Distributed photovoltaic power station equipment diagnosis method based on big data analysis

Abstract

The invention belongs to the technical field of photovoltaic power station equipment diagnosis, and discloses a distributed photovoltaic power station equipment diagnosis method based on big data analysis, and the method comprises the steps: S1, data collection: collecting data transmitted by a sensor on each piece of equipment of a distributed photovoltaic power station, and receiving the operation data of the equipment in real time; s2, data preprocessing: preprocessing the collected data; s3, feature extraction: extracting feature parameters capable of reflecting the running state of the equipment from the preprocessed data through data analysis and data mining; s4, model establishment: constructing an equipment fault diagnosis model based on historical fault data and corresponding characteristic parameters; and S5, fault diagnosis: inputting data acquired in real time into the fault diagnosis model, completing state judgment of each device of the distributed photovoltaic power station, and identifying potential faults. According to the invention, the operation data of the photovoltaic power station is collected, analyzed and processed, so that intelligent diagnosis and maintenance of equipment are realized.

Description

Technical Field

[0001] The present application belongs to the technical field of photovoltaic power station equipment diagnosis, and in particular, relates to a distributed photovoltaic power station equipment diagnosis method based on big data analysis. Background Art

[0002] With the continuous growth of global energy demand and the increasing prominence of environmental problems, the development and utilization of clean energy has become the focus of attention of all countries. Photovoltaic power generation, as an important renewable energy, has been widely used and developed rapidly due to its advantages such as cleanliness, environmental protection and abundant resources. In particular, distributed photovoltaic power stations have gradually become an important part of the photovoltaic power generation field due to their flexible deployment methods and friendliness to the power grid.

[0003] However, the widespread use of distributed photovoltaic power stations has also brought new challenges. Due to the dispersion of equipment, the wide variety of equipment and the complex operating environment, equipment fault diagnosis and maintenance have become particularly important. Traditional photovoltaic power station equipment diagnosis methods mainly rely on regular manual inspections and simple monitoring systems, which are not only inefficient, but also unable to detect and prevent potential equipment failures in a timely manner, resulting in high operation and maintenance costs and affecting the overall benefits of the power station. Summary of the invention

[0004] The purpose of this application is to overcome the problems of the prior art and disclose a distributed photovoltaic power station equipment diagnosis method based on big data analysis. This application collects, analyzes and processes the operating data of the photovoltaic power station to achieve intelligent diagnosis and maintenance of the equipment.

[0005] The purpose of this application is achieved through the following technical solutions: A distributed photovoltaic power station equipment diagnosis method based on big data analysis, the distributed photovoltaic power station equipment diagnosis method comprising: S1: Data collection, by collecting data transmitted by sensors on each device in the distributed photovoltaic power station, receiving the operating data of the device in real time; S2: Data preprocessing, preprocessing the collected data; S3: Feature extraction: extract characteristic parameters that can reflect the operating status of the equipment from the preprocessed data through data analysis and data mining; S4: Model establishment: building equipment fault diagnosis model based on historical fault data and corresponding characteristic parameters; S5: Fault diagnosis: input the real-time collected data into the fault diagnosis model to complete the status judgment of each device in the distributed photovoltaic power station and identify potential faults.

[0006] According to a preferred embodiment, the distributed photovoltaic power station equipment diagnosis method further includes: S6: Fault warning: for the equipment with potential faults identified in step S5, a warning message is issued and maintenance suggestions for the potential faults are provided; The diagnostic results and warning information obtained in step S5 are fed back to the operation and maintenance personnel to assist them in equipment maintenance and fault handling.

[0007] According to a preferred embodiment, the distributed photovoltaic power station equipment diagnosis method further includes: S7: Result feedback, generating a diagnostic report including fault cause analysis, treatment suggestions and subsequent maintenance plan.

[0008] According to a preferred embodiment, the data preprocessing process in step S2 includes: performing data cleaning, data interpolation and data normalization on the collected data; Outliers are removed through data cleaning, and the corresponding sample size is ensured through data interpolation. Then, data normalization is performed to eliminate dimensional differences.

[0009] According to a preferred embodiment, the data cleaning includes: data time stamp cleaning, data missing cleaning, data jump cleaning and data dead value cleaning; The data time stamp cleaning process includes: when checking the received data, presetting the range of the correct time stamp, and then judging the data time stamp according to the current time. If it exceeds the preset range, it is considered to be invalid data; The missing data cleaning process includes: first finding the missing data according to the time scale, and then filling the missing data; The data jump cleaning process includes: completing the jump data judgment by the interval growth judgment method or the preset condition judgment method, and then recovering the jump data by the deduction method or the blanking method or marking the corresponding jump data as an error or blanking; The data dead value cleaning process includes: first, manually presetting the dead value conditions, setting the data range and time interval of the dead value, and then, when it is found that the data does not exceed the specified value range within the specified time, the corresponding data has a dead value, and it is restored through deduction or blanking method, or the corresponding jump data is marked as an error or blanked.

[0010] According to a preferred embodiment, the data interpolation includes: completing the missing data supplementation by mathematical interpolation method, derivation method and data aggregation method; The mathematical interpolation method is to obtain the missing data in the middle by calculating the difference between the adjacent data before and after the loss for the accumulated increasing data. The deductive method is to obtain the missing data by analyzing the data related to the missing data and inferring them; The data aggregation method is to aggregate, calculate, and count the data according to the abstract standards of the business scenario to obtain the missing data.

[0011] According to a preferred embodiment, the feature extraction process in step S3 includes: The principal component analysis or factor analysis method is used to extract the principal components that can reflect the operating status of the equipment from the preprocessed data, and based on expert experience, the characteristic parameters that can reflect the operating status of the equipment are selected.

[0012] According to a preferred embodiment, the process of constructing the equipment fault diagnosis model in step S4 includes: Using machine learning algorithms, we build equipment fault diagnosis models. Based on historical fault data and characteristic parameters, we train and verify the equipment fault diagnosis models. We also use cross-validation methods to update and iterate the parameters of the equipment fault diagnosis models. At the same time, the model performance is evaluated through accuracy, recall rate, and F1 value, and is used for actual diagnosis when the equipment fault diagnosis model meets the preset indicators.

[0013] According to a preferred embodiment, the equipment fault diagnosis model uses a standard string effectiveness algorithm to perform equipment diagnosis. The standard string effectiveness algorithm diagnoses substandard equipment by combining equipment comparison compliance rate analysis; The validity algorithm of the standard string in the equipment fault diagnosis model performs power generation capacity deviation analysis and standard string power generation capacity comparison analysis on the standard string based on the data of the day, and compares the irradiation and power conversion rate of the standard string on the day with the daily average irradiation and power conversion rate of the standard string in history, or judges by the historical average ranking ratio of the standard string to determine whether the corresponding standard string meets the requirements as a standard string. If not, the standard string is marked as invalid. If it meets the requirements, it is marked as valid and the reason for the analysis is given.

[0014] According to a preferred embodiment, during the fault warning process of step S6, fault information processing is performed in a graded manner according to the severity of the fault.

[0015] The aforementioned main scheme of the present application and its further options can be freely combined to form multiple schemes, all of which are schemes that can be adopted and claimed for protection in the present application. After understanding the scheme of the present application, those skilled in the art can understand that there are multiple combinations based on the prior art and common knowledge, all of which are technical schemes to be protected by the present application, and they are not exhaustively listed here.

[0016] Beneficial effects of this application: The present application method can monitor the operating status of the equipment in real time and detect potential faults in time; the present application method adopts big data analysis and machine learning technology to improve the accuracy and timeliness of fault diagnosis, and improve the overall operating efficiency and economic benefits of photovoltaic power stations; through intelligent fault diagnosis and early warning, the operation and maintenance costs of the equipment are reduced; through timely fault handling and maintenance, the service life of the equipment is extended. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a processing flow diagram of the present application method; Figure 2 It is a schematic diagram of the process of data cleaning in an embodiment of the present application; Figure 3 It is a schematic diagram of the diagnostic process of an embodiment of the present application. DETAILED DESCRIPTION

[0018] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0019] refer to Figure 1 As shown, the present application discloses a distributed photovoltaic power station equipment diagnosis method based on big data analysis, and the distributed photovoltaic power station equipment diagnosis method includes the following steps.

[0020] Step S1: Data collection, by collecting data transmitted by sensors on each device of the distributed photovoltaic power station, receiving the operating data of the device in real time. The operating data includes: current, voltage, temperature, power and other parameters.

[0021] Step S2: Data preprocessing, preprocessing the collected data. The data preprocessing process includes: data cleaning, data interpolation and data normalization of the collected data; Through data cleaning, outliers are removed to ensure data quality. Data interpolation is used to ensure that there is a corresponding sample size to reduce the impact of incomplete data. Then, data normalization is performed to eliminate dimensional differences and improve the subsequent model training effect.

[0022] Preferably, the data cleaning includes: data time stamp cleaning, data missing cleaning, data jump cleaning and data dead value cleaning.

[0023] The data timestamp cleaning process includes: when checking the received data, presetting the range of the correct timestamp, and then judging the data timestamp according to the current time. If it exceeds the preset range, it is considered to be invalid data.

[0024] Specifically, there will be time stamp errors during data transmission. The reason for this kind of data is generally because the time on the acquisition device is incorrect. After the acquisition device collects the data, it needs to be marked with a time stamp so that it can be used in various subsequent processing of the system. If the time of the acquisition device is incorrect, then the time stamp marked on the data is also wrong, sometimes it is a past time stamp, and sometimes it is a future time stamp. The general hardware solution to solve this problem is to install a GPS clock on the acquisition device and sync the time of the acquisition device at any time. As long as the time of the acquisition device is correct, the time stamp marked when the data is collected is also correct. For the big data platform system, the time stamp in the data must be checked when the data is received and saved. Generally, it is judged based on its own system time, so first of all, it is necessary to ensure that its own time is correct, which is generally also synchronized through the GPS clock. When checking the data, the system needs to preset the range of the correct time stamp, and then judge the data time stamp based on the current time. If it exceeds the preset range, it is considered invalid data, and the platform refuses to store such data to avoid affecting the subsequent system functions. At the same time, this incorrect time stamp data needs to be saved to a fixed location so that the operation and maintenance personnel can view it at any time and perform corresponding processing.

[0025] The missing data cleaning process includes: first finding the missing data according to the time scale, and then filling the missing data.

[0026] Specifically, data loss may occur due to communication reasons during data collection and transmission. This communication problem can be divided into two situations. If the network between the collection device and the device is interrupted, the data loss caused by this reason cannot be recovered, because in general, the device collects data in a polling manner. When the communication is interrupted, the various data states generated during the operation of the device will not be cached. After the communication is restored, the operation data during this period will no longer be collected. If the network is interrupted during data transmission, the data may not be lost, but only delayed. This is because the gateway and data transmission service in the big data platform have the function of data caching, that is, when the network is interrupted, these services automatically cache the data locally and continuously detect the network. Once the network communication is restored, the cached data will be uploaded immediately. Therefore, in this kind of network interruption, the data is only delayed and will not be truly lost. But in either case, for the big data platform, data is always likely to be truly lost, so it is necessary to consider the cleaning of missing data. Data loss cleaning first needs to find the lost data according to the time stamp, and then fill the lost data according to the method mentioned in data interpolation.

[0027] The data jump cleaning process includes: completing the jump data judgment through the interval growth judgment method or the preset condition judgment method, and then recovering the jump data by the deduction method or the blanking method or marking the corresponding jump data as an error or blanking it.

[0028] Specifically, sometimes during the data receiving process, it is found that the uploaded data has a jump. There are many reasons for this problem, including equipment hardware reasons and acquisition configuration reasons. Equipment hardware reasons are generally caused by equipment maintenance leading to hardware replacement, in which the counters are updated, resulting in some measurement points being cleared, and the data jumping downward after being collected again. The reasons for the acquisition configuration are more complicated. If the address configuration of the acquisition measurement point is incorrect, resulting in the position error of the measurement point, then the data jump will not follow any rules, and may be large and small, and it is an "exaggerated" value that does not conform to the reasonable range of the data. For these jumping data, there is no use value and must be cleaned in time after entering the system. Sometimes it can be turned into a reasonable and valid value and continue to be used by other functional modules in the system; sometimes there is no way to restore it, and it can only be marked as an error or set to zero, so that it is no longer used by other functions, thereby avoiding affecting the overall statistical analysis. The judgment methods for data jumps include interval growth judgment method, preset condition judgment method, etc.

[0029] The interval growth judgment method mainly judges whether the data has jumped based on the reasonable interval change of the data. Because some data, especially the cumulative growth data, has a limited range of change per unit time and will not exceed its own electrical limit, it can be judged whether the data has jumped based on these device characteristics.

[0030] The preset condition judgment method generally requires manual intervention. Different value ranges are set in the system in advance according to the operating characteristics of each measuring point. After the system is running, these preset conditions are loaded and the collected data is monitored in real time. Once the data exceeds this value range, it can be considered that the data has jumped.

[0031] There are generally two methods for cleaning jump data: derivation and blanking. The derivation method will be introduced in detail in the following data interpolation. The blanking method is a situation where the jump data cannot be restored as described above, and it can only be simply blanked. The blanked data does not participate in various subsequent statistics, thereby minimizing its harmfulness.

[0032] The data dead value cleaning process includes: first, manually presetting the dead value conditions, setting the data range and time interval of the dead value, and then, when it is found that the data does not exceed the specified value range within the specified time, the corresponding data has a dead value, and it is restored through deduction or blanking method, or the corresponding jump data is marked as an error or blanked.

[0033] Specifically, during the data transmission process, it is sometimes found that some measuring points stop refreshing. This is because an abnormality occurs in the acquisition channel, resulting in some data changes not being sent up in time, so that the upper layer finds that the data has not changed for a long time, which does not conform to the actual physical characteristics of the measuring point. This kind of data also needs to be cleaned, because if it is not processed, the wrong data will be used for statistics, thereby providing users with wrong and distorted statistical results. The judgment of data dead values ​​first requires manual preset conditions, such as which data are dead value processed, the data range and time interval for judgment, where the data range refers to the interval of data changes up and down, and there are generally two settings, absolute interval and percentage interval. The absolute interval refers to the interval setting of the absolute value, while the percentage interval refers to the interval setting of the data change to the original value; the time interval refers to the length of time to judge whether the data has a dead value. During the operation of the system, if it is found that a certain data does not exceed the specified value range (absolute interval or percentage interval) within the specified time, it is considered that the data has a dead value and needs to be cleaned. The cleaning method is similar to the jump method, and generally there are two methods: deduction method and blanking method. The deduction method mainly analyzes the data related to the jump data and infers the normal value of the jump data. The blanking method cannot restore the data and can only blank it so that it no longer participates in various subsequent statistics.

[0034] Preferably, the data interpolation includes: completing the supplementation of missing data through mathematical interpolation method, deduction method and data aggregation method.

[0035] Preferably, the mathematical interpolation method is to obtain the lost data in the middle by calculating the difference between the adjacent data before and after the loss for the accumulated rising data.

[0036] Specifically, the interpolation method is mainly used for cumulative increasing data. When part of this data is lost, the missing data in the middle can be calculated by taking the difference between the most recent data before and after the loss. When using this interpolation method for statistics, if the statistical interval happens to include the missing interval, or there is a missing interval within the statistical interval, the statistical result will be completely correct without any impact of data loss.

[0037] Preferably, the derivation method is to obtain the missing data by inferring through analysis of data associated with the missing data.

[0038] Specifically, the deduction method is mainly used for data with correlation. By analyzing the data related to the lost data, the lost data is inferred. There are different types of inference methods according to different applications, such as theoretical deduction and reference deduction. Theoretical deduction is a method of calculating the data that should be generated during the loss time based on the device's own historical data, while reference deduction is a method of calculating the lost data based on the data generated by other devices in the same operating environment. This method is based on other data in the relevant environment, so it can only be approximate and cannot be completely consistent with the lost data, but it can basically meet the user's analysis needs and provide assistance for the correctness of the statistical results.

[0039] The data aggregation method is to aggregate, calculate, and count the data according to the abstract standards of the business scenario to obtain the missing data. Specifically, after completing the basic standard mapping and standard processing of the data, it is necessary to aggregate, calculate, count, and other light aggregation standardization of the data according to the abstract standards of the business scenario.

[0040] Step S3: Feature extraction: extract feature parameters that can reflect the operating status of the equipment from the preprocessed data through data analysis and data mining.

[0041] Preferably, the feature extraction process in step S3 includes: extracting the principal component that can reflect the operating status of the equipment from the preprocessed data using principal component analysis or factor analysis, and selecting characteristic parameters that can reflect the operating status of the equipment based on expert experience.

[0042] Step S4: Model establishment: constructing an equipment fault diagnosis model based on historical fault data and corresponding characteristic parameters.

[0043] Preferably, the process of building a device fault diagnosis model in step S4 includes: The equipment fault diagnosis model is constructed by using machine learning algorithms such as support vector machine (SVM), neural network (NN), and decision tree (DT).

[0044] Based on historical fault data and characteristic parameters, the equipment fault diagnosis model is trained and verified, and the parameter update and iterative optimization of the equipment fault diagnosis model is completed through the cross-validation method; At the same time, the model performance is evaluated through accuracy, recall rate, and F1 value, and is used for actual diagnosis when the equipment fault diagnosis model meets the preset indicators.

[0045] Step S5: Fault diagnosis, inputting the real-time collected data into the fault diagnosis model to complete the status judgment of each device in the distributed photovoltaic power station and identify potential faults.

[0046] Step S6: Fault warning. For the equipment with potential faults identified in step S5, a warning message is issued and maintenance suggestions for the potential faults are provided. The diagnostic results and warning information obtained in step S5 are fed back to the operation and maintenance personnel to assist them in equipment maintenance and fault handling.

[0047] Preferably, during the fault warning process, fault information processing is performed in a graded manner according to the severity of the fault, for example, minor faults are recorded and serious faults immediately sound an alarm.

[0048] Step S7: Feedback the results and generate a diagnostic report including fault cause analysis, treatment suggestions and subsequent maintenance plan.

[0049] The present application method can monitor the operating status of the equipment in real time and detect potential faults in time; the present application method adopts big data analysis and machine learning technology to improve the accuracy and timeliness of fault diagnosis, and improve the overall operating efficiency and economic benefits of photovoltaic power stations; through intelligent fault diagnosis and early warning, the operation and maintenance costs of the equipment are reduced; through timely fault handling and maintenance, the service life of the equipment is extended.

[0050] Example This embodiment discloses a distributed photovoltaic power station equipment diagnosis method based on big data analysis, which completes the distributed photovoltaic power station equipment diagnosis through data collection, data preprocessing, feature extraction, model building, fault diagnosis and result feedback.

[0051] refer to Figure 2 As shown, in this embodiment, the preprocessing of the collected data includes data cleaning, data interpolation and data aggregation.

[0052] Data cleaning includes: active power data cleaning, inverter daily power generation data cleaning, meter daily power generation data cleaning, string voltage and current data cleaning, environmental monitoring instrument irradiation data cleaning, box transformer voltage and current data cleaning.

[0053] Data cleaning is the process of re-examining and verifying data, with the goal of removing duplicate information, correcting existing errors, and providing data consistency. From two perspectives, data cleaning is to solve data quality problems and make data more suitable for analysis. There are corresponding solutions and methods for different purposes. The standard model of data cleaning is to input data into the data cleaning processor, "clean" the data through a series of steps, and then output the cleaned data in the desired format. The result of data cleaning is to process various dirty data accordingly, obtain standard, clean, and continuous data, and provide it to data statistics, data mining and other functions.

[0054] In various application systems, missing data often occurs and is even inevitable. Therefore, in most cases, information systems are incomplete to some extent. For business functions, the system loses a lot of useful information; the uncertainty components shown in the system are more difficult to grasp; the missing data will also cause the analysis process to fall into chaos, resulting in unreliable output. Therefore, missing data needs to be derived and filled by special methods to reduce the difference between data analysis and practical applications. In the process of data interpolation, it is extremely important to identify the mechanism of missing data. First of all, this involves the issue of representativeness. Statistically speaking, non-random missing data will produce biased estimates and therefore cannot represent the population well. Secondly, it determines the choice of data interpolation method. The processing of random missing data is relatively simple, but the processing of non-random missing data is more difficult because the degree of deviation is difficult to grasp. In general, data interpolation refers to the selection of reasonable data to replace missing data, and different interpolation methods will have a greater impact on overall inference.

[0055] In this embodiment, the data interpolation processing is mainly divided into the following steps: The user first sets the data that needs to be interpolated according to the operating characteristics of the equipment, that is, if data loss occurs, the data will be interpolated; during the data interpolation service, the integrity of the data is checked regularly, and if data loss is found, data interpolation is performed; there are many types of data interpolation methods, which need to be selected according to the preset conditions and the characteristics of the data, including: interpolation method and derivation method.

[0056] This embodiment uses an equipment fault diagnosis model using a standard string validity algorithm as an example for explanation. The equipment fault diagnosis model diagnoses substandard equipment by combining equipment comparison and standard compliance analysis.

[0057] The validity algorithm of the standard string in the equipment fault diagnosis model performs power generation capacity deviation analysis and standard string power generation capacity comparison analysis on the standard string based on the data of the day, and compares the irradiation and power conversion rate of the standard string on the day with the daily average irradiation and power conversion rate of the standard string in history, or judges by two methods such as the historical average ranking ratio of the standard string to see whether the standard string meets the requirements of being a standard string. If it does not meet the requirements, the standard string is marked as invalid, and if it meets the requirements, it is marked as valid, and the reason for the analysis is given.

[0058] The process of analyzing the deviation of power generation capacity of the benchmark string includes: (1) Obtaining data: Obtain valid data on the irradiation intensity, standard string branch current, standard string branch voltage, standard string capacity, inverter power, and inverter installed capacity corresponding to the irradiation intensity on the day being greater than the preset value.

[0059] (2) Calculate the single watt power, radiation intensity number and radiation and power conversion rate.

[0060] (3) Calculate the daily average radiation and power conversion rate.

[0061] The comparative analysis process of standard string power generation capacity includes: (1) If the standard string is a single branch: sort the daily standard string average current with the average current of other equipment in the power station. If the standard string is the entire inverter: sort the full power hours of the standard string with the full power hours of other equipment in the power station.

[0062] (2) Calculate the ranking ratio.

[0063] Ranking ratio = (standard string ranking / total ranking) * 100% If the string is a shared string, the total number should include the shared string.

[0064] (3) Calculate the historical average ranking ratio of the string.

[0065] The historical average ranking ratio of the string is the average ranking ratio of the historical comparison days of the string validity (7 days by default, configurable).

[0066] Validity judgment of the string Set the criteria for judging the validity of standard strings, including: Condition 1: The irradiation and power conversion rate of the day > the average of the daily average irradiation and power conversion rate within the historical comparison days of standard string validity (default 7 days, configurable) * (1-irradiation and power conversion rate deviation threshold) (default x%, configurable); Condition 2: The historical average ranking ratio of standard strings < the comparison threshold of standard string power generation capacity (default x%, configurable).

[0067] The process of judging the validity of the string includes: (1) Determine whether condition 2 (whether the ranking is high) is met.

[0068] ① No: The bid string is invalid (Reason: The bid string power generation capacity is ranked low, there may be unreasonable bid string selection or it has not been cleaned for a long time.) ② Yes: Proceed to the next step.

[0069] (2) Determine whether condition 1 (whether there are average daily irradiance and power conversion rate values ​​that can be compared) is met.

[0070] ① No: The standard string is valid (Reason: The standard string power generation capacity ranks high, but the irradiation data during the evaluation period is low or abnormal, and the power generation deviation cannot be evaluated.) ② Yes: Proceed to the next step.

[0071] (3) Determine whether condition 1 (whether there are radiation and power conversion rates on the same day that can be used for comparison) is met.

[0072] ① No: The string is valid (Reason: The string meets the top ranking conditions, but the irradiation data during the evaluation period of the day is low or abnormal, and the power generation deviation cannot be evaluated.) ② Yes: Proceed to the next step.

[0073] (4) Determine whether condition 1 (whether the conversion rate meets the standard) is met.

[0074] ① No: The tag string is invalid (Reason: The tag string may not be cleaned, has a continuous fault, or is continuously blocked.) ② Yes: The standard string is effective (reason: the standard string power generation capacity ranks high and the power generation deviation is small during the evaluation period).

[0075] refer to Figure 3 As shown, the equipment fault diagnosis model specifically completes equipment diagnosis through the following process.

[0076] (1) Determine whether the standard string data uses current data or full-power hours data.

[0077] (2) Calculate the daily current value or full-power hours of the standard string (the daily full-power hours of the inverter where the standard string is located). The data is x by default and can be configured.

[0078] (3) Determination of the validity of the string.

[0079] (4) Calculate the daily current value and daily full-power hours of the comparison equipment.

[0080] (5) Calculate the power generation compliance rate of the equipment on a daily basis.

[0081] (6) Calculate the monthly power generation compliance rate of the equipment.

[0082] (7) Compare the monthly power generation compliance rate of the equipment with the compliance rate threshold, and calculate the compliance rate deviation ratio.

[0083] (8) Output diagnostic results and treatment suggestions based on the equipment’s daily power generation compliance rate and monthly deviation ratio.

[0084] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A distributed photovoltaic power station equipment diagnosis method based on big data analysis, characterized in that: The distributed photovoltaic power station equipment diagnosis method comprises: S1: Data collection, by collecting data transmitted by sensors on each device in the distributed photovoltaic power station, receiving the operating data of the device in real time; S2: Data preprocessing, preprocessing the collected data; S3: Feature extraction: extract characteristic parameters that can reflect the operating status of the equipment from the preprocessed data through data analysis and data mining; S4: Model establishment: constructing equipment fault diagnosis model based on historical fault data and corresponding characteristic parameters; Step S4: The process of building a device fault diagnosis model includes: Using machine learning algorithms, we build equipment fault diagnosis models. Based on historical fault data and characteristic parameters, we train and verify the equipment fault diagnosis models. We also use cross-validation methods to update and iterate the parameters of the equipment fault diagnosis models. At the same time, the model performance evaluation is completed through accuracy, recall rate, and F1 value, and the equipment fault diagnosis model is used for actual diagnosis when it meets the preset indicators; The equipment fault diagnosis model adopts the standard string validity algorithm to perform equipment diagnosis. The standard string validity algorithm diagnoses the equipment that does not meet the standard by combining the equipment comparison compliance rate analysis; The validity algorithm of the standard string in the equipment fault diagnosis model performs power generation capacity deviation analysis and standard string power generation capacity comparison analysis on the standard string based on the data of the day, and compares the irradiation and power conversion rate of the standard string on the day with the daily average irradiation and power conversion rate of the standard string in history, or judges by the historical average ranking ratio of the standard string to determine whether the corresponding standard string meets the requirements of being a standard string. If it does not meet the requirements, the standard string is marked as invalid. If it meets the requirements, it is marked as valid and the reasons for the analysis are given; S5: Fault diagnosis: input the real-time collected data into the fault diagnosis model to complete the status judgment of each device in the distributed photovoltaic power station and identify potential faults.

2. The distributed photovoltaic power station equipment diagnostic method according to claim 1, characterized in that: The distributed photovoltaic power station equipment diagnosis method further includes: S6: Fault warning: for the equipment with potential faults identified in step S5, a warning message is issued and maintenance suggestions for the potential faults are provided; The diagnostic results and warning information obtained in step S5 are fed back to the operation and maintenance personnel to assist them in equipment maintenance and fault handling.

3. The distributed photovoltaic power station equipment diagnostic method according to claim 2, characterized in that: The distributed photovoltaic power station equipment diagnosis method further includes: S7: Result feedback, generating a diagnostic report including fault cause analysis, treatment suggestions and subsequent maintenance plan.

4. The distributed photovoltaic power station equipment diagnostic method according to claim 1, characterized in that: Step S2: data preprocessing includes: data cleaning, data interpolation and data normalization of the collected data; Outliers are removed through data cleaning, and the corresponding sample size is ensured through data interpolation. Then, data normalization is performed to eliminate dimensional differences.

5. The distributed photovoltaic power station equipment diagnostic method according to claim 4, characterized in that: The data cleaning includes: data time stamp cleaning, data missing cleaning, data jump cleaning and data dead value cleaning; The data time stamp cleaning process includes: when checking the received data, presetting the range of the correct time stamp, and then judging the data time stamp according to the current time. If it exceeds the preset range, it is considered to be invalid data; The missing data cleaning process includes: first finding the missing data according to the time scale, and then filling the missing data; The data jump cleaning process includes: completing the jump data judgment by the interval growth judgment method or the preset condition judgment method, and then recovering the jump data by the deduction method or the blanking method or marking the corresponding jump data as an error or blanking; The data dead value cleaning process includes: first, manually presetting the dead value conditions, setting the data range and time interval of the dead value, and then, when it is found that the data does not exceed the specified value range within the specified time, the corresponding data has a dead value, and it is restored through deduction or blanking method, or the corresponding jump data is marked as an error or blanked.

6. The distributed photovoltaic power station equipment diagnostic method according to claim 4, characterized in that: The data interpolation includes: completing the missing data supplementation through mathematical interpolation method, derivation method and data aggregation method; The mathematical interpolation method is to obtain the missing data in the middle by calculating the difference between the adjacent data before and after the loss for the accumulated increasing data. The deductive method is to obtain the missing data by analyzing the data related to the missing data and inferring them; The data aggregation method is to aggregate, calculate, and count the data according to the abstract standards of the business scenario to obtain the missing data.

7. The distributed photovoltaic power station equipment diagnostic method according to claim 1, characterized in that: Step S3: feature extraction process includes: The principal component analysis or factor analysis method is used to extract the principal components that can reflect the operating status of the equipment from the preprocessed data, and based on expert experience, the characteristic parameters that can reflect the operating status of the equipment are selected.

8. The distributed photovoltaic power station equipment diagnostic method according to claim 2, characterized in that: During the fault warning process in step S6, fault information processing is performed in different levels according to the severity of the fault.

Images