Method and system for detecting abnormal state of photovoltaic power generation

Through the abnormal state detection method of photovoltaic power generation, data preprocessing and abnormal judgment modules are used, combined with isolated forest algorithms and neural networks, the fault causes of distributed photovoltaic systems are identified, and the problem of strong concealment of distributed photovoltaic systems is solved and detection is improved. Accuracy and power generation efficiency.

CN119010788APending Publication Date: 2024-11-22SHENZHEN GUODIAN POWER SALES CO LTD

Patent Information

Application Number
CN202411075010.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Due to the lack of professional operation and maintenance, the faults are highly concealed, which makes the fault detection difficult, affecting the power generation efficiency and life cycle.

Method used

The abnormal state detection method of photovoltaic power generation is used to obtain and preprocess data, combine sunny solar irradiance characteristics screening, and use isolated forest algorithms and neural networks to judge the health status of the power station, calculate the correlation of incentives, and accurately identify abnormal states and incentives.

Benefits of technology

It improves the effectiveness and accuracy of detection, can promptly detect fault causes, helps the maintenance and power generation efficiency of photovoltaic power stations to fully utilize, and supports dual energy consumption control targets.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119010788A_ABST
    Figure CN119010788A_ABST
Patent Text Reader

Abstract

The invention discloses a photovoltaic power generation abnormal state detection method and system, and relates to the technical field of new energy, and the method comprises the steps: obtaining to-be-detected photovoltaic power station data and related solar irradiance data; performing data preprocessing on the to-be-detected photovoltaic power station data; carrying out screening processing on the related solar irradiance data; performing correlation analysis on different power stations to obtain photovoltaic power stations with similar to-be-detected data characteristics, and performing transverse comparison; judging the health state of the environment-similar photovoltaic power station group sample data; fitting a normal interval of the power generation power data of the photovoltaic power station; judging whether the proportion of the data in the normal interval in the to-be-detected data is higher than a set health threshold proportion or not; and calculating the correlation degree between each possible cause and the to-be-detected data, and outputting the possible cause with the highest correlation degree. By arranging the data processing module and the abnormity judgment module, the effectiveness and accuracy of detection are improved, and intelligent detection of user distributed photovoltaic fault abnormity is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of new energy technology, and in particular to a method and system for detecting abnormal state of photovoltaic power generation. Background Art

[0002] Photovoltaic arrays are an important part of photovoltaic power generation systems. Due to their long-term operation in a relatively harsh environment, factors such as heat, electricity, and external damage may cause open circuits, short circuits, hot spots in the array, as well as inverter overcurrent, overvoltage, power tube open circuits, and other faults, which seriously affect the power generation efficiency of the photovoltaic system. In addition, local shadows not only reduce the output power of photovoltaic modules, but also become one of the causes of hot spots and other faults. In order to avoid serious accidents caused by faults and improve the power generation efficiency of the system, domestic and foreign research institutions have started research on photovoltaic system fault diagnosis technology.

[0003] As a renewable energy power generation technology with high technical maturity and low resource cost. Affected by technological progress and policy support, the development of centralized photovoltaic power generation systems is slow, while distributed photovoltaic power generation systems have begun to accelerate. With the advantages of not being constrained by land use policies and flexible deployment, their share in newly built photovoltaic systems has increased significantly. Distributed photovoltaic power generation systems are scattered outdoors, lack professional operation and maintenance, have a large number of equipment, and have various forms of abnormal failures, which may manifest as array open circuit, short circuit, inverter overcurrent, overvoltage, power tube open circuit and hot spots. The probability of failure of photovoltaic systems is significantly higher than other forms of power generation. In addition, there are no rotating parts, and the failure is highly hidden. If the failure cannot be diagnosed and repaired early, the power generation of the entire life cycle will be significantly reduced. Summary of the invention

[0004] In order to solve the above technical problems, a method and system for detecting abnormal state of photovoltaic power generation are provided. This technical solution solves the problems raised in the above background technology.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for detecting abnormal state of photovoltaic power generation, comprising:

[0007] Obtain the data of the photovoltaic power station to be tested and the relevant solar irradiance data;

[0008] Performing data preprocessing on the photovoltaic power station data to be tested to obtain data to be tested;

[0009] A clear sky day screening mechanism is developed based on the solar irradiance characteristics of clear days, and the relevant solar irradiance data are screened and processed to obtain clear sky dates;

[0010] Conduct correlation analysis on different power stations to obtain photovoltaic power stations with similar data characteristics to be tested, and conduct horizontal comparison to screen out photovoltaic power station groups with similar environments;

[0011] The health status of sample data of photovoltaic power stations with similar environments is determined by the isolation forest algorithm;

[0012] Learning the features of the healthy sample data of the photovoltaic power station marked by the isolation forest algorithm, and fitting the normal interval of the power generation data of the photovoltaic power station using a neural network regression model;

[0013] Determine whether the proportion of data in the normal range in the data to be detected is higher than the set health threshold ratio. If so, output that the photovoltaic power station is normal; if not, output that the photovoltaic power station is in an abnormal state;

[0014] Obtain at least one possible cause of abnormal photovoltaic power generation status;

[0015] The correlation between each possible cause and the data to be measured is calculated, and the possible cause with the highest correlation is output as the abnormal cause of the photovoltaic power station.

[0016] Preferably, the step of preprocessing the data of the photovoltaic power station to be tested to obtain the data to be tested specifically includes:

[0017] Merge different data sources and store them on the same server;

[0018] Data cleaning is performed on the photovoltaic power station data to be tested by removing irrelevant data;

[0019] The cleaned data is uniformly transformed into the range of [0, 1] using the linear transformation formula;

[0020] The linear transformation formula is:

[0021]

[0022] In the formula, x i ′ is the value of the unified i-th data, x i is the value of the i-th data before unification, x min , x max are the minimum and maximum values ​​of all data respectively.

[0023] Preferably, the clear sky day screening mechanism is formulated in combination with the solar irradiance characteristics of clear days, and the relevant solar irradiance data is screened and processed to obtain the clear sky date, which specifically includes:

[0024] Obtain historical solar irradiance data and obtain the historical daily maximum irradiance value;

[0025] Sort the historical daily maximum irradiance values ​​and set the highest quarter as the clear sky threshold;

[0026] Determine whether the daily maximum irradiance of the data to be tested is higher than the clear sky threshold. If so, output the solar irradiance data for that day; if not, do not output;

[0027] Set the nighttime irradiance intensity to 0, and convert the solar irradiance data of the day into a solar irradiance sequence through a normalization formula;

[0028] Using the solar irradiance sequence, construct the solar irradiance fluctuation sequence per unit period;

[0029] Determine whether the product of the last segment of data and the previous segment of data in the solar irradiance fluctuation sequence per unit time period is positive. If so, no output is made; if not, the two segments of data are output as the fluctuation range;

[0030] Determine whether the number of fluctuation intervals contained in the solar irradiance data of the day is greater than 2. If so, set the day as a cloudy day; if not, set the day as a sunny day;

[0031] The normalization formula is:

[0032]

[0033] In the formula, x i,j ′ is the normalized value of the ith data on the jth day, x i,j is the value of the i-th data on the j-th day before normalization, x j,max is the maximum value of the data on the jth day;

[0034] The solar irradiance fluctuation sequence per unit period is:

[0035] b i,j =x i,j ′-x i-1,j ′,

[0036] Where b i,j is the solar irradiance fluctuation value from the i-1th moment to the i-th moment on the jth day.

[0037] Preferably, the performing correlation analysis on different power stations to obtain photovoltaic power stations with similar data characteristics to be detected, and performing horizontal comparison to screen out photovoltaic power station groups with similar environments specifically includes:

[0038] Obtain the power generation of each power station at each time period;

[0039] Use the correlation formula to calculate the correlation coefficient between each power station;

[0040] Determine whether the correlation coefficient is greater than 0.95. If so, the output is that the environment is similar to a photovoltaic power station group. If not, no output is made.

[0041] The correlation formula is:

[0042]

[0043] In the formula, ρ(P j ,P k ) is the correlation coefficient between the j-th photovoltaic power station and the k-th photovoltaic power station, P i,j is the power generation of the jth photovoltaic power station in the i-th period, P i,k is the power generation of the k-th photovoltaic power station in the i-th period, is the average power generation of the jth photovoltaic power station, is the average power generation of the kth PV power station, and n is the total number of time periods.

[0044] Preferably, the method of judging the health status of sample data of a photovoltaic power station group with a similar environment by using an isolation forest algorithm specifically includes:

[0045] S101: randomly select at least one sample point as the root node in the sample data of the photovoltaic power station group with similar environment; S102: randomly determine a dimension in all data dimensions, and randomly select a split point between the maximum and minimum values ​​of the data in the specified dimension as the split threshold; S103: divide the current data space into two subspaces according to the split threshold, put the data in the specified dimension that is greater than the split threshold into the right subspace, and put the data that is less than the split threshold into the left subspace; S104: determine whether the subspace contains only one data. If so, traverse each isolated tree for each data, record the number of layers of each isolated tree where the data falls, and calculate the average height value of the data in the isolated forest. If not, return to step S102;

[0046] S105: Calculate the standard average height value using the average search length formula;

[0047] S106: Substituting the average height value and the standard average height value into a scoring formula to obtain a score for the data outlier;

[0048] S107: Determine whether the data abnormal value score is higher than the set abnormal score threshold. If so, output the data as abnormal sample data; if not, output the data as healthy sample data; the average search length formula is:

[0049]

[0050] Where n is the total number of data, c is the standard average search length value, ξ is the Euler constant, which is approximately 0.5772; the scoring formula is:

[0051]

[0052] In the formula, s(i,n) is the outlier score of the i-th data. Preferably, the calculation of the correlation between each possible cause and the data to be measured and outputting the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station specifically includes:

[0053] Acquire the photovoltaic power station data when each possible inducement actually occurs; form the data to be measured and the photovoltaic power station data when the possible inducement actually occurs into a first time series and a second time series respectively;

[0054] According to the correlation coefficient formula, the correlation coefficient between the first time series and the second time series is obtained; the correlation coefficient is substituted into the correlation degree formula to obtain the correlation degree between each possible cause and the data to be measured; the possible cause with the highest correlation degree is output as the abnormal cause of the photovoltaic power station; the correlation coefficient formula is:

[0055]

[0056] Where, t i (k) is the correlation coefficient between the data to be tested and the photovoltaic power station data when the i-th possible cause actually occurs, Δ i k is the absolute value of the difference between the data to be tested and the kth data of the photovoltaic power station when the i-th possible cause actually occurs;

[0057] The correlation formula is:

[0058]

[0059] In the formula, r i is the correlation between the ith possible cause and the data to be tested, and n is the length of the time series.

[0060] Furthermore, a photovoltaic power generation abnormal state detection system is proposed, which is used to implement the photovoltaic power generation abnormal state detection method as described above, comprising:

[0061] A data processing module, which is used to obtain the data of the photovoltaic power station to be tested and the related solar irradiance data, perform data preprocessing on the data of the photovoltaic power station to be tested, obtain the data to be tested, formulate a clear sky day screening mechanism in combination with the solar irradiance characteristics of a clear day, screen and process the related solar irradiance data to obtain the clear sky date, perform correlation analysis on different power stations to obtain photovoltaic power stations with similar characteristics of the data to be tested, and perform horizontal comparison to screen out photovoltaic power stations with similar environments;

[0062] The abnormality judgment module is used to judge the health status of sample data of a photovoltaic power station group with similar environment through the isolation forest algorithm, statistically summarize the characteristics of healthy sample data of photovoltaic power stations marked by the isolation forest algorithm, fit the normal range of photovoltaic power station power generation data, judge whether the proportion of data in the normal range in the data to be detected is higher than the set health threshold ratio, obtain at least one possible cause of abnormal photovoltaic power generation, calculate the correlation between each possible cause and the data to be tested, and output the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station.

[0063] Optionally, the data processing module includes:

[0064] A data acquisition unit, the data acquisition unit is used to acquire the photovoltaic power station data to be tested and the related solar irradiance data;

[0065] A data processing unit, the data processing unit is used to perform data preprocessing on the photovoltaic power station data to be tested to obtain data to be tested;

[0066] A first data screening unit, the first data screening unit is used to formulate a clear sky day screening mechanism based on the solar irradiance characteristics of a clear day, screen and process relevant solar irradiance data, and obtain a clear sky date;

[0067] The second data screening unit is used to perform correlation analysis on different power stations to obtain photovoltaic power stations with similar data characteristics to be detected, and perform horizontal comparison to screen out photovoltaic power station groups with similar environments.

[0068] Optionally, the abnormality judgment module includes:

[0069] A data health judgment unit, the data health judgment unit is used to judge the health status of sample data of a photovoltaic power station group with a similar environment through an isolation forest algorithm;

[0070] A normal interval fitting unit, which is used to statistically summarize the features of the healthy sample data of the photovoltaic power station marked by the isolation forest algorithm, and fit the normal interval of the power generation data of the photovoltaic power station;

[0071] An abnormal state judgment unit, the abnormal state judgment unit is used to judge whether the proportion of data in the normal range in the data to be detected is higher than a set health threshold ratio;

[0072] The abnormal cause judgment unit is used to obtain at least one possible cause of the abnormal state of photovoltaic power generation, calculate the correlation between each possible cause and the data to be measured, and output the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] By setting up data processing modules and abnormality judgment modules, the effectiveness and accuracy of detection are improved. Based on the limited data of the system, intelligent detection of user distributed photovoltaic fault anomalies is achieved, and the root causes of the anomalies can be accurately and timely discovered, which is helpful for the subsequent maintenance and re-commissioning of photovoltaic power stations, and can give full play to the power generation efficiency and help achieve the dual control goals of energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 This is a flow chart of the photovoltaic power generation abnormal state detection method proposed by the present invention;

[0076] Figure 2 A flow chart of the data preprocessing method of the present invention;

[0077] Figure 3 A flow chart of a method for screening and processing relevant solar irradiance data according to the present invention;

[0078] Figure 4 It is a flow chart of the method for horizontal comparison of the present invention;

[0079] Figure 5 A flow chart of a method for judging the health status of sample data of a photovoltaic power station group in an environment similar to the present invention;

[0080] Figure 6 This is a flow chart of a method for calculating the correlation between each possible cause and the data to be measured, and outputting the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station. DETAILED DESCRIPTION

[0081] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0082] Reference Figure 1 As shown, a method for detecting abnormal state of photovoltaic power generation includes:

[0083] Obtain the data of the photovoltaic power station to be tested and the relevant solar irradiance data;

[0084] Performing data preprocessing on the photovoltaic power station data to be tested to obtain data to be tested;

[0085] A clear sky day screening mechanism is developed based on the solar irradiance characteristics of clear days, and the relevant solar irradiance data are screened and processed to obtain clear sky dates;

[0086] Conduct correlation analysis on different power stations to obtain photovoltaic power stations with similar data characteristics to be tested, and conduct horizontal comparison to screen out photovoltaic power station groups with similar environments;

[0087] The health status of sample data of photovoltaic power stations with similar environments is determined by the isolation forest algorithm;

[0088] Learning the features of the healthy sample data of the photovoltaic power station marked by the isolation forest algorithm, and fitting the normal interval of the power generation data of the photovoltaic power station using a neural network regression model;

[0089] Determine whether the proportion of data in the normal range in the data to be detected is higher than the set health threshold ratio. If so, output that the photovoltaic power station is normal; if not, output that the photovoltaic power station is in an abnormal state;

[0090] Obtain at least one possible cause of abnormal photovoltaic power generation status;

[0091] The correlation between each possible cause and the data to be measured is calculated, and the possible cause with the highest correlation is output as the abnormal cause of the photovoltaic power station.

[0092] Reference Figure 2 As shown, the data of the photovoltaic power station to be tested is preprocessed to obtain the data to be tested, which specifically includes:

[0093] Merge different data sources and store them on the same server;

[0094] Data cleaning is performed on the photovoltaic power station data to be tested by removing irrelevant data;

[0095] The cleaned data is uniformly transformed into the range of [0, 1] using the linear transformation formula;

[0096] The linear transformation formula is:

[0097]

[0098] In the formula, x i ′ is the value of the unified i-th data, x i is the value of the i-th data before unification, x min , x max are the minimum and maximum values ​​of all data respectively.

[0099] Without a unified standard, it is impossible to evaluate the experimental results. Therefore, dedimensionalization of data of different dimensions is the basis for evaluating the experimental results. Dedimensionalization can eliminate the impact of differences in value ranges and scale the data to a specific interval according to a certain ratio. Data unification is a commonly used method of dedimensionalization.

[0100] Reference Figure 3 As shown in the figure, a clear sky day screening mechanism is formulated in combination with the solar irradiance characteristics of clear days. The relevant solar irradiance data are screened and processed to obtain clear sky dates, including:

[0101] Obtain historical solar irradiance data and obtain the historical daily maximum irradiance value;

[0102] Sort the historical daily maximum irradiance values ​​and set the highest quarter as the clear sky threshold;

[0103] Determine whether the daily maximum irradiance of the data to be tested is higher than the clear sky threshold. If so, output the solar irradiance data for that day; if not, do not output;

[0104] Set the nighttime irradiance intensity to 0, and convert the solar irradiance data of the day into a solar irradiance sequence through a normalization formula;

[0105] Using the solar irradiance sequence, construct the solar irradiance fluctuation sequence per unit period;

[0106] Determine whether the product of the last segment of data and the previous segment of data in the solar irradiance fluctuation sequence per unit time period is positive. If so, no output is made; if not, the two segments of data are output as the fluctuation range;

[0107] Determine whether the number of fluctuation intervals contained in the solar irradiance data of the day is greater than 2. If so, set the day as a cloudy day; if not, set the day as a sunny day;

[0108] The normalization formula is:

[0109]

[0110] In the formula, x i,j ′ is the normalized value of the ith data on the jth day, x i,j is the value of the i-th data on the j-th day before normalization, x j,max is the maximum value of the data on the jth day;

[0111] The solar irradiance fluctuation sequence per unit period is:

[0112] b i,j =x i,j ′-x i-1,j ′,

[0113] Where b i,j is the solar irradiance fluctuation value from the i-1th moment to the i-th moment on the jth day.

[0114] In order to eliminate the influence of clouds on the judgment of abnormal faults in photovoltaic systems, it is necessary to define clear sky days. In the meteorological field, clear sky days are defined according to the cloud cover in the sky. When the cloud cover in the sky is less than 25%, it is considered sunny, when the cloud cover is between 25% and 50%, it is cloudy, and when the cloud cover is more than 50%, it is overcast. However, when the photovoltaic system components are abnormal, it may only cause the system to derate by 90%, and the power generation will be reduced by 10%. When there are 25% clouds in the sky, it will cause obvious fluctuations in the output of the photovoltaic system, and the fluctuation range is enough to cover up the abnormality of the photovoltaic system. Therefore, it is necessary to define clear sky days according to the status diagnosis needs.

[0115] Reference Figure 4 As shown in the figure, correlation analysis is performed on different power stations to obtain photovoltaic power stations with similar data characteristics to be detected, and horizontal comparison is performed to screen out photovoltaic power stations with similar environments, including:

[0116] Obtain the power generation of each power station at each time period;

[0117] Use the correlation formula to calculate the correlation coefficient between each power station;

[0118] Determine whether the correlation coefficient is greater than 0.95. If so, the output is that the environment is similar to a photovoltaic power station group. If not, no output is made.

[0119] The correlation formula is:

[0120]

[0121] In the formula, ρ(P j ,P k ) is the correlation coefficient between the j-th photovoltaic power station and the k-th photovoltaic power station, P i,j is the power generation of the jth photovoltaic power station in the i-th period, P i,k is the power generation of the k-th photovoltaic power station in the i-th period, is the average power generation of the jth photovoltaic power station, is the average power generation of the kth PV power station, and n is the total number of time periods.

[0122] According to the similar characteristics of the output power curves of photovoltaic power stations with close geographical locations, the positional relationship of two photovoltaic power stations can be quantitatively evaluated by calculating the correlation of the output power of the photovoltaic systems. The high correlation of the output power between different photovoltaic power stations indicates that the geographical locations of the installations are close. It is believed that the climate environments between the two are similar and can be used as horizontal references to each other, so as to eliminate the interference of continuous faults and temperature on abnormal detection in a certain power station and perform abnormal detection as the same photovoltaic system.

[0123] Reference Figure 5 As shown in the figure, the health status of sample data of photovoltaic power stations with similar environments is judged by the isolation forest algorithm, including:

[0124] S101: randomly selecting at least one sample point as a root node from sample data of a photovoltaic power station group with similar environment;

[0125] S102: randomly determine a dimension from all data dimensions, and randomly select a split point between the maximum value and the minimum value of the data in the specified dimension as a split threshold;

[0126] S103: Divide the current data space into two subspaces according to the segmentation threshold, put the data in the specified dimension that is greater than the segmentation threshold into the right subspace, and put the data that is less than the segmentation threshold into the left subspace;

[0127] S104: Determine whether the subspace contains only one data. If so, traverse each isolated tree for each data, record the number of layers of each isolated tree where the data falls, and calculate the average height of the data in the isolated forest. If not, return to step S102;

[0128] S105: Calculate the standard average height value using the average search length formula;

[0129] S106: Substituting the average height value and the standard average height value into a scoring formula to obtain a score for the data outlier;

[0130] S107: Determine whether the data abnormal value score is higher than the set abnormal score threshold, if so, output the data as abnormal sample data, if not, output the data as healthy sample data;

[0131] The average search length formula is:

[0132]

[0133] Where n is the total number of data, c is the standard average search length, and ξ is the Euler constant, which is approximately 0.5772;

[0134] The scoring formula is:

[0135]

[0136] In the formula, s(i,n) is the outlier score of the i-th data. is the average height value of the i-th data.

[0137] Anomalies refer to data points that are small in number and different. Anomalies are easily affected by a mechanism called isolation. The isolation forest algorithm can use subsampling methods to achieve lower linear time complexity and smaller memory requirements, as well as effectively handle false positives of normal samples and the influence of the characteristics of the data itself being easily masked.

[0138] Reference Figure 6As shown, the correlation between each possible cause and the data to be measured is calculated, and the possible cause with the highest correlation is output. The specific reasons for the abnormality of the photovoltaic power station include:

[0139] Obtain the photovoltaic power station data when each possible cause actually occurs;

[0140] The data to be tested and the data of the photovoltaic power station when the possible inducement actually occurs are respectively formed into a first time series and a second time series;

[0141] According to the correlation coefficient formula, the correlation coefficient between the first time series and the second time series is obtained;

[0142] Substitute the correlation coefficient into the correlation formula to obtain the correlation between each possible cause and the data to be tested;

[0143] The most likely cause of the output correlation is the abnormality of the PV power station;

[0144] The correlation coefficient formula is:

[0145]

[0146] In the formula, t i (k) is the correlation coefficient between the data to be tested and the photovoltaic power station data when the i-th possible cause actually occurs, Δ i k is the absolute value of the difference between the data to be tested and the kth data of the photovoltaic power station when the i-th possible cause actually occurs;

[0147] The correlation formula is:

[0148]

[0149] In the formula, r i is the correlation between the ith possible cause and the data to be tested, and n is the length of the time series.

[0150] Photovoltaic power generation anomalies are affected by many factors, which are mainly manifested in the process of energy conduction and photoelectric conversion. Listing various factors separately and collecting data changes of photovoltaic power stations when various factors occur for comparison can accurately discover the root causes of photovoltaic power generation anomalies.

[0151] Furthermore, the present solution also proposes a computer-readable storage medium on which a computer-readable program is stored. When the computer-readable program is called, the above-mentioned photovoltaic power generation abnormal state detection method is executed.

[0152] It is understandable that the storage medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid state drive (SSD).

[0153] To sum up, the advantages of the present invention are: by setting up a data processing module and an abnormality judgment module, the effectiveness and accuracy of detection are improved, and based on the limited data of the system, intelligent detection of user distributed photovoltaic fault anomalies is achieved, and the root cause of the anomaly can be accurately and timely discovered, which is helpful for the subsequent maintenance and re-commissioning of the photovoltaic power station, and can give full play to the power generation efficiency and help achieve the dual control goals of energy consumption.

[0154] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for detecting abnormal state of photovoltaic power generation, characterized in that: include: Obtain the data of the photovoltaic power station to be tested and the relevant solar irradiance data; Performing data preprocessing on the photovoltaic power station data to be tested to obtain data to be tested; A clear sky day screening mechanism is developed based on the solar irradiance characteristics of clear days, and the relevant solar irradiance data are screened and processed to obtain clear sky dates; Conduct correlation analysis on different power stations to obtain photovoltaic power stations with similar data characteristics to be tested, and conduct horizontal comparison to screen out photovoltaic power station groups with similar environments; The health status of sample data of photovoltaic power stations with similar environments is determined by the isolation forest algorithm; The features of the healthy sample data of the photovoltaic power station marked by the isolation forest algorithm are statistically summarized to fit the normal interval of the power generation data of the photovoltaic power station; Determine whether the proportion of data in the normal range in the data to be detected is higher than the set health threshold ratio. If so, output that the photovoltaic power station is normal; if not, output that the photovoltaic power station is in an abnormal state; Obtain at least one possible cause of abnormal photovoltaic power generation status; The correlation between each possible cause and the data to be measured is calculated, and the possible cause with the highest correlation is output as the abnormal cause of the photovoltaic power station.

2. A photovoltaic power generation abnormal state detection method according to claim 1, characterized in that: The performing data preprocessing on the photovoltaic power station data to be tested to obtain the data to be tested specifically includes: Merge different data sources and store them on the same server; Data cleaning is performed on the photovoltaic power station data to be tested by removing irrelevant data; The cleaned data is uniformly transformed into the range of [0, 1] using the linear transformation formula; The linear transformation formula is: In the formula, x i ′ is the value of the unified i-th data, x i is the value of the i-th data before unification, x min , x max are the minimum and maximum values ​​of all data respectively.

3. A photovoltaic power generation abnormal state detection method according to claim 2, characterized in that: The clear sky day screening mechanism is formulated in combination with the solar irradiance characteristics of clear days, and the relevant solar irradiance data is screened and processed to obtain the clear sky date, which specifically includes: Obtain historical solar irradiance data and obtain the historical daily maximum irradiance value; Sort the historical daily maximum irradiance values ​​and set the highest quarter as the clear sky threshold; Determine whether the daily maximum irradiance of the data to be tested is higher than the clear sky threshold. If so, output the solar irradiance data for that day; if not, do not output; Set the nighttime irradiance intensity to 0, and convert the solar irradiance data of the day into a solar irradiance sequence through a normalization formula; Using the solar irradiance sequence, construct the solar irradiance fluctuation sequence per unit period; Determine whether the product of the last segment of data and the previous segment of data in the solar irradiance fluctuation sequence per unit time period is positive. If so, no output is made; if not, the two segments of data are output as the fluctuation range; Determine whether the number of fluctuation intervals contained in the solar irradiance data of the day is greater than 2. If so, set the day as a cloudy day; if not, set the day as a sunny day; The normalization formula is: In the formula, x i,j ′ is the normalized value of the ith data on the jth day, x i,j is the value of the i-th data on the j-th day before normalization, x j,max is the maximum value of the data on the jth day; The solar irradiance fluctuation sequence per unit period is: b i,j =x i,j ′-x i-1,j ′, Where b i,j is the solar irradiance fluctuation value from the i-1th moment to the i-th moment on the jth day.

4. A photovoltaic power generation abnormal state detection method according to claim 3, characterized in that: The above-mentioned analysis of correlation between different power stations to obtain photovoltaic power stations with similar data characteristics to be detected, and the horizontal comparison to screen out photovoltaic power stations with similar environments specifically includes: Obtain the power generation of each power station at each time period; Use the correlation formula to calculate the correlation coefficient between each power station; Determine whether the correlation coefficient is greater than 0.

95. If so, the output is that the environment is similar to a photovoltaic power station group. If not, no output is made. The correlation formula is: In the formula, ρ(P j ,P k ) is the correlation coefficient between the j-th photovoltaic power station and the k-th photovoltaic power station, P i,j is the power generation of the jth photovoltaic power station in the i-th period, P i,k is the power generation of the k-th photovoltaic power station in the i-th period, is the average power generation of the jth photovoltaic power station, is the average power generation of the kth PV power station, and n is the total number of time periods.

5. A photovoltaic power generation abnormal state detection method according to claim 4, characterized in that: The method of judging the health status of sample data of photovoltaic power stations with similar environments by using the isolation forest algorithm specifically includes: S101: randomly selecting at least one sample point as a root node from sample data of a photovoltaic power station group with similar environment; S102: randomly determine a dimension from all data dimensions, and randomly select a split point between the maximum value and the minimum value of the data in the specified dimension as a split threshold; S103: Divide the current data space into two subspaces according to the segmentation threshold, put the data in the specified dimension that is greater than the segmentation threshold into the right subspace, and put the data that is less than the segmentation threshold into the left subspace; S104: Determine whether the subspace contains only one data. If so, traverse each isolated tree for each data, record the number of layers of each isolated tree where the data falls, and calculate the average height of the data in the isolated forest. If not, return to step S102; S105: Calculate the standard average height value using the average search length formula; S106: Substituting the average height value and the standard average height value into a scoring formula to obtain a score for the data outlier; S107: Determine whether the data abnormal value score is higher than the set abnormal score threshold, if so, output the data as abnormal sample data, if not, output the data as healthy sample data; The average search length formula is: Where n is the total number of data, c is the standard average search length, and ξ is the Euler constant, which is approximately 0.5772; The scoring formula is: In the formula, s(i,n) is the outlier score of the i-th data, is the average height value of the i-th data.

6. A photovoltaic power generation abnormal state detection method according to claim 5, characterized in that: The calculation of the correlation between each possible cause and the data to be measured and outputting the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station specifically includes: Obtain the photovoltaic power station data when each possible cause actually occurs; The data to be tested and the data of the photovoltaic power station when the possible inducement actually occurs are respectively formed into a first time series and a second time series; According to the correlation coefficient formula, the correlation coefficient between the first time series and the second time series is obtained; Substitute the correlation coefficient into the correlation formula to obtain the correlation between each possible cause and the data to be tested; The most likely cause of the output correlation is the abnormality of the PV power station; The correlation coefficient formula is: In the formula, t i (k) is the correlation coefficient between the data to be tested and the photovoltaic power station data when the i-th possible cause actually occurs, Δ i k is the absolute value of the difference between the data to be tested and the kth data of the photovoltaic power station when the i-th possible cause actually occurs; The correlation formula is: In the formula, r i is the correlation between the ith possible cause and the data to be tested, and n is the length of the time series.

7. A photovoltaic power generation abnormal state detection system, used to implement the photovoltaic power generation abnormal state detection method according to any one of claims 1 to 6, characterized in that: include: A data processing module, which is used to obtain the data of the photovoltaic power station to be tested and the related solar irradiance data, perform data preprocessing on the data of the photovoltaic power station to be tested, obtain the data to be tested, formulate a clear sky day screening mechanism in combination with the solar irradiance characteristics of a clear day, screen and process the related solar irradiance data to obtain the clear sky date, perform correlation analysis on different power stations to obtain photovoltaic power stations with similar characteristics of the data to be tested, and perform horizontal comparison to screen out photovoltaic power stations with similar environments; The abnormality judgment module is used to judge the health status of sample data of a photovoltaic power station group with similar environment through the isolation forest algorithm, statistically summarize the characteristics of healthy sample data of photovoltaic power stations marked by the isolation forest algorithm, fit the normal range of photovoltaic power station power generation data, judge whether the proportion of data in the normal range in the data to be detected is higher than the set health threshold ratio, obtain at least one possible cause of abnormal photovoltaic power generation, calculate the correlation between each possible cause and the data to be tested, and output the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station.

8. A photovoltaic power generation abnormal state detection system according to claim 7, characterized in that: The data processing module comprises: A data acquisition unit, the data acquisition unit is used to acquire the photovoltaic power station data to be tested and the related solar irradiance data; A data processing unit, the data processing unit is used to perform data preprocessing on the photovoltaic power station data to be tested to obtain data to be tested; A first data screening unit, the first data screening unit is used to formulate a clear sky day screening mechanism based on the solar irradiance characteristics of a clear day, screen and process relevant solar irradiance data, and obtain a clear sky date; The second data screening unit is used to perform correlation analysis on different power stations to obtain photovoltaic power stations with similar data characteristics to be detected, and perform horizontal comparison to screen out photovoltaic power station groups with similar environments.

9. A photovoltaic power generation abnormal state detection system according to claim 8, characterized in that: The abnormality judgment module comprises: A data health judgment unit, the data health judgment unit is used to judge the health status of sample data of a photovoltaic power station group with a similar environment through an isolation forest algorithm; A normal interval fitting unit, the normal interval fitting unit is used to statistically summarize the characteristics of the healthy sample data of the photovoltaic power station marked by the isolation forest algorithm, and fit the normal interval of the power generation data of the photovoltaic power station; An abnormal state judgment unit, the abnormal state judgment unit is used to judge whether the proportion of data in the normal range in the data to be detected is higher than a set health threshold ratio; The abnormal cause judgment unit is used to obtain at least one possible cause of the abnormal state of photovoltaic power generation, calculate the correlation between each possible cause and the data to be measured, and output the possible cause with the highest correlation as the abnormal cause of the photovoltaic power station.

Citation Information

Patent Citations

  • Method and device for judging fault type of photovoltaic module

    CN109271736A

  • Clear sky day judgment method based on photovoltaic power station output characteristic analysis

    CN115189421A

  • Photovoltaic operation anomaly detection method and system based on machine learning

    CN116383743A

Cited By

  • Distributed wind power plant on-grid energy estimation method and system

    CN119651614A

  • Electricity selling strategy optimization method based on market factors and resource configuration optimization

    CN119784175A

  • Abnormality detection method and device for solar panel and medium

    CN121000176A

  • A method, apparatus, and medium for anomaly detection of a solar panel

    CN121000176B