Distributed photovoltaic private capacity increase abnormity monitoring method and monitoring platform

Abstract

The invention relates to the technical field of photovoltaic power generation monitoring, and provides a distributed photovoltaic private capacity increasing abnormity monitoring method and platform, and the method comprises the following steps: obtaining the basic information and power generation data of a photovoltaic power generation equipment terminal; constructing a layer based on an SVM (Support Vector Machine) model for each factor of photovoltaic power generation, forming a multi-layer superposed structure, and obtaining a multi-layer superposed distributed photovoltaic generating capacity anomaly recognition algorithm model based on the SVM model; according to the obtained basic information and power generation data, a classification result is obtained through judgment of each layer based on an SVM model; adjusting the sample weight of each layer according to the classification performance of each layer of SVM model; and according to the obtained sample weight, carrying out weighted voting on classification results of all layers to obtain a final classification result. The theoretical generating capacity of the target user is determined by considering various factors, and whether the capacity of the client is increased privately is accurately judged.

Description

Technical Field

[0001] The present disclosure relates to the technical field related to photovoltaic power generation monitoring, and more specifically, to a distributed photovoltaic unauthorized capacity increase abnormal monitoring method and monitoring platform. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] In recent years, a large number of distributed clean energy sources have been connected to the power grid. Among them, distributed photovoltaics have the characteristics of clean and efficient, decentralized layout, easy installation and maintenance, etc., occupying an advantageous position among many distributed power sources, and have maintained a high-speed growth trend in recent years. After the distributed photovoltaics are connected to the grid, customers mostly adopt the operation mode of "self-generation and self-use, surplus power to the grid". Customers can obtain benefits through more power generation, so some customers will generate more power by privately increasing capacity. However, the output of distributed photovoltaics has typical time fluctuation characteristics, which will have an adverse impact on the voltage, frequency, and flow control of the normal operation of the power grid. The larger the access capacity, the more significant the impact, and even affect the safe and stable operation of the entire power grid, thereby threatening the normal power consumption of other customers of the power grid. Therefore, the access of distributed photovoltaics should fully consider the carrying capacity of the power grid, and distributed photovoltaics need to be connected according to the developable capacity of the power grid. Once the distributed photovoltaic customers privately increase the capacity to connect to the power grid, it will cause a serious impact on the power grid if it exceeds the carrying capacity of the power grid. Therefore, it is necessary to orderly control the access capacity of distributed photovoltaics.

[0004] At present, there are very few monitoring methods specifically for the abnormal capacity increase of distributed photovoltaic customers without permission, and the abnormality judgment method considers a single dimension. The current abnormality judgment method is to compare the unit capacity power generation of the target customer with the average unit capacity power generation of regional customers; or to consider the two factors of radiation and temperature to determine the theoretical power generation, and compare the actual power generation with the theoretical power generation. Both methods have problems such as simple considerations and single dimensions. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a distributed photovoltaic unauthorized capacity increase abnormal monitoring method and monitoring platform, which considers multiple factors to determine the theoretical power generation of the target user, accurately judges whether the customer has increased the capacity privately, and builds a corresponding monitoring platform for abnormal display and full-process control.

[0006] In order to achieve the above objectives, the present disclosure adopts the following technical solutions:

[0007] One or more embodiments provide a method for monitoring abnormalities of unauthorized capacity increase in distributed photovoltaic power generation, including the following steps:

[0008] Obtain basic information and power generation data of photovoltaic power generation equipment terminals;

[0009] A layer based on the SVM model is constructed for each factor of photovoltaic power generation to form a multi-layer superposition structure, and a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model is obtained;

[0010] According to the basic information and power generation data obtained, a classification result is obtained through each layer based on the SVM model;

[0011] Adjust the sample weight of each layer according to the classification performance of each layer of SVM model;

[0012] According to the obtained sample weights, the classification results of all layers are weighted voted to obtain the final classification result.

[0013] One or more embodiments provide a distributed photovoltaic unauthorized capacity increase abnormality monitoring platform, including:

[0014] Data acquisition module: configured to acquire basic information and power generation data of the photovoltaic power generation equipment terminal;

[0015] Construction module: configured to construct a layer based on the SVM model for each factor of photovoltaic power generation, forming a multi-layer superposition structure, and obtaining a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model;

[0016] Single-layer classification module: It is configured to obtain a classification result based on the SVM model through each layer according to the basic information and power generation data obtained;

[0017] Weight update module: configured to adjust the sample weight of each layer according to the classification performance of each layer of SVM model;

[0018] Weighted voting module: It is configured to obtain the final classification result by weighted voting on the classification results of all layers according to the obtained sample weights.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The abnormal monitoring method disclosed in the present invention takes into account multiple factors of photovoltaic power generation and makes judgments layer by layer through a multi-layer superposition structure, which can more comprehensively and accurately detect the abnormal capacity increase of distributed photovoltaic power generation. Compared with the traditional single factor judgment method, this embodiment takes into account multiple dimensional factors, making the judgment result more reliable and effectively improving the recognition rate of abnormal capacity increase.

[0021] In addition, by superimposing the SVM model layers and combining the weighted voting mechanism, the classification results of each layer are integrated into the final judgment, thus constructing a strong classifier. This combined classification method can balance the judgment results of each layer, reduce the probability of misjudgment due to a single factor, enhance the accuracy and robustness of the monitoring results, ensure the healthy development of new energy grid connection, and help improve the safety and stability of the power system.

[0022] The advantages of the present disclosure and the advantages of additional aspects will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The exemplary embodiments of the present disclosure and the description thereof are used to explain the present disclosure but do not constitute a limitation of the present disclosure.

[0024] Figure 1 It is a flow chart of the method for monitoring abnormal photovoltaic capacity increase without authorization according to Embodiment 1 of the present disclosure;

[0025] Figure 2 is a schematic diagram of the model training process of Embodiment 1 of the present disclosure;

[0026] Figure 3 It is the photovoltaic unauthorized capacity increase abnormal monitoring platform interface of the embodiment 1 of the present disclosure; DETAILED DESCRIPTION

[0027] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0028] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0029] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments in the present disclosure and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0030] Example 1

[0031] In the technical solutions disclosed in one or more embodiments, Figures 1 to 3As shown, a method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation includes the following steps:

[0032] Step 1: Obtain basic information and power generation data of the photovoltaic power generation equipment terminal;

[0033] Step 2: construct a layer based on the SVM model for each factor of photovoltaic power generation, form a multi-layer superposition structure, and obtain a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model;

[0034] Step 3: According to the basic information and power generation data obtained, a classification result is obtained through each layer based on the SVM model;

[0035] Step 4: Adjust the sample weight of each layer according to the classification performance of each layer of SVM model;

[0036] Step 5: According to the obtained sample weights, the classification results of all layers are weighted voted to obtain the final classification result;

[0037] In this embodiment, based on the basic information of the photovoltaic power generation equipment terminal (i.e., the client) and the power generation collection information data, comprehensive consideration is given to multiple factors such as the phased changes in the global power generation characteristic values, regional characteristics, lighting conditions, substation conditions, registered capacity, power generation, and collection power, to form a multi-layer superimposed abnormal three-dimensional inspection rule. Each layer generates a judgment result by constructing an SVM model. The judgment of each layer has a different weight in the final judgment. The AdaBoost technology is used to form the final strong classifier in the form of weighted voting, and this rule is used to perform capacity increase anomaly screening.

[0038] The abnormal monitoring method of this embodiment takes into account multiple factors of photovoltaic power generation and makes judgments layer by layer through a multi-layer superposition structure, which can more comprehensively and accurately detect abnormal capacity expansion of distributed photovoltaic power generation. Compared with the traditional single factor judgment method, this embodiment takes into account multiple dimensional factors, making the judgment result more reliable and effectively improving the recognition rate of abnormal capacity expansion.

[0039] In addition, this embodiment integrates the classification results of each layer into the final judgment by superimposing the SVM model layers and combining the AdaBoost weighted voting mechanism, thereby constructing a strong classifier. This combined classification method can balance the judgment results of each layer, reduce the probability of misjudgment due to a single factor, enhance the accuracy and robustness of the monitoring results, ensure the healthy development of new energy grid connection, and help improve the safety and stability of the power system.

[0040] In step 1, the power generation data of the photovoltaic power generation equipment terminal can collect the power generation load curve data of the photovoltaic client, mainly including: power, voltage, current, power factor, electric energy indication, etc.;

[0041] The basic information of photovoltaic power generation equipment terminals can be obtained through the basic file data in the marketing system, mainly including: reported capacity, installation date, settlement method, wiring method, metering method, etc.

[0042] In addition, the model training process also includes obtaining regional meteorological data, including: weather conditions, outdoor temperature, humidity, wind speed, etc.; as well as data on regional natural resources;

[0043] In some embodiments, the factors of photovoltaic power generation are specifically factors that may affect the power generation, including multiple factors such as changes in global power generation characteristic values, regional characteristics, lighting conditions, station area conditions, and collection power;

[0044] In view of the above factors, a global power generation characteristic value layer, a regional characteristic layer, a light condition layer, a substation situation layer, and a collection power layer are constructed in a layer-by-layer stacking manner. A trained SVM model is set for each layer to determine whether abnormal capacity increase has occurred based on the acquired data. Each layer outputs a classification result, and preliminary abnormal screening of photovoltaic power generation capacity increase is achieved based on each layer.

[0045] In order to obtain the classification model of photovoltaic data of each layer, this embodiment uses a mature SVM method with strong generalization ability to construct a classifier for the current layer. The construction method of the multi-layer superimposed distributed photovoltaic power generation anomaly recognition algorithm model (CVSVM model) based on the SVM model includes the following steps:

[0046] Step 21: Get training data samples, train the SVM model of each layer, identify the hyperplane, and obtain the classification results of each layer:

[0047] Given a mutually independent linear sample S = {(x1,y1),...,(x n ,y n )}, we can know the classification function:

[0048] f(x)=w T x+b

[0049] Among them, x i is the characteristic vector of distributed photovoltaic data, y i is the distributed photovoltaic category. w is the vector of the hyperplane, and the offset is b. After finding the photovoltaic data with the minimum interval, the interval needs to be maximized, as shown in the following formula:

[0050]

[0051] In reality, the more complex distributed photovoltaic classification is almost impossible to be completely linearly separable, and the slack variable constraint condition can be obtained:

[0052]

[0053] As long as all α are obtained, the photovoltaic data classification hyperplane of the current layer can be represented by α. In this embodiment, the SVM algorithm CvSVM (support vector machine) is used to calculate these α values, thereby obtaining the classification result;

[0054] Step 22, calculate error: evaluate the classification effect of the SVM model of the current layer on the training data and calculate its classification error rate;

[0055] Step 23: Update sample weight β i : Adjust the sample weight β according to the classification results of the SVM model in the current layer i , set higher weights for the misclassified samples so that they will be given more attention in the next round of repeated training of the SVM model; after the total weight change of all samples is less than the threshold X, the repeated training of the SVM model of this layer ends;

[0056] Step 24: Determine the classification weight a of each layer t :According to the classification error rate of each layer, determine the weight a of the SVM model trained on this layer in the entire boosting ensemble t .

[0057] Step 25, all trained SVM models are combined in a weighted voting manner to form a final strong classifier H(x), that is, a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model is obtained; photovoltaic abnormal capacity increase households can be determined by classification;

[0058]

[0059] Among them, a t represents the weight of layer t, h t (x) represents the classification result output by the corresponding layer t;

[0060] Furthermore, in step 21, the SVM model of each layer is trained to identify the training method of the hyperplane, such as Figure 2 As shown, the following steps are included:

[0061] Step 211, data acquisition: acquiring multi-dimensional data from the photovoltaic client and the power grid, and performing data preprocessing;

[0062] Data preprocessing can include data cleaning, feature extraction and standardization;

[0063] Data cleaning: Clean the collected data to remove noise and outliers. After processing, ensure the accuracy and consistency of the data.

[0064] Feature extraction and standardization: Extract key features from the cleaned data, perform dimensionality reduction and data standardization to facilitate subsequent model training.

[0065] Step 212: Use data sets of different layers to train the SVM model of each layer, and adjust the parameters of the model (such as kernel function, penalty parameter C, kernel parameter γ, etc.) to optimize the model performance;

[0066] Step 213, model evaluation: determining the optimal hyperparameters of a single SVM model through cross-validation and accuracy evaluation;

[0067] Based on the trained single SVM model, the SVM-based Boosting model consisting of multiple layers is trained and evaluated as a whole;

[0068] The layers based on the SVM model are described below;

[0069] (1) The global power generation characteristic value layer is configured to perform the following process:

[0070] Step 1-1, the global power generation and installed capacity within the statistical period are calculated to calculate the global theoretical power generation time characteristic value;

[0071] The global theoretical power generation time characteristic value represents the expected power generation time per unit capacity of distributed photovoltaic devices in the entire network under current meteorological conditions, which is used for subsequent abnormal monitoring.

[0072] Step 1-2: Calculate the theoretical power generation of each customer based on the global theoretical power generation time characteristic value and the registered capacity of each customer:

[0073] Theoretical power generation = global theoretical power generation time characteristic value × customer registered capacity;

[0074] Step 1-3: Compare the theoretical power generation with the actual power generation of the client to determine whether there is suspected unauthorized capacity increase. If the actual power generation is significantly higher than the theoretical value and exceeds the reasonable error range, it is determined to be suspected unauthorized capacity increase; otherwise, it is determined to be no capacity increase;

[0075] Step 1-4: The customer's registered capacity, customer's regional information, and actual power generation, etc., which are determined to be suspected of unauthorized capacity increase, are used as input data and input into the SVM model of the corresponding layer. The SVM model trained with historical data further classifies and distinguishes the abnormal data, and outputs the classification result of the SVM model to obtain classification result 1;

[0076] The classification result of the global power generation characteristic value layer is used as the base classifier of the SVM-based Boosting algorithm and input into the strong classifier to participate in weighted voting.

[0077] (2) Regional characteristics layer, configured to perform the following process:

[0078] Step 2-1: Divide the regions according to administrative districts and obtain the natural resource conditions of each region;

[0079] Specifically, the regions are divided into categories such as urban areas, mountainous areas, rural areas, etc. The natural resource conditions of each region, such as light intensity and sunshine duration, vary due to geographical location;

[0080] Step 2-2: Count the photovoltaic power generation characteristic data of each sub-region, including lighting conditions, installation density, etc., to determine the typical value of photovoltaic power generation in each sub-region.

[0081] Collect the registered capacity and actual power generation of each PV client, which will be used for abnormality identification later.

[0082] Step 2-3: Calculate the theoretical power generation in each region based on the photovoltaic power generation characteristic data and registered capacity:

[0083] Theoretical power generation = regional photovoltaic power generation characteristics (light conditions, sunshine duration, etc.) × customer registered capacity;

[0084] Step 2-4: The calculated theoretical power generation is used as the customer's benchmark power generation and compared with the actual power generation of the customer's meter reading. If the customer's actual power generation is significantly higher than the theoretical power generation and exceeds the reasonable error range, it can be marked as a suspected private capacity increase customer; otherwise, it is determined that the capacity has not been increased;

[0085] Step 2-5: The natural resource conditions, installation density, registered capacity and actual power generation of the above-mentioned customers suspected of privately increasing capacity are input into the SVM model trained with historical data for classification, and the output classification result 2 of the SVM model of the global power generation characteristic value layer is output;

[0086] In this embodiment, by taking the regional characteristics as a reference, abnormal power generation behaviors in different regions can be detected more accurately, avoiding misjudgments caused by differences in natural conditions.

[0087] The second classification result of the regional characteristic layer is used as the second base classifier of the SVM-based Boosting algorithm and input into the strong classifier to participate in weighted voting.

[0088] (3) The lighting condition layer is configured to perform the following process:

[0089] Step 3-1: Collect the lighting condition data of each month in the previous year, including sunshine duration, irradiance, etc., and analyze the correlation with the historical power generation to determine the impact of lighting conditions on power generation;

[0090] Step 3-2, collect and analyze the real-time light condition data of the month, and compare it with the data of the same period of the previous year to identify changes in light conditions;

[0091] Step 3-3: Calculate the theoretical power generation of the analysis month based on the lighting conditions and the customer's registered capacity information; the calculation formula can be:

[0092] Theoretical power generation = monthly light condition index × registered capacity;

[0093] The light conditions index is adjusted based on the light levels of the same period last year and the year-on-year change in light conditions this year to adjust the power generation expectations for the current month.

[0094] Step 3-4: According to the theoretical power generation and the actual power generation of the customer, mark the abnormal customers corresponding to the abnormal data, input the sunshine duration, irradiance, historical power generation, recorded capacity and actual power generation data of the marked abnormal customers into the SVM model of the light condition layer, and use the SVM model trained with historical data to classify, further determine whether there is unauthorized capacity increase, and obtain classification result three;

[0095] Classification result output: SVM outputs a classification label of "normal" or "abnormal", and customers judged to be abnormal are considered suspected of capacity increase.

[0096] The classification result of the illumination condition layer will be used as the base classifier 3 of the SVM-based Boosting algorithm and input into the strong classifier to participate in weighted voting.

[0097] (4) Collecting power layers, configured to perform the following process:

[0098] Step 4-1: For each distributed photovoltaic client, collect 24 or 96 points of power data (i.e., power values ​​every hour or every 15 minutes) to reflect the actual power output of the client at each time period during the day;

[0099] Step 4-2, compare the collected power values ​​of each time period with the registered capacity one by one;

[0100] Step 4-2: When the power value in any period exceeds the registered capacity, it is initially marked as suspected capacity increase;

[0101] For example, if a customer's registered capacity is 10kW, but the power reaches 12kW during a certain period of time, it will be recorded as an abnormal value.

[0102] Step 4-3: For the suspected capacity increase client, the power anomaly point ratio, registered capacity deviation, power fluctuation trend and other features marked as the suspected capacity increase client are counted and input into the SVM model of the corresponding layer, and the SVM classifier trained with historical data performs classification, and the SVM outputs the classification result to obtain the classification result 4;

[0103] Abnormal point statistics: Count the number and proportion of times the power exceeds the registered capacity in a day to capture the frequency and abnormal amplitude of power fluctuations. For example, record the total number of time periods when the power exceeds the registered capacity;

[0104] Recorded capacity deviation: record the difference between the actual power peak and the recorded capacity, calculate the average deviation or maximum deviation, and input it into the SVM model as a feature.

[0105] Power fluctuation trend: Analyze the power fluctuation trend in each period and extract other power-related features, such as the average power during the day and the power distribution variance, to further characterize the abnormal characteristics of power exceeding the registered capacity;

[0106] Based on the labeled historical data (normal customers and customers who are known to have increased capacity without permission), as well as the power anomaly ratio, registered capacity deviation, and power fluctuation trend feature data as input, the SVM classifier is trained with the output of whether to increase capacity. By adjusting the hyperparameters, it is ensured that the model can reduce misjudgment while capturing the overcapacity power data.

[0107] The classification result 4 of the power layer will be used as the base classifier 4 of the SVM-basedBoosting algorithm and input into the strong classifier to participate in weighted voting.

[0108] (5) Area situation layer

[0109] Step 5-1: Mark the client that is suspected of increasing capacity without permission, based on the situation where the customer's power exceeds the registered capacity.

[0110] Step 5-2: When the number of clients suspected of unauthorized capacity increase in the same area exceeds the set number, the voltage, current and other operating parameters of the power grid in the area are obtained;

[0111] Step 5-3: The power generation deviation between the suspected unauthorized capacity increase client and the client in the same substation area, the proportion of abnormal customers in the substation area, and grid parameters such as voltage and current are input as features into the SVM model, and the label of "unauthorized capacity increase" or "substation area abnormality" is output;

[0112] Step 5-4: When the output result is "unauthorized capacity increase", the customer is confirmed as a suspected target of unauthorized capacity increase; if it is "station area abnormality", the customer is marked as a station area abnormality exclusion target and removed from the capacity increase suspicion; this is the classification result five;

[0113] The SVM model of this layer is trained using the known "unauthorized capacity increase" and "station-specific abnormality" customer data in historical data; the SVM model input data is characteristic values ​​such as customer power generation deviation, proportion of abnormal customers in the station area, grid voltage and current anomalies, and outputs the "unauthorized capacity increase" or "station-specific abnormality" label as classification result five.

[0114] In this layer, when it is preliminarily determined that a distributed photovoltaic customer has privately increased the capacity on site, resulting in the equipment generating overcapacity, the customer's power generation will be compared with the power generation of distributed photovoltaic customers in the same area. If a large number of customers in the same area have abnormal power generation, it is possible that the abnormal data such as the voltage and current of the power grid in the area caused the power generation of customers in the entire area to deviate from the normal value, that is, the area is abnormal. In this case, the customer can be excluded from the suspected abnormal household and simultaneously included in the key monitoring pool for subsequent observation and analysis of the operation situation.

[0115] The classification result 5 of the substation situation layer will be used as the base classifier 5 of the SVM-basedBoosting algorithm and input into the strong classifier to participate in the weighted voting. This layer has the highest weight and can veto the overcapacity suspicion.

[0116] This embodiment adds a "substation situation layer" judgment and screening step in the abnormal judgment process, which can avoid the misjudgment of abnormal distributed photovoltaic customers' private capacity increase due to abnormal data such as the substation grid voltage and current, which causes the power generation of customers in the entire substation to deviate from the normal value, and can reduce unnecessary waste of manpower and material resources for on-site inspections.

[0117] In step 4, the classification weight of each layer is adjusted according to the classification performance of each layer of the SVM model; specifically: the weight of each layer is set in turn according to the classification accuracy of each layer, and the higher the classification accuracy, the higher the weight is assigned.

[0118] Finally, the classification results of each layer are weighted and voted with the corresponding weights to obtain the final classification result.

[0119] Example 2

[0120] Based on Example 1, this embodiment provides a distributed photovoltaic unauthorized capacity increase abnormal monitoring platform, including:

[0121] Data acquisition module: configured to acquire basic information and power generation data of the photovoltaic power generation equipment terminal;

[0122] Construction module: configured to construct a layer based on the SVM model for each factor of photovoltaic power generation, forming a multi-layer superposition structure, and obtaining a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model;

[0123] Single-layer classification module: It is configured to obtain a classification result based on the SVM model through each layer according to the basic information and power generation data obtained;

[0124] Weight update module: configured to adjust the sample weight of each layer according to the classification performance of each layer of SVM model;

[0125] Weighted voting module: It is configured to obtain the final classification result by weighted voting on the classification results of all layers according to the obtained sample weights;

[0126] It should be noted here that the above-mentioned modules in this embodiment correspond one-to-one to the steps in Example 1, and the specific implementation process is the same, which will not be repeated here.

[0127] Furthermore, it also includes a visualization module for visually displaying the abnormal monitoring results of distributed photovoltaic customers' unauthorized capacity increase in the form of multi-dimensional charts, map distributions, statistical data, etc., so as to facilitate managers to quickly understand the abnormal distribution situation.

[0128] Through intuitive charts and maps, the monitoring results are more readable, the efficiency of abnormal data analysis is improved, and managers can quickly find and lock abnormal areas and customers. The visual display interface is as follows: Figure 2 and 2 As shown; Figure 3 The display interface of the example area is only an example, and the display interface includes a map, a pie chart, a bar chart, and a display table of regional data. It is only an example and does not limit the solution.

[0129] It is feasible to further include a work order generation module for automatically generating an abnormal work order according to the final classification result output by the weighted voting module, and sending the customer information suspected of unauthorized capacity increase to the terminal of the relevant management unit.

[0130] The work order generation module can realize the automatic generation and distribution of work orders for abnormal customers, reduce manual intervention, improve management efficiency, and ensure that abnormal customers can be handled in a timely manner.

[0131] Optionally, a closed-loop management module can be set up to track and manage the entire process of abnormal work orders, including dispatch, verification, rectification, feedback and other processes, to ensure that each abnormal work order is effectively followed up and closed-loop managed.

[0132] This closed-loop management module uses the closed-loop management function to achieve full-chain control from problem discovery to problem solving, ensuring that measures to control unauthorized capacity increases are in place, forming a long-term and effective governance mechanism, and ensuring the healthy operation of the power grid.

[0133] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

[0134] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A method for monitoring abnormalities of unauthorized capacity increase in distributed photovoltaics, characterized in that: The steps include: Obtain basic information and power generation data of photovoltaic power generation equipment terminals; A layer based on the SVM model is constructed for each factor of photovoltaic power generation to form a multi-layer superposition structure, and a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model is obtained; According to the basic information and power generation data obtained, a classification result is obtained through each layer based on the SVM model; Adjust the sample weight of each layer according to the classification performance of each layer of SVM model; According to the obtained sample weights, the classification results of all layers are weighted voted to obtain the final classification result.

2. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation according to claim 1, characterized in that: A layer based on the SVM model is constructed for each factor of photovoltaic power generation, including the global power generation characteristic value layer, regional characteristic layer, lighting condition layer, substation condition layer and collected power layer.

3. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation as claimed in claim 2, characterized in that: The method for constructing a multi-layer overlay distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model includes the following steps: Get training data samples, train the SVM model of each layer, identify the hyperplane, and obtain the classification results of each layer: Evaluate the classification effect of the SVM model of the current layer on the training data and calculate its classification error rate; Adjust the sample weights according to the classification results of the SVM model, setting higher weights for misclassified samples; Determine the classification weight of each layer of the SVM model in this round of training according to the classification error rate; All trained SVM models are combined in a weighted voting manner to form the final strong classifier, that is, a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model is obtained.

4. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation according to claim 2, characterized in that: The global power generation characteristic value layer is configured to perform the following processes: The total power generation and installed capacity in the statistical period are calculated to calculate the characteristic value of the theoretical power generation time in the whole area; According to the global theoretical power generation time characteristic value and combined with the registered capacity of each client, the client's theoretical power generation is calculated: The theoretical power generation is compared with the actual power generation of the client to determine whether there is suspected unauthorized capacity increase. Those exceeding the set threshold are marked as abnormal, and the corresponding SVM model is used to further identify the abnormal data to obtain classification result one.

5. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation as claimed in claim 2, characterized in that: The regional characteristics layer is configured to perform the following processes: Divide the regions according to administrative divisions and obtain the natural resource conditions of each region; Collect statistics on photovoltaic power generation characteristics in each sub-region; Calculate theoretical power generation in different regions based on photovoltaic power generation characteristic data and registered capacity; The calculated theoretical power generation is used as the customer's benchmark power generation, which is compared with the client's actual meter reading power generation and marked as a customer suspected of privately increasing capacity. The corresponding SVM model is then used to classify the data of the suspected privately increased capacity customers in each area, and the classification result 2 is output.

6. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation as claimed in claim 2, characterized in that: The lighting condition layer is configured to perform the following process: Collect and analyze real-time light condition data for the month and compare it with the data from the same period of the previous year to identify changes in light conditions; Calculate and analyze the theoretical power generation of the month based on the lighting conditions and the customer's registered capacity information; According to the theoretical power generation and actual power generation, abnormal data are marked, and the marked abnormal data are input into the SVM model of the corresponding light condition layer. Combining the light condition data and power generation deviation, it is further determined whether there is unauthorized capacity increase, and the classification result three is obtained.

7. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation as claimed in claim 2, characterized in that: The power layer is collected and configured to perform the following processes: Collect power data for each distributed photovoltaic client; Compare the collected power values ​​of each time period with the registered capacity one by one; When the power value in any period exceeds the registered capacity, it is initially marked as suspected capacity increase; For the suspected capacity-increased clients, the features of the power anomaly point ratio, registered capacity deviation, and power fluctuation trend marked as suspected capacity-increased clients are statistically input into the SVM model of the collected power layer to obtain classification result four.

8. A method for monitoring abnormalities of unauthorized capacity increase of distributed photovoltaic power generation as claimed in claim 2, characterized in that: The station situation layer is configured to perform the following processes: According to the situation where the customer's power exceeds the registered capacity, mark the client suspected of increasing the capacity privately; When the number of clients suspected of unauthorized capacity increase in the same area exceeds the set number, the operation parameters of the power grid in the area are obtained; The power generation deviation between the suspected unauthorized capacity increase client and the client in the same substation area, the proportion of abnormal customers in the substation area, and the grid parameters of voltage and current are input as features into the SVM model to identify whether it is unauthorized capacity increase or substation abnormality. When the output result is unauthorized capacity increase, the customer is confirmed as a suspected target of unauthorized capacity increase; if it is a station-specific abnormality, the customer is marked as an excluded target of station-specific abnormality and removed from the list of suspected capacity increase; this is taken as classification result five.

9. A distributed photovoltaic unauthorized capacity increase abnormal monitoring platform, characterized in that: include: Data acquisition module: configured to acquire basic information and power generation data of the photovoltaic power generation equipment terminal; Construction module: configured to construct a layer based on the SVM model for each factor of photovoltaic power generation, forming a multi-layer superposition structure, and obtaining a multi-layer superposition distributed photovoltaic power generation anomaly recognition algorithm model based on the SVM model; Single-layer classification module: It is configured to obtain a classification result based on the SVM model through each layer according to the basic information and power generation data obtained; Weight update module: configured to adjust the sample weight of each layer according to the classification performance of each layer of SVM model; Weighted voting module: It is configured to obtain the final classification result by weighted voting on the classification results of all layers according to the obtained sample weights.

10. A distributed photovoltaic unauthorized capacity increase abnormal monitoring platform as claimed in claim 9, characterized in that: It also includes a visualization module for visually displaying the abnormal monitoring results of distributed photovoltaic customers' unauthorized capacity increase in the form of multi-dimensional charts, map distribution, and statistical data; Alternatively, it also includes a work order generation module, which is used to automatically generate an abnormal work order according to the final classification result output by the weighted voting module, and send the customer information determined to be suspected of private capacity increase to the relevant terminal; Alternatively, a closed-loop management module is also included to achieve full-process tracking of abnormal work orders.

Images