Method for complementing unknown measurement information of power distribution network based on data flow mapping

Abstract

The invention discloses a power distribution network unknown measurement information completion method based on data flow mapping, and relates to the technical field of power distribution network monitoring and control. The invention discloses a method for complementing unknown measurement information of a power distribution network based on data power flow mapping. The method comprises the following steps: step 1, constructing a data power flow mapping model; step 2, speculating and complementing missing data; 3, real-time updating and intelligent optimization are carried out; step 4, system integration and application; wherein the step 1 comprises the following sub-steps; step 1.1, data acquisition and preprocessing; step 1.2, feature extraction and analysis; 1.3, constructing a mapping model; wherein the step 2 comprises the following sub-steps; step 2.1, identification and speculation of missing data; step 2.2, constructing a speculative model; and step 2.3, data completion.

Description

Technical Field

[0001] The present invention discloses a method for completing unknown measurement information of a distribution network based on data flow mapping, which relates to the technical field of distribution network monitoring and control. Background Art

[0002] As an important part of the power system, the stability and reliability of the distribution network are directly related to the quality and safety of power supply. In the daily operation of the distribution network, measurement data is the basis for realizing key functions such as system status perception, optimized scheduling, and fault diagnosis. However, due to equipment failure, communication interruption, environmental interference and other reasons, measurement data is often missing or incomplete in the distribution network. This data loss phenomenon will lead to a decrease in the monitoring and control accuracy of the distribution network operation status, increase operation risks, and may even cause system failures or accidents.

[0003] Traditionally, power flow calculation is the main method to solve the missing measurement data in the distribution network. By establishing a power flow equation based on the physical model, the known system parameters and some measurement data are used to infer the status information of other nodes. However, this method has many problems in practical applications. First, the topological structure of the distribution network is complex and changes frequently, which makes it difficult for the power flow calculation based on the physical model to adapt to the changing actual situation. Secondly, the power flow equation is inherently nonlinear and has a high computational complexity. Especially when dealing with large-scale distribution networks, the computational efficiency is difficult to guarantee. In addition, due to the strong dependence of the physical model on the system parameters, once the parameters are inaccurate or change, the reliability of the calculation results will be seriously affected.

[0004] With the improvement of the digitalization of power grids, a large amount of historical measurement data has been accumulated, providing new possibilities for data-driven methods. By analyzing the relationship between these historical data and real-time operation data, data-driven methods can construct mapping relationships between system nodes without relying on complex physical models, thereby inferring and completing missing measurement data. Compared with traditional power flow calculation methods, methods based on data power flow mapping have higher flexibility and adaptability, and can play a better role in complex and changing distribution network environments.

[0005] With the development of smart grids, the types and coverage of sensors in distribution networks will continue to expand, and the acquisition of measurement data will become more abundant and diverse. How to effectively integrate data from different sources and build a more accurate and comprehensive data flow mapping model will become the focus of the next research. At the same time, the operating status of the distribution network is dynamically changing. How to achieve real-time updates of the data flow mapping model to ensure the timeliness and accuracy of data completion is also an urgent problem to be solved. With the advancement of artificial intelligence and machine learning technology, the intelligence level of distribution network data completion will continue to improve, and the completion effect will be further improved through autonomous learning and optimization. Summary of the invention

[0006] In view of this, the present invention provides a method for completing unknown measurement information of a distribution network based on data flow mapping, which is compared with the existing technical means; in application, the promotion and popularization of the data flow mapping method will depend on the effective integration with the existing distribution network monitoring and control system. Through deep integration with the dispatching system, fault diagnosis system, etc., the method will provide more comprehensive support for the efficient operation of the distribution network. In addition, in a complex actual environment, the measurement data of the distribution network will inevitably be disturbed by noise and abnormal values. These abnormal data can be identified and processed during the data flow mapping process to ensure the reliability and accuracy of the completion result. The method for completing unknown measurement information of a distribution network based on data flow mapping can effectively utilize existing data resources, and infer unknown measurement data by modeling the mapping relationship of power flow between nodes. Compared with traditional methods, the data flow mapping method has stronger adaptability and flexibility, and can better cope with complex and changeable distribution network environments. It can not only significantly reduce the computational complexity, but also improve the accuracy of data completion, thereby enhancing the real-time monitoring and control capabilities of the distribution network.

[0007] The present invention provides a method for completing unknown measurement information of a distribution network based on data flow mapping, comprising:

[0008] Step 1, construction of data flow mapping model;

[0009] Step 2: Inference and completion of missing data;

[0010] Step 3: Real-time update and intelligent optimization;

[0011] Step 4: System integration and application;

[0012] Wherein step 1 comprises:

[0013] Step 1.1. Data collection and preprocessing;

[0014] Step 1.2 feature extraction and analysis;

[0015] Step 1.3 Mapping model construction;

[0016] Wherein step 2 comprises:

[0017] Step 2.1. Identification and inference of missing data;

[0018] Step 2.2. Construction of inference model;

[0019] Step 2.3. Data completion.

[0020] Specifically, step 1.1. Data acquisition and preprocessing

[0021] Before building a data flow mapping model, data collection and preprocessing are required to ensure that the data used is accurate and available; this includes collecting historical measurement data, real-time operation data, and cleaning the historical measurement data and real-time operation data to achieve data standardization.

[0022] Historical measurement data (D hist ) is the historical data collected from the distribution network system, which includes node voltage (V i ), branch current (I ij ), active power (P i ) and reactive power (Q i ); The above historical data is used to provide status information of the distribution network at different time points.

[0023] Real-time operation data (D real ) To obtain data under the current operating state, which can reflect the real-time state and dynamic changes of the system;

[0024] Data cleaning: By removing outliers (D out ) and noise (D noise ), ensuring the quality of the data. Outliers may be caused by equipment failure or measurement errors, while noise may be due to environmental interference or transmission problems.

[0025] Data standardization: Standardize data from different sources to make them have the same scale. norm ) is convenient for subsequent processing and analysis, and avoids errors caused by different dimensions.

[0026] Specifically, step 1.2 feature extraction and analysis is as follows: after data preprocessing is completed, it enters the feature extraction and analysis stage, which aims to extract key information from a large amount of data and conduct in-depth analysis; it includes power flow feature extraction, principal component analysis, calculation of correlation coefficient matrix and pattern recognition.

[0027] Power flow feature extraction: Calculate the relationship between the voltage, current and power of each node; including calculating the power flow feature vector F between nodes i, obtain the basic characteristics of power flow;

[0028] Principal component analysis (PCA): Use principal component analysis to reduce the dimension of the data and extract the main features F pca , reducing computational complexity while retaining important information; principal component analysis can map high-dimensional data into low-dimensional space to reveal the main structure of the data;

[0029] Calculate the correlation coefficient matrix: By calculating the correlation coefficient matrix R between nodes, the power flow relationship between nodes is determined; the correlation coefficient between node voltage, current and power can be expressed as:

[0030]

[0031] Pattern recognition: Identify the main power flow patterns and find out the potential mapping relationships in the data through data clustering and association rule learning.

[0032] Specifically, step 1.3 mapping model construction

[0033] Based on feature extraction and analysis, a data flow mapping model is constructed, which can transform the power flow relationship in the data into a mathematical model for predicting and completing missing data; specifically, it includes selecting a suitable model, training a data-driven model, verifying and optimizing the data-driven model, and applying it to actual data.

[0034] Choose a suitable model: Select a data-driven model and use a graph neural network GNN or a multi-layer perceptron MLP. The graph neural network GNN is suitable for processing complex node and edge relationships, and is particularly suitable for data with complex topological structures such as distribution networks.

[0035] Training data-driven models: Using the training set (D train ) to train the data-driven model; the goal of the data-driven model is to minimize the difference between the predicted value and the true value, and the loss function l is set as:

[0036]

[0037] in is the voltage value predicted by the data-driven model, V i is the real voltage value; optimize the data-driven model: use the optimization algorithm to update the model parameters, and adjust the parameters Θ through an iterative process so that the loss function l reaches the minimum value.

[0038] Verification and optimization of data-driven models: Verify the performance of data-driven models: Use the validation set (D val )Evaluate the performance of the model and calculate the prediction error ε: The prediction accuracy of the model can be judged through the validation results, and the model can be adjusted as needed.

[0039] Application to actual data: Apply the trained and verified data-driven model to the actual distribution network data, predict and complete the missing measurement data, monitor the data in real time, and use the model to complete the data to improve the real-time monitoring and control capabilities of the distribution network.

[0040] Through the above steps, the constructed data flow mapping model can accurately capture the power flow relationship between nodes in the distribution network, and predict and complete missing data.

[0041] Specifically, step 2.1. Identification and inference of missing data

[0042] In the distribution network system, data missing problems usually come from equipment failure or communication interruption. First, the missing data needs to be accurately identified and inferred based on the existing data.

[0043] Assume that the observed data of the distribution network is D = {D i}, where D i is the ith observation. For missing data points We define an indicator function δ i :

[0044]

[0045] Missing data can be inferred from adjacent nodes or historical data. Assume that we have a set of known observation data D known and the inferred model for missing data f(D known ), the inference process can be expressed as:

[0046]

[0047] Specifically, step 2.2. Construction of the inference model

[0048] In order to accurately infer missing data, we can use the spatial interpolation inference method to build an inference model: for missing data points, we can use the observed data of neighboring nodes for interpolation. Let the spatial coordinates of the missing data point be (x i ,y i ), whose value can be estimated by Kriging interpolation or inverse distance weighted (IDW) method. The general formula of Kriging interpolation is:

[0049]

[0050] Among them, λ j is the Kriging weight, D j is the observation data of the neighboring points. Weight λj Determined by the semivariogram model, satisfying the constraints:

[0051]

[0052] Specifically, step 2.3. Data completion

[0053] After inferring the missing data, the inferred results need to be completed to improve the integrity of the distribution network data. Suppose the inferred data is D i , the completion process can be performed through the following optimization algorithms: Linear interpolation and polynomial interpolation are common completion methods. For linear interpolation, the missing data D i The neighboring data point D i-1 and D i+1 Perform the calculation:

[0054]

[0055] For polynomial interpolation, the Lagrange interpolation formula can be used:

[0056]

[0057] The data completion problem is transformed into an optimization problem, and an optimization algorithm (such as gradient descent) is used to minimize the error between the completed data and the actual data. For example, suppose the objective function is:

[0058]

[0059] Adjust and complete data through optimization algorithm So that l reaches the minimum value.

[0060] Through the above methods, the missing data in the distribution network can be effectively inferred and supplemented, thereby improving the monitoring and control capabilities of the system. These methods combine time series analysis, spatial interpolation and machine learning techniques to provide a comprehensive and efficient solution for data completion.

[0061] Specifically, step 3, real-time update and intelligent optimization.

[0062] In the distribution network, ensuring the real-time and accuracy of data completion is the key to improving the stability and reliability of the system. To this end, the establishment of a real-time update and intelligent optimization mechanism is crucial. The real-time update mechanism can quickly respond to changes in the system state and ensure the real-time and data integrity of the system by dynamically processing and completing the distribution network data. First, the real-time data stream processing system is responsible for receiving the latest data in the distribution network and integrating it into the data completion process in real time. These real-time data streams contain the observation values ​​of each node in the distribution network and can be used to update the current data status. By merging the newly arrived observation data with the existing data status, the real-time data processing system can keep the data up to date, so that the completion model can reflect the actual status of the system in a timely manner.

[0063] In order to achieve real-time data updates, incremental learning or online learning algorithms are introduced to dynamically adjust the parameters of the completion model. Through these algorithms, the model can continuously optimize its prediction ability based on the latest data. For example, in the incremental learning process, when new data arrives, the model will adjust the parameters based on the new data to minimize the prediction error. This dynamic adjustment can be expressed by the following formula

[0064]

[0065] Among them, Θ(t) is the model parameter at the current moment, η is the learning rate, is the gradient of the loss function. This updating process ensures that the model can adapt to the dynamic changes of the system and improves the accuracy and real-time performance of data completion.

[0066] In terms of intelligent optimization, the performance of data completion is further improved by introducing advanced optimization algorithms. These algorithms include gradient descent, genetic algorithms, and other heuristic optimization methods. The optimization goal is to minimize the error between the completed data and the actual observed data to achieve the best data completion effect.

[0067] In addition, in order to deal with noise and outliers in the distribution network, the intelligent optimization mechanism also needs to include anomaly detection and processing functions. Anomaly detection can identify abnormal fluctuations in the data and ensure the quality of the data through appropriate filtering and correction algorithms. The intelligent optimization algorithm monitors the quality of the data in real time and corrects the outliers, making the completion results more reliable and accurate.

[0068] In summary, the real-time update and intelligent optimization mechanism can effectively improve the real-time and accuracy of distribution network data completion by combining dynamic data processing, incremental learning, advanced optimization algorithms and anomaly detection. The implementation of these mechanisms ensures that the system can respond quickly to state changes while maintaining data integrity and reliability, providing solid support for the stable operation of the distribution network.

[0069] Specifically, step 4, system integration and application.

[0070] In the distribution network system, the successful implementation of data completion technology does not only rely on a single completion method, but also requires effective integration with the existing distribution network monitoring and control system. The process of system integration and application aims to seamlessly integrate data completion technology into the actual operation of the distribution network to improve the performance and reliability of the overall system.

[0071] First, the data completion method needs to be deeply integrated with the dispatching system and fault diagnosis system of the distribution network. The dispatching system relies on accurate real-time data to optimize power distribution and load scheduling, while the fault diagnosis system requires reliable data to identify and locate abnormal conditions in the system. By integrating data completion technology into these systems, it can be ensured that even in the absence of data, the system can still obtain accurate status information, thereby optimizing dispatching decisions and quickly diagnosing faults. For example, suppose the optimization goal of the dispatching system is:

[0072]

[0073] Among them, p k is the output of the kth generating unit, c k is the corresponding cost function. The accuracy of data completion technology directly affects the optimization effect of the objective function and ensures the rationality and economy of scheduling decisions.

[0074] During the system integration process, the output of the data completion algorithm needs to be connected to the existing data processing platform. To achieve this goal, the completion results can be integrated into the distribution network monitoring platform through an application programming interface or a data exchange format. For example, after data completion, the updated node voltage and branch current data need to be updated through the formula:

[0075]

[0076] Among them, V i (t) and I ij (t) represent the completed node voltage and branch current data, ε i and δ ij It is the completion error. Through effective data exchange and integration, it can be ensured that these completion data are delivered and applied to the actual system in a timely manner.

[0077] At the same time, system integration also needs to consider real-time data updates and intelligent optimization. In practical applications, the operating status and data flow of the distribution network are constantly changing, and the completion system must have real-time update capabilities. The real-time update process can be expressed by the following formula:

[0078] D current (t) = D provious ∪Dstream (t)

[0079] In this dynamic environment, data completion algorithms need to be able to process real-time data streams to ensure the timeliness and accuracy of completion results.

[0080] Finally, the data completion technology in the distribution network system should also be robust to noise and outliers. Data noise and outliers may affect the reliability of the completion results, so the integrated system needs to include anomaly detection and correction modules. Anomaly detection can be achieved through statistical methods or machine learning algorithms to ensure the quality of the data and the credibility of the completion results. The data anomaly detection function is:

[0081]

[0082] In the system integration and application stage, data completion technology needs to be closely integrated with the dispatching system, fault diagnosis system and data processing platform of the distribution network, and real-time updates and outlier processing need to be achieved. This not only improves the actual application effect of data completion, but also enhances the overall performance and reliability of the distribution network system, providing solid support for the intelligent and efficient operation of the power system.

[0083] The present invention aims at the common problem of missing measurement data in the distribution network, and proposes a method for completing unknown measurement information of the distribution network based on data flow mapping. When dealing with complex and changeable distribution networks, the traditional flow calculation method based on physical models often faces problems such as difficult model construction, high calculation complexity and strong dependence on system parameters, resulting in unsatisfactory data completion effect. The present invention utilizes a large amount of historical measurement data accumulated in the distribution network, and constructs an accurate data flow mapping model by mining the intrinsic relationship between the data and the current operation data. The model can realize efficient completion of missing measurement data without relying on complex power grid topology and accurate component parameters, thereby improving the operation monitoring capability of the distribution network and enhancing the real-time and reliability of the system. Through the method of the present invention, the impact of data missing on the operation of the power grid can be significantly reduced, providing solid technical support for the intelligent dispatching and optimization management of the distribution network.

[0084] Compared with the prior art, the method for completing unknown measurement information of distribution network based on data flow mapping proposed in this application has the following significant beneficial effects:

[0085] The method of the present application provides higher flexibility and adaptability than traditional power flow calculations. Through the technology based on data power flow mapping, the missing distribution network data can be accurately completed without relying on complex physical models and system parameters. This method overcomes the shortcomings of traditional physical models when facing dynamic and complex distribution networks, such as high computational complexity and strong dependence on system parameters. Compared with traditional power flow calculation methods, data power flow mapping can better adapt to the actual changes in the distribution network and improve the accuracy and reliability of the completion results.

[0086] The data flow mapping method shows superior computational efficiency when dealing with large-scale distribution networks. Traditional methods are often computationally complex and inefficient when faced with huge distribution network data, while data flow mapping technology can quickly complete data completion by utilizing the relationship between historical data and real-time data. This high efficiency significantly reduces computing time and resource consumption, and improves the system's real-time monitoring and control capabilities.

[0087] The method of the present application significantly improves the real-time performance of data completion. By implementing a real-time update mechanism, the data flow mapping method can dynamically adjust the completion model to ensure that the system can respond quickly when data is missing. The application of real-time data stream processing and online learning algorithms enables the completion model to adapt to data changes in a timely manner, thereby ensuring the timeliness and accuracy of the completion results. This feature is particularly suitable for the frequently changing operating conditions and data flow environments in distribution networks.

[0088] Safety is also an important advantage of this application. The data flow mapping method reduces the reliance on high-risk physical measurements through data-driven technology, thereby reducing potential safety hazards during operation. Compared with traditional methods that require high-altitude or complex environment measurements, this method reduces the risks in actual operations through data completion technology and improves the safety of system operation.

[0089] In terms of cost-effectiveness, although the data flow mapping method requires certain data processing and computing resources, its long-term maintenance and implementation costs are more economical than traditional manual measurements and complex calculations based on physical models. In addition, the versatility of this method also provides more application scenarios for the intelligent management and optimization of distribution networks, such as real-time monitoring and fault detection, further improving the comprehensive utilization rate and economic benefits of the system.

[0090] Compared with the existing technology, the distribution network unknown measurement information completion method based on data flow mapping has demonstrated many significant advantages and beneficial effects in terms of flexibility, computational efficiency, real-time performance, security and cost-effectiveness. It is an important innovation and progress in the field of distribution network data management and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0092] Figure 1 A flowchart of a method for completing unknown measurement information of a distribution network based on data flow mapping is provided in an embodiment of the present application.

[0093] Figure 2 A flow chart of constructing and training a data flow mapping model in step S103 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0094] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the following will be combined with the appended Figure 1-2 , the technical solutions in the embodiments of the present application are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0095] Therefore, the following Figure 1-2 The detailed description of the embodiments of the present application provided in the specification is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0096] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0097] The present application is further described in detail below in conjunction with embodiments and drawings.

[0098] See also Figure 1 , Figure 1 A flowchart of a method for completing unknown measurement information of a distribution network based on data flow mapping provided in an embodiment of the present application;

[0099] A method for completing unknown measurement information of a distribution network based on data flow mapping provided in an embodiment of the present application may include the following steps:

[0100] Step S101, comprehensively collect historical and real-time data of the distribution network. Collect historical measurement data and real-time operation data from each node of the distribution network. This includes key data such as voltage, current, and power factor. The data source can be sensors, measurement equipment, and communication systems in the distribution network. This step aims to obtain comprehensive system status information and provide a reliable data foundation for subsequent data processing and model building. Ensuring the comprehensiveness and accuracy of information during data collection is crucial for subsequent analysis.

[0101] Step S102, cleaning and standardizing the data. The collected raw data is preprocessed, including data cleaning and standardization. The data cleaning step involves removing noise from the data, correcting outliers, and filling missing data to improve data quality. Data standardization includes converting the data into a unified format and range to ensure comparability between different features. These processing steps help improve the accuracy and effectiveness of the subsequent modeling process.

[0102] Step S103, construct and train a data flow mapping model. Extract key features from the processed data and select a suitable mapping model for training. Feature extraction includes identifying and selecting important data features related to the state of the distribution network. Model selection involves selecting a suitable algorithm (such as a regression model or a graph neural network) and training the model. The training process uses a labeled data set to optimize the model parameters to build a model that can accurately map the data flow.

[0103] Step S104, system integration and practical application. Deeply integrate the data completion method with the distribution network monitoring system, dispatching system and fault diagnosis system. Through API or data exchange format, ensure that the completed data can be effectively connected with the actual system to support dispatching decisions and fault diagnosis. In addition, the system needs to have real-time update capabilities to keep the data up-to-date and complete, and provide support for the intelligent operation of the distribution network.

[0104] The embodiment of the present application provides a method for completing unknown measurement information of a distribution network based on data flow mapping. Figure 2 A flowchart of constructing and training a data flow mapping model provided in an embodiment of the present application.

[0105] The following is a detailed step-by-step description of feature extraction and selection, selecting a suitable model algorithm, model training and optimization, and model evaluation and validation, including the specific process and relevant formulas for each step:

[0106] Step S103-1, feature extraction and selection. Assume that the observed data of the distribution network is D = {D i}, where D i is the ith observation. For missing data points We define an indicator function δi :

[0107]

[0108] Missing data can be inferred from adjacent nodes or historical data. Assume that we have a set of known observation data D known and the inferred model for missing data f(D known ), the inference process can be expressed as:

[0109]

[0110] Step S103-2, select a suitable model algorithm. A suitable mapping model is trained. This involves evaluating various model algorithms (such as linear regression, support vector machine, graph neural network, etc.) and selecting the most suitable model. The criteria for model selection include prediction accuracy, computational complexity, and model interpretability. The application of graph neural network can describe the relationship between nodes by the following formula: In order to accurately infer missing data, we can use the spatial interpolation inference method to construct an inference model: For missing data points, the observed data of neighboring nodes can be used for interpolation. Let the spatial coordinates of the missing data point be (x i ,y i ), whose value can be estimated by Kriging interpolation or inverse distance weighted (IDW) method. The general formula of Kriging interpolation is:

[0111]

[0112] Among them, λ j is the kriging weight, D j is the observation data of the neighboring points. Weight λ j Determined by the semivariogram model, satisfying the constraints:

[0113]

[0114] Step S103-3, model training and optimization. Use the labeled data set to train the selected model. Convert the data completion problem into an optimization problem, and use an optimization algorithm (such as gradient descent) to minimize the error between the completed data and the actual data. For example, assume the objective function is:

[0115]

[0116] Adjust and complete data through optimization algorithm So that l reaches the minimum value.

[0117] Step S103-4, model evaluation and verification. Evaluate the performance of the model on the validation set and calculate the accuracy and reliability of the model. Evaluation indicators include mean square error, root mean square error, etc. MSE can be calculated using the following formula:

[0118]

[0119] in, is the value predicted by the model. According to the validation results, the model is adjusted and optimized as necessary to improve its performance and stability.

[0120] These steps ensure that the model building process, from feature extraction to final performance evaluation, goes through a rigorous process to ensure the accuracy of the data and the effectiveness of the model.

Claims

1. A method for completing unknown measurement information of a distribution network based on data flow mapping, characterized in that: include: Step 1, construction of data flow mapping model; Step 2: Inference and completion of missing data; Step 3: Real-time update and intelligent optimization; Step 4: System integration and application; Wherein step 1 comprises: Step 1.

1. Data collection and preprocessing; Step 1.2 feature extraction and analysis; Step 1.3 Mapping model construction; Wherein step 2 comprises: Step 2.

1. Identification and inference of missing data; Step 2.

2. Construction of inference model; Step 2.

3. Data completion.

2. According to claim 1, a method for completing unknown measurement information of a distribution network based on data flow mapping is characterized in that: Step 1.

1. Data collection and preprocessing Before building a data flow mapping model, data collection and preprocessing are required to ensure that the data used is accurate and available; this includes collecting historical measurement data, real-time operation data, and cleaning the historical measurement data and real-time operation data to achieve data standardization.

3. According to claim 2, a method for completing unknown measurement information of a distribution network based on data flow mapping is characterized in that: Historical measurement data (D hist ) is the historical data collected from the distribution network system, which includes node voltage (V i ), branch current (I ij ), active power (P i ) and reactive power (Q i ); the above historical data is used to provide status information of the distribution network at different time points; Real-time operation data (D real ) To obtain data under the current operating state, which can reflect the real-time state and dynamic changes of the system; Data cleaning: By removing outliers (D out ) and noise (D noise ), ensuring the quality of the data. Outliers may be caused by equipment failure or measurement errors, while noise may be due to environmental interference or transmission problems; Data standardization: Standardize data from different sources to make them have the same scale. norm ) is convenient for subsequent processing and analysis, and avoids errors caused by different dimensions.

4. The method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1 is characterized in that: Step 1.2 Feature extraction and analysis is as follows: After data preprocessing is completed, it enters the feature extraction and analysis stage. This stage aims to extract key information from a large amount of data and conduct in-depth analysis; it includes power flow feature extraction, principal component analysis, calculation of correlation coefficient matrix and pattern recognition.

5. A method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 4, characterized in that: Power flow feature extraction: Calculate the relationship between voltage, current and power at each node; Including calculating the power flow characteristic vector F between nodes i , obtain the basic characteristics of power flow; Principal component analysis (PCA): Use principal component analysis to reduce the dimension of the data and extract the main features F pca , reducing computational complexity while retaining important information; principal component analysis can map high-dimensional data into low-dimensional space to reveal the main structure of the data; Calculate the correlation coefficient matrix: By calculating the correlation coefficient matrix R between nodes, the power flow relationship between nodes is determined; the correlation coefficient between node voltage, current and power can be expressed as: Pattern recognition: Through data clustering and association rule learning, the main power flow patterns are identified and potential mapping relationships in the data are found.

6. The method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1 is characterized in that: Step 1.3 Mapping model construction Based on feature extraction and analysis, a data flow mapping model is constructed, which can transform the power flow relationship in the data into a mathematical model for predicting and completing missing data; specifically, it includes selecting a suitable model, training a data-driven model, verifying and optimizing the data-driven model, and applying it to actual data.

7. A method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 6, characterized in that: Choose a suitable model: Select a data-driven model and use a graph neural network or a multi-layer perceptron. Graph neural networks are suitable for processing complex node and edge relationships, and are particularly suitable for data with complex topological structures such as distribution networks. Training data-driven models: Using the training set (D train ) The data-driven model is trained; the goal of the data-driven model is to minimize the difference between the predicted value and the true value, and the loss function l is set as: in is the voltage value predicted by the data-driven model, V i is the real voltage value; optimize the data-driven model: use the optimization algorithm to update the model parameters, and adjust the parameters Θ through an iterative process so that the loss function l reaches the minimum value; Verification and optimization of data-driven models: Verify the performance of data-driven models: Use the validation set (D val )Evaluate the performance of the model and calculate the prediction error ε: Use the validation results to determine the model's predictive accuracy and adjust the model as needed; Application to actual data: Apply the trained and verified data-driven model to the actual distribution network data, predict and complete the missing measurement data, monitor the data in real time, and use the model to complete the data to improve the real-time monitoring and control capabilities of the distribution network; Through the above steps, the constructed data flow mapping model can accurately capture the power flow relationship between nodes in the distribution network, and predict and complete missing data.

8. The method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1 is characterized in that: Step 2.

1. Identification and inference of missing data In the distribution network system, data missing problems usually come from equipment failure or communication interruption. First, the missing data needs to be accurately identified and inferred based on the existing data; Assume that the observed data of the distribution network is D = {D i }, where D i is the ith observation value. For missing data points D i miss , define an indicator function δ i : Missing data can be inferred from adjacent nodes or historical data. Assume there is a set of known observation data D known and the inferred model for missing data f(D known ), the inference process is expressed as:

9. The method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1 is characterized in that: Step 2.

2. Construction of inference model In order to accurately infer missing data, we can use the spatial interpolation inference method to build an inference model: for missing data points, we can use the observed data of neighboring nodes for interpolation; let the spatial coordinates of the missing data points be (x i ,y i ), whose value can be estimated by Kriging interpolation or inverse distance weighted (IDW) method. The general formula of Kriging interpolation is: Among them, λ j is the kriging weight, D j is the observation data of the neighboring points, and the weight λ j Determined by the semivariogram model, satisfying the constraints:

10. A method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1, characterized in that: Step 2.

3. Data completion After inferring the missing data, the inferred results need to be completed to improve the integrity of the distribution network data. Suppose the inferred data is D i , the completion process is carried out through the following optimization algorithms: Linear interpolation and polynomial interpolation are common completion methods; for linear interpolation, the missing data D i The neighboring data point D i-1 and D i+1 Perform the calculation: For polynomial interpolation, the Lagrange interpolation formula can be used: The data completion problem is transformed into an optimization problem, and the error between the completed data and the actual data is minimized using an optimization algorithm (such as the gradient descent method); let the objective function be: Adjust and complete data through optimization algorithm So that l reaches the minimum value.

11. A method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1, characterized in that: Step 3: Real-time update and intelligent optimization First, the real-time data stream processing system is responsible for receiving the latest data from the distribution network and integrating it into the data completion process in real time. The real-time data stream contains the observation values ​​of each node in the distribution network, which is used to update the current data status. By merging the newly received observation data with the existing data status, the real-time data processing system can keep the data up to date, so that the completion model can reflect the actual status of the system in a timely manner. In order to achieve real-time data updates, incremental learning or online learning algorithms are introduced to dynamically adjust the parameters of the completion model. Through incremental learning or online learning algorithms, the model can continuously optimize its prediction ability based on the latest data. Including in the incremental learning process, when new data arrives, the model will adjust the parameters according to the new data to minimize the prediction error. This dynamic adjustment is expressed by the following formula Among them, Θ(t) is the model parameter at the current moment, η is the learning rate, is the gradient of the loss function. In terms of intelligent optimization, the performance of data completion is further improved by introducing advanced optimization algorithms; the optimization algorithms include gradient descent method and genetic algorithm; the optimization goal is to minimize the error between the completed data and the actual observed data to achieve the best data completion effect.

12. A method for completing unknown measurement information of a distribution network based on data flow mapping according to claim 1, characterized in that: Step 4: System Integration and Application First, the data completion method needs to be deeply integrated with the dispatching system and fault diagnosis system of the distribution network; the dispatching system relies on accurate real-time data to optimize power distribution and load scheduling, while the fault diagnosis system requires reliable data to identify and locate abnormal conditions in the system; by integrating data completion technology into these systems, it can be ensured that the system can still obtain accurate status information even when data is missing, thereby optimizing dispatching decisions and quickly diagnosing faults, including setting the optimization goal of the dispatching system as: Among them, p k is the output of the kth generating unit, c k is the corresponding cost function; the accuracy of data completion technology directly affects the optimization effect of the objective function, ensuring the rationality and economy of scheduling decisions; During the system integration process, the output of the data completion algorithm needs to be connected to the existing data processing platform. To achieve this goal, the completion results can be integrated into the distribution network monitoring platform through the application programming interface or data exchange format; after data completion, the updated node voltage and branch current data need to be updated through the formula: Among them, V i (t) and I ij (t) represent the completed node voltage and branch current data, ε i and δ ij It is to complete the error; through effective data exchange and integration, ensure that these completed data are delivered and applied to the actual system in a timely manner; At the same time, system integration also needs to consider real-time data update and intelligent optimization. In practical applications, the operating status and data flow of the distribution network are constantly changing. The completion system must have real-time update capabilities. The real-time update process can be expressed by the following formula: D current (t)=D provious ∪D stream (t) In this dynamic environment, data completion algorithms need to be able to process real-time data streams to ensure the timeliness and accuracy of completion results; Finally, the data completion technology in the distribution network system should also be robust to noise and outliers. Data noise and outliers may affect the reliability of the completion results. Therefore, the integrated system needs to include anomaly detection and correction modules. Anomaly detection can be achieved through statistical methods or machine learning algorithms to ensure the quality of data and the credibility of completion results. The data anomaly detection function is: During the system integration and application stage, data completion technology needs to be closely integrated with the distribution network's dispatching system, fault diagnosis system, and data processing platform, and real-time updates and outlier processing need to be achieved.

Images