Method for evaluating shadow shielding of photovoltaic string

Abstract

The invention provides a method for evaluating shadow shielding of a photovoltaic string, which comprises the following steps of: acquiring current and power time sequence data of a direct current end of the photovoltaic string from an intelligent photovoltaic platform, screening out data of night time periods (18: 00-6: 00) and cloudy and rainy days, filtering and normalizing the data to obtain n groups of current and power sequence data in an interval of [0, 1], a current sequence and a power sequence to be diagnosed are taken, characteristic quantities TM, Ka, Kb, SKM and eta are calculated, and whether a shadow phenomenon exists or not is judged by using fuzzy C-means clustering; based on the clustering center, dynamically setting light, medium and heavy shielding thresholds to obtain light, medium and heavy thresholds again; positioning a shielding occurrence time period through the SKM; and meanwhile, a sliding window is introduced to dynamically update daily sample data. According to the method, shadow fault assessment is realized only by using the detection data in the photovoltaic platform, maintenance personnel can conveniently troubleshoot the fault, other extra instruments and equipment are not needed, and the operation and maintenance cost of a photovoltaic power station is saved.

Description

Technical field:

[0001] The present invention relates to the technical field of fault diagnosis in distributed photovoltaic power stations, and in particular to a method for evaluating photovoltaic string shadowing phenomena. Background technology:

[0002] Distributed photovoltaics are deeply loved by people for the principles of adapting to local conditions, clean and efficient, decentralized layout, and nearby utilization. However, adapting to local conditions means that photovoltaic modules are often installed outdoors. After years of wind and rain, the harsh environment causes various failure problems of the modules to begin to appear. These problems will reduce the power generation efficiency of the modules, and in serious cases, they will also cause safety hazards. Among them, shadows caused by buildings, dust, etc. are one of the important reasons for reducing the power generation efficiency of the modules. In addition, long-term shadows may turn into hot spots, further reducing the power generation efficiency, and in serious cases, it may cause the photovoltaic modules to be scrapped.

[0003] The main methods for detecting shadow occlusion include image analysis and IV output characteristic curve. Image analysis mainly uses thermal imaging technology to compare the image differences between normal and abnormal components for diagnosis, but such methods often rely on expensive precision instruments to achieve high detection costs and are not suitable for distributed photovoltaic power stations. The IV output characteristic curve method uses the characteristics of the IV output characteristic curve for diagnosis, but there are many reasons for photovoltaic string mismatch, which are affected by other faults, and the acquisition of the IV output characteristic curve will affect the normal operation of photovoltaic components.

[0004] Therefore, it is necessary to propose a shadow fault assessment method that relies only on historical data. Through the rational use of the detection data of the photovoltaic platform, the characteristics of the current and power sequence are analyzed to realize the assessment of the shadow phenomenon, classify the shadow phenomenon, and locate the time period when the phenomenon occurs. This method not only uses historical data that contains a lot of information, but also does not affect the normal operation of the components, saves costs, conforms to the economic advantages of distributed photovoltaic power stations, facilitates maintenance personnel to troubleshoot faults, and has an important impact on improving the power generation efficiency of power stations. Summary of the invention:

[0005] The present invention is intended to provide a method for evaluating shadow shading of photovoltaic strings.

[0006] The present invention is implemented by adopting the following scheme, comprising the following steps:

[0007] Step 1: Obtain the current time series and power time series data of the DC end of the photovoltaic string in sunny weather in the past 7 days from the smart photovoltaic platform, select the data from 6:00 to 18:00, set the sampling period to 15 minutes / time, and collect 49 data of current and power series per day respectively, eliminate the influence of noise on the measurement, and perform data standardization so that the current and power series data are in the interval of [0,1], and obtain the current and power series to be diagnosed;

[0008] Step 2: Calculate the difference between the current series at adjacent moments, select the maximum value TM, which is the difference between the concave point of the current time series and the next moment, and record the moment of the concave point with the subscript i at this time, and calculate the slope K at moments i, i-1, and i+1 respectively. a and K b , TM, K a and K b As a diagnostic feature;

[0009] Step 3: Calculate the sliding skewness SK of the current time series, record the maximum sliding skewness SKM, and calculate the discrete area ratio η of the power sequence to be tested and the normal power sequence to characterize the degree of shadow blocking; wherein the sliding skewness is designed to be obtained by sliding the current time series of one day;

[0010] Step 4: According to the TM, K in step 3 a ,K b ,SKM,η constructs a sample set X={x1,x2...x i ...}, x i ={TM i ,K a,i ,K b,i ,SKM i ,η i}, cluster the sample set X to obtain normal, light, moderate, severe occlusion samples and non-occlusion fault samples; for occlusion samples and normal samples x i ={TM i ,K a,i ,K b,i ,SKM i ,η i}, based on the cluster center, the threshold is dynamically set to obtain the mild, moderate and severe thresholds;

[0011] Step 5: Compare the SKM with the corresponding time to locate the time period when the occlusion occurs;

[0012] Step 6: Introduce a sliding window to dynamically update daily sample data to adapt to daily data changes. Description of the drawings:

[0013] Figure 1A flow chart of a photovoltaic string shadowing assessment method according to an exemplary embodiment of this specification;

[0014] Figure 2 A schematic diagram of a current time series of a shadow shading phenomenon described in an exemplary embodiment of this specification;

[0015] Figure 3 A schematic diagram of skewness SK points described in an exemplary embodiment of this specification. Specific implementation method:

[0016] The present invention provides a method for evaluating shadow shading of photovoltaic strings. To make the purpose, technical solution and effect of the present invention clearer, further description is given. The specific examples described in the present invention are only used to explain the present invention and are not used to limit the present invention.

[0017] The present invention will be further described below in conjunction with the accompanying drawings. Figure 1 The figure is a flow chart of a photovoltaic string shadow shielding evaluation method according to an exemplary embodiment of the present specification, which specifically includes the following steps:

[0018] Step 1: Obtain the current time series and power time series data of the DC end of the photovoltaic string in sunny weather in the past 7 days from the smart photovoltaic platform. Since the output voltage cannot reach the starting voltage of the inverter under low irradiance, the data at night (18:00-6:00) is first removed. Secondly, the irradiance is low on rainy days, and the data fluctuates greatly, so the data on rainy days should be removed. The current time series and power time series data with high irradiance on sunny days and between 6:00 and 18:00 are selected. The sampling period is set to 15 minutes / time, and the current and power series are 49 data per day respectively. The influence of noise on the measurement is eliminated, and the data is standardized so that the current and power series data are in the interval [0,1], and the current and power series I to be diagnosed are obtained. diagnosis = {I(t i ),i=1,2,3...49};

[0019] Since distributed photovoltaics are generally located outdoors, they are very susceptible to sudden changes in the weather. The measurement noise caused by the scanning environment leads to inaccurate sampled data, which will affect the subsequent results. Therefore, arithmetic average filtering is required. The specific method is as follows: First, the data at the first moment, the second moment, and the third moment are averaged, and then the data at the fourth moment, the fifth moment, and the sixth moment are averaged. The calculations are continued in sequence, and finally a current sequence I with a sampling period of 15 minutes per time and a total of 49 data is obtained. diagnosis = {I(t i ),i=1,2,3...49}.

[0020] In order to ensure that the diagnosis results of different strings will not be affected by the size of the data, the data of each string is scaled to the same data interval and range. The data standardization process is as follows:

[0021]

[0022] In the formula, I′ is the normalized sequence value; I i is the value of the current sequence after filtering; max(I i ) and min(I i ) is the maximum and minimum value of the current sequence of a certain string in a day; i is the time.

[0023] Step 2: Take the normalized current time series in step 1, calculate the difference between the current series at adjacent moments, and select the maximum value TM, which is the difference between the concave point of the current time series and the next moment.

[0024] Subtract the current data at adjacent moments to get the maximum value TM:

[0025] TM={I(t i )-I(t i+1 )} max ,i=1,2,3...,48 (2)

[0026] And record I(t i ) and subscript i, calculate the slope K at time i, i+1 and i-1 respectively a and K b .

[0027]

[0028] TM、K a and K b As a diagnostic feature. Figure 2 It can be seen that when there is a shadow blocking the string, the current increases very slowly at the beginning, because the irradiance is small in the morning and the shadow blocks part of the sunlight. As the solar azimuth changes, the photovoltaic string begins to generate electricity normally. At this time, the shadow current changes greatly, while the normal current changes relatively little. The slope of the shadow current at time i and time i+1 is greater than the slope of the normal current sequence at time i and time i+1, and the slope of the shadow current at time i and time i-1 is less than the slope of the normal current sequence at time i and time i-1. Therefore, TM, K1 and K2 are selected as fault characteristic quantities.

[0029] Step 3: Calculate the sliding skewness SK of the current time series. The sliding skewness calculation diagram is as follows: Figure 3 As shown, record the maximum sliding deflection SKM and the corresponding time. There is a shadow shielding phenomenon. Current sequence: Idiagnosis = {I(t i ),i=1,2,3...49}, find the data I(t 25 ), respectively, with I(t 25 ) is the center point, the step length is 2, and 2 data are taken to the left and right to form a data set: {I(t 23 ),I(t 24 ),I(t 25 ),I(t 26 ),I(t 27 )} and calculate the skewness SK of the data set:

[0030]

[0031] Where I(t i ) represents the current sequence value at the i-th moment, n represents the number of sequences, represents the mean of the sequence, and σ is the standard deviation of the sequence. Then take I(t 24 ) and I(t 24 )The two current data on the left and I(t 26 ) and I(t 26 ) The two current data on the right, these six data are combined into a data set: {I(t 22 ), I(t 23 ), I(t 24 ), I(t 26 ), I(t 27 ), I(t 28 )}, and calculate SK, then use point I(t 23 ) and I(t 27 ) The step length is 2 and then the value is taken as follows Figure 2 As shown, SK is calculated until all points are taken. Then the largest SKM is found. The ratio of the discrete trapezoidal area of ​​the power sequence to be tested and the normal power sequence η is calculated to represent the degree of shadow occlusion. The expression of η is as follows

[0032]

[0033] Where P i is the power of the sample to be tested at the i-th moment, P nor,i Normal sample power at the i-th moment.

[0034] Step 4: According to the TM, K in step 3 a ,K b ,SKM,η constructs a sample set X={x1,x2...x i ...}, x i ={TM i ,K a,i ,Kb,i ,SKM i ,η i}, calculate each data sample x j With the membership degree of c0~c4, according to the size of the membership degree, the data sample x j Divide into normal, non-occlusion fault, mild occlusion, moderate occlusion and severe occlusion data clusters. Get normal, mild, moderate, severe occlusion samples and non-occlusion samples; the membership degree μ of each sample data and c0~c4 ij As shown in formulas (7) to (9):

[0035]

[0036] In the formula, c is the number of cluster centers; μ ij ——The membership degree of the jth sample to the i-th cluster center, μ ij The value range of is 0 to 1, and the sum of the membership of the sample to each cluster center is equal to 1. ; d ij ——Euclidean distance between the jth sample and the ith cluster center; m——fuzzy weighted index, ranging from 0 to +∞.

[0037] Eliminate non-occluded fault samples, for occluded samples and normal samples, based on the cluster center c of each light, medium and heavy occlusion i , dynamically set it to obtain the light, medium and heavy thresholds, and calculate its power loss:

[0038]

[0039] in is the fuzzy membership of the i-th sample point to the j-th category; η p (x i ) is the i-th sample and the power loss percentage; η p (c1),η p (c2),η p (c3),η p (c4) corresponds to normal, mild occlusion, moderate occlusion and severe occlusion respectively; the dynamic threshold is defined as:

[0040]

[0041] When there is no shadow: η p <ρ1; Mild occlusion: ρ1≤η p <ρ2; Moderate occlusion: ρ2≤η p <ρ3; severe occlusion: η p ≥ρ3.

[0042] Step 5: Compare the SKM with the corresponding time to locate the time period when the occlusion occurs;

[0043] Compare the SKM with the SK value at the symmetrical moment (if there is an SKM at 9:00 a.m., compare it with the corresponding SK value at 15:00 p.m.). If the data at one end is more than 20% less than that at the other end, it indicates that the shadow occurs in the time period of the subscript moment at that end.

[0044] Step 6: Introduce a sliding window to dynamically update daily sample data, which has adapted to daily data changes and can obtain updated thresholds every day:

[0045] X t ={x t-T+1 ,x t-T+2 ,...,x t} (12)

[0046] Among them, X t is the set of all samples on a certain day; x i is the feature value of the i-th sample of the day; T is the window size. As each day passes, the latest sample x t Add to the window, and move the earlier samples x t+T Move out of the window, recalculate the cluster center and sample distribution characteristics to update the thresholds ρ1, ρ2, ρ3 in real time.

[0047] Compared with the prior art, the effective effects of the embodiments of the present invention are at least one of the following:

[0048] 1) The present invention extracts data features from multiple dimensions, which can comprehensively reflect the shading status and degree of shading of photovoltaic strings.

[0049] 2) The present invention can locate the specific moment when shadow occlusion occurs by calculating the sliding deflection, providing a theoretical basis for operation and maintenance staff to eliminate instantaneous occlusion at a certain moment.

[0050] 3) The present invention introduces a sliding window to dynamically update the daily threshold to adapt to the daily changes in the external environment and update the latest data after excluding instantaneous occlusion. Improve the robustness of the photovoltaic power station

[0051] The above is a specific implementation method in engineering application, but the present invention is not limited to the described implementation method. The basic principle and method of the present invention lies in the above basic scheme. Changes, modifications, substitutions and deformations of the implementation methods without departing from the principles and spirit of the present invention still fall within the protection scope of the present invention.

Claims

1. A method for evaluating photovoltaic string shadowing, characterized in that: The following steps are involved: Step 1: Obtain the current time series and power time series data of the DC end of the photovoltaic string under sunny weather from the smart photovoltaic platform, select the data from 6:00 to 18:00, set the sampling period to 15 minutes / time, and collect 49 data of current and power series per day respectively, eliminate the influence of noise on the measurement, and perform data standardization so that the current and power series data are in the interval of [0,1], and obtain the current and power series data to be diagnosed; Step 2: Calculate the difference between the current series at adjacent moments, select the maximum value TM, which is the difference between the concave point of the current time series and the next moment, and record the moment of the concave point with the subscript i at this time, and calculate the slope K at moments i, i-1, and i+1 respectively. a and K b , TM, K a and K b As a diagnostic feature; Step 3: Calculate the sliding skewness SK of the current time series, record the maximum sliding skewness SKM, and calculate the discrete area ratio η of the power sequence to be tested and the normal power sequence to characterize the degree of shadow blocking; wherein the sliding skewness is designed to be obtained by sliding the current time series of one day; Step 4: According to the TM,,K in step 3 a ,,K b ,,SKM,,η constructs a sample set X={x1,x2...x i ...}, x i ={TM i ,K a,i ,K b,i ,SKM i ,η i }, clustering the sample set X to obtain normal, light, moderate, and heavy occlusion samples and non-occlusion fault samples; For occluded samples and normal samples x i ={TM i ,K a,i ,K b,i ,SKM i ,η i }, based on the cluster center, the threshold is dynamically set to obtain the mild, moderate and severe thresholds; Step 5: Compare the SKM with the corresponding time to locate the time period when the occlusion occurs; Step 6: Introduce a sliding window to dynamically update daily sample data to adapt to daily data changes.

2. A photovoltaic string shadowing assessment method according to claim 1, characterized in that: The slope K of the current time series point with the shadow concave and the two adjacent points a,i and K b,i The maximum value TM of the adjacent current difference is used as the diagnostic feature of shadow and non-shadow phenomena.

3. A photovoltaic string shadowing assessment method according to claim 1, characterized in that: Calculate the maximum sliding gradient SKM of the current time series to locate the moment when shading may occur: the shadow shading phenomenon current series I diagnosis = {I(t i ),i=1,2,3...49} in the middle value I(t 25 ), with I(t 25 ) as the center point, the step length is 2, and the data set is constructed: {I(t 23 ),I(t 24 ),I(t 25 ),I(t 26 ),I(t 27 )}, and calculate the skewness SK at this time, and then use I(t 24 ) as the starting point, take the data with a step length of 2 to the left and take I(t 26 ) is taken as the starting point to the right with a step size of 2 data points to form a data set, and the SK at this time is calculated. Next, each time the endpoint is taken as the starting point, data with a step size of 2 is taken to the left and right respectively. A total of 6 data points including the starting point are taken to construct a data set, and they are calculated in sequence until all the points are taken. The maximum value SKM is found, and it is compared with the SK value at the symmetrical moment at this time (if there is SKM at 9:00 am, it is compared with the SK value corresponding to 15:00 pm). If the data at one end is less than 20% of the other end at this time, it indicates that the shadow occurs in the time period of the subscript moment of that end.

4. A photovoltaic string shadowing assessment method according to claim 1, characterized in that: Based on each occlusion cluster center c j , dynamically set its threshold and calculate its power loss: in is the fuzzy membership of the i-th sample point to the j-th category; η p (x i ) is the i-th sample and the power loss percentage; η p (c1),,η p (c2),,η p (c3),,η p (c4) corresponds to normal, mild occlusion, moderate occlusion and severe occlusion respectively; the dynamic threshold is defined as: When there is no shadow: η p <ρ1; Mild occlusion: ρ1≤η p <ρ2; Moderate occlusion: ρ2≤η p <ρ3; severe occlusion: η p ≥ρ3.

5. The photovoltaic string shadowing assessment method according to claim 1, characterized in that: In order to adapt to the data changes of daily time series, a sliding window is introduced to dynamically update the daily sample data: X t ={x t-T+1 ,x t-T+2 ,...,x t } (3) Among them, X t is the set of all samples on a certain day; x i is the feature value of the i-th sample of the day; T is the window size. As each day passes, the latest sample x t Add to the window, and move the earlier samples x t+T Move out of the window, recalculate the cluster center and sample distribution characteristics to update the thresholds ρ1, ρ2, ρ3 in real time.

Images