Photovoltaic power station fault intelligent diagnosis method based on cloud platform

Abstract

The invention discloses an intelligent fault diagnosis method for a photovoltaic power station based on a cloud platform, relates to the technical field of photovoltaic power stations, and solves the problems that a unified'fault-characteristic curve 'mapping system is lacked, and the judgment difference of different operation and maintenance personnel on the same data is large. According to the method, the power generation capacity change characteristic curve is compared with the preset fault curve, the superposition proportion is quantified, multiple fault scenes are covered, missed judgment is avoided, the judgment standard is unified, and subjective experience is replaced; three types of core parameters of current, voltage and power are correlated, a standard trend section is extracted, a cross ratio mean value of a curve interval and a suspected signal is calculated, the signal with the highest correlation degree is locked, non-component interference is eliminated, and the diagnosis reliability is improved; flow output is clear and landing results, complex algorithm understanding is not needed, and the operation and maintenance threshold is lowered; rapid diagnosis depends on cloud platform data, response time is shortened, and high-efficiency operation and maintenance of a large-scale power station are adapted.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power stations, in particular to a photovoltaic power station fault intelligent diagnosis method based on a cloud platform. BACKGROUND

[0002] With the large-scale development of the global photovoltaic industry, photovoltaic power stations have the characteristics of expanded installed capacity, dispersed distribution, and a sharp increase in the number of components, which puts higher requirements on the real-time, accuracy, and operation and maintenance efficiency of fault diagnosis.

[0003] However, the current fault diagnosis technology of photovoltaic power stations still has significant limitations and is difficult to adapt to the fine management needs of large-scale power stations: there are obvious shortcomings in the suspected fault signal generation and screening link: first, the fault judgment standard is dependent on manual experience, there is a lack of a unified "fault-feature curve" mapping system, different operation and maintenance personnel have large differences in judging the same data, and it is difficult to cover component hidden cracks, PID effects and other hidden fault scenarios, which is prone to fault omission; second, when multiple suspected signals coexist, there is a lack of quantitative correlation evaluation methods, and only relying on subjective selection of " seemingly related " signals often misjudges non-component fault signals such as power grid voltage fluctuations and temporary sensor drift as core faults, leading to deviation of subsequent operation and maintenance direction.

[0004] In addition, the traditional diagnosis takes "manual on-site inspection" as the core, which requires operation and maintenance personnel to check components, inverters and other equipment one by one, which not only has a long response time (days) and high operation and maintenance cost in large-scale power stations, but also is difficult to quickly locate hidden faults in dispersed power stations, and cannot meet the needs of "real-time monitoring and efficient diagnosis" under cloud platform management.

[0005] In summary, the deficiencies of the prior art in abnormality recognition accuracy, fault scenario coverage, signal screening reliability, and operation and maintenance efficiency have become a key bottleneck restricting the stable operation and cost control of photovoltaic power stations, and an intelligent diagnosis method that combines the data advantages of the cloud platform and realizes "environmental interference stripping-standardized signal generation-multi-parameter quantitative screening" is urgently needed. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a photovoltaic power station fault intelligent diagnosis method based on a cloud platform, which solves the problem of a lack of a unified "fault-feature curve" mapping system, different operation and maintenance personnel having large differences in judging the same data, and difficulty in covering component hidden cracks, PID effects and other hidden fault scenarios, which is prone to fault omission.

[0007] To achieve the above purpose, the present application realizes the following technical scheme: a photovoltaic power station fault intelligent diagnosis method based on a cloud platform, comprising the following steps: Step one, confirm the current time of the light intensity and the associated power generation characteristics, and based on historical data characteristics, confirm whether the current time of photovoltaic module exists power generation abnormal situation, the specific way is: The current time associated with the light intensity is recorded as GQ, and the current time associated with the power generation characteristics is recorded as FD. The single-point intensity DQ is confirmed by using FD ÷ GQ = DQ; And take the current time as the reference time, confirm a group of traceable period, its traceable period is a preset period, record the light intensity with the same light intensity as GQ in the traceable period as the undetermined light intensity, and record the power generation data associated with the undetermined light intensity as the undetermined power generation value. Based on the confirmed undetermined light intensity and the undetermined power generation value, the undetermined characteristics are confirmed by using: undetermined power generation value ÷ undetermined light intensity = undetermined characteristics; From the confirmed several undetermined characteristics, lock the standard interval: from the minimum value Dd min Start, confirm the variable range, the range value of the variable range is Y1, Y1 is a preset value, and confirm the maximum value Dd max From the several undetermined characteristics, make the variable range change from [Dd min -Y1, Dd min +Y1] to [Dd max -Y1, Dd max ] and record the total number of undetermined characteristics included in each different variable range. From the recorded total number of undetermined characteristics, select the maximum value, and record the variable range associated with the maximum value as the standard range: If there is only one group of standard range, it is recorded as standard interval; If there are multiple groups of standard range, select the standard range with the initial value of the minimum value state as the standard range, and record the selected standard range as the standard interval; Identify whether the single-point intensity DQ confirmed at the current time is lower than the standard interval. If not, the subsequent monitoring process can be normally executed, if yes, directly generate the power generation abnormal signal for display; Step two, for the case that the photovoltaic module exists power generation abnormal situation, confirm the monitoring period, record the change characteristics between adjacent time points associated with the monitoring period, then generate the corresponding characteristic curve based on the change characteristics, and compare and verify the generated characteristic curve with the preset fault curve. According to the verification process, confirm the suspected signal, the specific way is: For the confirmed power generation abnormal signal, a group of monitoring periods is confirmed, the power generation of the photovoltaic module in the monitoring period is confirmed, and the power generation data of different time confirmed is sorted according to the time sequence, and the power generation change characteristics between adjacent time is confirmed, the change characteristics is the difference between the power generation of the next time and the power generation of the previous time in the adjacent time, and the feature curve of the power generation change characteristics in the monitoring period is generated according to the time sequence; The confirmed feature curve is compared with the preset fault curve: the feature curve and the fault curve are placed in the same two-dimensional coordinate system, and the feature curve is controlled to move forward and backward, in the moving process, the overlapping part of the feature curve and the preset fault curve is identified, and the related proportion of the overlapping part segment in the feature curve is recorded, the related proportion = the total length of the overlapping part segment ÷ the total length of the feature curve, in the moving process, the maximum value associated with several related proportions is confirmed, the confirmed maximum value is recorded as the calibrated feature of the current feature curve, if the calibrated feature ≥ Y2, the fault signal associated with the preset fault curve is recorded as the suspected signal, Y2 is a preset value, if the calibrated feature < Y2, no calibration is performed; The feature curve and the preset fault curve are compared in turn, according to the comparison process, the suspected signal associated with the feature curve is confirmed, and the confirmed suspected signal is displayed; Step three, identify the number of suspected signals, if there is only one group, directly display, if there are multiple groups, extract the change characteristics of other data characteristics in the monitoring period, confirm the suspected signal with the highest correlation degree of the photovoltaic module from the change characteristics, and display; Preferably, in step three, the specific way of confirming the suspected signal with the highest correlation degree is: The current photovoltaic module associated current, voltage and power data in the monitoring period are recorded, and the current change curve, voltage change curve and power change curve associated with the corresponding monitoring period are confirmed according to the time sequence; Confirming the standard trend segment existing in the interior from the confirmed different change curves: from the change curve, confirming the change characteristics between adjacent time points, and confirming the difference value between the change characteristics between adjacent time periods, the difference value is the interval value between the change characteristics of adjacent time periods, and confirming the characteristics to which the corresponding change curve belongs, confirming the characteristic threshold value preset by the characteristics to which the characteristics belong, if the interval value is less than or equal to the characteristic threshold value, the curve segment associated with the adjacent time period is marked as the same type of characteristic segment, and the same type of characteristic segment associated with the adjacent time period in the change curve is confirmed in turn, and the longest length value of the same type of characteristic segment is selected as the standard trend segment of the change curve, and the standard trend segment is recorded, the change characteristics associated with the standard trend segment are confirmed, and the maximum value and the minimum value are confirmed as the curve interval associated with the current change curve, if the interval value is greater than the characteristic threshold value, no marking is performed; And the curve intervals associated with the change curves of other characteristics are confirmed in turn; The curve interval confirmed by the characteristics is compared and verified with the numerical interval preset by the suspected signal, the intersection range is confirmed, and the range proportion value of the intersection range in the numerical interval is recorded, and the three range proportion values associated with the three curve intervals are confirmed, and the three range proportion values are processed by mean value, and the suspected characteristics associated with the current suspected signal are confirmed; Based on different suspected characteristics associated with different suspected signals, the maximum value is selected, the suspected signal associated with the maximum value is recorded as the suspected signal with the highest correlation degree, and is displayed.

[0008] The application provides a photovoltaic power station fault intelligent diagnosis method based on a cloud platform. The application calculates single-point intensity by "power generation / light intensity", removes light fluctuation interference, constructs a standard interval combined with historical similar light intensity data, accurately judges real power generation abnormalities of components, and reduces environmental factor misjudgment; The power generation change characteristic curve is compared with the preset fault curve, the overlap proportion is quantified, multiple fault scenarios are covered, misjudgment is avoided, the determination standard is unified, and subjective experience is replaced; The three types of core parameters of current, voltage and power are associated, the standard trend segment is extracted, the intersection proportion mean value of the curve interval and the suspected signal is calculated, the signal with the highest correlation degree is locked, non-component interference is excluded, and the diagnosis reliability is improved; The process outputs clear and practical results, does not need to understand complex algorithms, reduces the operation and maintenance threshold, relies on cloud platform data for rapid diagnosis, shortens the response time, and adapts to large-scale power station efficient operation and maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 The application provides a method flowchart. DETAILED DESCRIPTION

[0010] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of the present application.

[0011] Please refer to Figure 1 The present application provides a cloud platform-based intelligent fault diagnosis method for a photovoltaic power station, comprising the following steps: Step one, confirm the light intensity at the current time and the associated power generation characteristics, and based on the historical data characteristics, confirm whether the photovoltaic module at the current time has a power generation anomaly, specifically, under the corresponding light intensity, there is a corresponding power generation, such data can be directly collected, and based on the collected data, the power generation data analysis process is carried out, and whether the power generation data has a numerical anomaly is identified from the historical processing process, and the specific identification result is displayed for the designated personnel to check; Among them, the specific way of whether the photovoltaic module at the current time has a power generation anomaly is: The light intensity associated with the current time is denoted as GQ, and the power generation characteristics associated with the current time are denoted as FD, and the single-point intensity DQ is confirmed by using: FD ÷ GQ = DQ; And taking the current time as the reference time, a group of traceable periods is confirmed, and the traceable period is a preset period, which is determined in advance by the operator according to experience, and the current time is 0 time, and the traceable period is 1h in the past, so the corresponding traceable period is 23h-0h, the light intensity same as the light intensity GQ in the traceable period is denoted as the to-be-determined light intensity, and the power generation data associated with the to-be-determined light intensity is denoted as the to-be-determined power generation value, based on the to-be-determined light intensity and the to-be-determined power generation value, the to-be-determined characteristics are confirmed by using: to-be-determined power generation value ÷ to-be-determined light intensity = to-be-determined characteristics; From the confirmed several to-be-determined characteristics, the standard interval is locked: from the minimum value Dd min of the several to-be-determined characteristics, the variable range is confirmed, the range value of the variable range is Y1, Y1 is a preset value, and the specific value is determined by the operator according to experience, and the maximum value Dd max of the several to-be-determined characteristics is confirmed, and the variable range is changed from [Dd min , Dd min + Y1] to [Dd max - Y1, Dd maxThe range is changed, and the total number of pending features included in each different variable range is recorded. From the recorded total number of pending features, the maximum value is selected, and the variable range associated with the maximum value is recorded as the standard range. If there is only one group of standard ranges, it is recorded as a standard interval. If there are multiple groups of standard ranges, select the initial value in the standard range as the minimum value state, and record the selected standard range as the standard interval. For example: different standard ranges have different initial values. There are corresponding minimum values between multiple initial values. The corresponding standard range associated with the minimum value is the corresponding standard interval. Identify whether the single-point intensity DQ confirmed at the current time is lower than the standard interval. If not, it means that the photovoltaic module is generating electricity normally at the current time, and the subsequent monitoring process can be normally executed. If yes, it means that the photovoltaic module is generating electricity abnormally at the current time, and an electricity generation abnormality signal is directly generated for display. Step two, for the case that the photovoltaic module has electricity generation abnormality, confirm the monitoring period, record the change characteristics between adjacent time points associated with the monitoring period, and then generate the corresponding characteristic curve based on the change characteristics. Compare and verify the generated characteristic curve with the preset fault curve, and confirm the suspected signal according to the verification process. Specifically, when the specific abnormal reason is not clear, the closest change stage is evaluated according to the change characteristics between the curve values in the corresponding comparison process, so as to confirm the suspected signal. The specific way to confirm the suspected signal is: For the confirmed electricity generation abnormality signal, a group of monitoring periods is confirmed, the electricity generated by the photovoltaic module in the monitoring period is confirmed, and the electricity generation data at different times is sorted according to the time sequence. The change characteristics between adjacent time points are confirmed, the change characteristics are the difference between the electricity generation at the next time point and the electricity generation at the previous time point, and the characteristic curve of the electricity generation change characteristics in the monitoring period is generated according to the time sequence. The horizontal coordinate axis of this curve is the time line, the vertical coordinate axis is the change characteristics, the time line is the time period, the time period in the front is sorted in the front, the time period in the back is sorted in the back, and so on. According to the change relationship between time and electricity generation change characteristics, the corresponding characteristic curve is generated. The identified characteristic curve is compared with the preset fault curve: the characteristic curve and the fault curve are placed in the same two-dimensional coordinate system, and the characteristic curve is controlled to move forward and backward. In the moving process, the overlapping part of the characteristic curve and the preset fault curve is identified, and the relevant proportion of the overlapping part segment on the characteristic curve is recorded. The relevant proportion = the total length of the overlapping part segment ÷ the total length of the characteristic curve. From the moving process, the maximum value associated with several relevant proportions is identified. The identified maximum value is recorded as the calibrated feature of the current characteristic curve. If the calibrated feature ≥ Y2, the fault signal associated with the preset fault curve is recorded as the suspected signal. Otherwise, no calibration is performed. Y2 is a preset value determined by the operator in advance based on experience. The preset fault curve is based on the change characteristic curve generated under different fault states and is determined by the operator in advance based on specific practical experience. The characteristic curve and the preset fault curve are compared in turn. According to the comparison process, the suspected signal associated with the characteristic curve is identified, and the identified suspected signal is displayed. Specifically, in the actual comparison process, the characteristic curve and the preset fault curve are compared one by one. The corresponding characteristic curve and fault curve are placed in the same two-dimensional coordinate system. Then, by moving the corresponding characteristic curve, the overlapping part between the characteristic curve and the fault curve is identified. According to the specific proportion of the identified overlapping part, the specific calibration of the calibration feature is performed. From the identified specific process, the suspected signal is identified. Based on the corresponding identification process, the specific accuracy of the corresponding suspected signal identification process can be effectively guaranteed, and the overall judgment effect of the suspected signal can be effectively improved.

[0012] Step three, identify the number of suspected signals. If there is only one group, it is directly displayed. If there are multiple groups, extract the change characteristics of other data characteristics in the monitoring period. From the change characteristics, identify the suspected signal with the highest correlation degree associated with the photovoltaic module and display it. The specific way to identify the suspected signal with the highest correlation degree is as follows: The current current, voltage and power data associated with the photovoltaic module in the monitoring period are recorded, and the current change curve, voltage change curve and power change curve associated with the corresponding monitoring period are identified according to the time sequence. Confirm the standard trend segment existing in the interior from the confirmed different change curves: from the change curve, confirm the change characteristics between adjacent time points, which is consistent with the confirmation method of the power generation change characteristics, and is also confirmed according to the difference between the specific data of the next time point and the data of the previous time point, and the difference value of the change characteristics between adjacent time periods is confirmed, and the interval value between the change characteristics of the adjacent time periods is confirmed, and the corresponding characteristics (current, voltage or power data) of the change curve are confirmed, and the characteristic threshold of the corresponding characteristics is confirmed, and the characteristic threshold of the different corresponding characteristics is different, if the interval value is less than or equal to the characteristic threshold, the curve segment associated with the adjacent time period is marked as the same type of characteristic segment, otherwise, no marking is performed, the same type of characteristic segment associated with the adjacent time period in the change curve is confirmed in turn, and the longest length value of the same type of characteristic segment is selected as the standard trend segment of the change curve, and the standard trend segment is recorded, the change characteristics associated with the standard trend segment are confirmed, and the maximum value and the minimum value are confirmed as the curve interval associated with the current change curve; And the curve intervals associated with the change curves of other corresponding characteristics are confirmed in turn; Compare and verify the curve intervals confirmed by the corresponding characteristics with the preset numerical interval of the suspected signal, confirm the intersection range, record the range proportion value of the intersection range in the numerical interval, and confirm the three range proportion values associated with the three curve intervals, and the three range proportion values are processed by mean value, and the suspected characteristics associated with the current suspected signal are confirmed; Based on different suspected characteristics associated with different suspected signals, the maximum value is selected, the suspected signal associated with the maximum value is recorded as the suspected signal with the highest correlation degree, and is displayed for external personnel to view, so that the suspected signal confirmed at the current time is the specific signal with the highest correlation degree, which is the specific fault condition with a high probability; Specifically, the numerical change process of the corresponding curve is confirmed by confirming the numerical change characteristics associated with the corresponding signal one by one, and then the specific process with the most stable trend is selected from the corresponding change process, and the specific signal is displayed from the selected process, so as to achieve the optimal signal confirmation processing effect and ensure the specific accuracy in the evaluation process.

[0013] Part of the data in the above formula is dimensionless numerical calculation, and the contents not described in detail in the specification belong to the prior art known to those skilled in the art.

[0014] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A cloud-based intelligent fault diagnosis method for photovoltaic power stations, characterized in that: The following steps are involved: Step 1: Confirm the current light intensity and the associated power generation characteristics, and based on the historical data characteristics, confirm whether there is any abnormal power generation of the photovoltaic module at the current moment; Step 2: If there is abnormal power generation of the photovoltaic module, confirm the monitoring period, record the power generation associated with the monitoring period, confirm the change characteristics between adjacent moments, and then generate a corresponding characteristic curve based on the change characteristics. The generated characteristic curve is compared and verified with the preset fault curve. According to the verification process, the suspected signal is confirmed; Step 3: Identify the number of suspected signals. If there is only one group, display it directly. If there are multiple groups, extract the change characteristics of other data features during the monitoring period, identify the suspected signal with the highest degree of correlation with the photovoltaic module from the change characteristics, and display it.

2. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 1, characterized in that: In step 1, the specific method of determining whether there is power generation abnormality of the photovoltaic module at the current moment is: The light intensity associated with the current moment is recorded as GQ, and the power generation characteristic associated with the current moment is recorded as FD. The single point intensity DQ is determined by: FD ÷ GQ = DQ; Taking the current moment as the reference moment, a set of traceability periods is confirmed, where the traceability period is the preset period. The light intensity that is the same as the light intensity GQ within the traceability period is recorded as the pending light intensity, and the power generation data associated with the pending light intensity is recorded as the pending power generation value. Based on the confirmed pending light intensity and pending power generation value, the following formula is used: pending power generation value ÷ pending light intensity = pending feature, to confirm several pending features. From the confirmed features to be determined, lock the standard interval: from the minimum value Dd of several features to be determined min First, confirm the variable range, the range value of the variable range is Y1, where Y1 is the preset value, and confirm its maximum value Dd from several undetermined features max , so that the variable range is from [Dd min , Dd min +Y1] to [Dd max -Y1, Dd max ] to change the range, and record the total number of pending features included in each different variable range. From the total number of pending features recorded, select the maximum value, and record the variable range associated with the maximum value as the standard range. If there is only one set of standard ranges, it is recorded as the standard interval.

3. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 2, characterized in that: If there are multiple groups of standard ranges, the standard range whose initial value within the standard range is the minimum state is selected, and the selected standard range is recorded as the standard interval; Identify whether the single-point intensity DQ confirmed at the current moment is lower than the standard range. If not, execute the subsequent monitoring process normally. If so, directly generate a power generation abnormality signal for display.

4. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 1, characterized in that: In step 2, the specific method of confirming the suspected signal is: For the confirmed abnormal power generation signal, a set of monitoring periods is identified, and the power generation generated by the photovoltaic modules within the monitoring period is confirmed. The power generation data at different times are sorted according to the time sequence, and the power generation change characteristics between adjacent times are simultaneously identified. The change characteristics are the difference between the power generation at the next moment and the power generation at the previous moment within the adjacent moments. Based on the time sequence, a characteristic curve of the power generation change characteristics within this monitoring period is generated; Compare and verify the confirmed characteristic curve with the preset fault curve: place the characteristic curve and the fault curve in the same two-dimensional coordinate system, and control the characteristic curve to move back and forth. During the movement process, identify the overlapping part of the characteristic curve and the preset fault curve, and record the relevant proportion of the overlapping part in the characteristic curve. The relevant proportion = the total length of the overlapping part segment ÷ the total length of the characteristic curve. From the movement process, determine the maximum value associated with several relevant proportions, and record the confirmed maximum value as the calibration feature of the current characteristic curve. If the calibration feature ≥ Y2, then the fault signal associated with the preset fault curve is recorded as a suspected signal, and Y2 is the preset value; The characteristic curve is compared with the preset fault curve in turn. According to the comparison process, the suspected signal associated with the characteristic curve is confirmed, and the confirmed suspected signal is displayed accordingly.

5. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 4, characterized in that: If the calibration characteristic is less than Y2, no calibration is performed.

6. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 1, characterized in that: In step 3, the specific method of confirming the suspected signal with the highest correlation degree is: Record the current, voltage, and power data associated with the current photovoltaic module during the monitoring period, and confirm the current change curve, voltage change curve, and power change curve associated with the corresponding monitoring period based on the time sequence; Identify the standard trend segments existing within the identified different change curves, identify the change characteristics associated with the standard trend segments, and identify the maximum and minimum values ​​therefrom as the curve interval associated with the current change curve; And confirm the curve intervals associated with the corresponding change curves of other characteristics in turn; Compare and verify the curve interval confirmed by the feature with the preset numerical interval of the suspected signal to confirm the intersection range and record the range ratio of the intersection range in the numerical interval. Then confirm the three groups of range ratio values ​​associated with the three groups of curve intervals and average the three confirmed range ratio values ​​to confirm the suspected feature associated with the current suspected signal. Based on the different suspected features associated with different suspected signals, the maximum value is selected, and the suspected signal associated with the maximum value is recorded as the suspected signal with the highest correlation degree and displayed.

7. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 6, characterized in that: The confirmation process of the standard trend segment includes: confirming the change characteristics between adjacent moments from the change curve, and confirming the difference of the change characteristics between adjacent time periods, where the difference is the interval value between the change characteristics of the previous and next adjacent time periods, and confirming the corresponding characteristics of the change curve, confirming the preset characteristic threshold of the corresponding characteristics, if the interval value is ≤ the characteristic threshold, then the curve segments associated with the adjacent time periods are marked as similar characteristic segments, and the similar characteristic segments associated with the adjacent time periods in the change curve are confirmed in turn, and a group of similar characteristic segments with the longest length value are selected as the standard trend segment of this change curve, and this standard trend segment is recorded.

8. The cloud platform-based intelligent fault diagnosis method for photovoltaic power station according to claim 7, characterized in that: If the interval value is greater than the feature threshold, no marking is performed.

Images