Photovoltaic string fault diagnosis method based on multi-dimensional and multi-parameter numerical analysis

Abstract

The application discloses a photovoltaic string fault diagnosis method based on multi-dimension and multi-parameter numerical analysis, and comprises the following steps: acquiring voltage U DC , current I DC , module temperature data T model and corresponding meteorological data of a photovoltaic string; performing data cleaning on the data to obtain effective data; performing calculation on the effective data to obtain a state index of the photovoltaic string, and meanwhile, storing the state index of the photovoltaic string into historical data; performing multi-dimension and multi-parameter numerical analysis on the state index in combination with the state index data and the historical data of the photovoltaic string to calculate a fault characteristic index of the photovoltaic string; establishing a fault diagnosis model; and performing fault diagnosis on the photovoltaic string. The application formulates an analysis method of multiple fault types by using power data, meteorological data and historical analysis result data of a photovoltaic power station, and can instantaneously, efficiently and accurately judge and predict the fault of the photovoltaic string under the premise of not increasing cost, thereby providing powerful support for improving quality and increasing efficiency of the photovoltaic power station.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of photovoltaic power generation, and particularly relates to a photovoltaic string fault diagnosis method based on multi-dimensional and multi-parameter numerical analysis. BACKGROUND

[0002] With the implementation of energy structure reform and the "double carbon" target, photovoltaic power generation as a clean energy has been vigorously developed, and the capacity of power stations has also developed from kilowatt to hundreds of megawatts, even gigawatts. Large photovoltaic power stations not only reduce the cost, but also facilitate management. However, it also brings difficulties to the operation and maintenance of power stations. A hundred-megawatt photovoltaic power station has tens of thousands of photovoltaic strings, occupies thousands of acres, and it is difficult to find faults in time and accurately by relying on traditional on-site inspection and maintenance methods. It is even more difficult to predict faults. In addition, photovoltaic power stations are generally built in the wilderness far from the city, and it is difficult to recruit power station operation and maintenance personnel, and the personnel cost is high. In addition, with the development of China's economy, the labor cost is rising. In the future, less staff or even no staff will be on duty in photovoltaic power stations. Under this background, it is very important to do a good job in the operation and maintenance management of power stations. The core equipment of photovoltaic power stations is photovoltaic modules, and the component that has the greatest impact on power generation capacity is also photovoltaic modules. The smallest monitoring object is a photovoltaic string. Therefore, how to find or predict the fault of the photovoltaic string in time and accurately is of great significance to the operation and maintenance and quality improvement and efficiency increase of photovoltaic power stations.

[0003] At present, the fault diagnosis methods of photovoltaic strings mainly include the following:

[0004] The first is a photovoltaic module fault diagnosis method based on a circuit structure. This method needs to determine the number of sensors required according to the circuit structure of the photovoltaic array. Although the accuracy is high, the cost is also very high due to the use of a large number of sensors. For the photovoltaic power station that has been completed, increasing the sensors involves a series of engineering problems such as communication and wiring, and the economic efficiency and adaptability are not strong. The second is a photovoltaic module fault diagnosis method based on an IV output characteristic curve. The premise of data acquisition in this method is that the inverter has an IV scanning test function, or the IV scanning test is performed on site. The inverter of the existing photovoltaic power station rarely configures the IV scanning test function, and the method of measuring IV on site needs to consume a large amount of labor cost and is difficult to realize instantaneity. The other two are a photovoltaic module fault diagnosis method based on a mathematical model and a photovoltaic module fault diagnosis method based on an artificial intelligence algorithm. However, these methods have strong dependence on the environment, especially the photovoltaic module fault diagnosis method based on the artificial intelligence algorithm. The theoretical model is very complex, and is greatly affected by the outside world. A large amount of training and optimization is needed before the actual application, and the requirement for the data volume of the specific environmental data is higher. The dependence is stronger, and it is too complex for actual engineering diagnosis. The last one is a method based on historical data statistical analysis. This kind of analysis method is based on the statistical analysis of the data under the ideal irradiance condition. However, the irradiance data of the actual photovoltaic power station changes frequently and is unpredictable, so that the adaptability of this method is not strong in the complex weather environment. The above methods are based on on-site measurement or hardware configuration. In short, they all need professional equipment and technology, which is difficult and costly. SUMMARY

[0005] To solve the technical problems in the prior art, the purpose of the present application is to provide a photovoltaic string fault diagnosis method based on multi-dimensional and multi-parameter numerical analysis.

[0006] To achieve the above purposes and effects, the technical scheme adopted by the present application is as follows:

[0007] The photovoltaic string fault diagnosis method based on multi-dimensional and multi-parameter numerical analysis comprises the following steps:

[0008] S01: obtaining the voltage U DC , current I DC , module temperature data T model and corresponding meteorological data of the photovoltaic string;

[0009] S02: performing data cleaning on the data obtained in step S01 to obtain valid data;

[0010] S03: calculating the valid data after cleaning to obtain the state index of the photovoltaic string, and storing the state index of the photovoltaic string in the historical data;

[0011] S04: Combine photovoltaic string status index data and historical data to perform multi-dimensional and multi-parameter numerical analysis of status indexes and extract fault characteristic indicators of photovoltaic strings;

[0012] S05: Establish a fault diagnosis model;

[0013] S06: Based on the indicators obtained in step S04 and the fault diagnosis model obtained in step S05, perform fault diagnosis on the photovoltaic string.

[0014] Furthermore, in step S01, the voltage U of the photovoltaic string... DC Current I DC Component temperature data T model The meteorological data is for a complete day, and the meteorological data includes data related to the photovoltaic string voltage U. DC Current I DC Component temperature data T model Corresponding irradiance I r and ambient temperature data T temp .

[0015] Furthermore, in step S02, the steps of cleaning the data obtained in step S01 to obtain valid data specifically include:

[0016] S02-01: Extract sunshine data segment D from a full day's data, specifically the sunshine duration data, according to the defined sunshine duration interval. day ;

[0017] S02-02: Based on the inverter's operating status, obtain the grid connection and grid disconnection times of the inverters connected to the photovoltaic strings. If the inverter starts and stops multiple times within a day, record the times of each start and stop, denoted as t. load-k t unload-k k = {1, 2, 3, ...};

[0018] S02-03: Based on the inverter's first power-on time t load-1 and the last shutdown time t unload-k From D day The valid data after grid connection is extracted from the data and denoted as D. grid ;

[0019] S02-04: Regarding D grid Voltage data U in DC Perform cleaning;

[0020] S02-05: Regarding D grid Current data I in DC Perform cleaning;

[0021] S02-06: Regarding D gridThe ambient temperature data T temp Perform cleaning;

[0022] S02-07: Regarding D grid Component temperature data T model Perform cleaning;

[0023] S02-08: Regarding D grid The irradiance data Ir in the data is cleaned.

[0024] Furthermore, in steps S02-04, for D... grid Voltage data U in DC The rules for cleaning are as follows:

[0025] Voltage U in chronological order DC Perform a scan, if U DC <0 or U DC >nVoc, where n is the number of modules in the photovoltaic string and Voc is the open-circuit voltage of a single module, then U DC The time is recorded in the set Ut invalid , denoted as Ut invalid ={ut1, ut2, ..., ut N}; if ut i+1 -ut i >τ, where τ is the data granularity, determined based on the actual data collection and the workload of computer technology, then U DC (ut i )=(U DC (ut i-1 )+U DC (ut i+1 )) / 2, if ut i+1 -ut i If the expression equals τ, then continue the evaluation until ut is reached. j -ut i+m >τ, then U in this interval DC (ut i )=(U DC (ut i-1 )+U DC (ut j )) / 2.

[0026] Furthermore, in steps S02-05, for D... grid Current data I in DC The rules for cleaning are as follows:

[0027] Current I in chronological order DC Perform a scan, if I DC <0 or I DC >1.2Isc, where Isc is the short-circuit current of a single component, then IDC The moment is recorded in the collection It invalid , denoted as It invalid ={it1, it2, ..., it N}; if it i+1 -it i >τ, then I DC (it i ) = (I DC (it i-1 )+I DC (it i+1 )) / 2, if it n+1 -it n If the expression equals τ, then continue judging until it is found. j -it i+m >τ, then I in this interval DC (it i ) = (I DC (it j )+I DC (it i-1 )) / 2.

[0028] Furthermore, in steps S02-06, for D... grid The ambient temperature data T temp The rules for cleaning are as follows:

[0029] Obtain the lowest and highest temperatures over the past 30 years for the geographical location of the photovoltaic power station, denoted as T. temp-L T temp-H Obtain the lowest and highest temperatures from the day's meteorological data, denoted as T. temp-l T temp-h The minimum and maximum temperatures for the cleaning rules are calculated as T. temp-lowest =MIN(T) temp-L T temp-l ), T temp-hightest =MAX(T) temp-H T temp-h To avoid filtering out valid data due to regional temperature differences, a coefficient k = 1.1 is used to amplify the cleaning rules; furthermore, since temperature changes do not result in abrupt changes, T is calculated... temp Temperature change δT in the data series tmp =T tempi -T tempi-1 If δT tmp >5℃, then T tempi The time is recorded in the set Tt invalid , denoted as Tt invalid ={t1, t2, ..., t N}; if t i+1 -ti >τ, then T temp (t i ) = (T temp (t i-1 ) + T temp (t i+1 )) / 2, if t i+1 -t i >τ, then continue to judge, until t j -t i+m >τ, then T tempi (t i ) = (T temp (t i-1 ) + Ttemp(t j )) / 2, j = {1, 2, … i-1}.

[0030] Further, in step S02-07, the rule for cleaning the component temperature data T grid in D model is:

[0031] Scan T model in time sequence: because the component temperature changes continuously, the probability of large amplitude jump is very low, for the first data point, if |T model1 -T temp |>5℃, then T model1 =T temp , otherwise calculate T model in turn, the temperature change amount δT model =|T modeli -T modeli-1 | in data series, if δT model >5℃, then record the time of T modeli to the set t modelinvalid , recorded as t modelinvalid ={t1, t2, …, t N}; if t i+1 -t i >τ, then T model (t i ) = (T model (t i-1 ) + T model (t i+1 )) / 2, otherwise continue to judge, until t j -t i+m >τ, then T model (t i ) = (T model (t j ) + T model (t i-1 )) / 2.

[0032] Furthermore, in steps S02-08, for D... grid Irradiance data in r The rules for cleaning are as follows:

[0033] I in chronological order r Perform a scan: If I r <0 or I r >1500, then I r The corresponding time is recorded in set I. rinvalid In Chinese, it is denoted as It. invalid ={It t1 It t2 ,…It tn}; if I r (t i+1 )-I R (t i If )>τ, then I r (It i ) = (I r (It i-1 )+I r (It i+1 If the value is 2, then continue judging until it is 1. j -It i+m >τ, then I in this interval r (It i ) = (I r (It j )+I r (It i-1 )) / 2.

[0034] Furthermore, in step S03, the status indicators of the photovoltaic string include the photovoltaic string power generation, the dispersion rate of all strings connected to the inverter or combiner box, the following degree of string current and irradiance, and the power of the string under standard conditions. The steps for obtaining the status indicators of the photovoltaic string include:

[0035] S03-01: Utilize the effective voltage U obtained in step S02 DC and effective current I DC Calculate the power generation W of the photovoltaic string:

[0036]

[0037] S03-02: Calculate the dispersion rate of each data point of the string current connected to the inverter or combiner box, and obtain the dispersion rate Div of the strings of each inverter or combiner box, Div={d i , i∈N *}:

[0038]

[0039] Where S is the standard deviation of the string current values ​​at a certain data point; I ave The average value of the current in all strings at a given data point;

[0040] S03-03: Calculate the following degree of string current versus irradiance, and obtain the following degree data FL, FL = {flam, flpm}:

[0041]

[0042]

[0043] Among them, I M Ir is the nominal operating current of the component. am This refers to the irradiance data for the morning portion; Ir pm This is the irradiance data for the afternoon portion;

[0044] S03-04: From the effective irradiance data Ir and effective string current I DC and voltage U DC Obtain irradiance ≥700W / m 2 Ir irradiance meas Current I Dcmeas Voltage U DCmeas Calculate the string power under standard operating conditions using the following formula:

[0045]

[0046]

[0047] Among them, P stci P represents the nominal power at the i-th point in the data series. stc Ir is the corrected average nominal power. stc Irradiance under standard test conditions; T stc The component temperature under standard test conditions; Pm stc δ represents the nominal maximum operating power of the module; δ is the relative temperature coefficient of the module's power.

[0048] S03-05: From valid D grid Obtain irradiance Ir≥700W / m 2 The corresponding voltage U DC and component temperature data T model Calculate the string voltage under standard operating conditions using the following formula:

[0049] V stci =V measi +β(T modeli -T stcVoc stc

[0050]

[0051] wherein, V stci is the nominal power of the i-th point of the data series; V stc is the corrected average nominal power; V measi is the measured operating voltage of the photovoltaic string; T stc is the temperature of the module under standard test conditions; Pm stc is the nominal maximum operating power of the module; β is the relative temperature coefficient of the voltage of the module; Voc stc is the open-circuit voltage of the photovoltaic module under standard conditions, stc indicates standard conditions, and Voc indicates the open-circuit voltage.

[0052] Further, in step S04, the extraction of the fault features of the photovoltaic string includes abnormal identification and fault features of the string dispersion rate, abnormal identification and fault features of the string current following degree, abnormal identification and fault features of the string power, identification and features of the module fault, identification and features of the string orientation deviation, and abnormal identification and fault features of the string power generation capacity. The step of extracting the fault features of the photovoltaic string includes the following steps:

[0053] S04-01: Abnormal identification and fault feature extraction of the string dispersion rate is performed according to the following steps:

[0054] S04-01-01: The dispersion rate data series of each data point of the string connected to the string inverter or combiner box is calculated, and is recorded as Div={div i , i∈N *};

[0055] S04-01-02: The elements in the Div set are sequentially determined, and the data and serial numbers with div i >5% are added to the set UDiv, and are recorded as

[0056] S04-01-03: If UDiv is not empty, the string has a problem of large dispersion rate, and S04-04 is determined;

[0057] S04-01-04: The Hampel test method is used to identify abnormal strings;

[0058] S04-01-05: The Div state word in the historical record of the string is set as Div#Times#date, Times is the recording times, and is represented by four hexadecimal digits;

[0059] S04-02: Abnormal identification and fault feature extraction of the string current following degree is performed according to the following steps:

[0060] S04-02-01: Calculate the current following degree series of each group string in the morning and afternoon, respectively recorded as FL am = {fl ami , i ∈ N *} and FL pm = {fl pmj , j ∈ N *} according to the calculation formula of step S03-03.

[0061] S04-02-02: Determine the value of each element in FL am and FL pm in turn, put the elements with >1.25 into sets UFL am and UFL pm , recorded as UFL am = {(i, fl ami ), fl ami ∈ FL am} and UFL pm = {(j, fl pmj ), fl pmj ∈ FL pm}.

[0062] S04-02-03: If only UFL am is not empty and there is continuous data greater than 30 minutes, it is determined that there is a fixed shading or the group string is oriented westward in the morning. If the continuous time is less than 30 minutes, it is determined to be a temporary shading; if only UFL pm is not empty and there is continuous data greater than 30 minutes, it is determined that there is a fixed shading or the group string is oriented eastward in the afternoon; if UFL am and UFL pm are not empty and there is continuous data greater than 30 minutes, it is determined that there is a fixed shading or other low-efficiency problem. If UFLam and UFLpm are not empty and the FL state word in the group string historical data is empty, it is determined that there is a temporary shading or a new abnormality.

[0063] S04-02-04: Set the FL state word in the group string historical record to FL#Times#date.

[0064] S04-03: Group string power loss anomaly identification and fault feature extraction is performed according to the following steps:

[0065] S04-03-01: Calculate the output power PW stc of each group string in the standard state according to step S03-04.

[0066] S04-03-02: Take the group string inverter or bus box as a unit, and record the power of the group string connected therein as a set PW = {pwstci , i = 1, 2, … n}, n is the number of string groups;

[0067] S04-03-03: Add the elements of <nkPm> (where n is the number of series components in a string group; k = 1-ξ%-λ%, ξ is the annual attenuation rate of components, and λ is the correction factor) in set PW stci to set PW abnormal ;

[0068] S04-03-04: If the number of elements in PW abnormal is less than n, then the string group corresponding to the elements in set PW abnormal has a loss exceeding the standard, otherwise, perform the occlusion judgment, if there is no occlusion, then all string groups have a loss exceeding the standard, if there is occlusion, then evaluate after eliminating the occlusion;

[0069] If the power loss of the components exceeds the standard, set the PW state word in the historical record of the string group to PW#Times#date;

[0070] S04-04: Component fault identification and fault feature extraction are performed according to the following steps:

[0071] S04-04-01: Calculate the output voltage V stc of each string group under the standard state according to step S03-05;

[0072] S04-04-02: Take the string group inverter or DC bus as a unit, and record the output voltage of the string group connected thereto as a set V = {V stci , i = 1, 2, … n}, n is the number of string groups connected to the inverter or bus;

[0073] S04-04-03: Add the elements of <nkVm> in set V to set V abnormal , n is the number of series components in a string group; k = 1-ξ%-λ%, ξ is the annual attenuation rate of components, and λ is the correction factor;

[0074] S04-04-04: If the number of elements in set V abnormal is less than n, then the string group in set V abnormal has a component failure problem;

[0075] S04-04-05: If the number of elements in set V abnormal is equal to n, then perform the occlusion judgment, if there is no occlusion, then all components of the string group connected to the inverter or bus have a fault;

[0076] S04-04-06: If the components have a fault, set the V state word in the historical record of the string group to V#Times#date;

[0077] S04-05: Group string orientation deviation identification and fault feature extraction is performed according to the following steps:

[0078] S04-05-01: Obtain effective irradiance data, to the time of day t mid , and divide the irradiance data into the morning section Ir am and the afternoon section Ir pm for demarcation.

[0079] S04-05-02: Calculate the slope of the tangent line of each data point of Ir am and Ir pm , respectively, to obtain the slope data series, denoted as Ir slope-am = {Is ami , i ∈ N *}, Ir slope-pm = {Is pmi , i ∈ N *}.

[0080] S04-05-03: If the proportion of elements in Ir slope-am that are > 0 and the proportion of elements in Ir slope-pm that are < 0 are both > 95%, it is determined that the irradiance curve is smooth.

[0081] S04-05-04: Find the maximum value of irradiance Ir mid in the range of t max ± 30 minutes, and obtain the corresponding time t max .

[0082] S04-05-05: Take the string inverter or combiner box as the unit, calculate the time corresponding to the maximum current of each string connected thereto, denoted as t mi , i = 1, 2, … n, n being the number of strings.

[0083] S04-05-06: Determine whether t mi is contained in [t max - δ, t max + δ], if not, proceed to step S04-05-07.

[0084] S04-05-07: If T mi < t max - δ, it is determined that the string is eastward, otherwise it is westward.

[0085] S04-05-08: If the string is eastward, set the EW state word in the string history record to E#Times#date (Times is the number of records, represented by four hexadecimal digits), and if the string is westward, set the EW state word in the string history record to W#Times#date.

[0086] S04-06: Abnormal identification of string power generation and fault feature extraction is carried out according to the following steps:

[0087] S04-06-01: According to the power generation calculation formula in step S03-01, the power generation EG of the string connected to each inverter or combiner box is calculated i , denoted as a set EG={EG i , i=1, 2, 3, …n}, n is the number of strings;

[0088] S04-06-02: The abnormal string is identified by using the Hampel test method;

[0089] S04-06-03: The EG state word in the string history record is set as EG#Times#date.

[0090] Compared with the prior art, the beneficial effects of the present application are:

[0091] (1) The multi-parameter multi-dimensional, multi-parameter fault diagnosis and identification is used to exclude interference factors, and the anti-interference and stability are stronger, the misdiagnosis and missed diagnosis probability is reduced, and the fault type diagnosis is more accurate;

[0092] (2) The fault diagnosis has strong real-time, realizes the fault discovery at the first time, increases the photovoltaic power station operation instantaneity, and assists in improving the power station power generation;

[0093] (3) The string fault diagnosis is based on the operation data of the power station, can realize accurate identification and positioning of string fault, and has stronger practicability. BRIEF DESCRIPTION OF DRAWINGS

[0094] Figure 1 is a flow chart of the photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis of the present application;

[0095] Figure 2 is a flow chart of the string dispersion rate abnormal identification method of the present application;

[0096] Figure 3 is a flow chart of the string current following degree abnormal identification method of the present application;

[0097] Figure 4 is a flow chart of the string power abnormal identification method of the present application;

[0098] Figure 5 is a flow chart of the component fault identification method of the present application;

[0099] Figure 6 is a flow chart of the string orientation abnormal identification method of the present application;

[0100] Figure 7This is a flowchart of the string power generation anomaly identification method of the present invention;

[0101] Figure 8 This is a diagram of the historical data structure for string diagnosis in this invention;

[0102] Figure 9 This is a diagram of the string fault diagnosis model of the present invention;

[0103] Figure 10 This is a graph showing the string current irradiance data of Embodiment 1 of the present invention;

[0104] Figure 11 This is a graph showing the string voltage versus irradiance data from Embodiment 1 of the present invention.

[0105] Figure 12 This is a graph showing the ambient temperature versus component temperature data from Embodiment 1 of the present invention.

[0106] Figure 13 This is a graph showing the irradiance and string current data after cleaning in Embodiment 1 of the present invention.

[0107] Figure 14 This is a graph showing the string voltage data after cleaning in Embodiment 1 of the present invention;

[0108] Figure 15 This is a graph showing the ambient temperature and component temperature data after cleaning in Embodiment 1 of the present invention.

[0109] Figure 16 This is a series of curves showing the string current dispersion rate data of Embodiment 1 of the present invention;

[0110] Figure 17 This is a string current following curve diagram of Embodiment 1 of the present invention;

[0111] Figure 18 This is a dk / S data curve diagram of Embodiment 1 of the present invention;

[0112] Figure 19 The smooth irradiance and string current data are from Embodiment 1 of the present invention;

[0113] Figure 20 This is a slope curve of the irradiance curve tangent data from Embodiment 1 of the present invention.

[0114] Figure 21 This is a flowchart of photovoltaic string fault diagnosis according to Embodiment 1 of the present invention. Detailed Implementation

[0115] The present invention will now be described in detail so that its advantages and features can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0116] The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.

[0117] As shown in Figures 1-21 , a photovoltaic string fault diagnosis method based on multi-dimensional and multi-parameter numerical analysis includes the following steps:

[0118] S01: Obtain the voltage U DC , current I DC , component temperature data T model and corresponding meteorological data of the photovoltaic string;

[0119] S02: Data cleaning is performed on the data obtained in step S01 to obtain valid data;

[0120] S03: The valid data after cleaning is calculated to obtain the state index of the photovoltaic string, and the state index of the photovoltaic string is stored in the historical data;

[0121] S04: Multi-dimensional and multi-parameter numerical analysis is performed on the state index data and historical data of the photovoltaic string to extract the fault feature index of the photovoltaic string;

[0122] S05: Establish a fault diagnosis model;

[0123] S06: According to the index value obtained in step S04 and the fault diagnosis model obtained in step S05, the fault diagnosis of the photovoltaic string is performed.

[0124] In step S01, the voltage U DC , current I DC , component temperature data T model and meteorological data of the photovoltaic string are one complete day data; the data granularity τ is determined according to the actual data acquisition and computer technology task amount, preferably τ = 5 minutes; the meteorological data includes the irradiance I DC , ambient temperature data T DC corresponding to the voltage U model , current I r , and component temperature data T temp .

[0125] In step S02, the step of data cleaning on the data obtained in step S01 to obtain valid data specifically includes:

[0126] S02-01: According to the prescribed sunshine time interval, the data of the day with sunshine is intercepted from the data of the whole day, and the sunshine data segment D is obtained day ;

[0127] S02-02: According to the running state of the inverter, the grid-connected time and off-grid time of the inverter to which the photovoltaic string is connected are obtained, if the inverter is started and stopped multiple times within a day, the time of multiple start and stop is recorded, denoted as t load-k , t unload-k , k={1, 2, 3, …};

[0128] S02-03: According to the first start time t load-1 and the last shutdown time t unload-k of the inverter, the effective data after grid connection is intercepted from D day , denoted as D grid ;

[0129] S02-04: For the voltage data U grid in D DC , the cleaning rule is:

[0130] Scan the voltage U DC in time sequence, if U DC <0 or U DC >nVoc, n is the number of components in the photovoltaic string, Voc is the open circuit voltage of a single component, then the time of U DC is recorded to the set Ut invalid , denoted as Ut invalid ={ut1, ut2, …, ut N}; if ut i+1 -ut i >τ, then U DC (ut i )=(U DC (ut i-1 )+U DC (ut i+1 )) / 2, if ut i+1 -ut i =τ, then continue to judge, until ut j –ut i+m >τ, then U DC (ut i )=(U DC (ut i-1 )+U DC (ut j )) / 2;

[0131] S02-05: For the current data I grid in D DC , the cleaning rule is:

[0132] I(t) is the current in time sequence DC Scan, if I DC <0 or I DC >1.2Isc, Isc is the short-circuit current of the monolithic assembly, then I DC The moment is recorded to the set It invalid , denoted as It invalid ={it1, it2, …, it N}; if it i+1 -it i >τ, then I DC (it i )=(I DC (it i-1 )+I DC (it i+1 )) / 2, if it n+1 -it n =τ, then continue to judge, until it j -it i+m >τ, then I DC (it i )=(I DC (it j )+I DC (it i-1 )) / 2;

[0133] S02-06: For the ambient temperature data T grid in D temp , the cleaning rule is:

[0134] Get the lowest and highest temperature of the past 30 years of the geographical location of the photovoltaic power station, denoted as T temp-L , T temp-H , get the minimum and maximum temperature in the current weather data, denoted as T temp-l , T temp-h ; The minimum and maximum temperatures of the cleaning rule are calculated as T temp-lowest =MIN(T temp-L , T temp-l ), T temp-hightest =MAX(T temp-H , T temp-h ), in order to avoid regional temperature difference leading to effective data being filtered out, a coefficient k=1.1 is used to amplify the cleaning rule; In addition, since temperature change cannot jump, the temperature change amount δT temp = T tempi -T tempi-1 in the data series T tmp , if δT tempi >5℃, then Ttempi The time is recorded in the set Tt invalid , denoted as Tt invalid ={t1, t2, ..., t N}; if t i+1 -t i >τ, then T temp (t i )=(T temp (t i-1 )+T temp (t i+1 )) / 2, if t i+1 -t i If the value is τ, then continue judging until t. j -t i+m If T > τ, then T in this interval tempi (t i )=(T temp (t i-1 )+Ttemp(t j )) / 2, j={1,2、…i-1};

[0135] S02-07: Targeting D grid Component temperature data T model The cleaning rules are as follows:

[0136] T in chronological order model Perform a scan: Because component temperature changes are continuous, the probability of a large jump is very low. For the first data point, if |T model1 -T temp |>5℃, then T model1 =T temp Otherwise, calculate T sequentially. model Temperature change δT in the data series model =|T modeli -T modeli-1 |, if δT model >5℃, then T modeli The time is recorded in set t modelinvalid , denoted as t modelinvalid ={t1, t2, ..., t N}; if t i+1 -t i >τ, then T model (t i )=(T model (t i-1 )+T model (t i+1 If t is not found, the condition is repeated until t is reached. j -t i+m If >τ, then T in this interval model (ti ) = (T model j ) + T model i-1 ) ) / 2.

[0137] S02-08: For the irradiance data I grid in D r , the cleaning rule is:

[0138] Scan I r in time sequence: if I r < 0 or I r > 1500, record the time corresponding to I r to the set I rinvalid , denoted as It invalid = {It t1 , It t2 , …, It tn}; if I r (t i+1 ) - I R (t i ) > τ, then I r (It i ) = (I r (It i-1 ) + I r (It i+1 )) / 2, otherwise continue to judge, until It j - It i+m > τ, then I r (It i ) = (I r (It i ) + I r (It i-1 )) / 2.

[0139] In step S02-01, the rule for determining the sunshine time interval is as follows:

[0140] According to the sunrise time and sunset time in the meteorological data corresponding to the geographical position of the photovoltaic power station, the sunshine time interval of each day within the 25-year life cycle of the power station is determined, and the sunshine time interval of a day is defined as U day = [T rise , T set ].

[0141] In step S03, the state indicators of the photovoltaic string include the power generation of the photovoltaic string, the dispersion rate of all strings connected to the inverter or the combiner box, the following degree of string current and irradiance, and the power under the standard state of the string. The extraction steps of each state indicator are as follows:

[0142] ​​S03-01: Utilize the valid U obtained in step S02 DC and I DC The power generation of the photovoltaic string is calculated using the following formula to obtain the string power generation data W:

[0143]

[0144] S03-02: Calculate the dispersion rate of each data point of the string current connected to the inverter or combiner box, and obtain the dispersion rate of the strings of each inverter or combiner box, denoted as Div={d i , i∈N * Div is calculated using the following steps:

[0145] S03-02-01: Calculate the standard deviation of the current values ​​of all strings at a certain data point, denoted as S;

[0146] S03-02-02: Calculate the average value of the current in all strings at a given data point, denoted as I. ave ;

[0147] S03-02-03: Calculate the dispersion rate Div using the following formula:

[0148]

[0149] S03-03: Calculate the following degree of string current and irradiance tracking, obtain the tracking degree data, denoted as FL, and calculate it according to the following steps:

[0150] S03-03-01: Obtaining the midday time t based on meteorological data m , with t m To serve as a boundary, the irradiance data is divided into the morning portion (Ir). am and the afternoon part of Ir pm At the same time, the string current data is also divided into I... DCam I DCpm ;

[0151] S03-03-02: Calculate the string current I at each data point DC The degree of following the irradiance Ir is calculated using the following formula:

[0152]

[0153]

[0154] Among them, I M The nominal operating current of the component;

[0155] S03-04: From the effective irradiance data Ir and effective string current I DC and voltage U DCS03-04: Obtain the irradiance Ir≥700W / m2 from the effective D meas , current I Dcmeas , voltage U DCmeas , the string power in the standard working state is calculated according to the following formula:

[0156]

[0157]

[0158] Wherein, P stci is the nominal power of the i-th point of the data series; P stc is the corrected average nominal power; Ir stc is the irradiance under standard test conditions; T stc is the component temperature under standard test conditions; Pm stc is the nominal maximum working power of the component; δ is the relative temperature coefficient of the component power;

[0159] S03-05: Obtain the voltage U grid and the component temperature T DC corresponding to the irradiance Ir≥700W / m2 from the effective D model , the string voltage in the standard working state is calculated according to the following formula:

[0160] V stci = V measi + β (T modeli -T stc ) Voc stc

[0161]

[0162] Wherein, V stci is the nominal power of the i-th point of the data series; V stc is the corrected average nominal power; V measi is the measured working voltage of the photovoltaic string; T stc is the component temperature under standard test conditions; Pm stc is the nominal maximum working power of the component; β is the relative temperature coefficient of the component voltage.

[0163] In step S04, a multi-dimensional and multi-parameter fault identification method is constructed, and photovoltaic string fault feature indexes are extracted, including string dispersion rate anomaly identification and fault features, string current following degree anomaly identification and fault features, string power anomaly identification and fault features, component fault identification and features, string orientation deviation identification and features, and string power generation anomaly identification and fault features. The extraction steps of the fault feature indexes include:

[0164] S04-01: Abnormal identification of string discrete rate and fault feature extraction is performed in the following steps, see Figure 2 :

[0165] S04-01-01: Calculate the discrete rate data series of each data point of the string connected to the string inverter or combiner box, denoted as Div={div i , i∈N *};

[0166] S04-01-02: Determine the elements in the Div set one by one, and add the data and series number with div i >5% to the set UDiv, denoted as

[0167] S04-01-03: If UDiv is not empty, there is a problem of large discrete rate in the string, and S04-04 is determined;

[0168] S04-01-04: Identify abnormal strings using the Hampel test method; the Hampel test method is performed in the following steps:

[0169] S04-01-04-01: Calculate the median I DCj0 of the current of each string at time j in the UDiv set;

[0170] S04-01-04-02: Calculate the absolute value of the difference between the current of each string at time j and I DC0 , denoted as DI DCj ={d k |d k =|I DC -I DC0 |}, k=1,2,3,…n;

[0171] S04-01-04-03: Calculate the median d0 of DI DCj , and calculate the absolute deviation estimate S, the calculation formula is:

[0172] S=1.4826×d0;

[0173] S04-01-04-04: Determine whether the current of each string at time j is an outlier, the judgment formula is as follows:

[0174] d k >αS

[0175] Where, α is the detection threshold factor;

[0176] S04-01-05: Set the Div state word in the string history record to Div#Times#date (Times is the record number, represented by four hexadecimal digits);

[0177] S04-02: Abnormal identification of string current following degree and fault feature extraction is performed in the following steps, see Figure 3 :

[0178] S04-02-01: According to the calculation formula of S03-03, the current following degree series of each string in the morning and afternoon is calculated, and is recorded as FL am ={fl ami , i∈N *} and FL pm ={fl pmj , j∈N *};

[0179] S04-02-02: The values of each element in FL am and FL pm are sequentially judged, and the elements with a value greater than 1.25 are placed in sets UFL am and UFL pm , and are recorded as UFL am ={(i, fl ami ), fl ami ∈FL am} and UFL pm ={(j, fl pmj ), fl pmj ∈FL pm};

[0180] S04-02-03: If only UFL am is not empty and there is continuous data greater than 30 minutes, it is judged that there is a fixed shading or the string orientation is westward in the morning; if the continuous time is less than 30 minutes, it is determined to be a temporary shading; if only UFL pm is not empty and there is continuous data greater than 30 minutes, it is judged that there is a fixed shading or the string orientation is eastward in the afternoon; if both UFL am and UFL pm are not empty and there is continuous data greater than 30 minutes, it is judged that there is a fixed shading or other low-efficiency problem. If UFLam and UFLpm are not empty and the FL state word in the string historical data is empty, it is judged that there is a temporary shading or a new abnormality in the string;

[0181] S04-02-04: The FL state word in the string historical record is set to FL#Times#date (Times is the record number, represented by four hexadecimal digits);

[0182] S04-03: Abnormal identification of string power loss and fault feature extraction is performed in the following steps, see Figure 4 :

[0183] S04-03-01: Calculate the output power PW of each string under standard conditions according to the method shown in S03-04 stc ;

[0184] S04-03-02: Taking the string inverter or busbar box as a unit, record the power of the strings connected thereto as the set PW = {pw stci , i = 1, 2,... n}, where n is the number of strings;

[0185] S04-03-03: Add the elements in the set PW that are <nkPm (where n is the number of series-connected components in the string; k = 1 - ξ% - λ%, ξ is the annual attenuation rate of the component, and λ is the correction factor) to the set PW stci ; abnormal ;

[0186] S04-03-04: If the number of elements in PW abnormal is less than n, then there is a problem of excessive attenuation in the strings corresponding to the elements in the PW abnormal set; otherwise, perform occlusion judgment. If there is no occlusion, all strings have excessive attenuation; if there is occlusion, evaluate after eliminating the occlusion.

[0187] S04-03-04: If the component power loss exceeds the standard, set the PW status word in the string historical record to PW#Times#date (Times is the number of records, represented in four-digit hexadecimal);

[0188] S04-04: Component fault identification and fault feature extraction are carried out according to the following steps, see Figure 5 : [[ID=3I]]

[0189] S04-04-01: Calculate the output voltage V of each string under standard conditions according to the calculation formula described in S03-05 stc ;

[0190] S04-04-02: Taking the string inverter or DC busbar box as a unit, record the output voltage of the strings connected thereto as the set V = {V stci , i = 1, 2,... n}, where n is the number of strings connected to the inverter or busbar box;

[0191] S04-04-03: Add the elements in the set V that are <nkVm (n is the number of series-connected components in the string; k = 1 - ξ% - λ%, ξ is the annual attenuation rate of the component, and λ is the correction factor) to the set V abnormal ;

[0192] S04-04-04: If the number of elements in the set V abnormal is less than n, then there is a component fault problem in the strings in the V abnormal set;

[0193] S04-04-05: If the set V abnormal is equal to n, then the occlusion judgment is performed, and if there is no occlusion, then all components of the inverter or combiner box connected to the string are faulty;

[0194] S04-04-06: If the components are faulty, then the V status word in the string history record is set to V#Times#date (Times is the record number, represented by four hexadecimal digits);

[0195] S04-05: The string orientation deviation identification and fault feature extraction are performed according to the following steps, see Figure 6 :

[0196] S04-05-01: Obtain the effective irradiance data, and divide the irradiance data into morning segment Ir mid and afternoon segment Ir pm with midday time t am as the boundary;

[0197] S04-05-02: Calculate the slope of the tangent line of each data point of Ir am and Ir pm respectively, obtain the slope data series, denoted as Ir slope-am ={Is ami , i∈N *}, Ir slope-pm ={Is pmi , i∈N *};

[0198] S04-05-03: If the proportion of elements with >0 in Ir slope-am and <0 in Ir slope-pm is both >95%, then the irradiance curve is smooth;

[0199] S04-05-04: Find the maximum irradiance Ir max in the range of t mid ±30 minutes, and obtain the corresponding time t max ;

[0200] S04-05-05: Take the string inverter or combiner box as the unit, calculate the time corresponding to the maximum current of each string connected to it, denoted as t mi , i=1,2, …n, n is the number of strings;

[0201] S04-05-06: Determine whether t mi is contained in [t max -δ, t max +δ], if not, then proceed to step S04-05-07;

[0202] S04-05-07: If T mi <t max - δ, then determine that the string is eastward, otherwise westward;

[0203] S04-05-08: If the string is eastward, set the EW state word in the string history record to E#Times#date (Times is the record number, represented by four hexadecimal digits), if the string is westward, set the EW state word in the string history record to W#Times#date (Times is the record number, represented by four hexadecimal digits);

[0204] S04-06: Abnormal string power generation recognition and fault feature extraction is performed according to the following steps, see Figure 7 :

[0205] S04-06-01: According to the string power calculation formula described in S03-01, the power EG i of each string connected to the inverter or combiner box is calculated, denoted as a set EG = {EG i , i = 1, 2, 3, … n}, n is the number of strings;

[0206] S04-06-02: Use the Hampel test method to identify abnormal strings; add the abnormal strings to the set UEG, denoted as The Hampel test method is performed according to the following steps:

[0207] S04-06-02-01: Calculate the median EG0in the set EG;

[0208] S04-06-02-02: Calculate the absolute value of the difference between the power of each string and EG0, denoted as DEG = {d i |d i = |EG i - EG0|}, i = 1, 2, 3, … n;

[0209] S04-06-02-03: Calculate the median d0of DEG j , calculate the value of the absolute deviation estimate S, the calculation formula is:

[0210] S = 1.4826 × d0;

[0211] S04-06-02-04: Determine whether it is an outlier, the judgment formula is as follows:

[0212] d i > αS

[0213] Where, α is the detection threshold factor;

[0214] S04-06-03: Set the EG state word in the group string history record to EG#Times#date (Times is the record number, represented by four hexadecimal digits).

[0215] The fault diagnosis model, as shown in Figure 9 , includes fault feature names, fault feature parameters, and fault determination.

[0216] Embodiment 1

[0217] As shown in Figures 1-21 , a photovoltaic string fault diagnosis method based on multi-dimensional and multi-parameter numerical analysis includes the following steps:

[0218] S01: Obtain photovoltaic string voltage U DC , current I DC , module temperature data T model , and corresponding meteorological data, which includes irradiance I r and ambient temperature data T temp corresponding to photovoltaic string voltage U DC , current I DC , and module temperature data T model , as shown in Figure 10 , Figure 11 , Figure 12 .

[0219] S02: Data cleaning is performed on the obtained data to obtain valid data, and the cleaning steps are as follows:

[0220] S02-01: Obtain the sunshine time interval according to the sunrise time and sunset time in the meteorological data corresponding to the geographical location of the photovoltaic power station, denoted as U day =[06:47,17:23];

[0221] According to the specified sunshine time interval, the data during the day with sunshine is obtained, and the valid sunshine data data segment D day is obtained.

[0222] S02-02: According to the running state of the inverter, the grid-connected time and off-grid time of the inverter connected to the photovoltaic string are obtained, denoted as t load =07:08, t unload =17:07.

[0223] S02-03: According to the first start-up time t load and the last shutdown time t unload of the inverter, the valid data after grid connection is obtained from D day data, denoted as D grid .

[0224] S02-04: cleaning the voltage data U in D grid DC

[0225] S02-05: cleaning the current data I in D grid DC

[0226] S02-06: cleaning the ambient temperature data T in D grid temp

[0227] S02-07: cleaning the component temperature data T in D grid model

[0228] S02-08: cleaning the irradiance data Ir in D grid

[0229] The cleaned data curves are shown in Figure 13 , Figure 14 , Figure 15

[0230] S03: calculating the state indicators of the photovoltaic strings based on the cleaned effective data, and storing the state indicators of the photovoltaic strings in the historical data, which is specifically performed according to the following steps:

[0231] S03-01: calculating the power generation of the photovoltaic strings by using the effective U DC and I DC data obtained in step S02, and the power generation data of each string is shown in Table 1:

[0232] Table 1

[0233] String Number String 1 String 2 String 3 String 4 String 5 String Power / kWh 27.558 32.438 32.191 31.684 32.655 String Number String 6 String 7 String 8 String 9 String 10 String Power / kWh 33.867 33.904 33.800 32.148 9.695

[0234] S03-02: calculating the current dispersion rate of all strings connected to the inverter or combiner box, obtaining the dispersion rate of each data point of the strings in each inverter or combiner box, and the current dispersion rate data series of the strings is shown in Figure 16 ;

[0235] S03-03: calculating the following degree of the string current and irradiance, obtaining the morning irradiance following degree FL am data series and the afternoon irradiance following degree FL pm data series, and the string current following degree curve is shown in Figure 17 ;

[0236] S03-04: obtaining the effective irradiance data Ir, effective string current I DC and voltage U DC ​​​​​​​​​​The irradiance Ir≥700W / m2 is obtained meas , the current I DCmeas , the voltage U DCmeas The string power in the standard working state is calculated according to the following formula:

[0237]

[0238]

[0239] The string power calculation results are shown in Table 2:

[0240] Table 2

[0241]

[0242]

[0243] S03-05: From the effective D grid , the irradiance Ir≥700W / m2 is obtained DC , and the component temperature T model The string voltage in the standard working state is calculated according to the following formula, and β=0.32%:

[0244] V stci =V measi +β(T modeli -T stc )Voc stc

[0245]

[0246] The string working voltage calculation results are shown in Table 3:

[0247] Table 3

[0248] String Number String 1, 2 String 3, 4 String 5, 6 String 7, 8 String 9, 10 String Voltage / V 634.21 631.92 641.71 640.10 612.22

[0249] S04: Multi-dimensional numerical analysis is performed on the state indicators by combining the photovoltaic string state indicator data and historical data, and the fault characteristic indicators of the photovoltaic string are calculated;

[0250] S04-01: The string dispersion rate anomaly recognition and fault characteristic extraction are performed according to the following steps and methods:

[0251] S04-01-01: Taking the string inverter or combiner box as a unit, the dispersion rate data series of the string connected thereto is calculated, and the dispersion rate data curve is as shown in Figure 16 ;

[0252] S04-01-02: The elements in the Div set are sequentially determined, and div i>5% of data and serial number into the set UDiv, in the calculation, all the discrete rate series of the string inverter group are greater than 5%, all into the set UDiv:

[0253] S04-01-03: Because UDiv is not empty, the string has a problem of large discrete rate, proceed to S04-01-04 determination;

[0254] S04-01-04: Use Hampel test method to identify abnormal string, the specific steps are as follows:

[0255] S04-01-04-01: Calculate the median of the current of each string at time j in the set UDiv I DC0 ;

[0256] S04-01-04-02: Calculate the absolute value of the difference between the current of each string at time j and I DC0 , denoted as DI DCj ={d k |d k =|I DCk -I DC0 |},k=1,2,3,…n;

[0257] S04-01-04-03: Calculate the median d0 of DI DCj , calculate the absolute deviation estimate S, the calculation formula is:

[0258] S=1.4826×d0;

[0259] S04-01-04-04: Determine whether the current of each string at time j is an outlier, the judgment formula is as follows:

[0260] d k >αS

[0261] Wherein, α is the detection threshold factor, take α=3;

[0262] The data curve of dk / S at each time after calculation is shown in Figure 18 ;

[0263] S04-01-05: Set the Div state word in the string history record to Div#Times#date (Times is the record number, represented by four hexadecimal digits);

[0264] S04-02: String current following degree abnormality identification and fault feature extraction is carried out according to the following steps and methods:

[0265] S04-02-01: According to the calculation formula described in S03-03, the current following degree series of each string in the morning and afternoon is calculated, respectively denoted as FL am={fl ami , i ∈ N*}, FLpm = {fl pmj , j ∈ N*};

[0266] S04-02-02: Judging each element in FL am and FL pm in turn, and putting the elements greater than 1.25 into the sets UFL am and UFL pm . The abnormal periods of FL am and FL pm are shown in Table 4:

[0267] Table 4

[0268]

[0269] S04-02-03: According to the fact that the UFL am of strings 3, 4, 5, 6, 7, and 9 are all not empty and there is data for more than 30 consecutive minutes, it is judged that there is a fixed occlusion in the morning for all strings or the string orientation is westward; the UL am of string 1 is not empty and there is data for more than 30 consecutive minutes, then there is a fixed occlusion in the morning and afternoon for string 1 or the string orientation is eastward; the UFL am and UFL pm of string 10 are both not empty and there is data for more than 30 consecutive minutes, then there is a fixed occlusion or other inefficient problems with the string; <sch <sch

[0270] S04-02-04: Set the FL status word in the string historical record to FL#Times#Date (Times is the recording times, represented in four-digit hexadecimal); <sch <sch

[0271] S04-03: The abnormal identification of string power loss and the extraction of fault characteristics are carried out according to the following steps and methods: <sch <sch

[0272] S04-03-01: Calculate the output power PW stc under the standard state of each string according to the method shown in S03-04, as shown in Table 2; <sch <sch

[0273] S04-03-02: Taking the string inverter or busbar box as a unit, record the power of the strings connected to it as the set PW = {pw stc-i , i = 1, 2,... n}, where n is the number of strings; <sch <sch

[0274] S04-03-03: Add the elements in the set PW where PW stci <nkPm (where n = 22, k = 1 - ξ - λ, ξ = 3.9%, λ = 2%) to the set PW abnormal . The calculation of nkPm is as follows: <sch

[0275] nkPm = 22×(1 - 3.9% - 2%)×235 = 4864.97 W

[0276] S04-03-04: Calculate the PW after calculation abnormal Contains an element count, the number of elements is less than n, PW abnormal There is a problem of excessive loss in the strings corresponding to the elements in the set;

[0277] S04-03-04: Set the PW status word in the historical record of string 10 to PW#Times#Date (Times is the number of records, represented in four-digit hexadecimal);

[0278] S04-04: Component fault feature extraction and identification are carried out according to the following steps and methods:

[0279] S04-04-01: According to the calculation formula described in S03-05, calculate the production voltage Vstc of each string under standard conditions, as shown in Table 3;

[0280] S04-04-02: Taking the string inverter or DC busbar as a unit, record the output voltage of the strings connected to it as the set V = {Vstci, i = 1, 2,... n}, where n is the number of strings connected to the inverter or busbar;

[0281] S04-04-03: Add the elements in the set V that are < nkVm (where n = 22, k = 1 - ξ - λ, ξ = 3.9%, λ = 2%) to the set V abnormal where, nkVm is calculated as follows:

[0282] nkVm = 22×(1 - 3.9% - 2%)×30.2 = 625.2 W

[0283] S04-04-04: Calculate the V after calculation abnormal The number of elements in V is less than n, V abnormal There are component fault problems in the strings in the set;

[0284] S04-04-05: Set the V status word in the historical records of strings 9 and 10 to V#Times#Date (Times is the number of records, represented in four-digit hexadecimal);

[0285] S04-05: String orientation deviation identification and fault feature extraction are carried out according to the following steps and methods:

[0286] S04-05-01: Obtain the effective irradiance data, with the midday time t mid = 12:05 as the boundary, divide the irradiance data into the morning segment Ir am and the afternoon segment Irpm , irradiance data as shown in Figure 19

[0287] S04-05-02: Settle Ir am and Ir pm respectively slope-am = {Is ami , i ∈ N *}, Ir slope-pm = {Is pmi , i ∈ N *}, slope curve as shown in Figure 20

[0288] S04-05-03: The proportion of Ir slope-am > 0 is 96.6%, and the proportion of Ir slope-pm < 0 is 96.7%, both > 95%, and the irradiance curve is smooth

[0289] S04-05-04: Find the maximum irradiance Ir mid = 1048.9 W / ㎡ in the range of t max ± 30 minutes, and get the corresponding time t max = 12:00

[0290] S04-05-05: Calculate the time t mi corresponding to the maximum current of each group string connected to the string inverter or combiner box, as shown in Table 5 respectively

[0291] Table 5

[0292] String String 1 String 2 String 3 String 4 String 5 Peak Time 12:06 12:06 12:05 12:05 12:04 String String 6 String 7 String 8 String 9 String 10 Peak Time 12:33 12:33 12:26 12:26 12:33

[0293] S04-05-06: Determine whether t mi is contained in [t max - 15, t max + 15], and if groups 6, 7, 8, 9, and 10 are not contained in [t max - 15, t max + 15], proceed to step S04-05-07

[0294] S04-05-07: T mi > t max - 15 for groups 6, 7, 8, 9, and 10, and the group string is oriented west

[0295] S04-05-08: Set the EW state word in the historical record of groups 6, 7, 8, 9, and 10 to W#Times#date (Times is the number of records, represented by four hexadecimal digits)​​

[0296] S04-06: Abnormal identification of string power generation and fault feature extraction is performed according to the following steps and methods:

[0297] S04-06-01: According to the string power generation calculation formula in S03-01, the power generation EG of each string connected to the inverter or combiner box is calculated i , denoted as a set EG={EG i , i=1, 2, 3, …n}, n is the number of strings;

[0298] S04-06-02: The Hampel test method is used to identify abnormal strings, and the specific steps are as follows:

[0299] S04-06-02-01: The median EG0 of the power generation of each string in the EG set is calculated;

[0300] S04-06-02-02: The absolute value of the difference between the power generation of each string and EG0 is calculated, denoted as DEG={deg i |deg i =|EG i -EG0|}, i=1, 2, 3, …n;

[0301] S04-06-02-03: The median deg0 of DEG i is calculated, and the absolute deviation estimate S is calculated, and the calculation formula is:

[0302] S=1.4826×deg0;

[0303] S04-06-02-04: Determine whether the power generation of each string is an outlier, and the judgment formula is as follows:

[0304] deg>αS

[0305] Where, α is the detection threshold factor, and α=3 is taken;

[0306] The deg / S of each string after calculation is shown in Table 6:

[0307] Table 6

[0308] String String 1 String 2 String 3 String 4 String 5 deg / S 3.03 0.08 0.08 0.40 0.22 String String 6 String 7 String 8 String 9 String 10 deg / S 0.99 1.01 0.95 0.11 14.42

[0309] S04-06-03: String 1 and string 10 are judged as outliers, and the EG state word in the historical record of string 1 and string 10 is set to EG#Times#date (Times is the record number, represented by four hexadecimal digits).

[0310] S05: The S04 calculation data is input into the fault diagnosis model;

[0311] S06: Fault diagnosis according to the fault diagnosis model, the diagnosis process is as shown in Figure 21 It is finally known that the group string 1, 2, 8 exists in the morning, the internal fault of the assembly 10, the low efficiency or the attenuation of the group string 10 is over standard, and the group string 6, 7, 8, 9, 10 is westward.

[0312] Compared with the prior art, the present application has at least the following technical effects:

[0313] Using the photovoltaic power station electric quantity data, meteorological data, historical analysis result data, a multi-fault type analysis method is formulated, the fault characteristic parameter calculation considers the complex environment influence, the spatial and time dimension analysis is realized, the characteristic parameter calculation is more robust, under the premise of not increasing the cost, the photovoltaic group string fault can be judged and predicted instantly, efficiently and accurately, and the quality and efficiency of the photovoltaic power station are improved.

[0314] The part or structure not specifically described in the present application can adopt the prior art or existing product, and will not be repeated here.

[0315] The above is only an embodiment of the present application, and does not limit the patent range of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification, or direct or indirect application in other related technical fields, are also included in the patent protection range of the present application.

Claims

1. A photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis, characterized in that, Includes the following steps: S01: Obtain the voltage of the photovoltaic string U DC Current I DC Component temperature data T model And the corresponding meteorological data; the meteorological data includes data related to the photovoltaic string voltage. U DC Current I DC Component temperature data T model Corresponding irradiance I r and ambient temperature data T temp ; S02: Perform data cleaning on the data obtained in step S01 to obtain valid data; S03: Calculate the effective data after cleaning to obtain the status indicators of the photovoltaic string, and store the status indicators of the photovoltaic string in the historical data; the status indicators of the photovoltaic string include the photovoltaic string power generation, the dispersion rate of all strings connected to the inverter or combiner box, the following degree of string current and irradiance, and the power of the string under standard conditions. S04: Combine photovoltaic string status index data and historical data to perform multi-dimensional and multi-parameter numerical analysis of status indexes and extract fault characteristics of photovoltaic strings; S05: Establish a fault diagnosis model; S06: Based on the indicators obtained in step S04 and the fault diagnosis model obtained in step S05, perform fault diagnosis on the photovoltaic string; In step S04, extracting fault features of the photovoltaic string includes extracting abnormal string dispersion rate identification and fault features, abnormal string current following degree identification and fault features, abnormal string power identification and fault features, module fault identification and features, string orientation deviation identification and features, and abnormal string power generation identification and fault features. The steps for extracting fault features of the photovoltaic string include: S04-01: The identification of abnormal string dispersion rate and extraction of fault features shall be carried out according to the following steps: S04-01-01: Taking a string inverter or combiner box as a unit, calculate the discrete rate data series of each data point of the connected string, denoted as... Div={div i , i∈N * } ; S04-01-02: Sequentially Div Determine the elements in the set and... div i >5% of the data and serial numbers are added to the collection. UDi In v, denoted as UDiv={div j ,j⊆i} ; S04-01-03: If UDiv If it is not empty, then the string has a large dispersion rate, so S04-04 judgment is performed; S04-01-04: Use the Hampel test method to identify abnormal strings; S04-01-05: Retrieving strings from historical records Div The status word is set to Div#Times#Date, where Times is the number of records, represented by four hexadecimal digits. S04-02: String current tracking anomaly identification and fault feature extraction shall be performed according to the following steps: S04-02-01: Based on the component operating current and irradiance data, the current following series for each string in the morning and afternoon are calculated and denoted as follows: FL am ={fl ami , i∈N * }、FL pm ={fl pmj , j∈N * } ; S04-02-02: Judge sequentially FL am and FL pm Given the values ​​of each element, add elements with a value greater than 1.25 to the set. UFL am and UFL pm In the middle, it is denoted as UFL am ={( i , fl ami ), fl ami ∈ FL am }、 UFL pm ={( j , fl pmj ), fl pmj ∈ FL pm }; S04-02-03: If only UFL am If the data is not empty and exists for more than 30 consecutive minutes, it is determined that the cluster has a fixed obstruction in the morning or the cluster is oriented westward. If the consecutive time is less than 30 minutes, it is determined to be a temporary obstruction. UFL pm If the data is not empty and there are consecutive data points longer than 30 minutes, then the string has a fixed obstruction in the afternoon or the string is oriented eastward; if UFL am and UFL pm If all data points are non-empty and there are consecutive data points lasting longer than 30 minutes, then the string may have fixed occlusion or other inefficiencies; if UFLam, UFLpm There exists a non-empty string in the historical data. FL If the status word is empty, the string has temporary occlusion or new anomaly. S04-02-04: Retrieving strings from historical records FL The status word is set to FL#Times#Date; S04-03: String power loss anomaly identification and fault feature extraction shall be performed according to the following steps: S04-03-01: Calculate the output power of each string under standard conditions. PW stc ; S04-03-02: Taking a string inverter or combiner box as a unit, the string power connected to it is denoted as the set. PW= {pw stci , i=1,2,…n} , n This represents the number of strings. S04-03-03: Set PW middle PW stci <nkPm Add the elements to the collection PW abnormal In, among them, n This represents the number of components connected in series in the string; k=1-ξ%-λ% , ξ The annual degradation rate of the component. λ As a correction factor; S04-03-04: If PW abnormal The number of elements is less than n ,but PW abnormal If the string corresponding to the element in the set has a decay problem, otherwise an occlusion judgment is performed. If there is no occlusion, all strings will have a loss exceeding the limit. If there is occlusion, the evaluation is performed after the occlusion is removed. If the component power loss exceeds the limit, the string history will be updated. PW The status word is set to PW#Times#Date; S04-04: Component fault identification and fault feature extraction shall be performed according to the following steps: S04-04-01: Calculate the output voltage of each string under standard conditions. V stc ; S04-04-02: Taking a string inverter or DC combiner box as a unit, the output voltage of the strings connected to it is denoted as the set. V={V stci , i=1,2,…n} n is the number of strings connected to the inverter or combiner box; S04-04-03: Set V China nkVm Adding elements to the collection V abnormal middle; S04-04-04: If set V abnormal The number of elements is less than n ,but V abnormal The strings in the collection have component failure issues; S04-04-05: If set V abnormal Quantity equals n If there is no obstruction, then an obstruction judgment is performed. If there is no obstruction, then all components connected to the inverter or combiner box are faulty. S04-04-06: If a component malfunctions, the string history will be updated. V The status word is set to V#Times#Date; S04-05: String orientation deviation identification and fault feature extraction shall be performed according to the following steps: S04-05-01: Obtain effective irradiance data, based on midday time. t mid To serve as a boundary, the irradiance data is divided into morning and afternoon segments. Ir am and afternoon session Ir pm ; S04-05-02: Calculate separately Ir am and Ir pm The slope of the tangent line at each data point is used to obtain a slope data series, denoted as . Ir slope-am ={Is ami ,i∈N * } , Ir slope-pm ={Is pmi ,i∈N * } ; S04-05-03: If Ir slope-am The middle > 0 and Ir slope-pm If the proportion of elements with <0 is >95%, then the irradiance curve is considered to be smooth. S04-05-04: In t mid Irradiance within ±30 minutes Ir Maximum value Ir max Get the corresponding time t max ; S04-05-05: Taking a string inverter or combiner box as a unit, calculate the time corresponding to the maximum current of each connected string, denoted as... t mi , i=1,2,…n , n This represents the number of strings. S04-05-06: Judgment t mi Is it included in [ t max -δ , t max +δ If it is not included in the list, proceed to step S04-05-07; S04-05-07: If T mi <t max -δ If the string is true, it is determined that the string is biased to the east; otherwise, it is biased to the west. S04-05-08: If the string is biased towards the east, then the string history records will be updated. EW The status word is set to E#Times#Date. If the string is dated in the Western Hemisphere, the string history will be updated accordingly. EW The status word is set to W#Times#Date; S04-06: The identification of abnormal string power generation and extraction of fault features shall be carried out according to the following steps: S04-06-01: Calculate the power generation of each string connected to each inverter or combiner box. EG i , denoted as set EG={EG i ，i=1, 2,3，…n} , n This represents the number of strings. S04-06-02: Use the Hampel test method to identify abnormal strings; S04-06-03: Retrieving strings from historical records EG The status word is set to EG#Times#Date.

2. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 1, characterized in that, In step S01, the voltage of the photovoltaic string U DC Current I DC Component temperature data T model And the meteorological data is a complete day's data.

3. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 1, characterized in that, In step S02, the data cleaning process performed on the data obtained in step S01 to obtain valid data specifically includes: S02-01: Extract sunshine data segments from a full day's data, specifically those containing sunshine hours, according to a defined sunshine time interval. D day ; S02-02: Based on the inverter's operating status, obtain the grid connection and grid disconnection times of the inverters connected to the photovoltaic strings. If the inverter starts and stops multiple times within a day, record the multiple start and stop times, denoted as follows: t load-k , t unload-k , k={1、2、 3、…} ; S02-03: Based on the inverter's first power-on time t load-1 and the last shutdown time t unload-k ,from D day Extract the valid data after grid connection from the data. Let it be D grid ; S02-04: Regarding D grid Voltage data in U DC Perform cleaning; S02-05: Regarding D grid Current data in I DC Perform cleaning; S02-06: Regarding D grid Ambient temperature data T temp Perform cleaning; S02-07: Regarding D grid Component temperature data T model Perform cleaning; S02-08: Regarding D grid Irradiance data in Ir Clean it.

4. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 3, characterized in that, In steps S02-04, for D grid Voltage data in U DC The rules for cleaning are as follows: Voltage in chronological order U DC Perform a scan, if U DC <0 or U DC > nVoc n is the number of modules in the photovoltaic string. Voc If the open-circuit voltage of a single component is given, then... U DC Record the moment to the collection Ut invalid , recorded as Ut invalid ={ ut 1 ut 2 , ..., ut N };like ut i+1 - ut i > τ , τ The data granularity is determined based on the actual data collection and the workload of computer technology. U DC (ut) i ）= （U DC (ut) i-1 ）+U DC (ut) i+1 ）） / 2 ,like ut i+1 -ut i =τ Then continue judging until... ut j -ut i+m >τ Then this interval U DC (ut) i ）=（U DC (ut) i-1 ）+U DC (ut) j ）） / 2 .

5. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 3, characterized in that, In steps S02-05, for D grid Current data in I DC The rules for cleaning are as follows: Current in chronological order I DC Perform a scan, if I DC <0 or I DC >1.2 Isc , Isc If the short-circuit current is a single component, then... I DC Record the moment to the collection It invalid , recorded as It invalid ={it 1 it 2 ..., it N } ;like it i+1 -it i >τ ,but I DC (it) i ） =（I DC (it) i-1 ）+I DC (it) i+1 ）） / 2 ,like it n+1 -it n =τ Then continue judging until... it j -it i+m >τ Then this interval I DC (it) i ）=（I DC (it) j ）+I DC (it) i-1 ）） / 2 .

6. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 3, characterized in that, In steps S02-06, for D grid Ambient temperature data T temp The rules for cleaning are as follows: Obtain the lowest and highest temperatures over the past 30 years for the geographical location of the photovoltaic power station, denoted as follows: T temp-L 、T temp-H Obtain the lowest and highest temperatures from the day's meteorological data, and record them as follows: T temp-l , T temp-h The minimum and maximum temperatures for the cleaning rules are calculated as follows: T temp-lowest =MIN(T temp-L ，T temp-l ） , T temp-hightest =MAX(T) temp-H ，T temp-h ） To avoid filtering out valid data due to regional temperature differences, a coefficient is used. k =1.1 Enlarges the cleaning rules; T temp Temperature variation in the data series δT tmp =T tempi -T tempi-1 ,like δT tmp >5℃ Then T tempi Record the moment to the collection Tt invalid , recorded as Tt invalid ={t 1 ，t 2 ，…，t N } ; like t i+1 -t i >τ ,but T temp （t i ）=（T temp （t i-1 ）+ T temp （t i+1 ）） / 2 ,like t i+1 -t i =τ Then continue judging until... t j -t i+m >τ Then this interval T tempi （t i ）=（T temp （t i-1 ）+T temp （t j ）） / 2 , j = {1, 2, ..., i-1} .

7. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 3, characterized in that, In steps S02-07, for D grid Component temperature data T model The rules for cleaning are as follows: In chronological order T model Perform a scan: For the first data point, if |T model1 -T temp |>5℃ ,but T model1 =T temp Otherwise, calculate T sequentially. model Temperature variation in the data series δT model =|T modeli -T modeli-1 | ,like δT model >5℃ Then T modeli Record the moment to the collection t modelinvalid , recorded as t modelinvalid ={t1,t 2 ，…，t N } ; like t i+1 -t i >τ ,but T model （t i ）=（T model （t i-1 ）+T model （t i+1 ）） / 2 Otherwise, continue judging until... t j -t i+m >τ Then this interval T model （t i ） =（T model （t j ）+T model （t i-1 ）） / 2 .

8. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 3, characterized in that, In steps S02-08, for D grid Irradiance data in I r The rules for cleaning are as follows: In chronological order I r Perform a scan: If I r <0 or I r >1500, then I r The corresponding time is recorded in the set. I rinvalid In the middle, it is denoted as It invalid ={It t1 It t2 ,…It tn } ; like I r (t i+1 )-I R （t i )>τ ,but I r (It) i ）=（I r (It) i-1 ）+I r (It) i+1 ）） / 2 Otherwise, continue judging until... It j -It i+m >τ Then this interval I r (It) i ）=（I r (It) j ）+I r (It) i-1 ）） / 2 .

9. The photovoltaic string fault diagnosis method based on multi-dimensional, multi-parameter numerical analysis according to claim 1, characterized in that, In step S03, the steps for obtaining the status indicators of the photovoltaic string include: S03-01: Utilize the effective voltage obtained in step S02 U DC and effective current I DC Calculate the power generation of a photovoltaic string W : ； S03-02: Calculate the dispersion rate of each data point of the string current connected to the inverter or combiner box, and obtain the dispersion rate of the strings of each inverter or combiner box. Div , Div={d i , i∈N * } : ； in, S The standard deviation of all string current values ​​at a given data point; I ave The average value of the current in all strings at a given data point; S03-03: Calculate the tracking degree of string current versus irradiance to obtain tracking degree data. FL, FL ={ flam, flpm }: ； in, I M The nominal operating current of the component; Ir am This is the irradiance data for the morning portion; Ir pm This is the irradiance data for the afternoon portion; S03-04: From effective irradiance data Ir and effective string current I DC and voltage U DC Obtain irradiance ≥700W / m 2 irradiance Ir meas Current I Dcmeas ,Voltage U DCmeas Calculate the string power under standard operating conditions using the following formula: ； in, P stci The nominal power of the i-th point in the data series; P stc This is the corrected average nominal power; Ir measi This is the measured irradiance; Ir stc Irradiance under standard test conditions; T stc The component temperature under standard test conditions; Pm stc The nominal maximum operating power of the component; δ The relative temperature coefficient of component power; S03-05: From effective D grid Obtaining irradiance Ir ≥700W / m 2 corresponding voltage U DC and component temperature data T model Calculate the string voltage under standard operating conditions using the following formula: ； in, V stci For the data series number i The nominal power of each point; V stc This is the corrected average nominal power; V measi To measure the operating voltage of the photovoltaic string; β The relative temperature coefficient of component voltage; Voc stc This represents the open-circuit voltage of a photovoltaic module under standard conditions.

Images