A photovoltaic string aging fault degree assessment method and system

Abstract

The present invention discloses a method and system for evaluating the degree of aging fault of photovoltaic strings, including obtaining historical time series data of photovoltaic strings under normal working conditions of a target photovoltaic power station; calculating the curvature of a voltage series curve and the kurtosis of a current series curve respectively; performing equivalent integration processing on the current series curve and the voltage series curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively; constructing a sample data set according to the kurtosis, curvature, the first index and the second index, performing a first clustering processing on the sample data set to obtain a sample data set to be optimized; performing a second clustering processing on the sample data set to be optimized to obtain a target sample data set; and determining the evaluation result of the degree of aging fault of the target photovoltaic string based on the time series data, the sample data set to be optimized and the target sample data set obtained in real time. The method of the present invention can realize real-time monitoring and diagnosis of aging faults of photovoltaic strings.

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 and system for evaluating the degree of aging faults of photovoltaic strings. Background Art

[0002] Photovoltaic strings convert light energy into electricity and are an indispensable part of the photovoltaic power generation system. As the key to converting solar energy into electricity, their performance and stability are directly related to the efficiency and safety of the entire photovoltaic power generation system. As they work outdoors for a long time, performance degradation and various faults are inevitable as their service life increases. Among them, aging faults are one of the most common faults of photovoltaic strings. Their generation and evolution will gradually reduce the power generation efficiency of the photovoltaic power generation system and affect the safety of system operation.

[0003] Traditional aging fault judgment of photovoltaic strings often relies on manual inspection, which is prone to missed detection and misjudgment. For example, during manual inspection, inspectors may ignore certain details due to fatigue, distraction, etc., resulting in the omission of potential fault points; aging faults of photovoltaic strings are sometimes not intuitive, such as micro-cracks in the cell, aging of the welding strip, etc. These hidden faults are difficult to detect with the naked eye and are prone to missed detection. Summary of the invention

[0004] The present invention aims to provide a method and system for evaluating the degree of aging fault of photovoltaic strings, so as to solve the technical problem of how to improve the accuracy of aging fault detection of photovoltaic strings in distributed photovoltaic power stations.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for evaluating the degree of aging fault of a photovoltaic string, comprising:

[0006] Acquire historical time series data of photovoltaic strings in a target photovoltaic power station under normal working conditions, wherein the historical time series data includes a voltage series curve and a current series curve;

[0007] respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve, wherein the curvature is designed to be obtained by processing a characteristic triangle matching the voltage sequence curve;

[0008] Performing equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively;

[0009] constructing a sample data set according to the kurtosis, the curvature, the first index and the second index, and performing a first clustering process on the sample data set to obtain a sample data set to be optimized;

[0010] Performing a second clustering process on the sample data set to be optimized to obtain a target sample data set, wherein the target sample data set reflects the result after eliminating non-aging fault data of photovoltaic strings;

[0011] When evaluating the target photovoltaic string, an aging fault degree evaluation result of the target photovoltaic string is determined based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set.

[0012] As one preferred solution, respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve includes:

[0013] Taking three data points with the highest radiation in the voltage sequence curve as three vertices to construct a characteristic triangle;

[0014] Calculating the length and area of ​​each side of the characteristic triangle to obtain the curvature of the voltage sequence curve according to a discrete point curvature calculation method;

[0015] Constructing a current sample set based on the current sequence curve;

[0016] Calculating the fourth-order central moment of the current sample set according to the current sample set;

[0017] Calculate the sample variance of the current sample set based on the fourth-order central moment;

[0018] The kurtosis of the current sequence curve is calculated according to the sample variance and the fourth-order central moment.

[0019] As one of the preferred solutions, performing equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively include:

[0020] Using the sampling interval time as a division standard, the current sequence curve is divided into a plurality of first sub-intervals;

[0021] Calculating the area of ​​each of the first sub-intervals by trapezoidal integration method to obtain a first indicator reflecting the degree of current fault;

[0022] Using the sampling interval as a division standard, dividing the voltage sequence curve into a plurality of second sub-intervals;

[0023] The area of ​​each of the second sub-intervals is calculated by the trapezoidal integration method to obtain a second indicator reflecting the degree of voltage fault.

[0024] As one of the preferred solutions, the sample data set to be optimized includes a normal sample set, an aging fault sample set, and a non-aging fault sample set;

[0025] The performing a second clustering process on the sample data set to be optimized to obtain a target sample data set includes:

[0026] Constructing a second sample data set based on the normal sample set and the aging fault sample set;

[0027] Constructing a direct proportional function according to the first indicator and the second indicator;

[0028] Randomly selecting a point on the direct proportional function as an initial cluster center, and calculating the distance between the initial cluster center and each sample in the second sample data set;

[0029] Calculating the weight of each sample according to the distance;

[0030] In the weighted distance calculation process, the weighted distance from each of the aging data samples to the initial cluster center is calculated based on the weight and the distance;

[0031] In the process of determining the cluster center, the probability distribution of each of the samples is calculated based on the weighted distance, and a new cluster center is selected according to the probability distribution;

[0032] The weighted distance calculation process and the cluster center determination process are repeated until a target sample data set reflecting the degree of sample failure is obtained.

[0033] As one of the preferred solutions, the step of determining the aging fault degree assessment result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set includes:

[0034] Constructing a real-time sample according to the time series data acquired in real time;

[0035] Calculating a first Euclidean distance between the real-time sample and each cluster center in the sample data set to be optimized, and determining a fault type of the real-time sample based on the first Euclidean distance;

[0036] If the fault type is an aging fault sample type, calculating the second Euclidean distance between the real-time sample and each cluster center in the target sample data set;

[0037] The aging degree of the photovoltaic string is evaluated according to the second Euclidean distance.

[0038] Another embodiment of the present invention provides a photovoltaic string aging fault degree assessment system, comprising:

[0039] An acquisition module, used to acquire historical time series data of photovoltaic strings in a target photovoltaic power station under normal working conditions, wherein the historical time series data includes a voltage series curve and a current series curve;

[0040] A first calculation module, used for respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve, wherein the curvature is designed to be obtained by processing a characteristic triangle matching the voltage sequence curve;

[0041] A second calculation module is used to perform equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively;

[0042] a first clustering module, configured to construct a sample data set according to the kurtosis, the curvature, the first index and the second index, and perform a first clustering process on the sample data set to obtain a sample data set to be optimized;

[0043] A second clustering module is used to perform a second clustering process on the sample data set to be optimized to obtain a target sample data set, wherein the target sample data set reflects the result after eliminating non-aging fault data of photovoltaic strings;

[0044] The evaluation module is used to determine the aging fault degree evaluation result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set when evaluating the target photovoltaic string.

[0045] As one preferred solution, the first calculation module is used to:

[0046] Taking three data points with the highest radiation in the voltage sequence curve as three vertices to construct a characteristic triangle;

[0047] Calculating the length and area of ​​each side of the characteristic triangle to obtain the curvature of the voltage sequence curve according to a discrete point curvature calculation method;

[0048] Constructing a current sample set based on the current sequence curve;

[0049] Calculating the fourth-order central moment of the current sample set according to the current sample set;

[0050] Calculate the sample variance of the current sample set based on the fourth-order central moment;

[0051] The kurtosis of the current sequence curve is calculated according to the sample variance and the fourth-order central moment.

[0052] As one preferred solution, the second calculation module is used to:

[0053] Using the sampling interval time as a division standard, the current sequence curve is divided into a plurality of first sub-intervals;

[0054] Calculating the area of ​​each of the first sub-intervals by trapezoidal integration method to obtain a first indicator reflecting the degree of current fault;

[0055] Using the sampling interval as a division standard, dividing the voltage sequence curve into a plurality of second sub-intervals;

[0056] The area of ​​each of the second sub-intervals is calculated by the trapezoidal integration method to obtain a second indicator reflecting the degree of voltage fault.

[0057] As one of the preferred solutions, the sample data set to be optimized includes a normal sample set, an aging fault sample set, and a non-aging fault sample set;

[0058] The second clustering module is used for:

[0059] Constructing a second sample data set based on the normal sample set and the aging fault sample set;

[0060] Constructing a direct proportional function according to the first indicator and the second indicator;

[0061] Randomly selecting a point on the direct proportional function as an initial cluster center, and calculating the distance between the initial cluster center and each sample in the second sample data set;

[0062] Calculating the weight of each sample according to the distance;

[0063] In the weighted distance calculation process, the weighted distance from each of the aging data samples to the initial cluster center is calculated based on the weight and the distance;

[0064] In the process of determining the cluster center, the probability distribution of each of the samples is calculated based on the weighted distance, and a new cluster center is selected according to the probability distribution;

[0065] The weighted distance calculation process and the cluster center determination process are repeated until a target sample data set reflecting the degree of sample failure is obtained.

[0066] As one preferred solution, the evaluation module is used to:

[0067] Constructing a real-time sample according to the time series data acquired in real time;

[0068] Calculating a first Euclidean distance between the real-time sample and each cluster center in the sample data set to be optimized, and determining a fault type of the real-time sample based on the first Euclidean distance;

[0069] If the fault type is an aging fault sample type, calculating the second Euclidean distance between the real-time sample and each cluster center in the target sample data set;

[0070] The aging degree of the photovoltaic string is evaluated according to the second Euclidean distance.

[0071] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0072] 1) The present invention extracts data features from multiple dimensions. This multi-dimensional analysis can more comprehensively reflect the working status and potential faults of photovoltaic strings;

[0073] 2) The present invention can effectively eliminate non-aging fault data through two clustering processes, so as to focus on analyzing faults caused by aging. This data cleaning and screening process improves the accuracy of fault identification;

[0074] 3) The present invention can quickly provide the evaluation results of the aging fault degree of photovoltaic strings based on the time series data acquired in real time and the constructed sample data set. This real-time evaluation capability is of great significance for taking maintenance measures in time and preventing the expansion of faults;

[0075] 4) The present invention can timely discover and handle aging faults of photovoltaic strings, and can prevent the faults from expanding and affecting the overall performance of the power station, thereby improving the reliability and stability of the photovoltaic power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 is a flow chart of a method for evaluating the degree of aging fault of a photovoltaic string in one embodiment of the present invention;

[0077] Figure 2 It is a schematic diagram of a simulation curve of a voltage / current time series of an aging string in one embodiment of the present invention;

[0078] Figure 3 1 is a schematic diagram of the curvature that can be extracted from the concave characteristic of the aging fault voltage time series in one embodiment of the present invention, wherein the red curve is the aging fault voltage time series curve, the black curve is the normal voltage time series curve, the shaded part is the fault voltage degree index, and a, b, and c are the three sides of the characteristic triangle respectively;

[0079] Figure 4 It is a schematic diagram of extractable kurtosis of an aging fault current time series in one embodiment of the present invention, in which curves of different colors represent current time series curves at different aging degrees;

[0080] Figure 51 is a schematic diagram of extracting voltage fault degree index in one embodiment of the present invention, wherein the red curve is the aging fault voltage time series curve, the black curve is the normal voltage time series curve, and the shaded part is the fault voltage degree index;

[0081] Figure 6 1 is a schematic diagram of an extracted current fault degree index in one embodiment of the present invention, wherein the red curve is an aging fault current time series curve, the black curve is a normal current time series curve, and the shaded portion is a fault current degree index;

[0082] Figure 7 It is a structural flow chart of a photovoltaic string aging fault degree assessment system in one embodiment of the present invention. DETAILED DESCRIPTION

[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0085] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0086] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0087] An embodiment of the present invention provides a method for evaluating the degree of aging failure of a photovoltaic string. For details, see Figure 1 , Figure 1 The flowchart of the photovoltaic string aging fault degree assessment method provided in one embodiment of the present invention is shown. It includes steps S1 to S6:

[0088] S1. Acquire historical time series data of photovoltaic strings in a target photovoltaic power station under normal working conditions, wherein the historical time series data includes a voltage series curve and a current series curve;

[0089] S2. respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve, wherein the curvature is designed to be obtained by processing a characteristic triangle matching the voltage sequence curve;

[0090] S3, performing equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault, respectively;

[0091] S4, constructing a sample data set according to the kurtosis, the curvature, the first index and the second index, and performing a first clustering process on the sample data set to obtain a sample data set to be optimized;

[0092] S5, performing a second clustering process on the sample data set to be optimized to obtain a target sample data set, wherein the target sample data set reflects the result after eliminating non-aging fault data of photovoltaic strings;

[0093] S6. When evaluating the target photovoltaic string, determine the aging fault degree evaluation result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set.

[0094] The fault diagnosis sample data comes from Huawei FusionSolar Intelligent Photovoltaic Management Platform. To improve the accuracy of fault diagnosis, sunny weather time series data is selected as data samples. Several power stations in 2014, 2017 and 2020 are selected, with 540 groups of aging fault samples of varying degrees, 360 groups of non-aging fault samples, and 100 groups of normal samples. The sampling period and duration of the time series are 5 minutes, 6:00~18:00. To eliminate the influence of weather factors, all time series samples are first filtered and then normalized to achieve sample consistency.

[0095] Specifically, the SmartLogger data collector is used with a sampling interval of 5 minutes, and the data of the inverter and environmental monitor are transmitted to the Huawei Fusionsolar smart photovoltaic management system. In clear weather, the time period of 6:00~18:00 with strong light amplitude is selected for sampling, and a total of 145 data points are obtained as a group of data to be diagnosed; the original data to be diagnosed is filtered and normalized to obtain the voltage time series and current time series mapped in the [0,1] numerical interval.

[0096] Preferably, in one embodiment of the present invention, respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve comprises:

[0097] Taking three data points with the highest radiation in the voltage sequence curve as three vertices to construct a characteristic triangle;

[0098] Calculating the length and area of ​​each side of the characteristic triangle to obtain the curvature of the voltage sequence curve according to a discrete point curvature calculation method;

[0099] Constructing a current sample set based on the current sequence curve;

[0100] Calculating the fourth-order central moment of the current sample set according to the current sample set;

[0101] Calculate the sample variance of the current sample set based on the fourth-order central moment;

[0102] The kurtosis of the current sequence curve is calculated according to the sample variance and the fourth-order central moment.

[0103] Specifically, aging faults are obvious and mainly affect the series resistance Rs and the photocurrent Iph in the string equivalent circuit model parameters. Figure 2The figure shows a schematic diagram of the simulation curve of the voltage / current time series of the aging group string described in an exemplary embodiment of this specification. Since the photovoltaic group string works outdoors all year round, as the service life increases, the sealing of the component deteriorates, and water vapor penetrates into the interior of the component, causing the EVA and glass panels to decompose and discolor, and the transmittance decreases, reducing the photocurrent Iph of the component. At the same time, the viscosity between the packaging material and the battery cell decreases, and the stratification phenomenon occurs, causing the contact resistance to increase. That is, the series resistance increases. Therefore, in the voltage time series of the aging fault. When the degree of aging fault increases, the voltage time series curve will show a significant concave phenomenon under high irradiance. This is because when the irradiance increases, Iph increases, and the Rs loss voltage also increases accordingly. Since the light intensity reaches the maximum value at noon, the voltage change drops to the minimum value at this time, and the voltage time series shows a concave characteristic.

[0104] The voltage time series of aging faults has a concave characteristic, such as Figure 3 As shown, Figure 3 The curvature diagram is shown as the concave characteristic of the voltage time series of the aging fault. Other faults have no concave characteristics, so their curvature is calculated as a feature quantity. Since the current time series is approximately normally distributed, such as Figure 4 As shown, Figure 4 The figure shows that the kurtosis can be extracted from the time series of the aging fault current in a normal distribution, so the kurtosis is calculated as a supplementary feature. Although each fault current time series is approximately normally distributed, there are differences in the range of kurtosis values ​​of various faults. Therefore, the kurtosis KI can be extracted from the current time series for supplementation to ensure the identification of aging faults.

[0105] The curvature can indicate the degree of bending of the curve at a specific point, and the degree of concavity of the voltage time series is positively correlated with the magnitude of the curvature. The voltage time series is formed by connecting the maximum power point voltages at each moment. It is quite difficult to directly obtain the curvature of the voltage function by the function derivation method. Therefore, the discrete point curvature calculation method is adopted to characterize the concavity of the voltage time series by the curvature of the circumscribed circle of the characteristic triangle at (t2, U2). That is, the data point A (t2, U2) at 12:00 noon is used as one vertex of the triangle, and the two inflection points B, C (t1, U1) and (t3, U3) of the voltage time series are used as the other two vertices to form a characteristic triangle ABC. Calculate the lengths of each side a, b, c and the area S of the voltage characteristic triangle to obtain the curvature of the voltage characteristic triangle A (t2, U2). The calculation formula is as follows:

[0106] (1)

[0107] In the formula, a, b, c are the lengths of the three sides of the characteristic triangle, and S is the area of ​​the voltage characteristic triangle. According to Heron's formula, we get:

[0108] (2)

[0109] (3)

[0110] Wherein, U1 is the voltage value at the inflection point in the morning (6:00-12:00), U2 is the voltage value at 12:00 noon, U3 is the voltage value at the inflection point in the afternoon (12:00-18:00), t1 is the time corresponding to the voltage value at the inflection point in the morning (6:00-12:00), t2 is the time corresponding to the voltage value at 12:00 noon, t3 is the time corresponding to the voltage value at the inflection point in the afternoon (12:00-18:00).

[0111] The current time series is similar to a normal distribution, so the kurtosis is used to extract another feature quantity. The larger the kurtosis, the steeper the data point distribution and the higher the degree of aggregation. Its calculation is as shown in formula (4):

[0112] (4)

[0113] In the formula, ——current time series kurtosis; Ii and ——Current of aging strings and normal strings at the i-th moment, A; and ——Time series mean of current of aging strings and normal strings, A; and ——Standard deviation of current time series of aging strings and normal strings.

[0114] Preferably, in one embodiment of the present invention, performing equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault, respectively, comprises:

[0115] Using the sampling interval time as a division standard, the current sequence curve is divided into a plurality of first sub-intervals;

[0116] Calculating the area of ​​each of the first sub-intervals by trapezoidal integration method to obtain a first indicator reflecting the degree of current fault;

[0117] Using the sampling interval as a division standard, dividing the voltage sequence curve into a plurality of second sub-intervals;

[0118] The area of ​​each of the second sub-intervals is calculated by the trapezoidal integration method to obtain a second indicator reflecting the degree of voltage fault.

[0119] Based on the geometric characteristics of the aging fault graph, the first and second indicators reflecting the degree of voltage fault are extracted to perform degree classification and degree estimation. Figure 5 , Figure 6 As shown, Figure 5 The schematic diagram for extracting voltage fault degree index is shown. Figure 6 The figure shows a schematic diagram of extracting the current fault degree index. The first and second indexes refer to the ratio of the area enclosed by the fault voltage / current time series and the horizontal axis to the area enclosed by the normal time series and the horizontal axis. The increase of the first and second indexes indicates better string power generation performance. Since the time series is a discretely distributed data point, the voltage / current time series is divided into multiple sub-intervals using the sampling interval time as the division standard. The area of ​​each sub-interval is calculated by the trapezoidal integration method, and then the first and second indexes of the entire voltage / current time series are calculated. The calculation formulas of the first and second indexes are as follows:

[0120] (5)

[0121] (6)

[0122] Where, Unor,i is the normal voltage at the i-th moment, Inor,i is the normal current value at the i-th moment, ti is the i-th moment, Ui is the voltage of the string to be diagnosed at the i-th moment, and Ii is the current of the string to be diagnosed at the i-th moment.

[0123] Preferably, in one embodiment of the present invention, the sample data set to be optimized includes a normal sample set, an aging fault sample set, and a non-aging fault sample set;

[0124] The performing a second clustering process on the sample data set to be optimized to obtain a target sample data set includes:

[0125] Constructing a second sample data set based on the normal sample set and the aging fault sample set;

[0126] Constructing a direct proportional function according to the first indicator and the second indicator;

[0127] Randomly selecting a point on the direct proportional function as an initial cluster center, and calculating the distance between the initial cluster center and each sample in the second sample data set;

[0128] Calculating the weight of each sample according to the distance;

[0129] In the weighted distance calculation process, the weighted distance from each of the aging data samples to the initial cluster center is calculated based on the weight and the distance;

[0130] In the process of determining the cluster center, the probability distribution of each of the samples is calculated based on the weighted distance, and a new cluster center is selected according to the probability distribution;

[0131] The weighted distance calculation process and the cluster center determination process are repeated until a target sample data set reflecting the degree of sample failure is obtained.

[0132] The Euclidean distance between the j-th data sample xj=[KI,j, CU,j, FFU,j, FFI,j] and the initial cluster center ci=[KIo,i, CUo,i] (normal i=0, aging i=1, non-aging i=2) is calculated as shown in formula (7).

[0133] (7)

[0134] Where: di,j is the Euclidean distance between the jth sample and the cluster centers of normal c0, aging c1 and non-aging c2.

[0135] The membership degree of each data sample xj and c0~c2 is calculated, and the data sample xj is divided into normal, aging and non-aging data clusters according to the size of the membership degree to obtain the sample data set to be optimized.

[0136] The membership of each sample data with c0~c2 is as follows:

[0137] (8)

[0138] (9)

[0139] (10)

[0140] Where, c is the number of cluster centers; μij is the membership degree of the jth sample to the i-th cluster center, the value range of μij is 0~1, and the sum of the membership degrees of the sample and each cluster center is equal to 1; dij is the Euclidean distance between the jth sample and the i-th cluster center; m is the fuzzy weighted index, the value range is 0~+∞.

[0141] After obtaining the initial cluster centers and completing the initial classification, the membership matrix U = [μ0j, μ1j, μ2j]T and the normal, aging, and non-aging cluster centers c0, c1, and c2 are iteratively updated using formulas (8-10) until the number of iterations L = 800 or the objective function formula (10) converges to the threshold ε = 10-6, thereby obtaining the normal, aging, and non-aging cluster centers to obtain the target sample data set.

[0142] Among them, it is noted that the sample ( ) is roughly evenly distributed in the direct proportional function On both sides of , according to the sample distribution law, when performing C-means clustering, the selection of the initial cluster center is improved based on the K-means++ method to avoid subsequent iterations falling into the local optimum, including:

[0143] S501, in the direct proportional function A point is randomly selected as the initial cluster center c0, and all aging samples are calculated to a positive proportional function The vertical distance xj;

[0144] S502, calculate the weight wj of each data point:

[0145] ,

[0146] in, is a parameter that controls the influence of distance.

[0147] S503, calculating the weighted distance of all aging data samples:

[0148] ,

[0149] in, It is the minimum Euclidean distance from the data point xj to the current cluster center.

[0150] S504, select the next cluster center, and select the point with the largest probability as the next cluster center according to the probability distribution of the second center. The probability calculation method is:

[0151] ,

[0152] S505. Repeat steps S503 and S504 until four initial cluster centers are selected.

[0153] Preferably, in one embodiment of the present invention, determining the aging fault degree assessment result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set includes:

[0154] Constructing a real-time sample according to the time series data acquired in real time;

[0155] Calculating a first Euclidean distance between the real-time sample and each cluster center in the sample data set to be optimized, and determining a fault type of the real-time sample based on the first Euclidean distance;

[0156] If the fault type is an aging fault sample type, calculating the second Euclidean distance between the real-time sample and each cluster center in the target sample data set;

[0157] The aging degree of the photovoltaic string is evaluated according to the second Euclidean distance.

[0158] Specifically, after the fault level classification of aging faults is clearly defined, the aging degree is estimated by calculating the Euclidean distance between the aging sample and the normal cluster center c0 and the level cluster center ci, as shown in formula (11), where η is the estimated value of the fault degree.

[0159] (11)

[0160] Where: di,k is the Euclidean distance between the k-th data sample xk=[KI,k,CU,k,FFU,k,FFI,k] and the initial cluster center ci=[FFUo,i,FFIo,i], which is calculated as shown in formula (12):

[0161] (12)

[0162] Where: FFU,k is the second index of the k-th data sample, FFI,k is the first index of the k-th data sample, FFUo,i is the second index of the initial cluster center, and FFI,k is the first index of the initial cluster center.

[0163] Another embodiment of the present invention provides a photovoltaic string aging fault degree assessment system, for details, see Figure 7 , Figure 7 The schematic diagram of the structure of the photovoltaic string aging fault degree assessment system provided in one embodiment of the present invention is shown. It includes: an acquisition module 11, a first calculation module 12, a second calculation module 13, a first clustering module 14, a second clustering module 15, and an assessment module 16, wherein:

[0164] An acquisition module 11 is used to acquire historical time series data of photovoltaic strings in a target photovoltaic power station under normal working conditions, wherein the historical time series data includes a voltage series curve and a current series curve;

[0165] A first calculation module 12, used to calculate the curvature of the voltage sequence curve and the kurtosis of the current sequence curve respectively, wherein the curvature is designed to be obtained by processing a characteristic triangle matching the voltage sequence curve;

[0166] A second calculation module 13 is used to perform equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively;

[0167] A first clustering module 14 is used to construct a sample data set according to the kurtosis, the curvature, the first index and the second index, and perform a first clustering process on the sample data set to obtain a sample data set to be optimized;

[0168] A second clustering module 15 is used to perform a second clustering process on the sample data set to be optimized to obtain a target sample data set, wherein the target sample data set reflects the result after eliminating non-aging fault data of the photovoltaic strings;

[0169] The evaluation module 16 is used to determine the aging fault degree evaluation result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set when evaluating the target photovoltaic string.

[0170] Preferably, in one embodiment of the present invention, the first calculation module is used to:

[0171] Taking three data points with the highest radiation in the voltage sequence curve as three vertices to construct a characteristic triangle;

[0172] Calculating the length and area of ​​each side of the characteristic triangle to obtain the curvature of the voltage sequence curve according to a discrete point curvature calculation method;

[0173] Constructing a current sample set based on the current sequence curve;

[0174] Calculating the fourth-order central moment of the current sample set according to the current sample set;

[0175] Calculate the sample variance of the current sample set based on the fourth-order central moment;

[0176] The kurtosis of the current sequence curve is calculated according to the sample variance and the fourth-order central moment.

[0177] Preferably, in one embodiment of the present invention, the second calculation module is used to:

[0178] Using the sampling interval time as a division standard, the current sequence curve is divided into a plurality of first sub-intervals;

[0179] Calculating the area of ​​each of the first sub-intervals by trapezoidal integration method to obtain a first indicator reflecting the degree of current fault;

[0180] Using the sampling interval as a division standard, dividing the voltage sequence curve into a plurality of second sub-intervals;

[0181] The area of ​​each of the second sub-intervals is calculated by the trapezoidal integration method to obtain a second indicator reflecting the degree of voltage fault.

[0182] Preferably, in one embodiment of the present invention, the sample data set to be optimized includes a normal sample set, an aging fault sample set, and a non-aging fault sample set;

[0183] The second clustering module is used for:

[0184] Constructing a second sample data set based on the normal sample set and the aging fault sample set;

[0185] Constructing a direct proportional function according to the first indicator and the second indicator;

[0186] Randomly selecting a point on the direct proportional function as an initial cluster center, and calculating the distance between the initial cluster center and each sample in the second sample data set;

[0187] Calculating the weight of each sample according to the distance;

[0188] In the weighted distance calculation process, the weighted distance from each of the aging data samples to the initial cluster center is calculated based on the weight and the distance;

[0189] In the process of determining the cluster center, the probability distribution of each of the samples is calculated based on the weighted distance, and a new cluster center is selected according to the probability distribution;

[0190] The weighted distance calculation process and the cluster center determination process are repeated until a target sample data set reflecting the degree of sample failure is obtained.

[0191] Preferably, in one embodiment of the present invention, the evaluation module is used to:

[0192] Constructing a real-time sample according to the time series data acquired in real time;

[0193] Calculating a first Euclidean distance between the real-time sample and each cluster center in the sample data set to be optimized, and determining a fault type of the real-time sample based on the first Euclidean distance;

[0194] If the fault type is an aging fault sample type, calculating the second Euclidean distance between the real-time sample and each cluster center in the target sample data set;

[0195] The aging degree of the photovoltaic string is evaluated according to the second Euclidean distance.

[0196] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0197] 1) The present invention extracts data features from multiple dimensions. This multi-dimensional analysis can more comprehensively reflect the working status and potential faults of photovoltaic strings;

[0198] 2) The present invention can effectively eliminate non-aging fault data through two clustering processes, so as to focus on analyzing faults caused by aging. This data cleaning and screening process improves the accuracy of fault identification;

[0199] 3) The present invention can quickly provide the evaluation results of the aging fault degree of photovoltaic strings based on the time series data acquired in real time and the constructed sample data set. This real-time evaluation capability is of great significance for taking maintenance measures in time and preventing the expansion of faults;

[0200] 4) The present invention can timely discover and handle aging faults of photovoltaic strings, and can prevent the faults from expanding and affecting the overall performance of the power station, thereby improving the reliability and stability of the photovoltaic power station.

[0201] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for evaluating the degree of aging failure of photovoltaic strings, characterized in that: The following steps are involved: Acquire historical time series data of photovoltaic strings in a target photovoltaic power station under normal working conditions, wherein the historical time series data includes a voltage series curve and a current series curve; respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve, wherein the curvature is designed to be obtained by processing a characteristic triangle matching the voltage sequence curve; Performing equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively; constructing a sample data set according to the kurtosis, the curvature, the first index and the second index, and performing a first clustering process on the sample data set to obtain a sample data set to be optimized; Performing a second clustering process on the sample data set to be optimized to obtain a target sample data set, wherein the target sample data set reflects the result after eliminating non-aging fault data of photovoltaic strings; When evaluating the target photovoltaic string, determining an aging fault degree evaluation result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set; Wherein, determining the aging fault degree assessment result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set includes: Constructing a real-time sample according to the time series data acquired in real time; Calculating a first Euclidean distance between the real-time sample and each cluster center in the sample data set to be optimized, and determining a fault type of the real-time sample based on the first Euclidean distance; If the fault type is an aging fault sample type, calculating the second Euclidean distance between the real-time sample and each cluster center in the target sample data set; The aging degree of the photovoltaic string is evaluated according to the second Euclidean distance.

2. The photovoltaic string aging fault degree assessment method according to claim 1 is characterized in that: The respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve comprises: Taking three data points with the highest radiation in the voltage sequence curve as three vertices to construct a characteristic triangle; Calculating the length and area of ​​each side of the characteristic triangle to obtain the curvature of the voltage sequence curve according to a discrete point curvature calculation method; Constructing a current sample set based on the current sequence curve; Calculating the fourth-order central moment of the current sample set according to the current sample set; Calculate the sample variance of the current sample set based on the fourth-order central moment; The kurtosis of the current sequence curve is calculated according to the sample variance and the fourth-order central moment.

3. The photovoltaic string aging fault degree assessment method according to claim 1, characterized in that: The equivalent integration processing is performed on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault, respectively, including: Using the sampling interval time as a division standard, the current sequence curve is divided into a plurality of first sub-intervals; Calculating the area of ​​each of the first sub-intervals by trapezoidal integration method to obtain a first indicator reflecting the degree of current fault; Using the sampling interval as a division standard, dividing the voltage sequence curve into a plurality of second sub-intervals; The area of ​​each of the second sub-intervals is calculated by the trapezoidal integration method to obtain a second indicator reflecting the degree of voltage fault.

4. The photovoltaic string aging fault degree assessment method according to claim 1, characterized in that: The sample data set to be optimized includes a normal sample set, an aging fault sample set, and a non-aging fault sample set; The performing a second clustering process on the sample data set to be optimized to obtain a target sample data set includes: Constructing a second sample data set based on the normal sample set and the aging fault sample set; Constructing a direct proportional function according to the first indicator and the second indicator; Randomly selecting a point on the direct proportional function as an initial cluster center, and calculating the distance between the initial cluster center and each sample in the second sample data set; Calculating the weight of each sample according to the distance; In the weighted distance calculation process, the weighted distance from each of the aging data samples to the initial cluster center is calculated based on the weight and the distance; In the process of determining the cluster center, the probability distribution of each of the samples is calculated based on the weighted distance, and a new cluster center is selected according to the probability distribution; The weighted distance calculation process and the cluster center determination process are repeated until a target sample data set reflecting the degree of sample failure is obtained.

5. A photovoltaic string aging fault degree assessment system, characterized in that: Includes the following modules: An acquisition module, used to acquire historical time series data of photovoltaic strings in a target photovoltaic power station under normal working conditions, wherein the historical time series data includes a voltage series curve and a current series curve; A first calculation module, used for respectively calculating the curvature of the voltage sequence curve and the kurtosis of the current sequence curve, wherein the curvature is designed to be obtained by processing a characteristic triangle matching the voltage sequence curve; A second calculation module is used to perform equivalent integration processing on the current sequence curve and the voltage sequence curve to obtain a first index reflecting the degree of current fault and a second index reflecting the degree of voltage fault respectively; a first clustering module, configured to construct a sample data set according to the kurtosis, the curvature, the first index and the second index, and perform a first clustering process on the sample data set to obtain a sample data set to be optimized; A second clustering module is used to perform a second clustering process on the sample data set to be optimized to obtain a target sample data set, wherein the target sample data set reflects the result after eliminating non-aging fault data of photovoltaic strings; An evaluation module, used to determine an aging fault degree evaluation result of the target photovoltaic string based on the time series data acquired in real time, the sample data set to be optimized and the target sample data set when evaluating the target photovoltaic string; Wherein, the evaluation module is used for: Constructing a real-time sample according to the time series data acquired in real time; Calculating a first Euclidean distance between the real-time sample and each cluster center in the sample data set to be optimized, and determining a fault type of the real-time sample based on the first Euclidean distance; If the fault type is an aging fault sample type, calculating the second Euclidean distance between the real-time sample and each cluster center in the target sample data set; The aging degree of the photovoltaic string is evaluated according to the second Euclidean distance.

6. The photovoltaic string aging fault degree assessment system according to claim 5, characterized in that: The first computing module is used for: Taking three data points with the highest radiation in the voltage sequence curve as three vertices to construct a characteristic triangle; Calculating the length and area of ​​each side of the characteristic triangle to obtain the curvature of the voltage sequence curve according to a discrete point curvature calculation method; Constructing a current sample set based on the current sequence curve; Calculating the fourth-order central moment of the current sample set according to the current sample set; Calculate the sample variance of the current sample set based on the fourth-order central moment; The kurtosis of the current sequence curve is calculated according to the sample variance and the fourth-order central moment.

7. The photovoltaic string aging fault degree assessment system according to claim 5, characterized in that: The second computing module is used for: Using the sampling interval time as a division standard, the current sequence curve is divided into a plurality of first sub-intervals; Calculating the area of ​​each of the first sub-intervals by trapezoidal integration method to obtain a first indicator reflecting the degree of current fault; Using the sampling interval as a division standard, dividing the voltage sequence curve into a plurality of second sub-intervals; The area of ​​each of the second sub-intervals is calculated by the trapezoidal integration method to obtain a second indicator reflecting the degree of voltage fault.

8. The photovoltaic string aging fault degree assessment system according to claim 5, characterized in that: The sample data set to be optimized includes a normal sample set, an aging fault sample set, and a non-aging fault sample set; The second clustering module is used for: Constructing a second sample data set based on the normal sample set and the aging fault sample set; Constructing a direct proportional function according to the first indicator and the second indicator; Randomly selecting a point on the direct proportional function as an initial cluster center, and calculating the distance between the initial cluster center and each sample in the second sample data set; Calculating the weight of each sample according to the distance; In the weighted distance calculation process, the weighted distance from each of the aging data samples to the initial cluster center is calculated based on the weight and the distance; In the process of determining the cluster center, the probability distribution of each of the samples is calculated based on the weighted distance, and a new cluster center is selected according to the probability distribution; The weighted distance calculation process and the cluster center determination process are repeated until a target sample data set reflecting the degree of sample failure is obtained.

Images