A method and system for detecting anomalies in low-voltage residential photovoltaic power generation

Abstract

This invention provides a method for detecting anomalies in low-voltage residential photovoltaic (PV) power generation. The method includes acquiring low-voltage PV power generation user profile information and identifying user data for each user; calculating the electricity consumption data for each user based on their user data; establishing preliminary anomaly screening rules correlated with daily power generation and installed capacity; and filtering all PV users in the low-voltage PV user profile information by combining their user data and corresponding electricity consumption data to obtain PV users that meet the preliminary anomaly screening rules; and using a support vector machine (SVM) algorithm to perform cluster analysis on the PV users that meet the preliminary anomaly screening rules to further select anomalous PV users. This invention also provides a low-voltage residential PV power generation anomaly detection system. Implementing this invention overcomes the problems of complex operation, low efficiency, and low accuracy in existing low-voltage residential PV power generation anomaly detection methods.

Description

Technical Field

[0001] This invention relates to the field of power consumption anomaly monitoring technology, and in particular to a method and system for detecting anomalies in low-voltage residential photovoltaic power generation. Background Technology

[0002] The rapid development of new energy technologies has driven the rapid growth of photovoltaic (PV) power generation users, leading to a corresponding increase in the number of low-voltage PV power generation users. Given that the grid connection of low-voltage PV power generation users inevitably impacts the safe and reliable operation of the power grid, accurate monitoring of anomalies in low-voltage PV power generation can improve the data management capabilities for low-voltage PV grid-connected power and contribute to supporting the stable and reliable operation of the power grid.

[0003] There are several existing methods for detecting anomalies in low-voltage residential photovoltaic (PV) power generation. For example, one method calculates the ratio of PV power generation from each household in a given area to the total power generation of all residents during a specific time period on a sunny day, using this ratio as a benchmark to determine whether PV output is normal in the next time period. Another method establishes efficiency values ​​based on the PV power station's AC / DC conversion process, regional characteristics, and ambient light conditions, then compares these efficiency values ​​with actual detected efficiency values ​​to determine PV power generation anomalies. Yet another method involves on-site maintenance and repair when anomalies are detected during power generation data maintenance. However, these existing methods for detecting anomalies in low-voltage residential PV power generation are relatively complex to operate and rely on specific times or conditions to trigger anomaly monitoring, resulting in low efficiency. Furthermore, these existing methods do not consider the possibility of errors in manually entering PV user information, affecting the accuracy of anomaly monitoring results. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present invention is to provide a method and system for detecting anomalies in low-voltage residential photovoltaic power generation, which can overcome the problems of complex operation, low efficiency and low accuracy of existing methods for detecting anomalies in low-voltage residential photovoltaic power generation.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a method for detecting anomalies in low-voltage residential photovoltaic power generation, the method comprising the following steps:

[0006] Obtain low-voltage photovoltaic power generation user profile information, and determine the user data for each power generation user based on the low-voltage photovoltaic power generation user profile information; wherein, the user data includes the low-voltage photovoltaic power generation user's electricity meter relationship, photovoltaic installed capacity, user's distribution area, user's line, and user's substation;

[0007] Based on the user data of each power generation user, the electricity data of each power generation user is calculated; wherein, the electricity data includes daily power generation, daily on-grid electricity, monthly on-grid electricity and the on-grid settlement electricity of the previous month;

[0008] Establish preliminary screening rules for anomalies that are associated with daily power generation and installed capacity, and combine the user data and corresponding power data of each power generation user to screen all power generation users in the low-voltage photovoltaic user file information to obtain power generation users that meet the preliminary screening rules for anomalies.

[0009] Using a pre-defined support vector machine algorithm, cluster analysis is performed on power generation users that meet the preliminary anomaly screening rules to further select abnormal power generation users.

[0010] The preliminary screening rule for anomalies is that the daily power generation is greater than 8 times the installed capacity.

[0011] The specific steps for further selecting abnormal power generation users by using a preset support vector machine algorithm to perform cluster analysis on power generation users that meet the preliminary anomaly screening rules include:

[0012] Based on the K-means algorithm, dimensionality reduction clustering is performed on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules.

[0013] Using the aforementioned support vector machine algorithm, user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering are trained to obtain an SVM anomaly identification model for each category.

[0014] Based on the SVM anomaly identification model for each category, the power generation data of power generation users that meet the preliminary anomaly screening rules are all anomaly identified, and abnormal power generation users are further selected.

[0015] The step of performing dimensionality reduction and clustering on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules based on the K-means algorithm specifically includes:

[0016] For power generation users that meet the aforementioned preliminary anomaly screening rules, the user's distribution area, the user's line, and the user's substation are identified in the user data, and the number of categories to be clustered is set based on the identification results.

[0017] Based on daily power generation and installed capacity, the power generation data of power generation users that meet the preliminary anomaly screening rules are subjected to dimensionality reduction clustering. Then, a trained clusterer is used to perform cluster labeling on all power generation users that meet the preliminary anomaly screening rules to determine the power generation users that belong to each category after dimensionality reduction clustering.

[0018] The step of using the support vector machine algorithm to train the user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering, and obtaining the SVM anomaly identification model for each category, specifically includes:

[0019] Feature factors are expanded on the user data and corresponding electricity data of each category of power generation users, and each category of power generation users is labeled. Furthermore, a kernel mechanism is used to transform and add data dimensions for each tagged category of power generation users. The expanded feature factors include daily power generation / maximum historical 30-day power generation, daily power generation / installed capacity, and daily grid-connected electricity * number of days in the previous month / grid-connected settlement electricity in the previous month. The labels are categorized as normal or abnormal.

[0020] According to a certain proportion, all data after data dimension transformation and addition are divided into training sets and test sets corresponding to each category;

[0021] One-hot encoding is performed on the training set for each category, and all data in the training set for each category are normalized.

[0022] Initialize the SVM vector plane, and use this vector plane to detect and label all normalized data in the training set for each category;

[0023] The SVM vector loss function is defined using cross-entropy. The optimal parameters of the model are found by using grid search based on the SVM vector loss function, and an SVM anomaly recognition model for each category is constructed based on the optimal parameters.

[0024] This invention also provides a low-voltage residential photovoltaic power generation anomaly detection system, comprising:

[0025] The user data acquisition unit is used to acquire low-voltage photovoltaic power generation user profile information and determine the user data of each power generation user based on the low-voltage photovoltaic power generation user profile information; wherein, the user data includes the low-voltage photovoltaic power generation user's electricity meter relationship, photovoltaic installed capacity, user's distribution area, user's line, and user's substation.

[0026] The power data acquisition unit is used to calculate the power data of each power generation user based on the user data of each power generation user; wherein, the power data includes daily power generation, daily grid-connected power, monthly grid-connected power, and the grid-connected power settled in the previous month;

[0027] The abnormal power generation user preliminary selection unit is used to establish anomaly preliminary screening rules associated with daily power generation and installed capacity, and to screen all power generation users in the low-voltage photovoltaic user file information by combining the user data and corresponding power data of each power generation user, so as to obtain power generation users that meet the anomaly preliminary screening rules.

[0028] The abnormal power generation user final selection unit is used to perform cluster analysis on power generation users that meet the preliminary abnormal screening rules using a preset support vector machine algorithm, and further select abnormal power generation users.

[0029] The preliminary screening rule for anomalies is that the daily power generation is greater than 8 times the installed capacity.

[0030] The abnormal power generation user final selection unit includes:

[0031] The data dimensionality reduction and clustering module is used to perform dimensionality reduction and clustering on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules based on the K-means algorithm;

[0032] The SVM anomaly detection model training module is used to train the user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering using the support vector machine algorithm, so as to obtain the SVM anomaly detection model for each category.

[0033] The abnormal power generation user identification module is used to identify abnormalities in the power generation data of power generation users that meet the preliminary abnormal screening rules based on the SVM abnormal identification model of each category, and further select abnormal power generation users.

[0034] Implementing the embodiments of the present invention has the following beneficial effects:

[0035] 1. This invention uses anomaly screening rules and support vector machine algorithm, and reduces the involvement of external environmental data in the analysis based on the relationship of low-voltage photovoltaic users in the power grid architecture. It can flexibly cluster different installed capacities without the need for manual threshold determination, avoids interference from subjective factors of operators, improves the accuracy of characteristic threshold ranges, realizes the monitoring of low-voltage photovoltaic grid connection, and enhances the guarantee capability of safe and stable operation of low-voltage distribution network.

[0036] 2. Compared with existing technologies, this invention constructs anomaly preliminary screening rules based on daily power generation and installed capacity, realizes step-by-step dimensionality reduction clustering results based on grid topology, and trains an SVM anomaly identification model based on dimensionality reduction clustering. This enables step-by-step outlier analysis of low-voltage photovoltaic users and reduces the dependence of the analysis process on external information by replacing the need for external environmental data such as sunlight and temperature with the grid architecture affiliation. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, obtaining other drawings based on these drawings without creative effort still falls within the scope of the present invention.

[0038] Figure 1 A flowchart illustrating a method for detecting anomalies in low-voltage residential photovoltaic power generation, provided by an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of a low-voltage residential photovoltaic power generation anomaly detection system provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0041] like Figure 1 As shown in the figure, a method for detecting anomalies in low-voltage residential photovoltaic power generation is provided in an embodiment of the present invention. The method includes the following steps:

[0042] Step S1: Obtain low-voltage photovoltaic power generation user profile information, and determine the user data for each power generation user based on the low-voltage photovoltaic power generation user profile information; wherein, the user data includes the low-voltage photovoltaic power generation user's electricity meter relationship, photovoltaic installed capacity, user's distribution area, user's line, and user's substation.

[0043] The specific process involves obtaining low-voltage photovoltaic (PV) power generation user profile information and identifying the user data for each user within this profile. This data includes, but is not limited to, the user's electricity meter relationship, PV installed capacity, the user's transformer substation area, the user's power line, and the user's substation location. Essentially, the user data for each power generation user can be compiled into a wide table representing the low-voltage PV user profile for subsequent querying and analysis.

[0044] Step S2: Calculate the electricity data for each power generation user based on their user data; wherein the electricity data includes daily power generation, daily grid-connected electricity, monthly grid-connected electricity, and the electricity settled in the previous month.

[0045] The specific process involves querying the user data of each power generation user to calculate the daily power generation, daily on-grid power, and monthly on-grid power for each user, and then querying the on-grid settlement power for each user in the previous month according to the predefined settlement system.

[0046] Step S3: Establish preliminary screening rules for anomalies related to daily power generation and installed capacity, and combine the user data and corresponding power data of each power generation user to screen all power generation users in the low-voltage photovoltaic user file information to obtain power generation users that meet the preliminary screening rules for anomalies.

[0047] The specific process involves setting an initial screening rule for anomalies as daily power generation > 8 times the installed capacity, and then filtering out power generation users whose daily power generation > 8 times the installed capacity based on each power generation user's user data and corresponding power generation data.

[0048] Step S4: Using a preset support vector machine algorithm, perform cluster analysis on the power generation users that meet the preliminary screening rules for anomalies, and further select the abnormal power generation users.

[0049] The specific process is as follows: First, based on the K-means algorithm, dimensionality reduction clustering is performed on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules. The specific steps are as follows:

[0050] For power generation users that meet the preliminary anomaly screening rules, the user's distribution area, the user's line, and the user's substation are identified, and the number of categories to be clustered is set based on the identification results.

[0051] Based on daily power generation and installed capacity, the power generation data of power generation users that meet the initial anomaly screening criteria are subjected to dimensionality reduction clustering. A trained clusterer is then used to label all power generation users that meet the initial anomaly screening criteria to determine which power generation users belong to each category after dimensionality reduction clustering.

[0052] Secondly, using the Support Vector Machine (SVM) algorithm, the user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering are trained to obtain the SVM anomaly detection model for each category. The specific steps are as follows:

[0053] Feature factors are expanded on the user data and corresponding electricity data of each category of power generation users, and each category of power generation users is labeled. Furthermore, a kernel mechanism is used to transform and add data dimensions for each tagged category of power generation users. The expanded feature factors include daily power generation / maximum historical 30-day power generation, daily power generation / installed capacity, and daily grid-connected electricity * number of days in the previous month / grid-connected settlement electricity in the previous month. The labels are categorized as normal or abnormal.

[0054] According to a certain proportion, all data after data dimension transformation and addition are divided into training sets and test sets corresponding to each category;

[0055] One-hot encoding is performed on the training set for each category, and all data in the training set for each category are normalized.

[0056] Initialize the SVM vector plane, and use this vector plane to detect and label all normalized data in the training set for each category;

[0057] The SVM vector loss function is defined using cross-entropy. The optimal parameters of the model are found by using grid search based on the SVM vector loss function, and then an SVM anomaly recognition model for each category is constructed based on the optimal parameters.

[0058] Finally, based on the SVM anomaly detection model for each category, anomaly detection is performed on the power generation data of power generation users that meet the initial anomaly screening rules, and abnormal power generation users are further selected. The specific steps are as follows:

[0059] Anomalies in the daily power generation data of each category's test set are marked using an SVM anomaly detection model for each category. It is understood that the SVM anomaly detection model uses conventional techniques, which will not be elaborated upon here.

[0060] like Figure 2 As shown in the embodiment of the present invention, a low-voltage residential photovoltaic power generation anomaly detection system includes:

[0061] User data acquisition unit 110 is used to acquire low-voltage photovoltaic power generation user profile information and determine the user data of each power generation user based on the low-voltage photovoltaic power generation user profile information; wherein, the user data includes the low-voltage photovoltaic power generation user's electricity meter relationship, photovoltaic installed capacity, user's distribution area, user's line, and user's substation.

[0062] The power data acquisition unit 120 is used to calculate the power data of each power generation user based on the user data of each power generation user; wherein, the power data includes daily power generation, daily grid-connected power, monthly grid-connected power and the grid-connected settlement power of the previous month;

[0063] The abnormal power generation user preliminary selection unit 130 is used to establish anomaly preliminary screening rules associated with daily power generation and installed capacity, and to screen all power generation users in the low-voltage photovoltaic user file information by combining the user data and corresponding power data of each power generation user, so as to obtain power generation users that meet the anomaly preliminary screening rules.

[0064] The abnormal power generation user final selection unit 140 is used to perform cluster analysis on power generation users that meet the preliminary abnormal screening rules using a preset support vector machine algorithm, and further select abnormal power generation users.

[0065] The preliminary screening rule for anomalies is that the daily power generation is greater than 8 times the installed capacity.

[0066] The abnormal power generation user final selection unit 140 includes:

[0067] The data dimensionality reduction and clustering module is used to perform dimensionality reduction and clustering on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules based on the K-means algorithm;

[0068] The SVM anomaly detection model training module is used to train the user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering using the support vector machine algorithm, so as to obtain the SVM anomaly detection model for each category.

[0069] The abnormal power generation user identification module is used to identify abnormalities in the power generation data of power generation users that meet the preliminary abnormal screening rules based on the SVM abnormal identification model of each category, and further select abnormal power generation users.

[0070] Implementing the embodiments of the present invention has the following beneficial effects:

[0071] 1. This invention uses anomaly screening rules and support vector machine algorithm, and reduces the involvement of external environmental data in the analysis based on the relationship of low-voltage photovoltaic users in the power grid architecture. It can flexibly cluster different installed capacities without the need for manual threshold determination, avoids interference from subjective factors of operators, improves the accuracy of characteristic threshold ranges, realizes the monitoring of low-voltage photovoltaic grid connection, and enhances the guarantee capability of safe and stable operation of low-voltage distribution network.

[0072] 2. Compared with existing technologies, this invention constructs anomaly preliminary screening rules based on daily power generation and installed capacity, realizes step-by-step dimensionality reduction clustering results based on grid topology, and trains an SVM anomaly identification model based on dimensionality reduction clustering. This enables step-by-step outlier analysis of low-voltage photovoltaic users and reduces the dependence of the analysis process on external information by replacing the need for external environmental data such as sunlight and temperature with the grid architecture affiliation.

[0073] It is worth noting that in the above system embodiments, the various system units included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0074] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as ROM / RAM, disk, optical disk, etc.

[0075] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for detecting anomalies in low-voltage residential photovoltaic power generation, characterized in that, The method includes the following steps: Obtain low-voltage photovoltaic power generation user profile information, and determine the user data for each power generation user based on the low-voltage photovoltaic power generation user profile information; wherein, the user data includes the low-voltage photovoltaic power generation user's electricity meter relationship, photovoltaic installed capacity, user's distribution area, user's line, and user's substation; Based on the user data of each power generation user, the electricity data of each power generation user is calculated; wherein, the electricity data includes daily power generation, daily on-grid electricity, monthly on-grid electricity and the on-grid settlement electricity of the previous month; An anomaly screening rule is established that is associated with daily power generation and installed capacity. Combined with the user data and corresponding power generation data of each power generation user, all power generation users in the low-voltage photovoltaic user file information are screened to obtain power generation users that meet the anomaly screening rule. The anomaly screening rule is that the daily power generation is greater than 8 times the installed capacity. Using a preset support vector machine algorithm, cluster analysis is performed on power generation users that meet the preliminary anomaly screening rules to further select abnormal power generation users; The specific steps for further selecting abnormal power generation users by using a preset support vector machine algorithm to perform cluster analysis on power generation users that meet the preliminary anomaly screening rules include: Based on the K-means algorithm, dimensionality reduction clustering is performed on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules. Using the aforementioned support vector machine algorithm, user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering are trained to obtain an SVM anomaly identification model for each category. Based on the SVM anomaly identification model for each category, the power generation data of power generation users that meet the preliminary anomaly screening rules are all anomaly identified, and abnormal power generation users are further selected.

2. The method for detecting abnormalities in low-voltage residential photovoltaic power generation as described in claim 1, characterized in that, The step of performing dimensionality reduction and clustering on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules based on the K-means algorithm specifically includes: For power generation users that meet the aforementioned preliminary anomaly screening rules, the user's distribution area, the user's line, and the user's substation are identified in the user data, and the number of categories to be clustered is set based on the identification results. Based on daily power generation and installed capacity, the power generation data of power generation users that meet the preliminary anomaly screening rules are subjected to dimensionality reduction clustering. Then, a trained clusterer is used to perform cluster labeling on all power generation users that meet the preliminary anomaly screening rules to determine the power generation users that belong to each category after dimensionality reduction clustering.

3. The method for detecting abnormalities in low-voltage residential photovoltaic power generation as described in claim 2, characterized in that, The specific steps of using the Support Vector Machine (SVM) algorithm to train the user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering, and to obtain the SVM anomaly identification model for each category, include: Feature factors are expanded on the user data and corresponding electricity data of each category of power generation users, and each category of power generation users is labeled. Furthermore, a kernel mechanism is used to transform and add data dimensions for each tagged category of power generation users. Among them, the expanded feature factors include daily power generation / maximum historical power generation in 30 days, daily power generation / installed capacity, and daily grid-connected electricity. Number of days in the previous month / Electricity usage settled in the previous month; the label indicates whether it is normal or abnormal. According to a certain proportion, all data after data dimension transformation and addition are divided into training sets and test sets corresponding to each category; One-hot encoding is performed on the training set for each category, and all data in the training set for each category are normalized. Initialize the SVM vector plane, and use this vector plane to detect and label all normalized data in the training set for each category; The SVM vector loss function is defined using cross-entropy. The optimal parameters of the model are found by using grid search based on the SVM vector loss function, and an SVM anomaly recognition model for each category is constructed based on the optimal parameters of the model.

4. A low-voltage residential photovoltaic power generation anomaly detection system, characterized in that, include: The user data acquisition unit is used to acquire low-voltage photovoltaic power generation user profile information and determine the user data of each power generation user based on the low-voltage photovoltaic power generation user profile information; wherein, the user data includes the low-voltage photovoltaic power generation user's electricity meter relationship, photovoltaic installed capacity, user's distribution area, user's line, and user's substation. The power data acquisition unit is used to calculate the power data of each power generation user based on the user data of each power generation user; wherein, the power data includes daily power generation, daily grid-connected power, monthly grid-connected power, and the grid-connected power settled in the previous month; The abnormal power generation user initial selection unit is used to establish preliminary screening rules related to daily power generation and installed capacity. It combines user data and corresponding power generation data for each power generation user to screen all power generation users in the low-voltage photovoltaic user file information to obtain power generation users that meet the preliminary screening rules. The preliminary screening rule for abnormalities is a daily power generation exceeding 8 times the installed capacity. The abnormal power generation user final selection unit is used to perform cluster analysis on power generation users that meet the preliminary abnormal screening rules using a preset support vector machine algorithm, and further select abnormal power generation users. The abnormal power generation user final selection unit includes: The data dimensionality reduction and clustering module is used to perform dimensionality reduction and clustering on the user data and corresponding electricity data of power generation users that meet the preliminary anomaly screening rules based on the K-means algorithm; The SVM anomaly detection model training module is used to train the user data and corresponding electricity data of power generation users belonging to each category after dimensionality reduction and clustering using the support vector machine algorithm, so as to obtain the SVM anomaly detection model for each category. The abnormal power generation user identification module is used to identify abnormalities in the power generation data of power generation users that meet the preliminary abnormal screening rules based on the SVM abnormal identification model for each category, and further select abnormal power generation users.

Images