Distributed photovoltaic grid connection line loss diagnosis optimization method

Abstract

The invention relates to the technical field of power system data analysis, in particular to a distributed photovoltaic grid connection line loss diagnosis optimization method, which comprises the following steps of: acquiring power utilization data of a transformer area in real time; analyzing the power consumption data by adopting a line loss detection model constructed by a second-order clustering algorithm to obtain a first line loss diagnosis result; analyzing the first line loss diagnosis result by adopting a line loss abnormity judgment library, and sending out early warning information and a line loss decision when the abnormal data grade meets a preset standard of the line loss abnormity judgment library; according to the line loss decision, optimizing the state of equipment in the transformer area, and analyzing the optimized power consumption data by adopting a line loss detection model to obtain a second line loss diagnosis result; and based on the line loss evaluation index system, comparing and analyzing the first line loss diagnosis result and the second line loss diagnosis result to obtain an optimization effect and generate a line loss management report. According to the method, the problems that the initial classification number is fixed and the effect is poor when a traditional K-means clustering method is used for processing large-scale power data can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system data analysis, and in particular to a distributed photovoltaic grid-connected line loss diagnosis and optimization method. Background Art

[0002] With the country's emphasis on and vigorous promotion of clean energy, photovoltaic power generation, as an important form of distributed energy, has developed rapidly across the country. However, with the continuous expansion of the scale of photovoltaic grid connection, a series of new technical challenges have emerged in the distribution network, especially the problem of line loss in the substation area has become increasingly prominent. The power system is currently actively promoting and applying smart energy meters and full-event electricity consumption information collection systems. The system can upload the full amount of information of smart meters (including voltage, current, power and frozen power, etc.) and report abnormal power consumption events (such as power outages, meter cover opening events, voltage and current abnormalities, power flow reversal, etc.), providing a large amount of effective information for the analysis of line loss problems.

[0003] At present, the research work on line loss in substations at home and abroad focuses on theoretical line loss calculation and line loss management methods. There are few studies on diagnosing and reducing line losses. Traditional line loss analysis methods often rely on manual experience and judgment, which makes it difficult to achieve accurate identification. In addition, although the traditional K-means clustering method is more efficient in processing large-scale data, its initial classification number is fixed and the effect is poor, which cannot cope with complex and changeable power data.

[0004] Therefore, the present invention provides a distributed photovoltaic grid-connected line loss diagnosis optimization method in combination with a second-order clustering model to solve the above problems. Summary of the invention

[0005] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a distributed photovoltaic grid-connected line loss diagnosis optimization method to solve the problem that the initial number of classifications of the traditional K-means clustering method is fixed and the effect is poor when processing large-scale power data.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: A distributed photovoltaic grid-connected line loss diagnosis and optimization method, comprising: The power consumption data of the substation is collected in real time. The line loss detection model constructed by the second-order clustering algorithm is used to analyze the power consumption data to obtain the first line loss diagnosis result; specifically, the method includes: preprocessing the power consumption data to obtain preprocessed data; creating a cluster feature tree according to the preprocessed data; performing secondary clustering according to the cluster feature tree to determine the number of clusters and obtain secondary clustering results; analyzing the secondary clustering results to obtain the first line loss diagnosis result.

[0007] The line loss abnormality discrimination library constructed based on the line loss history archive library analyzes the first line loss diagnosis result to determine the abnormal data level. When the abnormal data level meets the preset standard of the line loss abnormality discrimination library, an early warning message is issued and a line loss decision obtained by matching the first line loss diagnosis result is issued.

[0008] The equipment status in the substation is optimized according to the line loss decision, and the optimized power consumption data is analyzed using the line loss detection model to obtain the second line loss diagnosis result.

[0009] Based on the line loss evaluation index system, the first line loss diagnosis result and the second line loss diagnosis result are compared and analyzed to obtain the optimization effect and generate a line loss management report.

[0010] Preferably, the creation of the cluster feature tree includes: determining a scanning radius; aggregating data points according to the scanning radius to form a plurality of cluster centers and a plurality of boundaries to form a cluster feature tree; and dynamically updating the cluster feature tree according to new preprocessed data.

[0011] Preferably, the secondary clustering is performed according to the cluster feature tree to determine the number of clusters and obtain the secondary clustering results, including: according to the cluster centers and boundaries in the cluster feature tree, the K-means algorithm is used to perform secondary clustering on the data points; the elbow rule, the BIC criterion or the AIC criterion is used to determine the number of clusters to obtain the secondary temporary clustering results; the secondary temporary clustering results are iteratively optimized, and when the secondary temporary clustering results converge, the secondary clustering results are obtained.

[0012] Preferably, a distributed photovoltaic grid-connected line loss diagnosis optimization method also includes: using the leave-one-out method or the K-fold cross-validation method to evaluate the secondary clustering results of the line loss detection model; adjusting the clustering parameters in the line loss detection model according to the evaluation results; and determining the model clustering parameter combination using a grid search algorithm or a random search algorithm.

[0013] Preferably, a distributed photovoltaic grid-connected line loss diagnosis optimization method also includes: using TSDAC algorithm and FTSDAC algorithm to losslessly compress various dynamic sequence data, using parallel execution method to batch query and process various dynamic sequence data, and using HBase cluster to store and process various dynamic sequence data.

[0014] Preferably, analyzing the secondary clustering result to obtain the first line loss diagnosis result includes: The local anomaly factor algorithm is used to analyze the data in the secondary clustering results to determine the abnormal data; the secondary abnormal data associated with the abnormal data is obtained based on the abnormal data; the abnormal data and the secondary abnormal data are analyzed by an expert system based on rules and knowledge base to determine the cause and type of the abnormality; the abnormal location of the abnormal data is determined based on the geographic information system and the topological structure of the power network; the abnormal data, abnormal location, abnormal type and abnormal cause are used as the first line loss diagnosis result.

[0015] Preferably, analyzing the secondary clustering result to obtain the first line loss diagnosis result includes: The local anomaly factor algorithm is used to analyze the data in the secondary clustering results to determine the abnormal data. The line loss anomaly diagnosis model constructed based on historical data and long short-term memory network is used to analyze the abnormal data to obtain the first line loss diagnosis result including abnormal data, abnormal location, abnormal type and abnormal cause.

[0016] Preferably, the process of obtaining the line loss decision includes: matching the initial response decision with the highest matching degree according to the first line loss detection result and the decision library constructed based on the line loss history archive library, and visually displaying the first line loss result and the initial response decision; the maintenance personnel adjust the initial response decision according to the data displayed on the visualization page, their own experience and the recorded data of the abnormal location to obtain the line loss decision.

[0017] Preferably, the process of determining the abnormal data level includes: analyzing each abnormal data according to the abnormal interval standard of each abnormal data in the line loss abnormality discrimination library to determine the level of each abnormal data; the abnormal interval standard is a mapping relationship between the interval of abnormal data and the abnormal data level.

[0018] Preferably, the process of obtaining preprocessed data includes: using a density-based spatial clustering algorithm to identify and remove noise data; using a sliding average method or a weighted average method to smooth time series data; using a box plot analysis method to identify and process outliers; determining feature data related to the line loss rate based on a line loss data feature library; and standardizing the feature data to obtain preprocessed data.

[0019] The beneficial effects of the present invention are: 1. Compared with the traditional K-means clustering method, the line loss detection model of the present invention adopts a second-order clustering algorithm to perform two-stage clustering on large-scale power data, which can improve the accuracy of large-scale power data clustering and achieve accurate classification; and the use of clustering feature trees can compress large-scale data, which is conducive to improving data processing efficiency, and the second-order clustering algorithm can dynamically adjust the number of clusters according to new data, which solves the problem that the initial number of classifications of the traditional K-means clustering method is fixed and the effect is poor when processing large-scale power data.

[0020] 2. The present invention uses a line loss detection model constructed by a second-order clustering algorithm to analyze the power consumption data and obtain a first line loss diagnosis result; then the line loss abnormality discrimination library constructed based on the line loss history archive is used to process the first line loss diagnosis result and determine the abnormal data level. When the abnormal data level meets the preset standard of the line loss abnormality discrimination library, an early warning message is issued to the monitoring personnel and a line loss decision obtained by matching the first line loss diagnosis result is issued; then the equipment status in the substation is optimized according to the line loss decision to reduce the line loss. And the line loss detection model is used to analyze the power consumption data collected after the equipment is optimized to obtain a second line loss diagnosis result; based on the line loss evaluation index system, the first line loss diagnosis result and the second line loss diagnosis result are compared and analyzed to obtain the optimization effect and generate a line loss management report, and the line loss management report is used as reference data for further optimizing the line loss. Through the above method, the present invention can optimize and adjust the line loss of the photovoltaic grid-connected substation, reduce the line loss, improve the power utilization rate, and thus save the power cost.

[0021] 3. The present invention adopts TSDAC algorithm and FTSDAC algorithm to perform lossless compression on various dynamic sequence data, adopts parallel execution method to batch query and process various dynamic sequence data, and adopts HBase cluster to store and process various dynamic sequence data; through the above methods, the present invention can process various dynamic sequence data in distributed photovoltaic grid-connected line loss diagnosis optimization method in a distributed manner, and improve the efficiency of lossless compression, storage and data processing of various dynamic sequence data. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a schematic flow chart of a distributed photovoltaic grid-connected line loss diagnosis and optimization method.

[0023] Figure 2 The present invention is a schematic flow chart of determining the number of clusters and obtaining secondary clustering results in the second-order clustering method.

[0024] Figure 3 The figure is a schematic flow chart of obtaining a line loss decision in one embodiment of the present invention. DETAILED DESCRIPTION

[0025] In the prior art, the second-order clustering algorithm is an efficient clustering method suitable for the analysis of large-scale data sets. The mathematical model and theoretical basis of the second-order clustering algorithm are deeply explored in the following way to provide theoretical support for the identification of line loss anomalies in photovoltaic grid-connected areas.

[0026] Overview of the second-order clustering algorithm: The second-order clustering algorithm operates in two stages: pre-clustering and cluster optimization. In the pre-clustering stage, the algorithm performs preliminary clustering division of the data by constructing a clustering feature tree to obtain the number of clusters. In the cluster optimization stage, the algorithm uses methods such as K-means to further refine and optimize the pre-clustering results.

[0027] In the mathematical model constructed by the second-order clustering algorithm, there are two stages: cluster feature tree and cluster optimization. Cluster feature tree stage: The cluster feature tree is a multi-level clustering data structure used to compress and summarize large-scale data sets. Each node of the cluster feature tree contains the following information: Cluster center: represents the center of the data point represented by the node. Radius: The maximum distance covering all data points, from the cluster center to these points. Height: The distance from the root of the tree to the node.

[0028] The construction of the clustering feature tree follows the following steps: Step 1: Select a suitable scanning radius, scan the data points and calculate their distance from the cluster center.

[0029] Step 2: Merge data points whose distance is less than the scanning radius into one cluster.

[0030] Step 3: Repeat the above process until all data points are merged into the cluster feature tree.

[0031] Clustering optimization stage: After the pre-clustering is completed, each cluster in the clustering feature tree is optimized using clustering algorithms such as K-means to obtain the final clustering results.

[0032] Theoretical basis: The theoretical basis of the second-order clustering algorithm is that it can handle high-dimensional data and large-scale data sets. The key to the second-order clustering algorithm is: ① Data compression: compress data through CF Tree to reduce the amount of calculation. ② Dynamic clustering: adaptively adjust the cluster center and the number of clusters to adapt to changes in data distribution. ③ Local sensitivity: identify outliers through methods such as local outlier factor (LOF) to improve the accuracy of anomaly identification.

[0033] The advantages of the second-order clustering algorithm in the identification of line loss anomalies in photovoltaic grid-connected areas include: ① High efficiency: suitable for processing large-scale data sets and fast calculation speed. ② High accuracy: through two-stage clustering, the accuracy of clustering is improved. ③ Robustness: able to identify and process outliers and noise data.

[0034] The following will refer to the attached Figure 1 To Attachment Figure 3 The embodiments of the present invention are described in detail. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0035] A distributed photovoltaic grid-connected line loss diagnosis optimization method, as shown in the attached Figure 1 As shown, the following steps are included: Step S11: Collect the power consumption data of the substation in real time.

[0036] In step S11, various power consumption data of the substation are collected in real time through various smart meters, and the power consumption data includes but is not limited to photovoltaic power generation, grid load, voltage level, current, power quality parameters, etc.

[0037] Step S12: Analyze the power consumption data using the line loss detection model constructed by the second-order clustering algorithm to obtain a first line loss diagnosis result.

[0038] The specific process of step S12 includes: preprocessing the power consumption data to obtain preprocessed data; creating a cluster feature tree based on the preprocessed data; performing secondary clustering based on the cluster feature tree to determine the number of clusters and obtain secondary clustering results; analyzing the secondary clustering results to obtain the first line loss diagnosis result.

[0039] Step S13: Analyze the first line loss diagnosis result based on the line loss abnormality discrimination library constructed based on the line loss history archive library to determine the abnormal data level.

[0040] Preferably, the line loss history archive is a database constructed by using a hierarchical clustering analysis method, which can store the power consumption data in the corresponding position in the line loss history archive according to the characteristics of the power consumption data collected in real time. The line loss abnormality discrimination library is a database established based on the line loss history archive and various power consumption data standards, and is provided with a mapping relationship between each power consumption data and the abnormal data level.

[0041] Step S14: When the abnormal data level meets the preset standard of the line loss abnormality judgment library, a warning message is issued and a line loss decision obtained by matching the first line loss diagnosis result is issued.

[0042] For example, when the abnormal data level of one of the abnormal data is greater than the preset level, an early warning message is sent to the monitoring personnel. And according to the first line loss diagnosis result, a line loss decision is obtained, which is adjusted by the maintenance personnel based on experience and historical data to reduce the line loss to the greatest extent.

[0043] Step S15: Optimize the equipment status in the substation according to the line loss decision, and use the line loss detection model to analyze the optimized power consumption data to obtain a second line loss diagnosis result.

[0044] Preferably, line loss decisions include but are not limited to equipment model, equipment location, voltage, current, power, power quality parameters, etc.

[0045] In step S15, after optimizing the equipment status in the photovoltaic grid-connected area according to the line loss decision, the line loss detection model is used to diagnose and analyze the electricity consumption data collected by the optimized smart meter to obtain a second line loss diagnosis result for comparison with the first line loss diagnosis result to obtain the line loss optimization effect.

[0046] Step S16: Based on the line loss evaluation index system, a first line loss diagnosis result and a second line loss diagnosis result are compared and analyzed to obtain an optimization effect and generate a line loss management report.

[0047] In this embodiment, compared with the traditional K-means clustering method, the line loss detection model of the present invention adopts a second-order clustering algorithm to perform two-stage clustering on large-scale power data, which can improve the accuracy of large-scale power data clustering and achieve accurate classification; and the use of a clustering feature tree can compress large-scale data, which is beneficial to improving data processing efficiency, and the second-order clustering algorithm can dynamically adjust the number of clusters according to new data, which solves the problem that the initial number of classifications of the traditional K-means clustering method is fixed and the effect is poor when processing large-scale power data.

[0048] The present invention uses a line loss detection model constructed by a second-order clustering algorithm to analyze the power consumption data and obtain a first line loss diagnosis result; then the line loss abnormality discrimination library constructed based on the line loss history archive library processes the first line loss diagnosis result to determine the abnormal data level, and when the abnormal data level meets the preset standard of the line loss abnormality discrimination library, a warning message is issued to the monitoring personnel and a line loss decision obtained by matching the first line loss diagnosis result is issued; then the equipment status in the substation is optimized according to the line loss decision to reduce the line loss. And the line loss detection model is used to analyze the power consumption data collected after the equipment is optimized to obtain a second line loss diagnosis result; based on the line loss evaluation index system, the first line loss diagnosis result and the second line loss diagnosis result are compared and analyzed to obtain the optimization effect and generate a line loss management report, and the line loss management report is used as reference data for further optimizing the line loss. Through the above method, the present invention can optimize and adjust the line loss of the photovoltaic grid-connected substation, reduce the line loss, improve the power utilization rate, and thus save the power operation cost.

[0049] In one embodiment of the present invention, the process of creating a cluster feature tree includes: determining a scanning radius; aggregating data points according to the scanning radius to form multiple cluster centers and multiple boundaries to form a cluster feature tree; and dynamically updating the cluster feature tree according to new preprocessed data.

[0050] In this embodiment, the present invention dynamically updates the clustering feature tree according to new preprocessed data, which can improve the accuracy of the clustering features of the clustering feature tree. Compared with the fixed clustering feature tree, the dynamically updated clustering feature tree can provide more accurate clustering features for complex and changeable power data.

[0051] In one embodiment of the present invention, as shown in the attached Figure 2 As shown, secondary clustering is performed according to the cluster feature tree to determine the number of clusters and obtain secondary clustering results, including the following steps: Step S21: Based on the cluster centers and boundaries in the cluster feature tree, the K-means algorithm is used to perform secondary clustering on the data points.

[0052] Step S22: using the elbow rule, the BIC criterion or the AIC criterion to determine the number of clusters to obtain a secondary temporary clustering result.

[0053] Step S23: iteratively optimize the secondary temporary clustering result, and obtain the secondary clustering result when the secondary temporary clustering result converges.

[0054] In step S23, after the cluster center is initialized, the k-means algorithm optimizes the cluster center and the number of clusters by iteration until the stopping criterion is met and convergence is completed. For example, when the cluster center change is less than a certain threshold or the number of iterations reaches the maximum number of iterations, the final secondary clustering result is obtained.

[0055] In each iteration, the algorithm first calculates the distance from each data point to each cluster center, and divides the data point into the class represented by the closest cluster center. Then, the k-means algorithm recalculates the average value of the data points in each class and updates the cluster center. This process will be repeated until the update amount of the cluster center is less than the set threshold or the preset number of iterations is reached, the algorithm stops, and the final clustering result is output.

[0056] Compared with the traditional K-means clustering method, the line loss detection model of the present invention adopts a second-order clustering algorithm to perform two-stage clustering on large-scale power data, which can improve the accuracy of large-scale power data clustering and achieve accurate classification; and the use of a clustering feature tree can compress large-scale data, which is conducive to improving data processing efficiency, and the second-order clustering algorithm can dynamically adjust the number of clusters according to new data, which solves the problem that the initial number of classifications of the traditional K-means clustering method is fixed and the effect is poor when processing large-scale power data.

[0057] In one embodiment of the present invention, a distributed photovoltaic grid-connected line loss diagnosis optimization method also includes: using the leave-one-out method or the K-fold cross-validation method to evaluate the secondary clustering results of the line loss detection model; adjusting the clustering parameters in the line loss detection model according to the evaluation results; and determining the model clustering parameter combination using a grid search algorithm or a random search algorithm.

[0058] In this embodiment, the present invention uses the leave-one-out method or the K-fold cross-validation method to verify and evaluate the secondary clustering results of the line loss detection model; according to the evaluation results, the clustering parameters in the line loss detection model, such as the scanning radius, the number of clusters, etc., are adjusted; and then a grid search algorithm or a random search algorithm is used to determine the model clustering parameter combination with the best clustering effect. Through the result verification and optimization method of this embodiment, the present invention can improve the clustering accuracy of the secondary clustering results of the line loss detection model, and improve the stability and accuracy of clustering.

[0059] In one embodiment of the present invention, a distributed photovoltaic grid-connected line loss diagnosis optimization method also includes: using TSDAC algorithm and FTSDAC algorithm to losslessly compress various dynamic sequence data, using parallel execution method to batch query and process various dynamic sequence data, and using HBase cluster to store and process various dynamic sequence data.

[0060] Preferably, the dynamic sequence data includes power consumption data sorted in chronological order, pre-processed data sorted in chronological order, and clustering data in a clustering feature tree sorted in chronological order.

[0061] In this embodiment, the present invention adopts TSDAC algorithm and FTSDAC algorithm to perform lossless compression on various dynamic sequence data, adopts parallel execution method to batch query and process various dynamic sequence data, and adopts HBase cluster to store and process various dynamic sequence data; through the above methods, the present invention can process various dynamic sequence data in distributed photovoltaic grid-connected line loss diagnosis optimization method in a distributed manner, and improve the efficiency of lossless compression, storage and data processing of various dynamic sequence data.

[0062] In an optional embodiment of the present invention, a specific real-time process of analyzing the secondary clustering results to obtain the first line loss diagnosis result includes: using a local anomaly factor algorithm to analyze the data in the secondary clustering results to determine the abnormal data; obtaining secondary abnormal data associated with the abnormal data based on the abnormal data; using an expert system based on rule and knowledge base construction to analyze the abnormal data and the secondary abnormal data to determine the cause and type of the abnormality; determining the abnormal location of the abnormal data based on the geographic information system and the power network topology; and using the abnormal data, abnormal location, abnormal type and abnormal cause as the first line loss diagnosis result.

[0063] In this embodiment, the present invention analyzes the data in the secondary clustering results through a local anomaly factor algorithm, can determine the abnormal data in the secondary clustering results, and separate the abnormal data from the secondary clustering results, so as to perform subsequent analysis and processing on the abnormal data.

[0064] According to the source of the abnormal data, secondary abnormal data associated with the abnormal data is obtained, and the secondary abnormal data includes but is not limited to all collectible data on the equipment of the abnormal data source. The abnormal data and secondary abnormal data are analyzed by an expert system based on rule and knowledge base construction, so that the cause of the abnormality and the type of abnormality can be determined. Then, based on the geographic information system and the topology of the power network, the abnormal location of the abnormal data is determined; the abnormal data, abnormal location, abnormal type and abnormal cause are used as the first line loss diagnosis result. Through the method of this embodiment, the present invention can comprehensively analyze the abnormal data in the secondary clustering results, diagnose the specific cause of the line loss and the specific geographical location of the abnormal data, so as to provide maintenance convenience for maintenance personnel, save the search for the specific location of the equipment corresponding to the abnormal data, and thus save maintenance time.

[0065] In an optional embodiment of the present invention, another specific real-time process of analyzing the secondary clustering results to obtain the first line loss diagnosis result includes: using a local anomaly factor algorithm to analyze the data in the secondary clustering results to determine abnormal data; using a line loss anomaly diagnosis model constructed based on historical data and a long short-term memory network to analyze the abnormal data to obtain a first line loss diagnosis result including abnormal data, abnormal location, abnormal type and abnormal cause.

[0066] In this embodiment, the historical data includes the accumulated abnormal power consumption data, the device model corresponding to each power consumption data, the geographical location of the device, and the mapping relationship between the abnormal data and the line loss cause.

[0067] The present invention uses a line loss abnormality diagnosis model constructed based on historical data and a long short-term memory network to analyze abnormal data to obtain a first line loss diagnosis result including abnormal data, abnormal location, abnormal type and abnormal cause. The present invention can comprehensively analyze the abnormal data in the secondary clustering result through the method of this embodiment, diagnose the specific line loss cause and the specific geographical location of the abnormal data, so as to provide maintenance convenience for maintenance personnel, save the time of searching for the specific location of the equipment corresponding to the abnormal data, and thus save maintenance time.

[0068] In one embodiment of the present invention, as shown in the attached Figure 3 As shown, the process of obtaining the line loss decision includes the following steps: Step S31: According to the first line loss detection result and the decision library constructed based on the line loss history archive library, an initial response decision with the highest matching degree is matched.

[0069] Step S32: Visually display the first line loss result and the initial response decision.

[0070] Step S33: The maintenance personnel adjust the initial response decision according to the data displayed on the visualization page, their own experience and the recorded data of the abnormal location to obtain a line loss decision.

[0071] In this embodiment, the present invention adds manual intervention operations based on matching the initial response decision with the highest matching degree according to the first line loss detection result and the decision library, that is, the maintenance personnel adjust the initial response decision according to the data displayed on the visualization page, their own experience, and the recorded data of the abnormal position to obtain the final line loss decision. The present invention can make the line loss decision more in line with the actual situation through the determination method of the line loss decision of this embodiment, improve the accuracy of the line loss decision, and thus reduce the line loss.

[0072] In one embodiment of the present invention, the process of determining the abnormal data level includes: analyzing each abnormal data according to the abnormal interval standard of each abnormal data in the line loss abnormality discrimination library to determine the level of each abnormal data; the abnormal interval standard is a mapping relationship between the interval of the abnormal data and the level of each abnormal data.

[0073] In this embodiment, the present invention determines the abnormal data level of the abnormal data acquired in real time by establishing a corresponding relationship between abnormal data and abnormal data level based on expert experience and historical data.

[0074] In one embodiment of the present invention, the process of obtaining preprocessed data includes: using a density-based spatial clustering algorithm to identify and remove noise data; using a sliding average method or a weighted average method to smooth time series data; using a box plot analysis method to identify and process outliers; determining feature data related to the line loss rate based on a line loss data feature library; and standardizing the feature data to obtain preprocessed data.

[0075] In this embodiment, the present invention uses a density-based spatial clustering algorithm to identify and remove noise data in electricity consumption data; uses a sliding average method or a weighted average method to smooth time series data to reduce the impact of random fluctuations; uses a box plot analysis method to identify and process outliers to ensure the overall quality and consistency of the data; determines feature data related to the line loss rate based on a line loss data feature library; uses a Z-score standardization method or a minimum-maximum standardization method to standardize feature data to eliminate the impact of dimensions, so that the data is on the same scale, and then obtains preprocessed data that is convenient for subsequent processing. The present invention uses the above-mentioned method to preprocess electricity consumption data, which can reduce noise interference, reduce data processing volume, and improve the accuracy of data analysis.

[0076] In one embodiment of the present invention, the distributed photovoltaic grid-connected line loss diagnosis optimization method of the present invention can effectively reduce the line loss rate, thereby reducing the operating cost of distributed photovoltaic grid-connected, and reducing the ineffective consumption of energy, which is helpful to achieve energy conservation and emission reduction. In addition, the present invention helps to timely discover and repair equipment failures and extend the service life of assets through accurate line loss anomaly identification, thereby improving the overall utilization rate of assets and the return on investment.

[0077] In one embodiment of the present invention, the system corresponding to the distributed photovoltaic grid-connected line loss diagnosis optimization method of the present invention includes a data acquisition layer, a data processing layer, an intelligent analysis layer, a decision support layer, and a user interaction layer.

[0078] The data collection layer is used to collect key information such as power load, power data, voltage level, etc. in the substation in real time through technical means such as power information collection system and automatic meter reading system. These data serve as the basis for system operation and provide raw materials for subsequent data processing and analysis.

[0079] The data processing layer is used to store, extract, process and analyze the collected electricity consumption data through distributed parallel processing technology. The core task is to pre-process the data, including outlier detection, data cleaning and format conversion to ensure the accuracy and availability of the data.

[0080] The intelligent analysis layer is used to perform intelligent diagnosis and analysis on the preprocessed data through the line loss detection model to obtain the line loss diagnosis results.

[0081] The decision support layer is used to analyze the first line loss diagnosis result based on the line loss abnormality discrimination library constructed by the line loss history archive library, determine the abnormal data level, and when the abnormal data level meets the preset standard of the line loss abnormality discrimination library, issue an early warning message and issue a line loss decision obtained by matching the line loss diagnosis result and adjusting by the maintenance personnel, so as to timely adjust the equipment status of the substation and reduce line loss.

[0082] The user interaction layer is built with a friendly interactive interface to support monitoring personnel or maintenance personnel to operate and manage the system. Maintenance personnel or monitoring personnel can adjust the initial response decision through the interactive interface to obtain the final line loss decision.

[0083] In addition, the functions of the user interaction layer include but are not limited to login profile information management, gateway information management, line loss management, power calculation and statistics, etc. Through API interface support, the deployment of the system physical structure and integration with other business systems are realized.

[0084] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A distributed photovoltaic grid-connected line loss diagnosis and optimization method, characterized in that: include: Collect electricity consumption data of the substation in real time; The line loss detection model constructed by the second-order clustering algorithm is used to analyze the power consumption data to obtain the first line loss diagnosis result; Specifically, the method includes: preprocessing the power consumption data to obtain preprocessed data; creating a cluster feature tree according to the preprocessed data; performing secondary clustering according to the cluster feature tree to determine the number of clusters and obtain a secondary clustering result; analyzing the secondary clustering result to obtain a first line loss diagnosis result; The line loss abnormality discrimination library constructed based on the line loss history archive library analyzes the first line loss diagnosis result to determine the abnormal data level. When the abnormal data level meets the preset standard of the line loss abnormality discrimination library, an early warning message is issued and a line loss decision obtained by matching the first line loss diagnosis result is issued; The equipment status in the substation is optimized according to the line loss decision, and the optimized power consumption data is analyzed using the line loss detection model to obtain the second line loss diagnosis result; Based on the line loss evaluation index system, the first line loss diagnosis result and the second line loss diagnosis result are compared and analyzed to obtain the optimization effect and generate a line loss management report.

2. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1 is characterized in that: The creation of the cluster feature tree includes: determining a scanning radius; aggregating data points according to the scanning radius to form multiple cluster centers and multiple boundaries to form a cluster feature tree; and dynamically updating the cluster feature tree according to new preprocessed data.

3. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 2 is characterized in that: The method of performing secondary clustering according to the cluster feature tree to determine the number of clusters and obtain secondary clustering results includes: performing secondary clustering on data points according to cluster centers and boundaries in the cluster feature tree using a K-means algorithm; determining the number of clusters using an elbow rule, a BIC criterion, or an AIC criterion to obtain a secondary temporary clustering result; iteratively optimizing the secondary temporary clustering result, and obtaining a secondary clustering result when the secondary temporary clustering result converges.

4. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1 is characterized in that: Also includes: The leave-one-out method or K-fold cross-validation method is used to evaluate the secondary clustering results of the line loss detection model; the clustering parameters in the line loss detection model are adjusted according to the evaluation results; and the grid search algorithm or random search algorithm is used to determine the model clustering parameter combination.

5. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1 is characterized in that: It also includes: using TSDAC algorithm and FTSDAC algorithm to losslessly compress various dynamic sequence data, using parallel execution method to batch query and process various dynamic sequence data, and using HBase cluster to store and process various dynamic sequence data.

6. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1 is characterized in that: The analyzing the secondary clustering result to obtain the first line loss diagnosis result includes: The local anomaly factor algorithm is used to analyze the data in the secondary clustering results to determine the abnormal data; the secondary abnormal data associated with the abnormal data is obtained based on the abnormal data; the abnormal data and the secondary abnormal data are analyzed by an expert system based on rules and knowledge base to determine the cause and type of the abnormality; the abnormal location of the abnormal data is determined based on the geographic information system and the topological structure of the power network; the abnormal data, abnormal location, abnormal type and abnormal cause are used as the first line loss diagnosis result.

7. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1, characterized in that: The analyzing the secondary clustering result to obtain the first line loss diagnosis result includes: The local anomaly factor algorithm is used to analyze the data in the secondary clustering results to determine the abnormal data. The line loss anomaly diagnosis model constructed based on historical data and long short-term memory network is used to analyze the abnormal data to obtain the first line loss diagnosis result including abnormal data, abnormal location, abnormal type and abnormal cause.

8. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1 is characterized in that: The process of obtaining the line loss decision includes: matching the initial response decision with the highest matching degree according to the first line loss detection result and the decision library constructed based on the line loss history archive library, and visually displaying the first line loss result and the initial response decision; the maintenance personnel adjust the initial response decision according to the data displayed on the visualization page, their own experience and the recorded data of the abnormal location to obtain the line loss decision.

9. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1, characterized in that: The process of determining the abnormal data level includes: analyzing each abnormal data according to the abnormal interval standard of each abnormal data in the line loss abnormality discrimination library to determine the level of each abnormal data; the abnormal interval standard is a mapping relationship between the interval of abnormal data and the abnormal data level.

10. The distributed photovoltaic grid-connected line loss diagnosis and optimization method according to claim 1, characterized in that: The process of obtaining the preprocessed data includes: using a density-based spatial clustering algorithm to identify and remove noise data; using a sliding average method or a weighted average method to smooth time series data; using a box plot analysis method to identify and process outliers; determining feature data related to the line loss rate based on a line loss data feature library; and standardizing the feature data to obtain the preprocessed data.

Images