Integrated learner-based photovoltaic court reverse heavy overload identification method

Abstract

The invention belongs to the technical field of distributed photovoltaic power generation, and particularly relates to a photovoltaic court reverse overload identification method based on an integrated learner. In order to solve the problem of low accuracy of reverse heavy overload early warning of a transformer area, reverse heavy overload identification is performed on a photovoltaic distribution transformer in a photovoltaic micro-grid from two aspects of data characteristics and an identification model by analyzing load data characteristics of the transformer when reverse heavy overload occurs. And the characteristics influencing the reverse load rate are used as input, and reverse heavy overload identification of the photovoltaic transformer in the micro-grid is realized by using an AdaBoost reverse heavy overload identification model.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed photovoltaic power generation, and in particular relates to a method for identifying reverse heavy overload in a photovoltaic area based on an integrated learner. Background Art

[0002] Distributed photovoltaics have gradually become a key component of smart distribution networks. While distributed photovoltaics are being vigorously developed, low-voltage distributed photovoltaics are mainly distributed in economically underdeveloped areas, and the problem of photovoltaic reverse overload is prominent. The dispatchability of electric vehicle charging and the controllability of photovoltaic power generation have become one of the effective ways to solve this problem. At the same time, the fault identification of reverse overload in microgrids has an important impact on the stability and efficiency of microgrids.

[0003] The existing reverse overload identification method determines the predicted time period of the distribution transformer area, obtains the corresponding original data such as historical load data and meteorological data, predicts the transformer load data, and issues photovoltaic reverse overload warning information by calculating the reverse load rate, and evaluates the operation of the area according to the distribution transformer load rate. However, this type of prediction method only focuses on the time series characteristics of the load data and ignores the dynamic correlation between the photovoltaic output, charging load and transformer capacity when the reverse overload occurs, resulting in low accuracy of the reverse overload warning in the area. Summary of the invention

[0004] In view of the above problems, the present invention provides a method for identifying reverse heavy overload in photovoltaic areas based on an integrated learner. The method is based on the reverse heavy overload identification model of the integrated learner, and simultaneously identifies reverse heavy overload of distribution transformers in photovoltaic microgrids from two perspectives: reverse load rate and training weights allocated by the learning model.

[0005] In a first aspect, the present invention provides a method for identifying reverse heavy overload in a photovoltaic area based on an integrated learner, comprising the following steps:

[0006] Step 1, select different characteristic data affecting the reverse load rate and collect them in real time;

[0007] Step 2, input the real-time feature data collected in step 1 into a pre-trained strong classifier, and output the recognition result of the reverse heavy overload of the photovoltaic area. The strong classifier is trained based on an integrated learning algorithm using a historical data set constructed with different feature data that affect the reverse load rate as training samples.

[0008] Furthermore, the calculation formula of the reverse load rate in step 1 is shown in formula (1):

[0009]

[0010] Where P d Expressed as photovoltaic power generation, PL It is expressed as the transformer equivalent load at the same time, S D Expressed as the rated capacity of the transformer.

[0011] Furthermore, the different characteristic data in step 1 are transformer load, electric vehicle charging load, photovoltaic power and transformer capacity.

[0012] Furthermore, the method further includes: training the strong classifier, specifically:

[0013] (1) Preprocess different feature data by quartile method and maximum and minimum normalization, then construct a data set and divide the data set into a training set and a test set;

[0014] (2) Construct a fault identification model based on AdaBoost. The input is the feature data in the training set, and the output is {-1, 1}, where -1 indicates that it belongs to the photovoltaic area reverse overload, and 1 indicates that it does not belong to the photovoltaic area reverse overload. The algorithm of the weak classifier in the fault identification model adopts the gradient boosting tree, the number of iterations of the weak classifier is K, and the final strong classifier f(x) is weighted integrated;

[0015] (3) Initialize the weight of the training set, as shown in formula (2):

[0016]

[0017] In the formula, w 1i represents the initial weight of each load feature in the training set, and m represents the number of samples in the training set;

[0018] (4) Using weight D k The training set is used to train different feature data to obtain the weak classifier G k (x):x→{-1,1}, k=1,2,…,K, calculate G k The classification error rate of (x) is shown in formula (3):

[0019]

[0020] In the formula, y i represents the sample label set {-1, 1}, w ki represents the weight of the kth basic classifier with the lowest error rate at present, and I represents the weight of the kth basic classifier with the lowest error rate at present. k (x i )The weight of misclassified samples;

[0021] (5) Calculate the weight coefficient of the weak classifier, as shown in formula (4):

[0022]

[0023] In the formula, R is the number of categories;

[0024] (6) Update the weight distribution of the training set, as shown in formula (5):

[0025] w k+1,i =w ki Z K exp(-α k y i G k (x i )),i=1,2,…,m (5)

[0026] In the formula, Z K is the normalization factor, as shown in formula (6):

[0027]

[0028] (7) The final classifier f(x) of the weighted ensemble is shown in formula (7):

[0029]

[0030] (8) Use accuracy, recall, precision and F1 score to evaluate different feature data of the test set.

[0031] In a second aspect, the present invention further provides a photovoltaic area reverse heavy overload identification device based on an integrated learner, comprising:

[0032] A characteristic data selection module is used to select different characteristic data that affect the reverse load rate and collect them in real time;

[0033] The reverse heavy overload identification module is used to input the real-time collected feature data into a pre-trained strong classifier and output the identification result of the reverse heavy overload in the photovoltaic area. The strong classifier is trained based on an integrated learning algorithm using a historical data set constructed with different feature data that affect the reverse load rate as training samples.

[0034] Furthermore, the calculation formula of the reverse load rate is shown in formula (1):

[0035]

[0036] Where P d Expressed as photovoltaic power generation, P L It is expressed as the transformer equivalent load at the same time, S D Expressed as the rated capacity of the transformer.

[0037] Furthermore, the different characteristic data are transformer load, electric vehicle charging load, photovoltaic power and transformer capacity.

[0038] Furthermore, the device also includes a strong classifier training module, specifically including:

[0039] The data preprocessing unit performs data preprocessing on different feature data through quartile method and maximum and minimum normalization, and then constructs a data set, and divides the data set into a training set and a test set;

[0040] The model building unit builds a fault identification model based on AdaBoost. The input is the feature data in the training set, and the output is {-1, 1}, where -1 indicates that it belongs to the photovoltaic area reverse heavy overload, and 1 indicates that it does not belong to the photovoltaic area reverse heavy overload. The algorithm of the weak classifier in the fault identification model adopts the gradient boosting tree, the number of iterations of the weak classifier is K, and a strong classifier f(x) is weighted integrated;

[0041] Model training unit, training fault identification model based on AdaBoost;

[0042] The result evaluation unit uses accuracy, recall, precision and F1 score to evaluate different feature data of the test set.

[0043] Furthermore, the process of training the fault identification model based on AdaBoost includes:

[0044] Initialize the weight of the training set, as shown in formula (2):

[0045]

[0046] In the formula, w 1i represents the initial weight of each load feature in the training set, and m represents the number of samples in the training set;

[0047] (4) Using weight D k The training set is used to train different feature data to obtain the weak classifier G k (x):x→{-1,1}, k=1,2,…,K, calculate G k The classification error rate of (x) is shown in formula (3):

[0048]

[0049] In the formula, y i represents the sample label set {-1, 1}, w ki represents the weight of the kth basic classifier with the lowest error rate at present, and I represents the weight of the kth basic classifier with the lowest error rate at present. k (x i )The weight of misclassified samples;

[0050] (5) Calculate the weight coefficient of the weak classifier, as shown in formula (4):

[0051]

[0052] In the formula, R is the number of categories;

[0053] (6) Update the weight distribution of the training set, as shown in formula (5):

[0054] w k+1,i =w ki Z K exp(-α k y i G k (x i )),i=1,2,…,m (5)

[0055] In the formula, Z K is the normalization factor, as shown in formula (6):

[0056]

[0057] (7) The final classifier f(x) of the weighted ensemble is shown in formula (7):

[0058]

[0059] In a third aspect, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, and when the instructions are executed, the photovoltaic station reverse heavy overload identification method based on the integrated learner is implemented.

[0060] In a fourth aspect, the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the photovoltaic station reverse heavy overload identification method based on the integrated learner is implemented.

[0061] Compared with the prior art, the present invention has the following advantages:

[0062] (1) The reverse overload identification method of the present invention analyzes the transformer load data characteristics when reverse overload occurs, and identifies the reverse overload of the photovoltaic distribution transformer in the photovoltaic microgrid from two perspectives: data characteristics and identification model. The characteristics that affect the reverse load rate are used as input, and the reverse overload identification model of AdaBoost is used to realize the reverse overload identification of the photovoltaic transformer in the microgrid.

[0063] (2) Use the reverse load rate to filter the feature data that affects the reverse overload event. The existing reverse overload identification method obtains the original data of the distribution transformer, such as historical load data and meteorological data, in terms of feature selection. However, meteorological data mainly affects the size of photovoltaic power generation and has little relevance to the transformer equivalent load. Therefore, the reverse load rate that determines whether the distribution transformer has a reverse overload is used as the starting point. By analyzing the impact of distributed photovoltaic output and various load demands on the transformer equivalent load in the microgrid, photovoltaic power generation, electric vehicle charging load, transformer capacity and total load of the substation are selected as the input of the reverse overload identification model.

[0064] (3) This method takes into account the ability of the AdaBoost model to handle unbalanced feature data, iteratively trains multiple classifiers on the same training data, and integrates them together to build a more powerful final classifier. When processing multi-dimensional feature data, it tends to give greater weight to data with reverse overload, so that during iteration, AdaBoost will focus on the data characteristics of the overload period, thereby improving the accuracy of the recognition model. The advantage of this method is that there is no need to manually adjust the weights of each classifier and the reverse overload recognition result can be easily obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of the photovoltaic area reverse heavy overload identification method based on the integrated learner;

[0066] Figure 2 It is the load data of photovoltaic distribution transformer;

[0067] Figure 3 It is a schematic diagram of a photovoltaic area reverse heavy overload identification device based on an integrated learner;

[0068] Figure 4 For identification comparison infographic. DETAILED DESCRIPTION

[0069] In order to further illustrate the technical solution of the present invention, the present invention is further described below through embodiments.

[0070] like Figure 1 As shown, a photovoltaic area reverse heavy overload identification method based on an integrated learner in this embodiment includes the following steps:

[0071] Step 1, select different characteristic data affecting the reverse load rate and collect them in real time;

[0072] The reverse load rate assessment standard is a thermal stability assessment standard published in the Guidelines for the Assessment of the Carrying Capacity of Distributed Generation Access to the Grid issued by my country's Electric Power Industry Standards. It refers to the ratio of the transmission power flowing through the transmission and transformation equipment from the low voltage level to the high voltage level grid to the equipment capacity. The calculation formula is shown in formula (1):

[0073]

[0074] Where P d Expressed as photovoltaic power generation, P L It is expressed as the transformer equivalent load at the same time, S D Expressed as the rated capacity of the transformer.

[0075] Taking the 15-minute load data as an example, if the reverse load rate exceeds 80% for 8 consecutive points, it is considered that a reverse heavy overload has occurred. Figure 2 This is the actual load data of a distribution station in Hebei Province. It can be seen that photovoltaic reverse power transmission occurs at noon and lasts for more than 2 hours. If photovoltaic power generation is insufficient during this period, a reverse heavy overload will occur.

[0076] The selection of characteristic data is determined according to the reverse load rate, and the different characteristic data selected are transformer load, electric vehicle charging load, photovoltaic power and transformer capacity.

[0077] Step 2, input the feature data collected in real time in step 1 into a pre-trained strong classifier, and output the recognition result of the reverse heavy overload in the photovoltaic area. The strong classifier is trained based on an integrated learning algorithm using a historical data set constructed with different feature data that affect the reverse load rate as training samples.

[0078] This embodiment uses AdaBoost, which is a kind of ensemble learning algorithm. Its basic concept is to iteratively train multiple weak classifiers on the same training set and integrate them together to build a more powerful final classifier. Because of its better effect when processing unbalanced data sets, a fault identification model based on AdaBoost is proposed. When processing fault load data features, AdaBoost tends to give greater weight to data with reverse heavy overload.

[0079] The method of this embodiment further includes: training the strong classifier, specifically:

[0080] (1) Preprocess different feature data by quartile method and maximum and minimum normalization, then construct a data set and divide the data set into a training set and a test set;

[0081] (2) Construct a fault identification model based on AdaBoost. The input is the feature data in the training set, and the output is {-1, 1}, where -1 indicates that it belongs to the photovoltaic area reverse overload, and 1 indicates that it does not belong to the photovoltaic area reverse overload. The algorithm of the weak classifier in the fault identification model adopts the gradient boosting decision tree (GBDT). The number of iterations of the weak classifier is K, and the final strong classifier f(x) is obtained by weighted integration;

[0082] (3) Initialize the weight of the training set, as shown in formula (2):

[0083]

[0084] In the formula, w 1i represents the initial weight of each load feature in the training set, and m represents the number of samples in the training set;

[0085] (4) Using weight D k The training set is used to train different feature data to obtain the weak classifier G k (x):x→{-1,1}, k=1,2,…,K, calculate G k The classification error rate of (x) is shown in formula (3):

[0086]

[0087] In the formula, y i represents the sample label set {-1, 1}, w ki represents the weight of the kth basic classifier with the lowest error rate at present, and I represents the weight of the kth basic classifier with the lowest error rate at present. k (x i )The weight of misclassified samples;

[0088] (5) Calculate the weight coefficient of the weak classifier, as shown in formula (4):

[0089]

[0090] In the formula, R is the number of categories;

[0091] (6) Update the weight distribution of the training set, as shown in formula (5):

[0092] w k+1,i =w ki Z K exp(-α k y i G k (x i )),i=1,2,…,m (5)

[0093] In the formula, Z Kis the normalization factor, as shown in formula (6):

[0094]

[0095] (7) The final classifier f(x) of the weighted ensemble is shown in formula (7):

[0096]

[0097] (8) Use accuracy, recall, precision and F1-score to evaluate different feature data of the test set. When the reverse overload recognition model is used for fault recognition, it is inevitable that there will be misjudgment behavior. The accuracy (Accuracyscore), recall (Recall), precision (Precision) and F1-score (F1-score) of the recognition results are the key basis for evaluating the recognition effect of the model, and can reflect the specific classification performance of the constructed recognition model.

[0098] Accuracy: refers to the comparison between the identified operating status of the photovoltaic equipment and the actual operating status in binary or multi-classification. The accuracy calculation formula is shown in (8).

[0099]

[0100] Recall: A measure of how accurately the model predicts all actual correct examples. It is calculated as the ratio of the total number of true positive cases to the number of true positive predictions made by the classifier. That is, the percentage of true positive examples that the classifier correctly detects. The formula for calculating recall is shown in (9).

[0101]

[0102] Precision: refers to the proportion of samples whose transformer operating state is 1 and whose actual operating state is also 1 in all samples whose identified operating state is 1. The formula for calculating the precision is shown in (10).

[0103]

[0104] The F1 score is defined as the harmonic mean of precision and recall, and is also the most representative evaluation indicator of the performance of the classification model. The formula for calculating the F1 score is shown in (11).

[0105]

[0106] Among them, TP is a true positive example, TN is a true negative example, FP is a false positive example, and FN is a false negative example.

[0107] In addition, if Figure 3As shown, the embodiment of the present invention also provides a photovoltaic area reverse heavy overload identification device based on an integrated learner, comprising:

[0108] A characteristic data selection module is used to select different characteristic data that affect the reverse load rate and collect them in real time;

[0109] The reverse heavy overload identification module is used to input the feature data collected in real time into a pre-trained strong classifier and output the identification result of the reverse heavy overload in the photovoltaic area. The strong classifier is trained based on an integrated learning algorithm using a historical data set constructed with different feature data that affect the reverse load rate as training samples.

[0110] The different characteristic data in the characteristic data selection module are transformer load, electric vehicle charging load, photovoltaic power and transformer capacity.

[0111] The device also includes a strong classifier training module, specifically including:

[0112] The data preprocessing unit performs data preprocessing on different feature data through quartile method and maximum and minimum normalization, and then constructs a data set, and divides the data set into a training set and a test set;

[0113] The model building unit builds a fault identification model based on AdaBoost. The input is the feature data in the training set, and the output is {-1, 1}, where -1 indicates that it belongs to the photovoltaic area reverse heavy overload, and 1 indicates that it does not belong to the photovoltaic area reverse heavy overload. The algorithm of the weak classifier in the fault identification model adopts the gradient boosting tree, the number of iterations of the weak classifier is K, and a strong classifier f(x) is weighted integrated;

[0114] Model training unit, training fault identification model based on AdaBoost, including:

[0115] Initialize the weight of the training set, as shown in formula (2):

[0116]

[0117] In the formula, w 1i represents the initial weight of each load feature in the training set, and m represents the number of samples in the training set;

[0118] Use the weight D k The training set is used to train different feature data to obtain the weak classifier G k (x):x→{-1,1}, k=1,2,…,K, calculate G k The classification error rate of (x) is shown in formula (3):

[0119]

[0120] In the formula, y i represents the sample label set {-1, 1}, w ki represents the weight of the kth basic classifier with the lowest error rate at present, and I represents the weight of the kth basic classifier with the lowest error rate at present. k (x i )The weight of misclassified samples;

[0121] Calculate the weight coefficient of the weak classifier, as shown in formula (4):

[0122]

[0123] In the formula, R is the number of categories;

[0124] Update the weight distribution of the training set, as shown in formula (5):

[0125] w k+1,i =w ki Z K exp(-α k y i G k (x i )),i=1,2,…,m (5)

[0126] In the formula, Z K is the normalization factor, as shown in formula (6):

[0127]

[0128] The weighted integration produces a strong classifier f(x), as shown in formula (7):

[0129]

[0130] The result evaluation unit uses accuracy, recall, precision and F1 score to evaluate different feature data of the test set.

[0131] In addition, an embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed, the above-mentioned photovoltaic station reverse heavy overload identification method based on the integrated learner is implemented.

[0132] An embodiment of the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the photovoltaic station reverse heavy overload identification method based on the integrated learner is implemented.

[0133] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The schemes in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0134] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0137] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0138] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0139] Example 2

[0140] The data of photovoltaic distribution transformers in a certain reverse overload area in Hebei Province is used as experimental data. It contains the transformer load, capacity, photovoltaic power in the area and EV charging load data of photovoltaic reverse overload events that occurred from January to July 2023, and the sampling interval is 1h.

[0141] All load data features are normalized and then input into various classifiers for training. The input vector dimension of each classifier is the same as the dimension of the input load feature data, that is, 4-dimensional input features, but the output dimension is 1-dimensional. Train each classifier, output the features belonging to photovoltaic reverse heavy overload as -1, and output the features that do not belong to photovoltaic reverse heavy overload as 1. Through error weight allocation, a strong classifier is finally iterated, and the output transformer operation status is 1, indicating that the photovoltaic equipment is operating normally at this time; the output transformer operation status is -1, indicating that the photovoltaic equipment is reversely overloaded at this time. 80% of the data set is set as the training set, and 20% is set as the test set. The above-mentioned evaluation indicators are used to evaluate the test data results. The reverse overload identification information of the method of this embodiment is as follows. Figure 4 shown.

[0142] from Figure 4 It can be seen that the fault identification method based on AdaBoost is effective. In the 400 test sets, there are actually 60 reverse overload sample points, and the method in this section identifies 43 reverse overload sample points. The method in this section can effectively identify the reverse overload of the transformer.

[0143] The recognition effect is evaluated by calculating four indicators such as accuracy. The AdaBoost-based fault recognition algorithm is compared with the random forest algorithm, which represents the bagging algorithm. The test results of four indexes such as F1-score are shown in Table 1.

[0144] Table 1 Comparison of evaluation indicators

[0145]

[0146] As can be seen from Table 1, the AdaBoost method has a high accuracy rate in identifying photovoltaic reverse overload. The F1-score is 7.6% higher than the random forest algorithm representing bagging. The recall rate, accuracy rate and precision rate are significantly higher than the random forest algorithm. This is because this method can automatically select the features with the greatest classification impact, so as to better identify the fault state, and this method can adjust the weight and strength of the weak classifier according to the classification results of each iteration, so that the model is more suitable for the reverse overload situation. The random forest algorithm is composed of multiple decision trees, and its adaptability to some special situations is relatively weak. On the contrary, the error feedback mechanism of the AdaBoost method can continuously adjust the weak classifier during the training process, so that the model can accurately distinguish outliers, thereby improving the accuracy and reliability of fault identification.

Claims

1. A method for identifying reverse heavy overload in photovoltaic areas based on an integrated learner, characterized in that: The following steps are involved: Step 1, select different characteristic data affecting the reverse load rate and collect them in real time; Step 2, input the feature data collected in real time in step 1 into a pre-trained strong classifier, and output the recognition result of the reverse heavy overload in the photovoltaic area. The strong classifier is trained based on an integrated learning algorithm using a historical data set constructed with different feature data that affect the reverse load rate as training samples.

2. According to claim 1, a photovoltaic area reverse heavy overload identification method based on an integrated learner is characterized in that: The calculation formula of the reverse load rate in step 1 is shown in formula (1): Where P d Expressed as photovoltaic power generation, P L It is expressed as the transformer equivalent load at the same time, S D Expressed as the rated capacity of the transformer.

3. The photovoltaic area reverse overload identification method based on an integrated learner according to claim 1 is characterized in that: The different characteristic data in step 1 are transformer load, electric vehicle charging load, photovoltaic power and transformer capacity.

4. The photovoltaic area reverse overload identification method based on an integrated learner according to claim 1 is characterized in that: The method further includes: training the strong classifier, specifically: (1) Preprocess different feature data by quartile method and maximum and minimum normalization, then construct a data set and divide the data set into a training set and a test set; (2) Construct a fault identification model based on AdaBoost. The input is the feature data in the training set, and the output is {-1, 1}, where -1 indicates that it belongs to the photovoltaic area reverse overload, and 1 indicates that it does not belong to the photovoltaic area reverse overload. The algorithm of the weak classifier in the fault identification model adopts the gradient boosting tree, the number of iterations of the weak classifier is K, and the final strong classifier f(x) is weighted integrated; (3) Initialize the weight of the training set, as shown in formula (2): In the formula, w 1i represents the initial weight of each load feature in the training set, and m represents the number of samples in the training set; (4) Using weight D k The training set is used to train different feature data to obtain the weak classifier G k (x):x→{-1,1}, k=1,2,…,K, calculate G k The classification error rate of (x) is shown in formula (3): In the formula, y i represents the sample label set {-1, 1}, w ki represents the weight of the kth basic classifier with the lowest error rate at present, and I represents the weight of the kth basic classifier with the lowest error rate at present. k (x i )The weight of misclassified samples; (5) Calculate the weight coefficient of the weak classifier, as shown in formula (4): In the formula, R is the number of categories; (6) Update the weight distribution of the training set, as shown in formula (5): w k+1,i =w ki Z K exp(-α k y i G k (x i )),i=1,2,…,m (5) In the formula, Z K is the normalization factor, as shown in formula (6): (7) The final classifier f(x) of the weighted ensemble is shown in formula (7): (8) Use accuracy, recall, precision and F1 score to evaluate different feature data of the test set.

5. A photovoltaic area reverse heavy overload identification device based on an integrated learner, characterized in that: include: A characteristic data selection module is used to select different characteristic data that affect the reverse load rate and collect them in real time; The reverse heavy overload identification module is used to input the feature data collected in real time into a pre-trained strong classifier and output the identification result of the reverse heavy overload in the photovoltaic area. The strong classifier is trained based on an integrated learning algorithm using a historical data set constructed with different feature data that affect the reverse load rate as training samples.

6. A photovoltaic area reverse heavy overload identification device based on an integrated learner according to claim 5, characterized in that: The different characteristic data are transformer load, electric vehicle charging load, photovoltaic power and transformer capacity.

7. A photovoltaic area reverse heavy overload identification device based on an integrated learner according to claim 5, characterized in that: The device also includes a strong classifier training module, specifically including: The data preprocessing unit performs data preprocessing on different feature data through quartile method and maximum and minimum normalization, and then constructs a data set, and divides the data set into a training set and a test set; The model building unit builds a fault identification model based on AdaBoost. The input is the feature data in the training set, and the output is {-1, 1}, where -1 indicates that it belongs to the photovoltaic area reverse heavy overload, and 1 indicates that it does not belong to the photovoltaic area reverse heavy overload. The algorithm of the weak classifier in the fault identification model adopts the gradient boosting tree, the number of iterations of the weak classifier is K, and a strong classifier f(x) is weighted integrated; Model training unit, training fault identification model based on AdaBoost; The result evaluation unit uses accuracy, recall, precision and F1 score to evaluate different feature data of the test set.

8. A photovoltaic area reverse heavy overload identification device based on an integrated learner according to claim 7, characterized in that: The process of training the fault identification model based on AdaBoost includes: Initialize the weight of the training set, as shown in formula (2): In the formula, w 1i represents the initial weight of each load feature in the training set, and m represents the number of samples in the training set; Use the weight D k The training set is used to train different feature data to obtain the weak classifier G k (x):x→{-1,1}, k=1,2,…,K, calculate G k The classification error rate of (x) is shown in formula (3): In the formula, y i represents the sample label set {-1, 1}, w ki represents the weight of the kth basic classifier with the lowest error rate at present, and I represents the weight of the kth basic classifier with the lowest error rate at present. k (x i )The weight of misclassified samples; Calculate the weight coefficient of the weak classifier, as shown in formula (4): In the formula, R is the number of categories; Update the weight distribution of the training set, as shown in formula (5): w k+1,i =w ki Z K exp(-α k y i G k (x i )),i=1,2,…,m (5) In the formula, Z K is the normalization factor, as shown in formula (6): The final classifier f(x) of the weighted ensemble is shown in formula (7):

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the photovoltaic station reverse heavy overload identification method based on the integrated learner according to any one of claims 1 to 4 is implemented.

10. An electronic device, characterized in that: It includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the photovoltaic station area reverse heavy overload identification method based on an integrated learner as claimed in any one of claims 1 to 4 is implemented.

Images