Photovoltaic string global dust accumulation phenomenon diagnosis method

Abstract

The invention provides a photovoltaic string global dust accumulation phenomenon diagnosis method, which can judge whether an abnormal power generation string and a global dust accumulation phenomenon exist in a photovoltaic string or not by utilizing the characteristics in the current and voltage time sequence of the photovoltaic string, and also can calculate the dust accumulation degree of the photovoltaic string under a dust accumulation layer, so that the diagnosis accuracy of the global dust accumulation phenomenon of the photovoltaic string is improved. Therefore, the method is of great guiding significance to the operation and maintenance work of the power station.

Description

TECHNICAL FIELD

[0001] The present application relates to a method for diagnosing global dust accumulation phenomenon of photovoltaic string in time series, belonging to the field of photovoltaic power generation. BACKGROUND

[0002] At present, under the background of global energy structure transformation, photovoltaic power generation, as an important part of clean energy system, its installed capacity is showing a rapid expansion trend in China. However, in the process of technology iteration, photovoltaic power station operation faces complex external environmental influence, especially the systematic performance degradation problem caused by component surface pollution gradually highlights.

[0003] In the actual operation and maintenance of distributed power station, global dust accumulation phenomenon has become a key factor restricting power generation performance. When photovoltaic array is exposed to open space for a long time, its surface will form dense global dust periodically, causing about 15%-25% loss of incident light energy. This homogeneous pollution not only directly reduces the photoelectric conversion efficiency, but also causes secondary faults such as hot spot effect, ultimately leading to the decline of power output stability of the whole power generation unit. The intelligent monitoring and cleaning maintenance mechanism for this problem has become an important breakthrough to improve the operation quality of new energy infrastructure.

[0004] The detection methods for global dust accumulation phenomenon mainly include: 1) computer vision-based diagnosis method: this method belongs to the emerging application field of artificial intelligence technology, which uses intelligent algorithm to learn computer vision of photovoltaic panel pictures, but the cost of taking pictures for large distributed photovoltaic power station is too high, and even some areas cannot be photographed, and because the working environment of different photovoltaic power stations is different, the generalization effect of the learned model also becomes a problem; 2) data-driven diagnosis method: this method can use the historical operation data of distributed photovoltaic power station, analyze and calculate the historical operation meteorological and irradiation conditions, current and voltage time series and other data, diagnose the global dust accumulation phenomenon and its degree, which has low cost and good accuracy.

[0005] The present application provides a photovoltaic string global dust accumulation phenomenon diagnosis method, which analyzes the current and voltage time series of photovoltaic power station obtained by photovoltaic intelligent operation and maintenance platform, can not only diagnose whether there is global dust accumulation in photovoltaic string, but also calculate the dust accumulation degree of photovoltaic string under global dust accumulation, to judge whether to take maintenance measures, which has strong guiding significance for the operation and maintenance of power station. SUMMARY

[0006] A time series-based photovoltaic string global dust accumulation phenomenon diagnosis method, characterized by: extracting the current and voltage time series output by the photovoltaic intelligent operation and maintenance platform of the photovoltaic power station, screening the data through irradiance, and comparing the fault characteristic parameters obtained through the steps of the patent with the threshold value to diagnose whether there is a global dust accumulation phenomenon in the photovoltaic string, and calculate the dust accumulation degree, the specific steps are as follows:

[0007] Step one: Extract the current-voltage (I-V) time series data output by the inverter through the photovoltaic intelligent operation and maintenance platform, and the sampling frequency is 15 minutes / time. The data screening is limited to the daytime light effective interval (06:00-18:00), and the irradiance is stable and the component is in the generating state in this period, excluding the influence of morning and evening weak light or night no output data. In order to ensure the data quality, further screen the high-quality data segment (500W / m 2 The peak value of solar irradiance G is more than 500W / m 2 The critical irradiance for normal operation of the component is 500W / m ratio , below which the component output power is significantly affected by irradiance fluctuations, and the dust characteristics are drowned in noise). For random noise in the original time series data, a moving average filtering algorithm (window size is set to 5 sampling points, i.e. 75 minutes, balance denoising effect and data lag) is used for smoothing processing to eliminate abnormal fluctuations and retain the time series characteristics reflecting the true working state of the component.

[0008] Step two: In order to solve the problem of "subjectivity of reference string selection" in traditional methods, the size of power loss is used as the standard to determine abnormal sequence, and the power maximum value matching method is used to determine the reference string, the specific process includes: calculating the average power of each string, and selecting the normal string with the maximum power peak value as the reference sequence. Based on the baseline data, a theoretical power curve is constructed, and the instantaneous ratio Q(i) of the actual power of each sampling point of the photovoltaic string to the power of the normal string is calculated as a characteristic quantity. Calculate the mean μ and standard deviation σ of all sampling points, and set the dynamic threshold interval [μ-3σ, μ+3σ] (consistent with the normal distribution assumption, covering 99.7% of normal fluctuations), and the sampling points exceeding the interval are determined as abnormal state, and the corresponding string is marked as suspected fault unit.

[0009] Step three: The suspected fault string screened out in step two needs to be further verified whether it is caused by global dust accumulation. This step extracts the dust characteristics through the ratio analysis of voltage-current time series: according to step two, the normal and abnormal current and voltage time series are obtained, and the data is normalized to map the current and voltage time series to the [0,1] interval; then calculate the voltage ratio V ratio and current ratio I ratio of the abnormal string and the reference string. Global dust accumulation will cause I ratio to decrease significantly (current is suppressed by shading), while Vratio Basically maintain 1 (the voltage is less affected by the shielding), so the combination of the two can effectively distinguish dust accumulation from other faults. In order to quantify the degree of dust accumulation, a global dust accumulation index T is proposed. The global dust accumulation index T is defined as satisfying V ratio (m)>0.85 and I ratio The number of sampling points with (m)>0.85 (i.e., points with significant dust accumulation characteristics) is calculated, and finally the global dust accumulation index T is calculated for fault diagnosis.

[0010] Step 4: Set the diagnostic threshold and calculate the daily irradiance G>500W / m 2 The number of effective sampling points k. According to historical data statistics, when there is no dust accumulation on the component, the 95% confidence interval of T is [0.8, 1]; when dust accumulation causes performance degradation of more than 5%, T drops below 0.8. Therefore, the dynamic diagnosis threshold T is set th =0.8k. If the measured T>T th (i.e., the dust accumulation feature points exceed the 95% percentile value of the normal range), it is determined that the string has global dust accumulation.

[0011] Step 5: According to step 4, the PV string with global dust accumulation is obtained, and the power discrete area difference S between the normal PV string and the PV string is calculated. d , the area difference S d As a characteristic quantity to diagnose the degree of dust accumulation, if the discrete area difference S d The larger the value, the more serious the dust accumulation is and the more cleaning is needed. In practical applications, the classification standards can be set based on the experience of power station operation and maintenance, such as S d <10kWh is mild dust accumulation, 10kWh≤S d <30kWh is moderate, S d ≥30kWh is severe and needs to be cleaned immediately. Description of the drawings:

[0012] Figure 1 This is a flowchart for diagnosing dust accumulation in the entire photovoltaic string.

[0013] Figure 2 This is a time series diagram of the current of an exemplary photovoltaic string global dust accumulation phenomenon.

[0014] Figure 3 This is a voltage time series diagram of an exemplary photovoltaic string global dust accumulation phenomenon. Specific implementation method:

[0015] The present invention will be further described below with reference to the accompanying drawings and examples. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0016] The present invention provides a method for diagnosing global dust accumulation in photovoltaic strings. The method uses the time series of the output voltage and current of the photovoltaic strings to identify the global dust accumulation phenomenon and estimate the degree of dust accumulation. The diagnostic process is as follows: Figure 1 The specific diagnostic methods are as follows:

[0017] The photovoltaic intelligent operation and maintenance platform extracts the current and voltage time series output by the photovoltaic power station inverter, filters the data by irradiance, and compares the obtained fault characteristic parameters with the threshold through the steps proposed in this patent. It diagnoses whether there is global dust accumulation in the photovoltaic strings and calculates the degree of dust accumulation. The specific steps are as follows:

[0018] Step 1: Extract the inverter output current-voltage (IV) time series data through the photovoltaic intelligent operation and maintenance platform, with a sampling frequency of 15 minutes / time. Data screening is limited to the effective daylight period (06:00-18:00), during which the irradiance is stable and the components are in the power generation state, excluding the influence of weak light in the morning and evening or no output data at night. To ensure data quality, further screening is carried out for solar irradiance G peak exceeding 500W / m 2 High-quality data segment (500W / m 2 The critical irradiance for normal module power generation. Below this value, module output power is significantly affected by irradiance fluctuations, and dust accumulation characteristics are overwhelmed by noise. A sliding average filter algorithm (with a window size of 5 sampling points, or 75 minutes, to balance denoising and data lag) is used to smooth the random noise in the original time series data, eliminating abnormal fluctuations and retaining the time series characteristics that reflect the actual working status of the module.

[0019] Step 2: In order to solve the problem of "high subjectivity in the selection of reference strings" in traditional methods, the size of power loss is used as the standard for determining abnormal sequences, and the reference strings are determined by the maximum power matching method. The specific process includes: calculating the average power of each string, and selecting the normal string with the largest power peak as the reference sequence. Construct a theoretical power curve based on the baseline data, and calculate the instantaneous ratio Q(i) of the actual power to the normal string power at each sampling point of the photovoltaic string as the characteristic quantity. Calculate the mean μ and standard deviation σ of all sampling points, set the dynamic threshold range [μ-3σ, μ+3σ] (in line with the normal distribution assumption, covering 99.7% of normal fluctuations), and the sampling points outside this range are judged to be abnormal, and the corresponding strings are marked as suspected fault units. The specific process is as follows:

[0020] ①Calculate the average power of each sampling point of each photovoltaic string

[0021]

[0022] The average power of all PV strings By comparison, the photovoltaic string with the largest average power is regarded as the normal string, and its power is P pro ;

[0023] ② Calculate the instantaneous ratio Q(i) of the actual power to the normal power of the PV string at each sampling point and the average value μ and standard deviation σ of the ratio of the actual power to the normal power at all sampling points:

[0024]

[0025] Among them, P rea (i) is the actual power of the ith sampling point of the PV string, P pro (i) is the normal power of the ith sampling point of the PV string;

[0026] ③Calculate the upper limit HL and lower limit LL of the power ratio when the photovoltaic string is operating normally:

[0027] HL=μ+3σ (5)

[0028] LL=μ-3σ (6)

[0029] According to the ratio Q(i), the closer the value of Q(i) is to 1, the closer the actual power is to the normal power. HL and LL are used as the upper and lower limits of normal PV string operation. If the number of sampling points where the ratio Q(i) is not in the interval [LL, HL] is greater than 90%, the PV string is abnormal, and the time series of the PV string is defined as an abnormal series.

[0030] Step 3: The suspected faulty strings selected in step 2 need to be further verified to see if they are caused by global dust accumulation. This step extracts dust accumulation features by analyzing the ratio of voltage-current time series: according to step 2, the normal and abnormal current and voltage time series are obtained, and the data is normalized and mapped into current and voltage time series in the [0,1] interval. Then, the abnormal and normal voltage and current time series are calculated respectively when the solar irradiance exceeds 500W / m 2 The ratio of time periods V ratio and I ratio , and V ratio and I ratio As the characteristic quantity for diagnosing the global dust accumulation phenomenon of the photovoltaic string, the global dust accumulation index T is finally calculated for fault diagnosis. The calculation method is as follows:

[0031] S1. Select the time series with irradiance greater than 500W / m 2 Sampling points of the normal current time series are set as B1={I normal (1),I normal (2)…I normal(k)}, the sample point set of normal voltage time series is B2 = {U normal (1), U normal (2) … U normal (k)}, the sample point set of abnormal current time series is C1 = {I abnormal (1), I abnormal (2) … I abnormal (k)}, the sample point set of abnormal voltage time series is C2 = {U abnormal (1), U abnormal (2) … U abnormal (k)} ;

[0032] Wherein k is the number of sampling points with irradiance greater than 500 W / m 2 , in the above formula, I normal (k) is the normal current corresponding to the kth sampling point with irradiance greater than 500 W / m 2 , U normal (k) is the normal voltage corresponding to the kth sampling point with irradiance greater than 500 W / m 2 , I abnormal (k) is the abnormal current corresponding to the kth sampling point with irradiance greater than 500 W / m 2 , and U abnormal (k) is the abnormal voltage corresponding to the kth sampling point with irradiance greater than 500 W / m 2 .

[0033] S2. According to the sampling points selected from S1, the values of I ratio (m) and V ratio (m) are calculated from the first sampling point (m = 1), and the calculation formula is as follows:

[0034]

[0035] Wherein, E i is the normal current error, and E v is the normal voltage error.

[0036] Global dust accumulation can cause I ratio to decrease significantly (current is suppressed by shielding), and V ratio is basically maintained at 1 (voltage is less affected by shielding), so the combination of the two can effectively distinguish dust from other faults. To quantify the degree of dust, a global dust index T is proposed, and the global dust index T is defined as the number of sampling points (i.e. points with significant dust characteristics) that satisfy V ratio (m) > 0.85 and I ratio (m) > 0.85. If the sampling point I ratio (m) > 0.85 and V ratio(m) > 0.85, the sampling point meets the global dust accumulation condition, and the global dust accumulation index T is added by 1;

[0037] S3. Repeat the steps of S2 to calculate the next sampling point until k sampling points are calculated;

[0038] Step four: set the diagnostic threshold, and count the number of effective sampling points k of daily irradiance G > 500 W / m 2 2. Through the statistical history data, when the components have no dust accumulation, the 95% confidence interval of T is [0.8, 1]; when the dust accumulation causes the performance degradation of more than 5%, T decreases to below 0.8. Therefore, the dynamic diagnostic threshold T th = 0.8k is set. If the measured T > T th (i.e., the dust feature point exceeds the 95% quantile value of the normal interval), it is determined that the global dust accumulation phenomenon exists in the photovoltaic string.

[0039] Step five: according to step four, the photovoltaic string with global dust accumulation is obtained, and the power dispersion area difference S d of the normal photovoltaic string and the photovoltaic string is calculated. d The area difference S

[0040] The dispersion area difference is used to diagnose the dust accumulation degree of the photovoltaic string, and the calculation formula is as follows:

[0041]

[0042] Wherein, N is the sampling point interval of the photovoltaic string; if the dispersion area difference S d is larger, the dust accumulation degree is more serious, and cleaning is required. In actual application, the grading standard can be set in combination with the operation and maintenance experience of the power station, for example, S d < 10 kWh is light dust, 10 kWh ≤ S d < 30 kWh is moderate, and S d ≥ 30 kWh is severe, and immediate cleaning is required.

[0043] The above gives the specific implementation in engineering application, but the present application is not limited to the described embodiments. The basic principle and method of the present application is in the above basic scheme, and the changes, modifications, replacements and deformations of the embodiments without departing from the principles and spirits of the present application still fall within the protection scope of the present application.

Claims

1. A method for diagnosing global dust accumulation in photovoltaic strings, characterized by: The data is filtered by irradiance, and the fault characteristic parameters obtained are compared with the threshold through the steps proposed in this patent to diagnose whether there is global dust accumulation in the photovoltaic strings and calculate the degree of dust accumulation. The specific steps are as follows: Step 1: Extract the inverter output current-voltage (IV) time series data through the photovoltaic intelligent operation and maintenance platform, with a sampling frequency of 15 minutes / time. Data screening is limited to the effective daylight period (06:00-18:00), during which the irradiance is stable and the components are in the power generation state, excluding the influence of weak light in the morning and evening or no output data at night. To ensure data quality, further screening is carried out for solar irradiance G peak exceeding 500W / m 2 High-quality data segment (500W / m 2 The critical irradiance for normal power generation of the module is irradiance below which the module output power is significantly affected by irradiance fluctuations, and the dust accumulation characteristics are submerged by noise. To address the random noise in the original time series data, a sliding average filtering algorithm (with a window size of 5 sampling points, or 75 minutes, balancing the denoising effect with data lag) is used to smooth the data, eliminating abnormal fluctuations and retaining the time series characteristics that reflect the actual working status of the module. Step 2: To address the subjective nature of baseline string selection in traditional methods, the power loss magnitude is used as the criterion for determining abnormal sequences. The baseline string is determined using the maximum power matching method. The specific process involves calculating the average power of each string and selecting the normal string with the highest peak power as the baseline sequence. A theoretical power curve is constructed based on baseline data, and the instantaneous ratio Q(i) of the actual power to the normal string power at each sampling point is calculated as a characteristic quantity. The mean μ and standard deviation σ of all sampling points are calculated, and a dynamic threshold range of [μ-3σ, μ+3σ] is set (consistent with the normal distribution assumption and covering 99.7% of normal fluctuations). Sampling points outside this range are considered abnormal, and the corresponding strings are marked as suspected faulty units. Step 3: The suspected faulty strings screened out in step 2 need to be further verified to see if they are caused by global dust accumulation. This step extracts dust accumulation features by analyzing the ratio of voltage-current time series: Based on the normal and abnormal current and voltage time series obtained in step 2, the data is normalized and mapped into current and voltage time series in the [0,1] interval; then the voltage ratio V of the abnormal string to the reference string is calculated. ratio and current ratio I ratio Global dust accumulation will cause I ratio drops significantly (the current is blocked and suppressed), while V ratio Basically maintain 1 (the voltage is less affected by the shielding), so the combination of the two can effectively distinguish dust accumulation from other faults. In order to quantify the degree of dust accumulation, a global dust accumulation index T is proposed. The global dust accumulation index T is defined as satisfying V ratio (m)>0.85 and I ratio The number of sampling points with (m)>0.85 (i.e., points with significant dust accumulation characteristics) is calculated, and finally the global dust accumulation index T is calculated for fault diagnosis. Step 4: Set the diagnostic threshold and calculate the daily irradiance G>500W / m 2 The number of effective sampling points k. According to historical data statistics, when there is no dust accumulation on the component, the 95% confidence interval of T is [0.8, 1]; when dust accumulation causes performance degradation of more than 5%, T drops below 0.

8. Therefore, the dynamic diagnosis threshold T is set th =0.8k. If the measured T>T th (i.e., the dust accumulation feature points exceed the 95% percentile value of the normal range), it is determined that the group of strings has global dust accumulation. Step 5: According to step 4, the PV string with global dust accumulation is obtained, and the power discrete area difference S between the normal PV string and the PV string is calculated. d , the degree of dust accumulation in the photovoltaic string is diagnosed by the discrete area difference, and the calculation formula is as follows: Where N is the sampling point interval of the photovoltaic string; if the discrete area difference S d The larger the value, the more serious the dust accumulation is and the more necessary it is to clean it.

Images