Photovoltaic performance evaluation method and system based on data analysis

Abstract

The invention discloses a photovoltaic performance evaluation method and system based on data analysis, and relates to the technical field of photovoltaic performance evaluation, and the method comprises the steps: collecting the photovoltaic performance data of a power station in a distributed manner, and obtaining a first excess index according to an abnormal value; performing data processing and weight calculation on the integrated data and the first excess index array, and outputting optimal integrated data; and calculating a current actual prediction integrated feature array according to the actual photovoltaic performance of the hardware equipment, and evaluating the photovoltaic performance of the power station in a distributed manner. Understanding of characteristics of different photovoltaic modules is further improved through multi-level data processing, calculation and comparison of a plurality of excess index arrays and based on a stacking model and a feature weighted summation process, an optimal performance combination is finally obtained, the system has a real-time updating capability, and the optimal performance combination is obtained by integrating data analysis of each hardware device. The performance of each component is comprehensively evaluated, so that long-term efficient and stable operation of the power station is guaranteed, and the relation between photovoltaic performance data and the power station is deeply, completely and comprehensively evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic performance evaluation, and particularly to a photovoltaic performance evaluation method and system based on data analysis. Background Art

[0002] In recent years, the performance evaluation of photovoltaic power generation systems is crucial to ensure that photovoltaic systems maximize their efficiency throughout their life cycle. The performance of photovoltaic systems directly affects their power generation efficiency, economy, and the stability of the entire energy system. With the increasing scale of photovoltaic power stations, how to efficiently and accurately evaluate and optimize the performance of photovoltaic systems has become an urgent problem to be solved. Traditional photovoltaic performance evaluation methods mainly rely on manual inspections and expert experience-based analyses. However, with the increasing number of photovoltaic power stations and the diversification of operation data, the efficiency and accuracy of traditional methods can no longer meet the requirements of modern photovoltaic power generation.

[0003] Currently, the Chinese invention patent with the application number CN202311678181.9 discloses a method and system for comprehensive evaluation of the performance of photovoltaic power stations, which collects the operation data of photovoltaic power stations to form a feature matrix and performs optimization processing. The invention can better handle the outliers and noise in the data, improve the robustness and reliability of clustering, and the hierarchical clustering that integrates multiple methods can better adapt to different data characteristics and discover hidden rules and patterns. However, in reality, the performance of photovoltaic systems is not only affected by the photovoltaic modules themselves, but also by other system components, such as inverters, storage devices, and system integration methods. Existing methods often ignore this complexity, and there is also a problem of efficiency decay over time, making it impossible to deeply and comprehensively evaluate the relationship between photovoltaic performance data and power stations. Summary of the Invention

[0004] The technical problem solved by the present invention is that in reality, the performance of photovoltaic systems is not only affected by the photovoltaic modules themselves, but also by other system components, such as inverters, storage devices, and system integration methods. Existing methods often ignore this complexity, and there is also a problem of efficiency decay over time, making it impossible to deeply and comprehensively evaluate the relationship between photovoltaic performance data and power stations.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: A photovoltaic performance evaluation method based on data analysis, comprising the following steps: Step S1: Distributedly collect the photovoltaic performance data of the power station, obtain abnormal values, obtain a first excess index according to the abnormal values, and generate a first excess index array; Step S2: Collect the integration data of each hardware device; Step S3: Output the optimal integrated data by performing data processing and weight calculation on the integrated data and the first excess index array; Step S4: Calculate the current actual predicted integrated feature array according to the actual photovoltaic performance of the hardware device, and distributively evaluate the photovoltaic performance of the power station.

[0006] Preferably, step S1 includes: Collect photovoltaic performance data from components such as photovoltaic modules, inverters, and system controllers of each power station. The photovoltaic performance data includes voltage, current, temperature, irradiance, and output power. Perform data processing on the photovoltaic performance data. The data processing includes data cleaning and normalization calculation. Classify the processed photovoltaic performance data into a first photovoltaic performance dataset, a second photovoltaic performance dataset, a third photovoltaic performance dataset, a fourth photovoltaic performance dataset, and a fifth photovoltaic performance dataset according to the classifications of voltage, current, temperature, irradiance, and output power respectively; Data cleaning includes deleting negative values of the photovoltaic performance data. Normalization calculation includes transforming the photovoltaic performance data after data cleaning to a unified scale.

[0007] Preferably, calculate the median, average value, and standard deviation of the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset respectively. Weightedly combine the median, average value, and standard deviation to obtain a universal value. Use the Z-distribution to calculate the confidence intervals of the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset respectively, to obtain a first confidence interval, a second confidence interval, a third confidence interval, a fourth confidence interval, and a fifth confidence interval. Compare the data in the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset with the first confidence interval, the second confidence interval, the third confidence interval, the fourth confidence interval, and the fifth confidence interval respectively, and put all the data not within the confidence interval into a preset abnormal data queue; Subtract the data stored in the abnormal data queue from the corresponding confidence intervals respectively to obtain a first excess index. Arrange all the first excess indices in an array according to the corresponding photovoltaic performance datasets to obtain a first excess index array. The row meaning of the first excess index array is the photovoltaic performance dataset, and the column meaning is the first excess index corresponding to each photovoltaic performance dataset.

[0008] Preferably, step S2 includes: The hardware device includes an analog front-stage DC-DC DC power supply, an inverter main circuit, an isolation transformer, a controller, and a drive circuit module; Measure the input voltage and output voltage of the DC-DC DC power supply using an oscilloscope, measure the AC voltage waveform output by the inverter, detect the waveform, phase and amplitude of the output voltage of the isolation transformer, and record the switching frequency and duty cycle of the DC-DC DC power supply and the inverter; Use an oscilloscope and a logic analyzer to collect the signals of the controller and the drive circuit; The integrated data includes the input voltage and output voltage of the DC-DC DC power supply, the AC voltage waveform output by the inverter, the waveform, phase and amplitude of the output voltage of the isolation transformer, the switching frequency and duty cycle of the DC-DC DC power supply and the inverter, the signals of the controller and the signals of the drive circuit; Allocate software modules to each hardware device, and write the service functions corresponding to each software module in an independent form.

[0009] Preferably, the service function includes an initialization program, a running program and a port definition program: The initialization program is used for the automatic initialization of the hardware device and the start of the interrupt service program. After all data is initialized, a fault judgment is performed. If there is a fault in the system, the system will be shut down. If there is no fault in the system, it will enter a loop detection and start grid connection work; The running program is used to calculate the second excess index of each power station based on the photovoltaic performance data under normal conditions; The second excess index is represented by comparing the difference between the actual output and the theoretical output, and its mathematical expression is: ; Among them, is the second excess index, is the actual output efficiency corresponding to the th integrated data, is the theoretical output efficiency corresponding to the th integrated data, is the total number of integrated data; Arrange all the second excess indices according to the array arrangement method of the first excess index array to obtain a second excess index array; The port definition program is used to perform Huffman port coding on all hardware, and extract integrated data from the hardware through the coding of the port.

[0010] Preferably, taking the fourth excess index array as input data, performing PCA principal component analysis after Z-score standardization to obtain a fifth excess index array corresponding to the characteristic data of the array nodes corresponding to the data of the fourth excess index array, and respectively passing the fifth excess index array through a ridge regression model, a CART decision tree model, a random forest model, a support vector machine model, and an XGBoost gradient boosting decision tree model to obtain a sixth excess index array, a seventh excess index array, an eighth excess index array, a ninth excess index array, and a tenth excess index array.

[0011] Preferably, combining the sixth excess index array, the seventh excess index array, the eighth excess index array, the ninth excess index array, and the tenth excess index array by column to form a new unified feature array, standardizing the node feature data values of the unified feature array to obtain an eleventh excess matrix, taking the eleventh excess matrix as input feature data and inputting it into a linear regression model to train the linear regression model to obtain the optimal weights corresponding to each excess index array, and using the weights of the trained linear regression model to perform weighted summation on the input feature data to output an optimal integrated feature array.

[0012] Preferably, the detected integrated data is updated in real time to obtain a sixth excess index array, a seventh excess index array, an eighth excess index array, a ninth excess index array, and a tenth excess index array updated in real time, and an updated optimal integrated feature array is obtained according to the updated sixth excess index array, seventh excess index array, eighth excess index array, ninth excess index array, and tenth excess index array. The node data in the optimal integrated feature array includes the optimal performance values obtained from the optimal distribution of the integrated data and the photovoltaic performance data corresponding to each hardware device.

[0013] Preferably, a distributed evaluation of the photovoltaic performance of the current power station is performed, and the distributed evaluation logic includes: Calculating the actual photovoltaic performance mean value of each hardware device of each power station according to each unit time period respectively, normalizing the light, temperature, load, and electrical parameters respectively and using them as actual weight values to input into the trained linear regression model, performing weighted summation on the currently input feature data, and outputting the currently actual predicted integrated feature array; Calculating the current first excess index according to the currently actual predicted integrated feature array and the optimal integrated feature array based on the calculation method of the first excess index, and inputting the current first excess index into an SVR support vector regression model to analyze the attenuation trend and establish an attenuation model of the device.

[0014] A photovoltaic performance evaluation system based on data analysis, including an acquisition module, a data processing module, and an evaluation module: The acquisition module is used to distributively acquire the photovoltaic performance data of the power station, obtain abnormal values, and acquire the integrated data of each hardware device; The data processing module includes obtaining a first excess index according to the abnormal value, generating a first excess index array, and outputting optimal integrated data by performing data processing and weight calculation on the integrated data and the first excess index array; The evaluation module includes calculating a current actual prediction integrated feature array according to the actual photovoltaic performance of the hardware device, and distributively evaluating the photovoltaic performance of the power station.

[0015] Advantages of the present invention: Through multi-level data processing, calculating and comparing multiple excess index arrays, and combining with a machine learning model, real-time optimization of device performance is achieved. The process of the stacking model and feature weighted summation further improves the understanding of the characteristics of different photovoltaic components, and finally obtains the optimal performance combination. The system has the ability to update in real time, can continuously obtain the latest data of the photovoltaic power station and hardware devices, and accordingly generate an optimal integrated feature array. By analyzing the attenuation trend through the SVR model, it provides an accurate prediction of the future state of the device, discovers the risk of potential faults or performance degradation in advance. Adopting a distributed evaluation strategy, it can calculate the performance for each power station and hardware device respectively and conduct a comprehensive evaluation, enabling the performance of each component to be independently monitored and dynamically adjusted, which helps to reduce the overall maintenance cost of the system and improve the operation efficiency of the power station. By comprehensively analyzing the data of hardware devices such as photovoltaic panels, inverters, controllers, and drive circuits, the system can comprehensively evaluate the performance of each component, timely adjust the operation strategy, thereby ensuring the long-term efficient and stable operation of the power station, and deeply and completely comprehensively evaluating the relationship between photovoltaic performance data and the power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 FIG. is a schematic diagram of the basic process of a photovoltaic performance evaluation method based on data analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention is provided in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments.

[0018] Referring to Figure 1 , an embodiment of the present invention provides a photovoltaic performance evaluation method based on data analysis, including the following steps: Step S1: Distributively acquire the photovoltaic performance data of the power station, obtain abnormal values, obtain a first excess index according to the abnormal values, and generate a first excess index array; Step S2: Acquire the integrated data of each hardware device; Step S3: Output the optimal integrated data by performing data processing and weight calculation on the integrated data and the first excess index array; Step S4: Calculate the current actual predicted integrated feature array according to the actual photovoltaic performance of the hardware device, and distributively evaluate the photovoltaic performance of the power station.

[0019] The said Step S1 includes: Collect photovoltaic performance data from components such as photovoltaic modules, inverters, and system controllers of each power station. The photovoltaic performance data includes voltage, current, temperature, irradiance, and output power. Perform data processing on the photovoltaic performance data. The data processing includes data cleaning and normalization calculation. Classify the data-processed photovoltaic performance data into the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset respectively according to the classifications of voltage, current, temperature, irradiance, and output power; Data cleaning includes deleting the negative values of the photovoltaic performance data, and normalization calculation includes transforming the data-cleaned photovoltaic performance data to a unified scale.

[0020] Calculate the median, average value, and standard deviation of the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset respectively. Weightedly combine the median, average value, and standard deviation to obtain a universal value. Calculate the confidence intervals of the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset respectively using the Z-distribution to obtain the first confidence interval, the second confidence interval, the third confidence interval, the fourth confidence interval, and the fifth confidence interval. Compare the data in the first photovoltaic performance dataset, the second photovoltaic performance dataset, the third photovoltaic performance dataset, the fourth photovoltaic performance dataset, and the fifth photovoltaic performance dataset with the first confidence interval, the second confidence interval, the third confidence interval, the fourth confidence interval, and the fifth confidence interval respectively, and put all the data not within the confidence intervals into a preset abnormal data queue; Subtract the data stored in the abnormal data queue from the corresponding confidence intervals respectively to obtain the first excess index. Arrange all the first excess indexes in an array according to the corresponding photovoltaic performance datasets to obtain the first excess index array. The row meaning of the first excess index array is the photovoltaic performance dataset, and the column meaning is the first excess index corresponding to each photovoltaic performance dataset.

[0021] The said Step S2 includes: The hardware device includes an analog front-stage DC-DC DC power supply, an inverter main circuit, an isolation transformer, a controller, and a drive circuit module; Measure the input voltage and output voltage of the DC-DC DC power supply using an oscilloscope, measure the AC voltage waveform output by the inverter, detect the waveform, phase and amplitude of the output voltage of the isolation transformer, and record the switching frequency and duty cycle of the DC-DC DC power supply and the inverter; Use an oscilloscope and a logic analyzer to collect the signals of the controller and the drive circuit to ensure that the control and drive signals work according to the design requirements; The integrated data includes the input voltage and output voltage of the DC-DC DC power supply, the AC voltage waveform output by the inverter, the waveform, phase and amplitude of the output voltage of the isolation transformer, the switching frequency and duty cycle of the DC-DC DC power supply and the inverter, the signals of the controller and the signals of the drive circuit; Allocate software modules to each hardware device, and write the service functions corresponding to each software module in an independent form.

[0022] The service functions include an initialization program, a running program and a port definition program: The initialization program is used for the automatic initialization of the hardware device and the start of the interrupt service program. After all data is initialized, it enters the fault judgment. If there is a fault in the system, it will stop. If there is no fault in the system, it will enter the loop detection and start the grid connection work; The running program is used to calculate the second excess index of each power station based on the photovoltaic performance data under normal conditions; The second excess index is represented by comparing the difference between the actual output and the theoretical output, and its mathematical expression is: ; Among them, is the second excess index, is the actual output efficiency corresponding to the th integrated data, is the theoretical output efficiency corresponding to the th integrated data, is the total number of integrated data; Arrange all the second excess indices according to the array arrangement method of the first excess index array to obtain the second excess index array; Bubble sort the rows of the first excess index array and the second excess index array respectively based on the time series to obtain the third excess index array and the fourth excess index array sorted based on the time series; The port definition program is used to perform Huffman port coding on all hardware, and extract integrated data from the hardware through the coding of the port.

[0023] Taking the fourth excess index array as input data, performing PCA principal component analysis after Z-score standardization, to obtain a fifth excess index array corresponding to the characteristic data of the array nodes corresponding to the data of the fourth excess index array, and respectively passing the fifth excess index array through a ridge regression model, a CART decision tree model, a random forest model, a support vector machine model, and an XGBoost gradient boosting decision tree model to obtain a sixth excess index array, a seventh excess index array, an eighth excess index array, a ninth excess index array, and a tenth excess index array.

[0024] The ridge regression model, CART decision tree model, random forest model, support vector machine model, and XGBoost gradient boosting decision tree model have different learning methods and independently learn different parts of the integrated data so that they can capture the characteristics of the integrated data from different perspectives, and these characteristics are also the predictions of the integrated data by these machine learning models from different aspects respectively.

[0025] Combining the sixth excess index array, seventh excess index array, eighth excess index array, ninth excess index array, and tenth excess index array by column to form a new unified feature array, standardizing the node feature data values of the unified feature array to obtain an eleventh excess matrix, taking the eleventh excess matrix as input feature data and inputting it into a linear regression model to train the linear regression model. The linear regression model is used to learn how to combine the characteristics of each excess index array into the final predicted value and give the weight of each feature, to obtain the optimal weights corresponding to each excess index array, and using the weights of the trained linear regression model to perform weighted summation on the input feature data to output the optimal integrated feature array.

[0026] The optimal integrated feature array contains the expected output and optimal operating state of each system component under specific environmental conditions.

[0027] Updating the detected integrated data in real time, obtaining the sixth excess index array, seventh excess index array, eighth excess index array, ninth excess index array, and tenth excess index array updated in real time, and obtaining the updated optimal integrated feature array according to the updated sixth excess index array, seventh excess index array, eighth excess index array, ninth excess index array, and tenth excess index array. The node data in the optimal integrated feature array includes the optimal performance values obtained from the optimal distribution of the integrated data corresponding to each hardware device and the photovoltaic performance data.

[0028] Through the stacking model technology, an "optimal integrated data" can be obtained, that is, by combining the prediction results of multiple base models, the overall prediction accuracy can be maximized, and the errors and overfitting of a single model can be reduced.

[0029] Perform a distributed evaluation of the photovoltaic performance of the current power station. The distributed evaluation logic includes: Calculate the actual photovoltaic performance mean value of each hardware device of each power station according to each unit time period respectively. After normalizing the light, temperature, load and electrical parameters respectively, use them as the actual weight values to input into the trained linear regression model, perform weighted summation on the currently input feature data, and output the current actual predicted integrated feature array; Calculate the current first excess index according to the calculation method of the first excess index based on the current actual predicted integrated feature array and the optimal integrated feature array. Input the current first excess index into the SVR support vector regression model to analyze the attenuation trend and establish the attenuation model of the device. Through these models, the performance attenuation situation of the device in the future period of time can be predicted.

[0030] A photovoltaic performance evaluation system based on data analysis, including a collection module, a data processing module and an evaluation module: The collection module is used to distributively collect the photovoltaic performance data of the power station, obtain abnormal values, and collect the integrated data of each hardware device; The data processing module includes obtaining the first excess index according to the abnormal value, and generating a first excess index array. Through data processing and weight calculation of the integrated data and the first excess index array, output the optimal integrated data; The evaluation module includes calculating the current actual predicted integrated feature array according to the actual photovoltaic performance of the hardware device, and distributively evaluating the photovoltaic performance of the power station.

[0031] Through multi-level data processing, calculating and comparing multiple excess index arrays, and combining machine learning models, the present invention realizes the real-time optimization of device performance. The processes of the stacking model and feature weighted summation further improve the understanding of the characteristics of different photovoltaic components, and finally obtain the optimal performance combination. The system has the ability to update in real time, can continuously obtain the latest data of the photovoltaic power station and hardware devices, and generate the optimal integrated feature array accordingly. Analyze the attenuation trend through the SVR model to provide an accurate prediction for the future state of the device, discover the risk of potential faults or performance degradation in advance, adopt a distributed evaluation strategy, can calculate the performance of each power station and hardware device respectively and conduct a comprehensive evaluation, so that the performance of each component can be independently monitored and dynamically adjusted, which helps to reduce the overall maintenance cost of the system and improve the operation efficiency of the power station. Through comprehensive data analysis of hardware devices such as photovoltaic panels, inverters, controllers, and drive circuits, the system can comprehensively evaluate the performance of each component, adjust the operation strategy in time, so as to ensure the long-term efficient and stable operation of the power station, and deeply and completely comprehensively evaluate the relationship between photovoltaic performance data and the power station.

[0032] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or a plurality of processes and / or Figure 1 boxes or a plurality of boxes.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A photovoltaic performance evaluation method based on data analysis, characterized in that: The following steps are involved: Step S1: Distributedly collect photovoltaic performance data of the power station, obtain abnormal values, obtain a first excess index according to the abnormal values, and generate a first excess index array; Step S2: Collecting integrated data of each hardware device; Step S3: Processing the integrated data and the first excess index array and calculating weights to output optimal integrated data; Step S4: Calculate the current actual predicted integrated characteristic array according to the actual photovoltaic performance of the hardware device, and distribute the photovoltaic performance of the power station.

2. The photovoltaic performance evaluation method based on data analysis according to claim 1, characterized in that: The step S1 comprises: Collecting photovoltaic performance data from photovoltaic modules, inverters, system controllers and other components of each power station, the photovoltaic performance data including voltage, current, temperature, irradiance and output power, performing data processing on the photovoltaic performance data, the data processing including data cleaning and normalization calculation, and dividing the processed photovoltaic performance data into a first photovoltaic performance data set, a second photovoltaic performance data set, a third photovoltaic performance data set, a fourth photovoltaic performance data set and a fifth photovoltaic performance data set according to the classification of voltage, current, temperature, irradiance and output power; Data cleaning includes deleting negative values ​​of the photovoltaic performance data, and normalization calculation includes transforming the photovoltaic performance data after data cleaning to a uniform scale.

3. The photovoltaic performance evaluation method based on data analysis according to claim 2, characterized in that: Calculate the intermediate value, average value and standard deviation of the first photovoltaic performance data set, the second photovoltaic performance data set, the third photovoltaic performance data set, the fourth photovoltaic performance data set and the fifth photovoltaic performance data set respectively, obtain the universal value after weighted combination of the intermediate value, average value and standard deviation, calculate the confidence interval of the first photovoltaic performance data set, the second photovoltaic performance data set, the third photovoltaic performance data set, the fourth photovoltaic performance data set and the fifth photovoltaic performance data set respectively by using Z distribution to obtain the first confidence interval, the second confidence interval, the third confidence interval, the fourth confidence interval and the fifth confidence interval, compare the data in the first photovoltaic performance data set, the second photovoltaic performance data set, the third photovoltaic performance data set, the fourth photovoltaic performance data set and the fifth photovoltaic performance data set with the first confidence interval, the second confidence interval, the third confidence interval, the fourth confidence interval and the fifth confidence interval respectively, and put all the data not in the confidence interval into a preset abnormal data queue; The data stored in the abnormal data queue are respectively subtracted from the corresponding confidence interval to obtain a first excess index, and all first excess indices are arranged in an array according to the corresponding photovoltaic performance data sets to obtain a first excess index array, wherein the rows of the first excess index array mean the photovoltaic performance data sets, and the columns mean the first excess index corresponding to each photovoltaic performance data set.

4. The photovoltaic performance evaluation method based on data analysis according to claim 1, characterized in that: The step S2 comprises: The hardware device includes an analog front-stage DC-DC power supply, an inverter main circuit, an isolation transformer, a controller and a drive circuit module; Use an oscilloscope to measure the input voltage and output voltage of the DC-DC power supply, measure the AC voltage waveform output by the inverter, detect the waveform, phase and amplitude of the output voltage of the isolation transformer, and record the switching frequency and duty cycle of the DC-DC power supply and inverter; Use oscilloscope and logic analyzer to collect signals from controller and drive circuit; The integrated data includes the input voltage and output voltage of the DC-DC power supply, the AC voltage waveform output by the inverter, the waveform, phase and amplitude of the output voltage of the isolation transformer, the switching frequency and duty cycle of the DC-DC power supply and inverter, the signal of the controller and the signal of the drive circuit; A software module is allocated to each hardware device respectively, and the service function corresponding to each software module is written in an independent form.

5. The photovoltaic performance evaluation method based on data analysis according to claim 4, characterized in that: The service function includes an initialization program, a running program and a port definition program: The initialization program is used for the automatic initialization of hardware devices and the startup of interrupt service programs. After all data are initialized, the system enters the fault judgment. If there is a fault in the system, the system is shut down. If there is no fault in the system, the system enters the cycle detection and starts the grid connection work. The operating program is used to calculate the second excess index of each power station based on the photovoltaic performance data under normal conditions; The second excess index is expressed by comparing the difference between the actual output and the theoretical output, and its mathematical expression is: ； in, is the second excess index, For the The actual output efficiency corresponding to the integrated data is For the The theoretical output efficiency corresponding to the integrated data is: is the total number of integrated data; Arranging all second excess indexes in an array according to the array arrangement method of the first excess index array to obtain a second excess index array; Bubble sort each row of the first excess index array and the second excess index array based on the time series to obtain a third excess index array and a fourth excess index array sorted based on the time series; The port definition program is used to perform Huffman port coding on all hardware, and to extract integrated data from the hardware through port coding.

6. The photovoltaic performance evaluation method based on data analysis according to claim 5, characterized in that: The fourth excess index array is used as input data, and PCA principal component analysis is performed after Z-score standardization to obtain a fifth excess index array corresponding to the feature data of the fourth excess index array data, and the fifth excess index array is respectively subjected to ridge regression model, CART decision tree model, random forest model, support vector machine model and XGBoost gradient boosting decision tree model to obtain a sixth excess index array, a seventh excess index array, an eighth excess index array, a ninth excess index array and a tenth excess index array.

7. The photovoltaic performance evaluation method based on data analysis according to claim 6, characterized in that: The sixth excess index array, the seventh excess index array, the eighth excess index array, the ninth excess index array and the tenth excess index array are merged by columns to form a new unified feature array, the node feature data values ​​of the unified feature array are standardized to obtain an eleventh excess matrix, the eleventh excess matrix is ​​input into a linear regression model as input feature data, the linear regression model is trained to obtain the optimal weights corresponding to each excess index array, the weights of the trained linear regression model are used to perform weighted summation on the input feature data, and an optimal integrated feature array is output.

8. The photovoltaic performance evaluation method based on data analysis according to claim 7, characterized in that: The integrated data obtained by real-time update detection is obtained to obtain the real-time updated sixth excess index array, seventh excess index array, eighth excess index array, ninth excess index array and tenth excess index array, and an updated optimal integrated feature array is obtained according to the updated sixth excess index array, seventh excess index array, eighth excess index array, ninth excess index array and tenth excess index array, wherein the node data in the optimal integrated feature array includes the optimal performance value obtained by the optimal distribution of the integrated data corresponding to each hardware device and the photovoltaic performance data.

9. The photovoltaic performance evaluation method based on data analysis according to claim 8, characterized in that: Conduct distributed evaluation of the photovoltaic performance of the current power station. The distributed evaluation logic includes: Calculate the actual photovoltaic performance average of each hardware device in each power station according to the unit time period, normalize the light, temperature, load and electrical parameters respectively and input them into the trained linear regression model as the actual weight value, perform weighted summation on the current input feature data, and output the current actual prediction integrated feature array; The current first excess index is calculated according to the calculation method of the first excess index based on the current actual predicted integrated feature array and the optimal integrated feature array, and the current first excess index is input into the SVR support vector regression model to analyze the attenuation trend and establish an attenuation model of the equipment.

10. A photovoltaic performance evaluation system based on data analysis, characterized in that: Including acquisition module, data processing module and evaluation module: The acquisition module is used to collect photovoltaic performance data of the power station in a distributed manner, obtain abnormal values, and collect integrated data of various hardware devices; The data processing module includes obtaining a first excess index according to the abnormal value, generating a first excess index array, processing the integrated data and the first excess index array through data processing and weight calculation, and outputting optimal integrated data; The evaluation module includes calculating the current actual prediction integrated characteristic array according to the actual photovoltaic performance of the hardware device, and distributively evaluating the photovoltaic performance of the power station.

Images