a "short-term abnormal string", which is suspected to be a temporary component failure, and drone inspection is initiated. If a string is simultaneously marked as a "candidate string for daily inefficiency" and a "candidate string for long-term degradation," and it is not a string subject to power rationing, then further judgment is made: if other strings under the inverter are normal and only this string is abnormal, it is determined to be a "string with individual component defects," and drone inspection is initiated; if multiple strings in the same area are abnormal, it is determined to be a "string covered by dust in the area," and a string cleaning plan is initiated.

[0010] Furthermore, step S4 specifically includes: (1) Calculation of power loss For strings identified as "area dust-covered strings", calculate the theoretical power loss for the next 3 days:

[0011] Where R_imax is the maximum value of R_i within the past 3 months, R_current is the current regional string average relative power generation R_i, k is the k-th day in the next 3 days, and P0 mk The power generation of the m-th benchmark string on the k-th day is calculated based on the irradiance forecast from the weather forecast. (2) Set threshold Set a power loss threshold ΔE0. If ΔE ≥ ΔE0, then trigger the cleaning plan. (3) Cleaning judgment Access weather forecast data to obtain weather data for the next 3 days: a. If a sandstorm occurs within the next 3 days, the cleaning will be delayed until 1 day after the sandstorm ends; b. If there is heavy rain in the next 3 days, this cleaning plan will be cancelled; c. If there are no special weather conditions, immediately start the cleaning robot to clean the "dust-covered clusters in the area".

[0012] A low-cost photovoltaic inefficient string identification, analysis, inspection, cleaning, and maintenance device applying the above method is characterized by mainly comprising: The benchmark string configuration module is used to select benchmark strings and ensure their cleanliness. The power generation analysis module is used to analyze the power generation of the selected benchmark strings, including: anomaly analysis, benchmark calibration, and relative power generation calculation. The anomaly identification and linkage module is used to identify non-benchmark strings with abnormal power generation based on a set threshold, and to link drone inspection or string cleaning according to the identification results. The cleaning plan optimization module is used to optimize the cleaning plan.

[0013] Compared with the prior art, the beneficial effects of this disclosure are: 1) Improved accuracy of inefficient string identification: Through multi-layer logic of “benchmark string anomaly pre-verification + benchmark value calculation + relative power generation”, the interference of the benchmark string’s own anomalies on the analysis results is further eliminated. Combined with horizontal analysis, vertical analysis and power curtailment identification algorithm, the false judgment rate of inefficient strings is greatly reduced, and short-term anomalies, long-term decay and power curtailment strings are accurately distinguished. 2) Reduced operation and maintenance costs: On-demand drone inspection (only for abnormal clusters), reducing more than 90% of invalid inspection work; Cleaning plans are formulated based on weather and power loss to avoid "over-cleaning" and greatly reduce the energy consumption of cleaning robots; 3) Improved power generation efficiency: By timely identifying component defects and avoiding power loss due to dust, the annual power generation of a 100MW photovoltaic power station can be increased by 2%-3%; 4) Strong cost controllability: By adopting "low-cost cleaning robots + existing SCADA system", no new high-cost hardware equipment is required, and the annual operation and maintenance cost of a single 100MW power plant is reduced by RMB150,000 to RMB200,000. Attached Figure Description

[0014] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments of this disclosure taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same components.

[0015] Figure 1 This is an overall flowchart of an exemplary embodiment of the present disclosure. Detailed Implementation

[0016] Preferred embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0017] This disclosure provides a low-cost method for identifying, analyzing, inspecting, cleaning, and maintaining inefficient photovoltaic (PV) strings. In one exemplary embodiment, the method and system for identifying, analyzing, inspecting, cleaning, and maintaining inefficient PV strings in 100MW and above PV power plants includes four modules: a benchmark string configuration module, a power generation analysis module, an anomaly identification and linkage module, and a cleaning plan optimization module. The flowchart of this module is attached. Figure 1 As shown: 1. Benchmark string configuration First, benchmark strings are selected. Among the inverter strings in the photovoltaic power station, three strings are chosen as benchmark strings based on the core principle of "uniform coverage across the entire area and avoidance of local environmental interference." In practice, the power station's layout is first divided into three differentiated areas: east, central, and west (or south, central, and north), ensuring that the number of inverters and string distribution density in each area are basically consistent. Then, from each area, one inverter string with no historical fault records, the same installation age, and far from shading sources (such as trees, buildings, or utility poles) is randomly selected as the benchmark string for that area. This ensures that the three benchmark strings can represent the normal operating status of different areas throughout the entire site, avoiding distortion of the baseline value due to local environmental differences (such as dust accumulation in edge areas and more stable sunlight in central areas).

[0018] Then, the benchmark strings are cleaned and maintained. One low-cost cleaning robot adapted to the photovoltaic module type is configured for each of the three benchmark strings, with a high-frequency cleaning frequency of "once a day." The cleaning time is fixed at 2:00-4:00 AM during the off-peak hours (no power generation during this period, completely avoiding the impact of the cleaning process on daytime power generation). Simultaneously, dust sensors are installed at the end of the cleaning robots to automatically detect the dust coverage on the module surface after each cleaning. The coverage must be ≤0.5%, with no stubborn stains such as bird droppings or oil residue. This ensures that the benchmark strings are always in a standardized clean operating state, and their power generation can serve as a reliable benchmark for comparing the efficiency of all strings in the field.

[0019] 2. Power Generation Analysis 1) Data Acquisition. Through the power plant's SCADA system, the actual power generation of 3 benchmark strings and the actual power generation of all other strings in the plant are collected daily (denoted as actual value Pi, i=1,2,...n, where n is the total number of strings in the plant - 3). At the same time, daily weather data (sunlight intensity, temperature), maintenance records, and other data are also collected.

[0020] 2) Perform anomaly analysis on benchmark strings. The process involves three steps: verifying the effectiveness of cleanliness status, verifying the deviation of power generation baseline, and cross-validating multiple benchmarks. This process determines whether the three benchmark strings are abnormal.

[0021] a. Verification of cleaning status validity. The system first retrieves the cleaning robot operation log of the corresponding benchmark group. If there are records such as "not cleaning once a day" or "component surface cleanliness <95% after cleaning", it is directly marked as "cleaning failure abnormal", its qualification for benchmark value calculation is suspended, and it is judged as an abnormal group.

[0022] b. Power generation baseline deviation verification. A performance baseline for benchmark strings is constructed based on historical data. Power generation data from the past 90 days (excluding maintenance and extreme weather days) is used, combined with irradiance and module temperature, to construct a "standard power generation baseline model" through linear regression fitting: P_standard=a×G+b×T+c (where G is irradiance in W / m²; T is module temperature in °C; a, b, and c are fitting coefficients that are dynamically adjusted according to the module model).

[0023] Collect the actual power generation of the benchmark string on the day of the test, P_actual, and calculate the deviation rate from the actual power generation, P_standard: Deviation rate = |P_actual - P_standard| / P_standard × 100%. If the deviation rate is >8% (5% for new modules, 12% for modules that have been in operation for more than 5 years, which can be adjusted according to the degradation of the power plant modules), it is marked as "abnormal performance deviation", its eligibility for benchmark value calculation is suspended, and it is judged as an abnormal string.

[0024] c. Multi-benchmark cross-validation.

[0025] The abnormal marking of the three benchmark strings is statistically analyzed. If only one benchmark string is abnormal (cleaning failure or performance deviation), it is judged as "individual abnormality". The remaining two normal benchmark strings are included in the subsequent benchmark value calculation.

[0026] If two or more benchmark strings are abnormal, prioritize investigating common interference factors (such as regional power grid voltage fluctuations, SCADA data acquisition errors, and sudden shading in the same area): retrieve power station power grid monitoring data (voltage / current fluctuation values), SCADA system communication logs, and benchmark string area images (check for temporary shading, such as ribbons or accumulated fallen leaves); if common interference factors are found, re-collect data after the interference is eliminated; if no common interference is found, activate the "historical benchmark value replacement scheme": use the average power generation of benchmark strings under the same weather conditions (light intensity and temperature deviation ≤10%) over the past 7 days as a temporary benchmark value, and trigger an early warning on the operation and maintenance terminal to prompt manual on-site verification of the benchmark string status.

[0027] 3) Benchmark string reference value calibration If all three benchmark strings are normal, the average of the actual power generation of the three strings on that day is taken as the benchmark value P0; if only one benchmark string is abnormal, the average of the actual power generation of the remaining two normal strings on that day is taken as the benchmark value P0; if the historical benchmark value replacement scheme is used, the calculated historical average power generation is taken as the temporary benchmark value P0_temp, and the benchmark value model is recalibrated after the benchmark strings return to normal.

[0028] 4) Calculation of relative power generation The daily relative power generation of each non-benchmark string is calculated according to the formula to eliminate the common influence of weather and temperature on power generation: R_i=Pi / P0 (if a temporary benchmark value P0_temp is used, then R_i=Pi / P0_temp).

[0029] 3. Anomaly Detection and Linkage 1) Horizontal comparison Sort all non-benchmark strings by R_i from largest to smallest, and filter out strings with R_i < 0.9 (the threshold can be adjusted according to the power plant operation and maintenance accuracy requirements, such as 0.85 in the Gobi Desert due to sand and dust). When the number of filtered strings is small, they can be marked as "inefficient candidate strings for the day". If the number of filtered results is large, it may indicate long-term decay, and further longitudinal comparative analysis should be carried out.

[0030] 2) Vertical comparison a. For each non-benchmark string, collect R_i data continuously for 30 days to construct a "daily relative power generation - time" variation curve.

[0031] b. Use the “3σ principle” to remove outliers from the curve: Calculate the average value μ and standard deviation σ of R_i over 30 days, and remove values ​​<μ-3σ or >μ+3σ (most of which are sudden drops / rises caused by temporary maintenance or sudden blockages).

[0032] c. For the curve after removing outliers, calculate the downward slope k by linear regression: k=ΔR / Δt (ΔR is the difference of R_i between two adjacent days, Δt is the time interval, in days).

[0033] d. Sort all strings in descending order of the absolute value of k, and select strings with |k|>0.01 / day (i.e., monthly relative power generation decay exceeds 0.3, threshold can be adjusted), and mark them as "long-term decay candidate strings".

[0034] 3) Power rationing identification From the "strings with low power generation", strings whose power generation has decreased due to grid curtailment are precisely screened out, and their status as "abnormal candidate strings" is excluded to avoid wasting ineffective operation and maintenance costs. If a string R_i < 0.9 and meets the conditions of "there is a grid curtailment instruction on the day" and "the string has no maintenance record", then the string is determined to be a "curtailed string" and excluded from the abnormal candidate list.

[0035] 4) Judgment of abnormal string types If a string is only marked as a "single-day inefficient candidate string" and is not a power-limited string, it is determined to be a "short-term abnormal string", which is presumed to be a temporary component failure, such as loose wiring or obstruction, and is then inspected by drones.

[0036] If a string is simultaneously marked as a "candidate string for daily inefficiency" and a "candidate string for long-term degradation," and it is not a string subject to power rationing, then further judgment is made: if other strings under the inverter are normal and only this string is abnormal, it is determined to be a "string with individual component defects," and a drone inspection is initiated; if multiple strings in the same area are abnormal, it is determined to be a "string covered by dust in the area," and the string cleaning plan optimization module is initiated.

[0037] 4. Cleaning plan optimization 1) Calculation of power loss For strings identified as "area dust-covered strings", calculate the theoretical power loss for the next 3 days using the formula.

[0038] Where R_imax is the maximum value of R_i within the past 3 months, R_current is the current regional string average relative power generation R_i, k is the k-th day in the next 3 days, and P0 mk The power generation of the m-th benchmark string on the k-th day is calculated based on the irradiance forecast from the weather.

[0039] 2) Set threshold Set a power loss threshold ΔE0 (e.g., 500kWh, determined based on the power plant's cost per kilowatt-hour). If ΔE ≥ ΔE0, then trigger the cleaning plan.

[0040] 3) Cleaning determination Use weather forecast data to obtain weather data for the next 3 days.

[0041] a. If a sandstorm (wind force ≥ level 5 and air humidity < 30%) is expected in the next 3 days, cleaning will be delayed until 1 day after the sandstorm ends; b. If there is heavy rain (rainfall ≥ 20mm) in the next 3 days, the cleaning plan will be cancelled (natural cleaning will be achieved by using rainwater); c. If there are no special weather conditions, immediately start the cleaning robot to clean the "dust-covered clusters in the area".

[0042] In this embodiment, by calculating the daily relative power generation (actual value / benchmark string baseline value) and combining horizontal sorting and vertical slope analysis, abnormal strings are identified after eliminating curtailed strings, thus achieving accurate identification of short-term anomalies, long-term degradation, and curtailed photovoltaic strings. Drone inspections are linked on demand based on component defects, sending inspection commands only to strings with short-term anomalies or individual component defects, avoiding ineffective operations and reducing maintenance costs. A scientific cleaning plan is developed based on future power loss due to dust cover and weather forecasts, dynamically adjusting cleaning times or canceling cleaning to further reduce maintenance costs and improve the overall power generation efficiency of the power station.

[0043] Application Examples Taking a 100MW photovoltaic power station as an example, the specific implementation steps are as follows: 1. Implementation Preparation 1) Benchmark string selection: Select one inverter string (numbered B1, B2, and B3) in each of the three areas of the power station (east, central, and west) as benchmark strings. The distance between the three strings is 1km to avoid local shading interference. 2) Equipment configuration: One low-cost cleaning robot is configured for each of B1, B2 and B3, and it is set to automatically clean once a day at 3 am (when there is no light); 3) Data interface integration: Connect the power plant's SCADA system to the analysis server and enable the third-party meteorological API interface to ensure that the previous day's power generation, weather, maintenance and other data are automatically collected before 8:00 a.m. every day; 4) Power generation baseline construction: Collect power generation data of B1, B2, and B3 over the past 90 days (excluding rainy days and maintenance days), and combine them with the solar irradiance and module temperature during the same period to fit the baseline model: P_standard=a×G+b×T+c (where G is solar irradiance in W / m²; T is module temperature in °C; a, b, and c are fitting coefficients that are dynamically adjusted according to the module model).

[0044] 2. Power Generation Analysis 1) Benchmark string anomaly analysis Cleaning status check: Retrieve the logs of cleaning robots B1, B2, and B3. If there are records such as "not cleaning once a day" or "cleanliness of component surface <95% after cleaning", they will be directly marked as "cleaning failure abnormal", their baseline value calculation qualification will be suspended, and they will be judged as abnormal clusters. Power generation baseline deviation verification: Collect the actual power generation P_actual of the benchmark string on the day, and calculate P_standard according to the formula based on the day's irradiance and module temperature, as well as the deviation rate from P_standard: Deviation rate = |P_actual - P_standard| / P_standard × 100%. If the deviation rate > 8% (5% for new modules, 12% for modules that have been in operation for more than 5 years, which can be adjusted according to the degradation of the power plant modules), it is marked as "performance deviation abnormal", its qualification for benchmark value calculation is suspended, and it is judged as an abnormal string; Multi-benchmark cross-validation: Statistically analyze the abnormal marking of benchmark strings B1, B2, and B3. If only one benchmark string is abnormal (cleaning failure or performance deviation), it is judged as "individual abnormality", and the remaining two normal benchmark strings are included in the subsequent benchmark value calculation.

[0045] If two or more benchmark strings are abnormal, prioritize investigating common interference factors (such as regional power grid voltage fluctuations, SCADA data acquisition errors, and sudden shading in the same area): retrieve power station power grid monitoring data (voltage / current fluctuation values), SCADA system communication logs, and benchmark string area images (check for temporary shading, such as ribbons or accumulated fallen leaves); if common interference factors are found, re-collect data after the interference is eliminated; if no common interference is found, activate the "historical benchmark value replacement scheme": use the average power generation of benchmark strings under the same weather conditions (light intensity and temperature deviation ≤10%) over the past 7 days as a temporary benchmark value, and trigger an early warning on the operation and maintenance terminal to prompt manual on-site verification of the benchmark string status.

[0046] 2) Benchmark string reference value calibration If all benchmark strings B1, B2, and B3 are normal, the average of the actual power generation of the three strings on that day is taken as the benchmark value P0. If only one benchmark string is abnormal, the average of the actual power generation of the remaining two normal strings on that day is taken as the benchmark value P0. If the historical benchmark value replacement scheme is used, the calculated historical average power generation is taken as the temporary benchmark value P0_temp. After the benchmark strings return to normal, the benchmark value model is recalibrated.

[0047] 3) Calculation of relative power generation The daily relative power generation of each non-benchmark string is calculated according to the formula to eliminate the common influence of weather and temperature on power generation: R_i=Pi / P0 (if a temporary benchmark value P0_temp is used, then R_i=Pi / P0_temp).

[0048] 3. Anomaly Detection and Linkage 1) Horizontal comparison Sort all non-benchmark strings of the day in descending order of R_i, and filter out strings with R_i < 0.9. When the number of filtered strings is small, they can be marked as "inefficient candidate strings of the day"; if the number of filtered results is large, it may indicate long-term decay, and further longitudinal analysis should be carried out.

[0049] 2) Vertical comparison 3) Power rationing identification From the "strings with low power generation", strings whose power generation has decreased due to grid curtailment are precisely screened out, and their status as "abnormal candidate strings" is excluded to avoid wasting ineffective operation and maintenance costs. If a string R_i < 0.9 and meets the conditions of "there is a grid curtailment instruction on the day" and "the string has no maintenance record", then the string is determined to be a "curtailed string" and excluded from the abnormal candidate list.

[0050] 4) Judgment of abnormal string types If a string is only marked as a "single-day inefficient candidate string" and is not a power-limited string, it is determined to be a "short-term abnormal string", which is presumed to be a temporary component failure, such as loose wiring or obstruction, and is then inspected by drones.

[0051] If a string is simultaneously marked as a "candidate string for daily inefficiency" and a "candidate string for long-term degradation," and it is not a string subject to power rationing, then further judgment is made: if other strings under the inverter are normal and only this string is abnormal, it is determined to be a "string with individual component defects," and a drone inspection is initiated; if multiple strings in the same area are abnormal, it is determined to be a "string covered by dust in the area," and the string cleaning plan optimization module is initiated.

[0052] 4. Cleaning plan optimization 1) Calculation of power loss For strings identified as "area dust-covered strings", calculate the theoretical power loss for the next 3 days using the formula:

[0053] Where R_imax is the maximum value of R_i within the past 3 months, R_current is the current regional string average relative power generation R_i, k is the k-th day in the next 3 days, P01 K P02 K P03 K These represent the power generation of the benchmark strings B1, B2, and B3 on the k-th day in the future, calculated based on the irradiance forecast from the weather forecast.

[0054] 2) Set threshold Set a power loss threshold ΔE0 (e.g., 500kWh, determined based on the power plant's cost per kilowatt-hour). If ΔE ≥ ΔE0, then trigger the cleaning plan.

[0055] 3) Cleaning determination Use weather forecast data to obtain weather data for the next 3 days.

[0056] a. If a sandstorm (wind force ≥ level 5 and air humidity < 30%) is expected in the next 3 days, cleaning will be delayed until 1 day after the sandstorm ends; b. If there is heavy rain (rainfall ≥ 20mm) in the next 3 days, the cleaning plan will be cancelled (natural cleaning will be achieved by using rainwater); c. If there are no special weather conditions, immediately start the cleaning robot to clean the "dust-covered clusters in the area".

[0057] The above technical solutions are merely exemplary embodiments of the present invention. For those skilled in the art, based on the application methods and principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the methods described in the specific embodiments of the present invention. Therefore, the methods described above are merely preferred and not restrictive.

Claims

1. A low-cost method for identifying, analyzing, inspecting, cleaning, and maintaining inefficient photovoltaic strings, characterized in that... Includes the following steps: S1, select benchmark strings to represent the normal operating status of different areas of the field; Clean and maintain the benchmark string; S2, perform power generation analysis on the selected benchmark string, including: anomaly analysis, benchmark calibration, and relative power generation calculation; S3, based on a set threshold, identifies non-benchmark strings with abnormal power generation, and coordinates with drones for inspection or string cleaning based on the identification results; S4, Cleaning plan optimization.

2. The method according to claim 1, characterized in that, In step S1, the specific method for selecting the benchmark string includes: Based on the power station's layout, three differentiated areas are divided into east, central and west, or south, central and north, to ensure that the number of inverters and string distribution density covered by each area are basically the same. At least one inverter string with no historical fault records, the same component installation age, and far away from the source of obstruction is randomly selected from each region and used as the benchmark string for that region.

3. The method according to claim 1 or 2, characterized in that, In step S1, the specific methods for ensuring the cleaning of the benchmark string include: Each benchmark string is equipped with a low-cost cleaning robot adapted to the photovoltaic module type, and a high-frequency cleaning frequency is set; the cleaning time is fixed in the early morning when there is no sunlight and no power generation operation. Meanwhile, a dust sensor is installed at the end of the cleaning robot to automatically detect the dust coverage rate on the component surface after each cleaning. It is necessary to ensure that the coverage rate is ≤ the set threshold and that there are no stubborn stains. This ensures that the benchmark string is always in a standardized clean operating state, and its power generation serves as a reliable benchmark value for comparing the efficiency of all strings in the field.

4. The method according to claim 2, characterized in that, In step S2, the specific steps of the power generation analysis include: The power plant's SCADA system collects the actual power generation of the benchmark string and the actual power generation of all other strings in the plant, Pi, i=1,2,...n, where n is the total number of strings in the plant minus the number of benchmark strings. It also collects the daily weather data and maintenance record data. The system determines whether the benchmark strings are abnormal by verifying the cleanliness status, checking the baseline deviation of power generation, and cross-verifying multiple benchmarks. If two or more benchmark strings in different areas are abnormal, the system retrieves power grid monitoring data, SCADA system communication logs, and images of the benchmark string areas to prioritize identifying common interference factors. If common interference factors are identified, data is re-collected after the interference is eliminated. If no common interference is identified, the "historical benchmark value replacement scheme" is activated, which uses the average power generation of benchmark strings under the same weather conditions over a past period as a temporary benchmark value. At the same time, an early warning is triggered on the operation and maintenance terminal, prompting manual on-site verification of the benchmark string status. Benchmark string baseline calibration: If all benchmark strings are normal, the average of the actual power generation of each benchmark string on the day is taken as the baseline value P0; if only one benchmark string in a region is abnormal, the average of the actual power generation of the remaining normal benchmark strings on the day is taken as the baseline value P0; if the historical baseline replacement scheme is used, the calculated historical average power generation is taken as the temporary baseline value P0_temp, and the baseline value is recalibrated after the benchmark strings return to normal. Relative power generation calculation: Calculate the daily relative power generation of each non-benchmark string: R_i = Pi / P0; if a temporary base value P0_temp is used, then R_i = Pi / P0_temp.

5. The method according to claim 4, characterized in that, Step S3 specifically includes: (1) Horizontal comparison Sort all non-benchmark strings in descending order of R_i, filter out strings with R_i < set threshold, and record them as "strings with low power generation". When the number of selected strings is small, they are marked as "inefficient candidate strings for the day"; if the number of selected results is large, longitudinal analysis is performed. (2) Vertical comparison For each non-benchmark string, R_i data was continuously collected for 30 days to construct a "daily relative power generation - time" variation curve; Outliers in the curve are removed using the "3σ principle": calculate the mean μ and standard deviation σ of R_i over 30 days, and remove values ​​that are <μ-3σ or >μ+3σ. For the curve after removing outliers, the downward slope k is calculated by linear regression: k = ΔR / Δt, where ΔR is the difference between R_i on two adjacent days, and Δt is the time interval in days; Sort all strings in descending order of the absolute value of k, filter out strings whose |k|> set threshold, and mark them as "long-term decay candidate strings"; (3) Power rationing identification From the "strings with low power generation", select the strings whose power generation has decreased due to grid curtailment, and exclude them as abnormal candidate strings to avoid wasting unnecessary operation and maintenance costs: If a string R_i < 0.9 and satisfies "there is a power grid rationing instruction on the day" and "there is no maintenance record for the string", then the string is determined to be a "power rationing string". (4) Judgment of abnormal string type If a string is only marked as a "single-day inefficient candidate string" and is not a power-limited string, it is determined to be a "short-term abnormal string", which is suspected to be a temporary component failure, and drone inspection is initiated. If a string is simultaneously marked as a "candidate string for daily inefficiency" and a "candidate string for long-term degradation," and it is not a string subject to power rationing, then further judgment is made: if other strings under the inverter are normal and only this string is abnormal, it is determined to be a "string with individual component defects," and drone inspection is initiated; if multiple strings in the same area are abnormal, it is determined to be a "string covered by dust in the area," and a string cleaning plan is initiated.

6. The method according to claim 5, characterized in that, Step S4 specifically includes: (1) Calculation of power loss For strings identified as "area dust-covered strings", calculate the theoretical power loss for the next 3 days: Where R_imax is the maximum value of R_i within the past 3 months, R_current is the current regional string average relative power generation R_i, k is the k-th day in the next 3 days, and P0 mk The power generation of the m-th benchmark string on the k-th day is calculated based on the irradiance forecast from the weather forecast. (2) Set threshold Set a power loss threshold ΔE0. If ΔE ≥ ΔE0, then trigger the cleaning plan. (3) Cleaning judgment Access weather forecast data to obtain weather data for the next 3 days: a. If a sandstorm occurs within the next 3 days, the cleaning will be delayed until 1 day after the sandstorm ends; b. If there is heavy rain in the next 3 days, this cleaning plan will be cancelled; c. If there are no special weather conditions, immediately start the cleaning robot to clean the "dust-covered clusters in the area".

7. A low-cost photovoltaic inefficient string identification, analysis, inspection, cleaning, and maintenance device applying the method described in any one of claims 1-6, characterized in that, include: The benchmark string configuration module is used to select benchmark strings and ensure their cleanliness. The power generation analysis module is used to analyze the power generation of the selected benchmark strings, including: anomaly analysis, benchmark calibration, and relative power generation calculation. The anomaly identification and linkage module is used to identify non-benchmark strings with abnormal power generation based on a set threshold, and to link drone inspection or string cleaning according to the identification results. The cleaning plan optimization module is used to optimize the cleaning plan.

Images