Correction module and correction method for abnormal data of photovoltaic power station

Abstract

The invention provides a correction module and a correction method for abnormal data of a photovoltaic power station, the correction module comprises a data acquisition module, a parameter classification module, an anomaly detection module and an anomaly correction module, and the data acquisition module, the parameter classification module, the anomaly detection module and the anomaly correction module are connected in sequence. According to the method, the problems of distortion, accumulated value abnormity and the like caused by jumping, data missing and short-time fluctuation of the power station data can be effectively solved, and the accuracy of electric quantity calculation and the effectiveness and integrity of the data are improved.

Description

Technical Field

[0001] The present application relates to the field of photovoltaic power stations, and in particular to the field of abnormal data point detection and correction in photovoltaic power stations. Background Art

[0002] With the popularity and development of solar-storage charging stations, real-time data of the operation of various equipment in power stations is crucial for energy management, efficiency optimization and fault diagnosis. However, data anomalies or loss are often encountered in power station operation. For example, the metering value of the meter or battery may suddenly change in a short period of time, network interruption or equipment replacement may cause data interruption, and data fluctuation may cause distortion. These problems not only affect the accuracy of power calculation, but also lead to inconsistencies in power station operation, which in turn affects the reliability of the power station alarm system.

[0003] Existing technologies are usually unable to effectively deal with the complex situations of various types of data anomalies in photovoltaic power plants, especially failing to consider the characteristics of different types of data and their corresponding detection and correction methods.

[0004] In view of this, this application is proposed. Summary of the invention

[0005] The present application provides a correction module and correction method for abnormal data points in a photovoltaic power station. The module adopts corresponding detection methods based on abnormal data points of different parameter types, can efficiently and accurately identify abnormal data points, and adopts a time offset-based patching method to correct the abnormal data points, which can significantly improve the integrity and accuracy of the data.

[0006] The present application provides a correction module for abnormal data points of a photovoltaic power station, comprising:

[0007] A data acquisition module, used to collect operating data of the photovoltaic power station; the operating data includes one or more of grid load forecast data, multi-parameter monitoring data of photovoltaic power generation equipment, battery voltage data, daytime photovoltaic power generation, photovoltaic power generation data or load current data;

[0008] A parameter classification module, connected to the data acquisition module, for classifying the operation data;

[0009] an anomaly detection module, connected to the parameter classification module, for detecting abnormal data points using different anomaly detection algorithms for different categories of the operating data, and marking the abnormal data points;

[0010] An abnormality correction module is connected to the abnormality detection module and is used to correct the abnormal data points.

[0011] Furthermore, a completion module is included to complete the missing data.

[0012] On the other hand, the present application also provides a method for correcting abnormal data points based on the above correction module, comprising the following steps:

[0013] Step 1: Collect the operation data of the photovoltaic power station;

[0014] Step 2: classifying the operation data;

[0015] Step 3: using different abnormal data point detection methods for different types of operating data to detect and mark the abnormal data points;

[0016] Step 4: Correct the abnormal data points.

[0017] Furthermore, in step 3, using different abnormal data point detection methods for different types of operating data to detect the abnormal data points includes:

[0018] For the power grid load forecast data, an automatic encoder is used to detect abnormal data points;

[0019] For the multi-parameter monitoring data of the photovoltaic power generation equipment, an isolation forest algorithm is used to detect abnormal data points;

[0020] For the battery voltage data, a box plot method is used to detect abnormal data points;

[0021] For the daytime photovoltaic power generation, an autoregressive integrated moving average model is used to detect abnormal data points;

[0022] For photovoltaic power generation data or load current data, the Z-Score method is used to detect abnormal data points.

[0023] Furthermore, the multi-parameter monitoring data of the photovoltaic power generation equipment includes one or more of voltage, current, power and temperature.

[0024] Furthermore, the correction of the abnormal data points in step 4 includes correcting the abnormal data points using a repair method based on time offset.

[0025] Furthermore, the time offset-based repair method includes:

[0026] Step A: Calculate the compensation value B of the Nth abnormal data point using formula (1): N :

[0027] B N =B N-1 +T N (1)

[0028] Among them, BN is the compensation value of the Nth abnormal data point, T N is the jump value of the Nth abnormal data point, N is the number of the abnormal data point, N is a positive integer greater than or equal to 0, X0 represents the origin of the data, and the value of T0 is 0;

[0029] Step B: The corrected value X of the data points between the M-1th abnormal data point and the Mth abnormal data point after compensation N Calculated by formula (2):

[0030] X N =B N +Y N (2)

[0031] Among them, Y N is the original value of the Nth abnormal data point.

[0032] Furthermore, it also includes step 5, a completion step: completing the missing data.

[0033] Furthermore, the completion step in step 5: completing the missing data includes:

[0034] For the power grid load prediction data, a generative adversarial network is used to complete the missing data;

[0035] For the multi-parameter monitoring data of the photovoltaic power generation equipment, K-nearest neighbor interpolation is used to complete the missing data.

[0036] For the battery voltage data, a sliding interpolation method is used to complete the missing data;

[0037] For the daytime photovoltaic power generation, a long short-term memory network is used to complete the missing data;

[0038] For photovoltaic power generation data or load current data, a mean filling method or a median filling method is used to fill in the missing data.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] The present invention can effectively solve the problems of power station data jump, data loss, distortion caused by short-term fluctuations, abnormal accumulated values, etc., and improve the accuracy of power calculation and the validity and integrity of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 Schematic diagram of the structure of the correction module of this embodiment. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Figure 1 FIG. 4 shows a schematic diagram of the structure of the abnormal data correction module of this embodiment. Figure 1 As shown, the correction module 10 includes a data acquisition module (not shown in the figure), a parameter classification module 100, an anomaly detection module 200, an anomaly correction module 300 and a completion module 400, and the data acquisition module, the parameter classification module 100, the anomaly detection module 200, the anomaly correction module 300 and the completion module 400 are connected in sequence.

[0045] The data acquisition module is used to collect the operation data of the photovoltaic power station; the operation data includes one or more of grid load forecast data, multi-parameter monitoring data of photovoltaic power generation equipment, battery voltage data and daytime photovoltaic power generation.

[0046] The parameter classification module 100 is used to classify the operation data, such as Figure 1As shown, this embodiment divides the operation data into five categories, namely A1, B1, C1, D1 and E1. For the convenience of illustration, this embodiment further limits the above five types of operation data to A1 as grid load prediction data, B1 as voltage, current, power, temperature and other multi-parameter monitoring data of photovoltaic power generation equipment, C1 as battery voltage data, D1 as daytime photovoltaic power generation data and E1 as photovoltaic power generation or load current of photovoltaic power station. In this regard, the above five types of data each have their own characteristics, such as A1 presents a complex nonlinear data relationship, B1 presents a large number of parameters and interrelationships, C1 presents an asymmetric distribution, D1 is a time series data with obvious trend and E1 presents a normal distribution. However, in the prior art, an abnormal data detection method is usually used to detect all types of data. Obviously, the prior art fails to consider the characteristics of different data types. For this reason, this embodiment proposes to adopt a suitable abnormal data detection method according to the characteristics of different types of operation data, thereby significantly improving the accuracy and detection efficiency.

[0047] The anomaly detection module 200 of this embodiment uses different anomaly detection algorithms for the five types of operation data mentioned above:

[0048] For the power grid load prediction data A1, the autoencoder algorithm A2 is used to detect abnormal data points to obtain the first abnormal data point set and mark them.

[0049] For the multi-parameter monitoring data B1 of the photovoltaic power generation equipment, the isolation forest algorithm B2 is used to detect abnormal data points, obtain the second normal data point set, and mark it.

[0050] For the battery voltage data C1, the box plot method C2 is used to detect abnormal data points, obtain the third abnormal data, and mark it.

[0051] For the daytime photovoltaic power generation D1, the autoregressive integrated moving average model D2 is used to detect abnormal data points to obtain a fourth abnormal data point set and mark it.

[0052] For the photovoltaic power generation data or the load current data E1, the Z-Score method E2 is used to detect abnormal data points, and the fifth normal data point set is obtained and marked.

[0053] The abnormal correction module 300 is used to correct the value of the abnormal data point. The abnormal correction module 300 of this embodiment uses a repair method based on time offset to correct the value of the abnormal data point.

[0054] The patching method based on time offset is specifically as follows:

[0055] Step A: Calculate the compensation value B of the Nth abnormal data point using formula (1): N :

[0056] B N =B N-1 +T N (1)

[0057] Among them, B N is the compensation value of the Nth abnormal data point, T N is the jump value of the Nth abnormal data point, N is the number of the abnormal data point, N is a positive integer greater than or equal to 0, X0 represents the origin of the data, and the value of T0 is 0;

[0058] Step B: For the Mth abnormal data point, use formula (2) to compensate and obtain the correction value

[0059] X N =B N +Y N (2)

[0060] Among them, Y N is the original value of the Nth abnormal data point.

[0061] For example, the jump abnormality is detected for the first time at 24-09-13 08:00:00:001, and the jump value is 50, then the correction value is 0+50=50.

[0062] The second jump abnormality was detected at 23-09-13 12:00:00:151, and the jump value was -30, so the correction value was 50+(-30)=20.

[0063] The jump abnormality was detected for the third time at 23-09-13 16:23:00:458, and the jump value was -80, so the correction value was 20+(-80)=-60.

[0064] Then, before 23-09-13 08:00:00:001, the revised value is 0, the original value is 200, and the revised value is 200.

[0065] Between 23-09-13 08:00:00:001 and 24-09-13 12:00:00:151, the correction value is 50, the original value is 200, and the corrected value is 250.

[0066] Between 23-09-13 12:00:00:151 and 24-09-13 16:23:00:458, the correction value is 20, the original value is 200, and the corrected value is 220.

[0067] After 23-09-13 16:23:00:458, the correction value is -80, the original value is 200, and the corrected value is 120.

[0068] This implementation corrects the abnormal data points in the first normal data point set, the second normal data point set, the third normal data point set, the fourth normal data point set and the fifth normal data point set through a time offset-based patching method to obtain the first corrected set A3, the second corrected value B3, the third corrected value C3, the fourth corrected value D3 and the fifth corrected value E3 respectively.

[0069] The completion module 400 is used to complete missing data. Similarly, in the prior art, a completion algorithm is used to complete missing data of all types of data. Obviously, it does not take into account that different data types have different characteristics. In this regard, the completion module 400 of this embodiment uses different completion algorithms for different types of data, which can effectively fill in the missing data and ensure data continuity.

[0070] For the above five types of operation data, the completion module 400 of this implementation adopts five corresponding completion algorithms:

[0071] For the first revised set A3 of grid load forecast data, a generative adversarial network A4 is used to complete the missing data;

[0072] For the second revised set B3 of multi-parameter monitoring data of photovoltaic power generation equipment, K-nearest neighbor interpolation B4 is used to complete the missing data.

[0073] For the third correction set C3 of battery voltage data, the sliding interpolation method C4 is used to complete the missing data;

[0074] For the fourth correction set D3 of daytime photovoltaic power generation, the long short-term memory network D4 is used to complete the missing data;

[0075] For the fifth revised set E3 of photovoltaic power generation data or load current data, mean filling E4 is used to complete the missing data.

[0076] Optionally, this embodiment further includes a manual parameter setting unit 2001 connected to the anomaly detection module 200 and configured to manually adjust various parameters in the anomaly detection module 200 .

[0077] Optionally, the present embodiment further includes a power plant strategy unit 400, which is connected to the completion module 400 and is used to provide a trend reference of the operating data to the completion module 400, so that the completion data can effectively avoid the occurrence of data distortion.

[0078] On the other hand, this embodiment also provides a correction method using the above correction module, comprising the following steps:

[0079] Step 1: Collect the operation data of the photovoltaic power station;

[0080] Step 2: Classify the operation data;

[0081] Step 3: Using different abnormal data point detection methods for different types of operating data to detect abnormal data points and mark them;

[0082] Step 4: Correct abnormal data points.

[0083] Optionally, in step 3, using different abnormal data point detection methods to detect abnormal data points for different types of operating data includes:

[0084] For grid load forecast data, an autoencoder is used to detect abnormal data points;

[0085] For the multi-parameter monitoring data of photovoltaic power generation equipment, the isolation forest algorithm is used to detect abnormal data points;

[0086] For battery voltage data, the box plot method is used to detect abnormal data points;

[0087] For the daytime photovoltaic power generation, the autoregressive integrated moving average model is used to detect abnormal data points;

[0088] For photovoltaic power generation data or load current data, the Z-Score method is used to detect abnormal data points.

[0089] Optionally, the multi-parameter monitoring data of the photovoltaic power generation equipment includes one or more of voltage, current, power and temperature.

[0090] Optionally, correcting the abnormal data points in step 4 includes correcting the abnormal data points using a repair method based on time offset.

[0091] Optional time-shift based patching methods include:

[0092] Step A: Calculate the compensation value B of the Nth abnormal data point using formula (1): N :

[0093] B N =B N-1 +T N (1)

[0094] Among them, B N is the compensation value of the Nth abnormal data point, T N is the jump value of the Nth abnormal data point, N is the number of the abnormal data point, N is a positive integer greater than or equal to 0, X0 represents the origin of the data, and the value of T0 is 0;

[0095] Step B: The corrected value X of the data points between the M-1th abnormal data point and the Mth abnormal data point after compensation NCalculated by formula (2):

[0096] X N =B N +Y N (2)

[0097] Among them, Y N is the original value of the Nth abnormal data point.

[0098] Optionally, a completion module is also included to complete missing data, including:

[0099] For power grid load forecasting data, a generative adversarial network is used to complete missing data;

[0100] For the multi-parameter monitoring data of photovoltaic power generation equipment, K-nearest neighbor interpolation is used to complete the missing data.

[0101] For battery voltage data, the sliding interpolation method is used to complete the missing data;

[0102] For the daytime photovoltaic power generation, the long short-term memory network is used to complete the missing data;

[0103] For photovoltaic power generation data or load current data, the mean filling method or median filling method is used to fill in the missing data.

[0104] The "equal" or "same" or "equal" disclosed in the present invention must take into account the distribution of engineering parameters, and the error distribution is within ±30%; the definition of "parallel" of two line segments or two straight lines is that the angle between the two line segments or two straight lines is less than or equal to 45 degrees; the definition of "perpendicular" of two line segments or two straight lines is that the angle between the two line segments or two straight lines is within the range of [60, 120] degrees; the definition of "phase mismatch" also needs to take into account the distribution of engineering parameters, and the error distribution of the degree of mismatch is within ±30%. In addition, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more constraints, an element defined by the phrase "comprising a ..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0105] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0106] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A correction module for abnormal data of a photovoltaic power station, characterized in that: include: A data acquisition module, used to collect operating data of the photovoltaic power station; the operating data includes one or more of grid load forecast data, multi-parameter monitoring data of photovoltaic power generation equipment, battery voltage data, daytime photovoltaic power generation, photovoltaic power generation data or load current data; A parameter classification module, connected to the data acquisition module, for classifying the operation data; an anomaly detection module, connected to the parameter classification module, for detecting abnormal data points using different anomaly detection algorithms for different categories of the operating data, and marking the abnormal data points; An abnormality correction module is connected to the abnormality detection module and is used to correct the abnormal data points.

2. The correction module according to claim 1, characterized in that It also includes a completion module to complete missing data.

3. A method for correcting abnormal data of a photovoltaic power station using the correction module according to any one of claims 1 to 2, characterized in that: The following steps are involved: Step 1: Collect the operation data of the photovoltaic power station; Step 2: classifying the operation data; Step 3: using different abnormal data point detection methods for different types of operating data to detect and mark the abnormal data points; Step 4: Correct the abnormal data.

4. The correction method according to claim 3, characterized in that: In step 3, different abnormal data point detection methods are used for different types of operation data to detect the abnormal data points, including: For the power grid load forecast data, an automatic encoder is used to detect abnormal data points; For the multi-parameter monitoring data of the photovoltaic power generation equipment, an isolation forest algorithm is used to detect abnormal data points; For the battery voltage data, a box plot method is used to detect abnormal data points; For the daytime photovoltaic power generation, an autoregressive integrated moving average model is used to detect abnormal data points; For photovoltaic power generation data or load current data, the Z-Score method is used to detect abnormal data points.

5. The correction method according to claim 4, characterized in that: The multi-parameter monitoring data of the photovoltaic power generation equipment includes one or more of voltage, current, power and temperature.

6. The correction method according to claim 3, characterized in that: The correction of the abnormal data points in step 4 includes correcting the abnormal data points using a repair method based on time offset.

7. The correction method according to claim 6, characterized in that: The time offset-based repair method includes: Step A: Calculate the compensation value B of the Nth abnormal data point using formula (1): N : B N =B N-1 +T N (1) Among them, B N is the compensation value of the Nth abnormal data point, T N is the jump value of the Nth abnormal data point, N is the number of the abnormal data point, N is a positive integer greater than or equal to 0, X0 represents the origin of the data, and the value of T0 is 0; Step B: The corrected value X of the data points between the M-1th abnormal data point and the Mth abnormal data point after compensation N Calculated by formula (2): X N =B N +Y N (2) Among them, Y N is the original value of the Nth abnormal data point.

8. The correction method according to claim 3, characterized in that: It also includes step 5, the completion step: completing the missing data.

9. The correction method according to claim 8, characterized in that: The completion step in step 5: completing the missing data includes: For the power grid load prediction data, a generative adversarial network is used to complete the missing data; For the monitoring data of multiple parameters of the photovoltaic power generation equipment, K-nearest neighbor interpolation is used to complete the missing data; For the battery voltage data, a sliding interpolation method is used to complete the missing data; For the daytime photovoltaic power generation, a long short-term memory network is used to complete the missing data; For photovoltaic power generation data or load current data, a mean filling method or a median filling method is used to fill in the missing data.

Images