Distribution network measurement data repair method and device

Abstract

The present invention provides a method and device for repairing distribution network measurement data, which can be applied to the fields of power monitoring technology and power data repair technology. The method includes: in the case of determining that there is missing data in the distribution network measurement data set, determining the data type of the missing data and the data type of each of the multiple measurement data in the distribution network measurement data set; in the case of determining that the data types of the multiple measurement data are different and there is only one missing data at the same monitoring time, determining the first initial repair data of the missing data based on the multiple measurement data at the same monitoring time as the missing data and the electrical correlation function; based on the distribution network measurement data set, generating a distribution network measurement data tensor; performing data repair of tensor decomposition on the distribution network measurement data tensor to obtain second initial repair data matching the missing data; based on the first initial repair data and the second initial repair data, obtaining repair data matching the missing data.

Description

Technical Field

[0001] The present invention relates to the field of power data processing, specifically to the field of power monitoring technology and power data repair technology, and more specifically to a distribution network measurement data repair method and device. Background Art

[0002] With the rapid growth of social economy and the rapid development of the power industry, users have higher requirements for the quality and stability of power supply. In this context, distributed generators (DG), energy storage, and electric vehicles are connected to the distribution network in large numbers, which changes the flow direction in the traditional distribution network from unidirectional to bidirectional, while increasing the complexity of the system grid structure. The distribution network presents normalized random fluctuations and intermittency. As a platform to support demand-side response management, the distribution network is the last link of the power system that directly connects users. Its operating status is directly related to the reliability of power supply and the residents' electricity experience. Therefore, it is very important to efficiently and accurately obtain the real-time status information of the distribution network.

[0003] However, the data quality of distribution network measurement generally has many problems such as low accuracy, data transmission delay, and bad data. In addition, when the communication channel is unstable or the terminal sensor fails, data loss will occur. At this time, data completion, as an effective supplementary means, can simulate or predict system state information that is not directly measured by using existing measurement data and prediction models. Compared with real-time measurement, the measurement data after data completion has a larger error, and the error size directly affects the observability of state estimation. Therefore, it is of great significance to study the data repair method of distribution network measurement. Summary of the invention

[0004] In view of the above problems, the present invention provides a method and device for repairing distribution network measurement data.

[0005] According to a first aspect of the present invention, a method for repairing distribution network measurement data is provided, comprising: when it is determined that there is missing data in a distribution network measurement data set, determining the data type of the missing data and the data type of each of multiple measurement data in the distribution network measurement data set; when it is determined that the data types of the multiple measurement data are different and there is only one missing data at the same monitoring time, determining first initial repair data of the missing data based on multiple measurement data at the same monitoring time as the missing data and an electrical correlation function; based on the distribution network measurement data set, generating a distribution network measurement data tensor, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data; performing data repair by tensor decomposition on the distribution network measurement data tensor to obtain second initial repair data matching the missing data; and obtaining repair data matching the missing data based on the first initial repair data and the second initial repair data.

[0006] According to an embodiment of the present invention, a distribution network measurement data tensor is generated based on a distribution network measurement data set, including: determining a distribution network measurement data sequence including a monitoring moment corresponding to missing data from the distribution network measurement data set, wherein the distribution network measurement data sequence also includes measurement data of multiple other monitoring moments, and the multiple other monitoring moments and the monitoring moment corresponding to the missing data constitute a monitoring cycle; generating a distribution network measurement data matrix based on multiple distribution network measurement data sequences of different monitoring cycles; and generating a distribution network measurement data tensor based on multiple distribution network measurement data matrices of different data types, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data.

[0007] According to an embodiment of the present invention, data repair by performing tensor decomposition on a distribution network measurement data tensor to obtain second initial repair data matching the missing data includes: decomposing the distribution network measurement data tensor according to a data type dimension, a monitoring period dimension, and a monitoring time dimension to obtain a plurality of decomposition factor matrices; and repairing the missing data with the objective function of minimizing the sum of the ranks of the plurality of decomposition factor matrices to obtain second initial repair data matching the missing data.

[0008] According to an embodiment of the present invention, the objective function is to minimize the sum of the ranks of multiple decomposition factor matrices to repair the missing data and obtain second initial repair data matching the missing data, including: using the augmented Lagrangian function to transform the objective function to obtain a relaxed expression function; based on the auxiliary distribution network measurement parameter matrix, the relaxed expression function is transformed to obtain a simplified function; and based on the simplified function, the missing data is repaired to obtain the second initial repair data.

[0009] According to an embodiment of the present invention, the missing data is repaired based on a simplified function to obtain second initial repair data, including: based on the Lagrange multiplier matrix, the simplified function is transformed to obtain a multi-parameter function, wherein the multi-parameter function includes a sub-function corresponding to the auxiliary distribution network measurement parameter matrix, a sub-function corresponding to the Lagrange multiplier matrix, a sub-function corresponding to the distribution network measurement data tensor and a sub-function corresponding to a plurality of decomposition factor matrices; and the variables of other sub-functions except the current sub-function in the fixed multi-parameter function are iterated according to an iteration method matching the current sub-function type; the above iteration process is repeated until a predetermined condition is met, and the second initial repair data is obtained based on the distribution network measurement data tensor of the current round that meets the predetermined condition.

[0010] According to an embodiment of the present invention, the predetermined condition includes: the ratio of the norm of the difference between the current round of distribution network measurement data tensor and the previous round of distribution network measurement data tensor to the norm of the distribution network measurement data is less than or equal to a preset norm threshold.

[0011] According to an embodiment of the present invention, the distribution network measurement data repair method also includes: when it is determined that the data types of multiple measurement data in the distribution network measurement data set are the same and there are measurement data of different multiple usage objects in the distribution network measurement data set, generating an object distribution network measurement data tensor based on the distribution network measurement data set, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data; performing data repair by tensor decomposition on the object distribution network measurement data tensor to obtain third initial repair data matching the missing data; and obtaining repair data matching the missing data based on the third initial repair data.

[0012] According to an embodiment of the present invention, based on a distribution network measurement data set, an object distribution network measurement data tensor is generated, including: determining from the distribution network measurement data set an object distribution network measurement data sequence of a single user object including a monitoring moment corresponding to missing data, wherein the object distribution network measurement data sequence also includes measurement data of multiple other monitoring moments, and the multiple other monitoring moments and the monitoring moment corresponding to the missing data constitute a monitoring cycle; based on multiple object distribution network measurement data sequences of different monitoring cycles, an object distribution network measurement data matrix is ​​generated; and based on multiple distribution network measurement data matrices of different user objects, an object distribution network measurement data tensor is generated.

[0013] According to an embodiment of the present invention, based on the first initial repair data and the second initial repair data, repair data matching the missing data is obtained, including: determining the data difference between the first initial repair data and the second initial repair data; and when it is determined that the data difference is less than or equal to a preset threshold, performing weighted summation on the first initial repair data and the second initial repair data to obtain the repair data.

[0014] A second aspect of the present invention provides a distribution network measurement data repair device, comprising:

[0015] A type determination module, for determining the data type of the missing data and the data types of the plurality of measurement data in the distribution network measurement data set when it is determined that there is missing data in the distribution network measurement data set;

[0016] A first data determination module, configured to determine first initial repair data of the missing data based on the plurality of measurement data at the same monitoring time as the missing data and the electrical correlation function, when it is determined that the data types of the plurality of measurement data are different and there is only one missing data at the same monitoring time;

[0017] A first tensor generation module, configured to generate a distribution network measurement data tensor based on the distribution network measurement data set, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data;

[0018] A second data determination module is used to perform data repair of tensor decomposition of the distribution network measurement data tensor to obtain second initial repair data matching the missing data; and

[0019] The first data repair module is used to obtain repair data matching the missing data based on the first initial repair data and the second initial repair data.

[0020] The distribution network measurement data repair method and device adopted by the present invention can utilize the electrical correlation characteristics of the distribution network measurement data set to repair the missing data through the electrical correlation function to obtain the first initial repair data. In addition, the distribution network measurement data set has the characteristics of periodicity, temporal randomness and electrical correlation, and the distribution network measurement data tensor is subjected to tensor decomposition data repair to obtain the second initial repair data. Multiple repair data obtained by different methods and principles are fused to obtain the final repair data, thereby improving the repair accuracy, accuracy and effectiveness of the repair data while improving the processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0022] Figure 1 A diagram showing an application scenario of a method and device for repairing distribution network measurement data according to an embodiment of the present invention is shown.

[0023] Figure 2 A flow chart of a method for repairing distribution network measurement data according to an embodiment of the present invention is shown.

[0024] Figure 3 A schematic diagram of decomposing a distribution network measurement data tensor according to an embodiment of the present invention is shown.

[0025] Figure 4 A schematic diagram comparing repair results of a distribution network measurement data repair method provided according to an embodiment of the present invention and a traditional repair method is shown.

[0026] Figure 5 A structural block diagram of a distribution network measurement data repair device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0027] Below, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of concepts of the present invention.

[0028] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0029] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0030] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0031] In the technical solution of the present invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0032] In the scenario of using personal information for automated decision-making, the methods, devices, and systems provided by the embodiments of the present invention provide users with corresponding operation portals for users to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered. The expression "automated decision-making" here refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests and hobbies, or economic, health, credit status, etc. through computer programs, and making decisions. The expression "expert decision-making" here refers to the activity of making decisions by people who specialize in a certain field, have specialized experience, knowledge and skills, and have reached a certain level of professionalism.

[0033] State Estimation (SE) is the process of inferring system state variables by using real-time measurement data obtained from traditional SCADA and wide-area measurement systems and historical data obtained from smart power consumption information collection systems. In essence, it is the process of mapping measurement data to the numerical value of state variables. In distribution networks, too few measurement configurations will lead to unobservable problems and state estimation cannot be performed normally. Measurement data is an important data source for state estimation, and the quality and redundancy of its data directly affect the observability, controllability and measurability of distribution network state estimation. Improving the observability of distribution networks can improve the accuracy and precision of distribution network state estimation. SE, as the key to maintaining safe and stable operation of the system, provides reliable data for power system voltage regulation technology, distributed energy output and active load management technology, active distribution network active / reactive coordinated optimization technology, and active distribution network self-healing technology.

[0034] SE has been developed in the transmission network for more than 30 years, and its technology is relatively mature, which is of reference significance for the distribution network. However, unlike the transmission network, the distribution network has the characteristics of radial network, small number of measurement configuration nodes, communication system limiting the accuracy and rate of data exchange, and three-phase load imbalance, which makes the distribution network state estimation need to be expanded and innovated on the basis of the transmission network. At present, the traditional distribution network SE is based on SCADA and AMI, but the distribution system node scale is large, and the measurement device can only be configured in a few nodes, which affects the accuracy of SE to a certain extent.

[0035] Figure 1 A diagram showing an application scenario of a method and device for repairing distribution network measurement data according to an embodiment of the present invention is shown.

[0036] like Figure 1As shown in the figure, in the distribution network, common measurement devices include traditional measurement data acquisition and monitoring control systems (SCADA) and advanced measurement systems (AMI), as well as new measurement systems represented by synchronized phasor measurement units (DPMU). Among them, the SCADA system is an important subsystem of the energy management system, which is widely used and has a high degree of maturity. It includes distribution terminal units (DTU), feeder terminal units (FTU), and transformer terminal units (TTU). Its data collection frequency is from a few seconds to a few minutes. The measurement data includes node injection power, branch power, node voltage amplitude, and branch current amplitude. AMI is generally installed on smart meters of low-voltage users and some transformers and medium-voltage feeders. Its measurement data generally includes node load and branch power, as well as node voltage and branch current amplitude. AMI provides a large number of redundant measurements for the distribution network, breaking through the unobservable limitations of the low-voltage distribution network and improving the measurement accuracy of the medium-voltage distribution network. Compared with traditional measurements, DPMU is installed in power plants and substations, and can also collect the voltage phasor of the node and the current phasor of the connected branch. It also has a high sampling frequency (sampling frequency 25-100Hz) and can achieve time synchronization of sampling data through the global navigation satellite system.

[0037] With the advancement of power system monitoring technology, the application of smart meters and special transformer acquisition terminals has become very common, which provide rich power consumption data for low-voltage area topology networks. However, in the actual power system data collection and processing process, data loss is still an inevitable problem due to various reasons. Especially in SCADA systems and AMI systems, the reasons for data loss include the following:

[0038] 1) Failure of the data collection equipment itself: Smart meters or data collection terminals in SCADA and AMI systems may have incomplete or missing data due to hardware failure or software defects. This failure may be sporadic or continuous, depending on the nature of the failure and the time it takes to repair it. Since SCADA and AMI systems have high requirements for real-time data collection, once data is missing, it is difficult to recover the data through the system itself, which increases the difficulty of data completion.

[0039] Specifically, if the fault causes random or occasional missing measurement values, the method provided by the present invention can effectively estimate the missing values ​​by utilizing the global structure of the data.

[0040] 2) Unstable communication network: SCADA systems and AMI systems rely on stable communication networks to transmit data. When the communication transmission network between the information collection system and the smart meter fails, data transmission will be interrupted, resulting in the terminal being unable to receive valid user data. The data loss caused by this failure usually occurs randomly and has an irregular duration. Random data loss requires the completion model to be more robust to handle irregular missing patterns.

[0041] Specifically, the fault usually leads to random missing of measurement, which may cause the local structure of the data to be destroyed. However, the method provided by the present invention predicts the missing value through the global structure information of the measurement data, so that a relatively high completion result can still be maintained.

[0042] 3) Emergency events: such as large-scale power outages, load shedding or line maintenance, etc., may cause the data acquisition system to temporarily fail, resulting in the loss of continuity of the measured data of the SCADA system and AMI system. The data loss caused by this kind of failure has a certain regularity and lasts for a long time. Long-term continuous missing data will affect the accuracy of the completed data, and the missing values ​​need to be completed with the help of the adjacent data of the missing point or historical data information.

[0043] The continuous missing data caused in this case may pose a challenge to the traditional low-rank tensor completion algorithm, because the continuous missing data may destroy the intrinsic structure of the data. However, the method provided by the present invention can better reconstruct the destroyed measurement parameter structure by introducing tensor CP decomposition, and can eliminate the negative impact of continuous missing data structure to a certain extent.

[0044] 4) Imperfect system design: During the design and implementation of the SCADA system and AMI system, there may be design defects or implementation errors, resulting in incomplete or missing data records. The data loss caused by this kind of failure is usually sporadic and scattered. Although this missing pattern is irregular, the amount of data loss is usually small and relatively easy to complete.

[0045] Since the amount of missing measurement data is usually small and the missing values ​​are usually occasional and scattered, the method provided by the present invention can effectively estimate the missing values ​​by utilizing surrounding data and overall structure.

[0046] 5) Environmental factors: Unpredictable environmental factors, such as magnetic field interference, earthquakes and other natural disasters, may affect the data acquisition equipment in the SCADA system and AMI system, resulting in abnormal or missing measurements. At this time, the location and amount of missing data are random and highly unpredictable, which requires a high level of robustness of the completion model.

[0047] The following will be based on Figure 1 The scene described by Figure 2~Figure 4 The distribution network measurement data repair method according to the embodiment of the present invention is described in detail.

[0048] Figure 2 A flow chart of a method for repairing distribution network measurement data according to an embodiment of the present invention is shown.

[0049] like Figure 2 As shown, the distribution network measurement data repair method of this embodiment includes operations S210 to S250.

[0050] In operation S210, when it is determined that there is missing data in the power distribution network measurement data set, a data type of the missing data and a data type of each of the plurality of measurement data in the power distribution network measurement data set are determined.

[0051] The distribution network measurement data set may include multiple measurement data. Each measurement data may be obtained by Figure 1 The measurement data is collected by the measuring device shown. The type of the measurement data is not limited, and may include, for example, one or more of: node injection power, branch power, node voltage amplitude, and branch current amplitude.

[0052] The distribution network measurement data set may also include missing data, which can be understood as the missing data value corresponding to a certain monitoring moment.

[0053] When it is determined that there is missing data in the distribution network measurement data set, the missing data can be repaired, such as completed, by executing operations S210-S250 to obtain repair data corresponding to the monitoring time. When it is determined that there is no missing data in the distribution network measurement data set, the distribution network system can be directly analyzed using the distribution network measurement data set, so as to perform subsequent regulation on the distribution network system based on the analysis results, thereby improving the accuracy and effectiveness of power regulation.

[0054] The distribution network measurement data set may include measurement data and missing data of the same data type. However, it is not limited thereto. It may also include measurement data and missing data of different data types. For example, the distribution network measurement data set includes power, current, and power supply of distribution network nodes. At monitoring time A, the missing data is power, while the measurement data current and power supply are known.

[0055] In operation S220 , when it is determined that the data types of the plurality of measurement data are different and there is only one missing data at the same monitoring time, first initial repair data of the missing data is determined based on the plurality of measurement data at the same monitoring time as the missing data and the electrical correlation function.

[0056] When it is determined that the data types of the multiple measurement data are different and there is only one missing data at the same monitoring time, the characteristic that the parameters of the distribution network measurement data have electrical correlation can be used to determine the first initial repair data of the missing data based on the multiple measurement data at the same monitoring time as the missing data and the electrical correlation function.

[0057] For example, when an inductive electrical device is started, the active power and current increase at the same time, the power factor decreases, and the active power, current, and power factor change at the same time. Based on the characteristics of the electrical correlation, multiple historical distribution network measurement data can be fitted to obtain an electrical correlation function related to active power, current, and power factor that can reflect the electrical correlation. Thus, multiple measurement data are substituted into the electrical correlation function to obtain the first initial repair data of the missing data.

[0058] For example, the electrical correlation function can also be obtained by fitting in other ways. For example, it can be obtained by calculating according to existing electrical principles. Or it can be obtained by training using a deep learning model. As long as it is an electrical correlation function that can reflect the electrical correlation of multiple measurement data as parameters.

[0059] By using the electrical correlation between multiple measurement data to obtain the electrical correlation function, the accuracy of data repair can be improved by using the effective information of multiple measurement data for calculation. In the data repair of single-user multi-measurement data, the electrical correlation between multiple measurement data such as voltage, current, and power is used to improve the repair accuracy and efficiency of the repaired data.

[0060] In operation S230, a power distribution network measurement data tensor is generated based on the power distribution network measurement data set. The power distribution network measurement data tensor includes data elements corresponding to the missing data.

[0061] The third-order distribution network measurement data tensor can be constructed by utilizing the periodicity, temporal randomness and electrical correlation between measurement parameters of the distribution network measurement data set. The low-rank characteristic of the distribution network measurement data tensor is utilized, and the rank of the distribution network measurement data tensor is minimized as the optimization goal to iterate and obtain the second initial repair data that matches the missing data.

[0062] In operation S240, data repair of tensor decomposition is performed on the distribution network measurement data tensor to obtain second initial repair data matching the missing data.

[0063] However, considering that directly obtaining the minimum rank of the distribution network measurement data tensor will increase the computational difficulty, the distribution network measurement data tensor can be decomposed, and the data of the tensor decomposition can be repaired by using the principle that the sum of the ranks of the decomposition factor matrices after decomposition is the minimum, thereby obtaining the second initial repair data.

[0064] By executing operations S230 and S240, not only are various characteristics of the distribution network measurement data set, such as periodicity, time series randomness, and electrical correlation between measurement parameters, reasonably utilized to obtain the second initial repair data, but also data repair by performing tensor decomposition of the distribution network measurement data tensor reduces the computational complexity and shortens the processing time of data repair.

[0065] In operation S250, repair data matching the missing data is obtained based on the first initial repair data and the second initial repair data.

[0066] Exemplarily, the first initial repair data may be directly used as the final repair data, but this is not limited thereto. The final repair data may also be obtained based on the first initial repair data and the second initial repair data.

[0067] By using two methods to repair missing data, the repair accuracy of the repaired data can be improved, and the effectiveness and accuracy of missing data completion can be improved.

[0068] In summary, the present invention utilizes the electrical correlation characteristics of the distribution network measurement data set, and repairs the missing data through the electrical correlation function to obtain the first initial repair data. In addition, the distribution network measurement data set has the characteristics of periodicity, temporal randomness and electrical correlation, and performs data repair by tensor decomposition of the distribution network measurement data tensor to obtain the second initial repair data. Multiple repair data obtained by different methods and principles are fused to obtain the final repair data, thereby improving the repair accuracy, accuracy and effectiveness of the repair data while improving the processing efficiency.

[0069] Exemplarily, when it is determined that the data types of the plurality of measurement data are different or there is more than one missing data at the same monitoring time, operations S230 and S240 may be directly performed to obtain second initial repair data, and the second initial repair data may be used as repair data.

[0070] In another example, when there are missing data at different multiple monitoring moments in the distribution network measurement data set, the first initial repair data obtained can be directly used as the repair data at the monitoring moment to repair the distribution network measurement data tensor, and obtain the distribution network measurement data tensor repaired at some monitoring moments. Using the distribution network measurement data tensor repaired at some monitoring moments, operations S230 and S240 are performed to deal with the situation where it is determined that the data types of the multiple measurement data are different or there is not only one missing data at the same monitoring moment, thereby improving the processing accuracy of different processing scenarios and expanding the scope of application.

[0071] According to an embodiment of the present invention, Figure 2 The operation S230 shown, based on the distribution network measurement data set, generates a distribution network measurement data tensor, which may include: determining a distribution network measurement data sequence including the monitoring time corresponding to the missing data from the distribution network measurement data set. The distribution network measurement data sequence also includes measurement data of multiple other monitoring times, and the multiple other monitoring times and the monitoring time corresponding to the missing data constitute a monitoring cycle. Based on multiple distribution network measurement data sequences of different monitoring cycles, a distribution network measurement data matrix is ​​generated. Based on multiple distribution network measurement data matrices of different data types, a distribution network measurement data tensor is generated. The distribution network measurement data tensor includes data elements corresponding to the missing data.

[0072] Taking the distribution network measurement data set including measurement data of data types such as active power, current value, voltage value and power factor as an example, the high-order coupling correlation between multiple measurement data can be used to construct a three-order multi-measurement data tensor based on three characteristic orders, namely the distribution network measurement data tensor. The three characteristic orders are the number of measurement days, i.e. multiple monitoring cycles, the number of sampling points within 24 hours (i.e. multiple monitoring moments within a cycle), and the specific number of measurement data parameters. The three characteristic orders correspond to the periodicity of power data, the randomness of time series, and the electrical correlation between measurement parameters.

[0073] Specifically, the smart meter parameters with missing data for 21 consecutive days within 73 days can be selected to form a third-order distribution network measurement data tensor, where the measurement data selected are active power P, voltage U, current I, power factor . Represents the number of measurement days, Represents the number of sampling points in a day, Represents the number of measured data parameters.

[0074] According to an embodiment of the present invention, Figure 2The operation S240 shown, performing data repair of tensor decomposition on the distribution network measurement data tensor to obtain the second initial repair data matching the missing data, may include: decomposing the distribution network measurement data tensor according to the data type dimension, the monitoring cycle dimension and the monitoring time dimension respectively to obtain multiple decomposition factor matrices. Taking the minimum sum of the ranks of the multiple decomposition factor matrices as the objective function, repairing the missing data to obtain the second initial repair data matching the missing data.

[0075] Figure 3 A schematic diagram of decomposing a distribution network measurement data tensor according to an embodiment of the present invention is shown.

[0076] like Figure 3 As shown, the distribution network measurement data tensor is a third-order tensor, which can be expressed as Figure 3 The cube representation shown in the figure shows that the distribution network measurement data tensor is decomposed by CP (Canonical Polyadic Decomposition, CPD) according to three different dimensions: data type dimension, monitoring cycle dimension and monitoring time dimension.

[0077] Decompose into multiple slice vectors consisting of three dimensions. For example, the slice vectors (a1, b1, c1), ..., (a R , b R 、c R ). The decomposition factor matrix can include U1=[a1, a2, …, a R ]、U2=[ b1、b2、……、b R ] and U3=[ c1, c2, …, c R ].

[0078] For distribution network measurement data tensors with missing data , a weighted least squares model with missing data is proposed as follows (1):

[0079] (1)

[0080] Where: represents the Hadamard product, It is called the decomposition factor matrix of the distribution network measurement data tensor, which comes from the combination of a rank tensor, such as U1=[a1, a2, …, a R ],like Figure 3 As shown. R is a positive integer, which is called the rank of CP decomposition. is a A non-negative weight tensor of the same size, see formula (2), which stores the location information of the observable measurement data and missing data in the distribution network measurement data tensor.

[0081] (2)

[0082] This can be done by factoring the matrix sequence To approximate the original distribution network measurement data tensor ,In general, the approximation requires much less storage than the ,distribution network measurement data tensor.

[0083] For example, a distribution network measurement data tensor with missing data can be defined as The rank of Since the complete distribution network measurement data tensor has low rank, the objective function is to continuously minimize the rank of the distribution network measurement data tensor, and establish the trace norm regularized tensor CP decomposition (Trace Norm Regularized CANDECOMP / PARAFAC Decomposition, TNCP) model as follows (3):

[0084] (3)

[0085] Where: and They are respectively the distribution network measurement data tensor with missing data and the target distribution network measurement data tensor after repair; For tensors The set of observable distribution network measurement data locations (a, b, c) in .

[0086] Since it is a difficult problem to calculate the tensor rank, the weighted sum of the rank of the distribution network measurement data tensor is used to replace the rank of the distribution network measurement data tensor. Formula (3) can be transformed into formula (4):

[0087] (4)

[0088] Where: is the weighting factor, which is a constant and satisfies as well as ; is the distribution network measurement data tensor Expand the matrix along the i-th mode. Formula (4) not only solves the non-convex discrete problem of solving the tensor rank of the distribution network measurement data, but also integrates the different characteristics of the distribution network measurement data, such as time series, periodicity, spatial correlation, electrical correlation, etc., through the tensor expansion operation, which is more conducive to the mining of the internal structure of the tensor. The n-th mode expansion matrix of the distribution network measurement data tensor is expressed as formula (5):

[0089] (5)

[0090] From formula (5), we can get the distribution network measurement data tensor: The three mode expansion matrices are shown in formula (6):

[0091] (6)

[0092] set up is the distribution network measurement data tensor of rank r The mode-n expands the matrix, is the decomposition factor matrix of the distribution network measurement data tensor, .Sure The rank of The upper bound of the rank of is, see formula (7).

[0093] (7)

[0094] So it is available to replace , the objective function is transformed into formula (8):

[0095] (8)

[0096] Where: , yes The upper bound of is an integer.

[0097] By using the processing method of the present invention, the decomposition factor matrix can be used to reduce the difficulty of repairing the distribution network measurement data tensor with the minimum rank as the objective function, thereby ensuring the processing accuracy while improving the processing efficiency and reducing the processing difficulty.

[0098] According to an embodiment of the present invention, taking the minimum sum of the ranks of the respective decomposition factor matrices as the objective function, repairing the missing data to obtain the second initial repair data matching the missing data may include: using the augmented Lagrangian function to transform the objective function to obtain a relaxed expression function. Based on the auxiliary distribution network measurement parameter matrix, the relaxed expression function is transformed to obtain a simplified function. Based on the simplified function, the missing data is repaired to obtain the second initial repair data.

[0099] Due to the discrete nature of the rank, formula (8) is a non-convex optimization problem. It can be transformed into a convex optimization problem by replacing the rank function with the trace norm, as shown in the following formula (9):

[0100] (9)

[0101] In the formula, represents the outer product of U1, U2 and U3.

[0102] The augmented Lagrangian function can be established to transform the objective function such as formula (9), remove the equality constraint, and obtain the relaxed expression function as follows formula (10):

[0103] (10)

[0104] Formula (10) can be called the TNCP result of simultaneously decomposing and completing the distribution network measurement data tensor. It is only necessary to perform SVD operations on some smaller decomposition factor matrices to obtain the repair data of the missing data. For example, the distribution network measurement data repair method provided in this embodiment reduces the algorithm complexity and shortens the iteration time.

[0105] For example, the relaxation expression function in formula (10) has interdependent matrix nuclear norm terms, that is, when optimizing the sum of multiple matrix nuclear norms, these matrices share the same terms and cannot be optimized separately.

[0106] The auxiliary distribution network measurement parameter matrix can be used to transform the relaxation expression function to obtain a simplified function as shown in formula (12).

[0107] By splitting these interdependent terms, they can be solved independently. Specifically, the auxiliary distribution network measurement parameter matrix is ​​introduced The equivalent formula (11) of formula (10) is obtained, namely:

[0108] (11)

[0109] By using the auxiliary distribution network measurement parameter matrix to further transform the relaxed expression function, a simplified function can be obtained, which further reduces the difficulty of repairing missing data using the simplified function and improves the repair accuracy of the second initial repair data.

[0110] According to an embodiment of the present invention, repairing the missing data based on a simplified function to obtain the second initial repair data may include: transforming the simplified function based on the Lagrange multiplier matrix to obtain a multi-parameter function. The multi-parameter function includes a sub-function corresponding to the auxiliary distribution network measurement parameter matrix, a sub-function corresponding to the Lagrange multiplier matrix, a sub-function corresponding to the distribution network measurement data tensor, and a sub-function corresponding to a plurality of decomposition factor matrices. The variables of other sub-functions in the multi-parameter function except the current sub-function are fixed, and the current sub-function is iterated in an iteration method matching the current sub-function type. The above-mentioned iterative process is repeated until a predetermined condition is met, and the second initial repair data is obtained based on the distribution network measurement data tensor of the current round that meets the predetermined condition.

[0111] Specifically, the simplified function is an optimization problem with equality constraints. The Lagrange multiplier matrix can be constructed to further transform the simplified function to obtain the multi-parameter function shown in formula (12).

[0112] (12)

[0113] Where: is the Lagrange multiplier matrix, where n = 1, 2, 3; is the inner product of the matrix; is the norm of the matrix; is the regularization parameter; is the penalty parameter.

[0114] Optionally, you can Continuous Minimization , then update Therefore, the complex optimization problem in formula (12) can be simplified into four sub-problems: iteratively solving the sub-function corresponding to the auxiliary distribution network measurement parameter matrix, the sub-function corresponding to the Lagrange multiplier matrix, the sub-function corresponding to the distribution network measurement data tensor, and the sub-function corresponding to multiple decomposition factor matrices.

[0115] Specifically, solve Sub-problems can be fixed , and , then the original problem is transformed into the following formula (13):

[0116] (13)

[0117] set up , then formula (13) can be expressed as formula (14):

[0118] (14)

[0119] at last The solution is as follows:

[0120] (15)

[0121] Solution Subproblem, fixed , and , then the original problem is transformed into the following formula (16):

[0122] (16)

[0123] The closed-form solution of the subproblem of formula (16) is obtained by the singular value thresholding algorithm (SVT), see formula (17):

[0124] (17)

[0125] Where: Indicates Perform soft thresholding on the threshold.

[0126] Similarly, solve Subproblem, fixed , and By introducing the Karush-Kuhn-Tucker (KKT) condition, the distribution network measurement data tensor is obtained The solution is as follows:

[0127] (18)

[0128] Where: yes The complement of is the index set of the unobserved entries in the distribution network measurement data tensor, is the projection operator.

[0129] Lagrange multiplier matrix The iterative form of is as follows:

[0130] (19)

[0131] According to an embodiment of the present invention, the predetermined condition may include: the ratio of the norm of the difference between the current round of distribution network measurement data tensor and the previous round of distribution network measurement data tensor to the norm of the distribution network measurement data is less than or equal to a preset norm threshold.

[0132] The above four sub-processes are solved alternately and iteratively. When The tensor completion result of the distribution network measurement data is used to determine the value corresponding to the monitoring time of the missing data from the tensor completion result of the distribution network measurement data as the second initial repair data.

[0133] The preset norm threshold may include 10 -4 , but it is not limited to this, and it can also be adjusted according to actual conditions, as long as the preset norm threshold can be determined to determine reasonable predetermined conditions.

[0134] By utilizing the repair method provided in the embodiment of the present invention, the processing difficulty can be simplified, the iteration efficiency can be improved, and the processing accuracy can be guaranteed by transforming the function multiple times.

[0135] According to an embodiment of the present invention, Figure 2 The operation S250 shown, based on the first initial repair data and the second initial repair data, obtaining repair data matching the missing data may include: determining a data difference between the first initial repair data and the second initial repair data. When it is determined that the data difference is less than or equal to a preset threshold, weighted summing the first initial repair data and the second initial repair data is performed to obtain the repair data.

[0136] The first initial repair data and the second initial repair data can be directly weighted and summed to obtain the repair data, but it is not limited to this. The first initial repair data and the second initial repair data can also be compared to determine the data difference between them. When the absolute value of the data difference is less than or equal to the preset threshold, it can be determined that the repair effects of the two are accurate and effective. The first initial repair data or the second initial repair data can be used as the repair data. The result of the weighted sum of the first initial repair data and the second initial repair data can also be used as the repair data.

[0137] The first initial repair data and the second initial repair data are used as judgments to obtain repair data, which can be combined through a variety of means to further improve processing accuracy and ensure that the repair result is accurate and effective.

[0138] According to an embodiment of the present invention, Figure 2 Before the operation S210 shown, the distribution network measurement data repair method may further include: when it is determined that the data types of the plurality of measurement data in the distribution network measurement data set are the same and there are measurement data of different plurality of use objects in the distribution network measurement data set, generating a distribution network measurement data tensor for the object based on the distribution network measurement data set. The distribution network measurement data tensor includes data elements corresponding to the missing data. Performing data repair of tensor decomposition on the distribution network measurement data tensor for the object to obtain third initial repair data matching the missing data. Based on the third initial repair data, obtaining repair data matching the missing data.

[0139] Optionally, the distribution network measurement data set may include a situation where a single measurement data is missing, for example, multiple users collect measurement data of the same data type in the same monitoring cycle and at the same monitoring time of the same monitoring cycle. For example, the measurement data includes active power consumption.

[0140] Specifically, based on the distribution network measurement data set, the method of generating the object distribution network measurement data tensor is similar to the method of generating the distribution network measurement data tensor based on the distribution network measurement data set. The difference is that the data type dimension is replaced by the object dimension. For example, the active power value of electricity is composed of three dimensions: the number of measurement days, the number of sampling points within 24 hours, and the number of users to form a third-order tensor. The periodicity, temporal randomness and spatial correlation of electricity data are respectively dealt with. In order to deeply explore the high-order correlation of multi-user active power data, a comprehensive strategy is adopted, which combines the tensor model with the singular value decomposition (SVD) technology, aiming to effectively deal with the missing value problem by capturing the potential structure in the data.

[0141] The data of 30 days in a month are selected to form the third-order distribution network measurement tensor. Represents the number of days in the month, Represents the number of time segments in a day. Represents the number of users in the area.

[0142] According to an embodiment of the present invention, generating an object distribution network measurement data tensor based on a distribution network measurement data set may include: determining an object distribution network measurement data sequence of a single user object including a monitoring moment corresponding to missing data from the distribution network measurement data set. The object distribution network measurement data sequence also includes measurement data of multiple other monitoring moments, and the multiple other monitoring moments and the monitoring moment corresponding to the missing data constitute a monitoring cycle. Based on multiple object distribution network measurement data sequences of different monitoring cycles, an object distribution network measurement data matrix is ​​generated. Based on multiple distribution network measurement data matrices of different users, an object distribution network measurement data tensor is generated.

[0143] According to an embodiment of the present invention, performing data repair of tensor decomposition on the object distribution network measurement data tensor to obtain the third initial repair data matching the missing data may include: decomposing the object distribution network measurement data tensor according to the object dimension, the monitoring cycle dimension and the monitoring time dimension respectively to obtain multiple object decomposition factor matrices. Taking the minimum sum of the ranks of the multiple object decomposition factor matrices as the objective function, the missing data is repaired to obtain the third initial repair data matching the missing data.

[0144] Specifically, the missing data is repaired by taking the minimum sum of the ranks of multiple object decomposition factor matrices as the objective function to obtain the third initial repair data matching the missing data. For the specific repair method, please refer to formula (1) to formula (19), which will not be repeated here.

[0145] The third initial repair data can be directly used as the repair data. However, it is not limited thereto. The third initial repair data can also be adjusted, for example, the third initial repair data and the measurement data at adjacent monitoring moments are weighted and summed to obtain the repair data that matches the missing data.

[0146] When it is determined that the data types of multiple measurement data in the distribution network measurement data set are the same and there are measurement data of different multiple usage objects in the distribution network measurement data set, an object distribution network measurement data tensor is generated based on the distribution network measurement data set, and a third initial repair data is generated to expand the application scope. By utilizing the data characteristics of the same distribution network measurement data set and using similar tensor construction methods and processing algorithms, the protection scope is expanded while improving the targeted processing and the accuracy of repair.

[0147] For example, random data anomalies or missing values ​​due to environmental factors may increase the difficulty of the TNCP algorithm because these outliers may interfere with the algorithm's recognition of the low-rank structure of the data. Figure 2 Before the operation S210 shown, outlier detection and processing can be performed on the measurement data in the distribution network measurement data set to ensure the accuracy of the completion. Outlier detection can include comparing the measurement data with an outlier threshold, determining that the measurement data is an outlier if it does not meet the outlier threshold, and determining that the measurement data is a normal value if it meets the outlier threshold.

[0148] The processing of the outliers may include: deleting the outliers, but is not limited thereto, and may also include replacing the outliers with adjacent normal measurement data.

[0149] By identifying outliers, the impact of noise data on repair can be reduced and the accuracy of subsequent repairs can be improved.

[0150] In general, the traditional low-rank tensor completion algorithm relies on the inherent low-rank structure of the data when dealing with the problem of missing distribution network measurement data sets. In order to improve the completion effect, the present invention proposes a TNCP algorithm, which introduces tensor CP decomposition on the basis of the traditional low-rank tensor completion algorithm and uses the decomposition factor matrix instead of the tensor module expansion matrix, which not only reduces the computational complexity and shortens the calculation time, but more importantly, compared with the traditional low-rank tensor completion algorithm, the proposed algorithm has a stronger ability to mine the internal structure of distribution network measurement and is more robust.

[0151] Figure 4 A schematic diagram comparing repair results of a distribution network measurement data repair method provided according to an embodiment of the present invention and a traditional repair method is shown.

[0152] The distribution network measurement data repair method provided by the present invention is referred to as the TNCP method. Traditional repair methods include: traditional HaLRTC method, cubic spline interpolation method, Kalman filtering method and low-rank matrix completion (LRMC) method.

[0153] The TNCP method and the traditional repair method are used to process the same distribution network measurement data including missing data with different missing rates to complete the missing measurement data of the distribution network. The relative squared error (RSE) is used to analyze the completion effect of the algorithm on the distribution network measurement data with different missing rates.

[0154] like Figure 4 Through comparison, it is concluded that the method proposed in the present invention has higher completion accuracy than the traditional HaLRTC method in processing missing measurement data of distribution network compared with other completion algorithms; and the method proposed in the present invention has better completion performance and pays less computational cost compared with the traditional HaLRTC method.

[0155] Specifically, the proposed algorithm and the traditional HaLRTC algorithm have significantly better completion effects than other traditional completion algorithms at all missing rates, so only the proposed algorithm and the traditional HaLRTC algorithm are compared in completion speed. Compared with the traditional HaLRTC algorithm, the proposed algorithm has roughly the same completion error at low missing rates, but the completion error of the proposed algorithm is significantly lower than that of the traditional HaLRTC algorithm at high missing rates; and it can be seen from Table 1 that the proposed algorithm takes less time than the traditional HaLRTC algorithm to complete the distribution network measurement data with missing values ​​at any missing rate, which proves that the proposed algorithm has better performance and higher efficiency.

[0156] It can be concluded that the algorithm proposed in the present invention can efficiently, accurately and reasonably fill in the missing values ​​in the measurement of AMI system quantities in the distribution network, thereby improving the quality of the data, thereby improving the observability of the distribution network, and improving the accuracy and precision of the distribution network state estimation.

[0157] Table 1 Comparison of the repair results of the distribution network measurement data repair method and the traditional repair method

[0158]

[0159] Based on the above distribution network measurement data repair method, the present invention also provides a distribution network measurement data repair device. Figure 5 The device is described in detail.

[0160] Figure 5 A structural block diagram of a distribution network measurement data repair device according to an embodiment of the present invention is shown.

[0161] like Figure 5 As shown, the distribution network measurement data repairing device 500 of this embodiment includes a type determination module 510 , a first data determination module 520 , a first tensor generation module 530 , a second data determination module 540 and a first data repairing module 550 .

[0162] The type determination module 510 is used to determine the data type of the missing data and the data type of each of the multiple measurement data in the distribution network measurement data set when it is determined that there is missing data in the distribution network measurement data set. In one embodiment, the type determination module 510 can be used to perform the operation S210 described above, which will not be repeated here.

[0163] The first data determination module 520 is used to determine the first initial repair data of the missing data based on the multiple measurement data at the same monitoring time as the missing data and the electrical correlation function when it is determined that the data types of the multiple measurement data are different and there is only one missing data at the same monitoring time. In one embodiment, the first data determination module 520 can be used to perform the operation S220 described above, which will not be repeated here.

[0164] The first tensor generation module 530 is used to generate a distribution network measurement data tensor based on the distribution network measurement data set, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data. In one embodiment, the first tensor generation module 530 can be used to perform the operation S230 described above, which will not be repeated here.

[0165] The second data determination module 540 is used to perform data repair of tensor decomposition of the distribution network measurement data tensor to obtain second initial repair data matching the missing data. In one embodiment, the second data determination module 540 can be used to perform the operation S240 described above, which will not be repeated here.

[0166] The first data repair module 550 is used to obtain repair data matching the missing data based on the first initial repair data and the second initial repair data. In one embodiment, the first data repair module 550 can be used to perform the operation S250 described above, which will not be described in detail here.

[0167] According to an embodiment of the present invention, the first tensor generation module 530 includes a first sequence determination submodule, a first matrix generation submodule and a first tensor generation submodule.

[0168] The first sequence determination submodule is used to determine a distribution network measurement data sequence including the monitoring time corresponding to the missing data from the distribution network measurement data set, wherein the distribution network measurement data sequence also includes the measurement data of multiple other monitoring times, and the multiple other monitoring times and the monitoring time corresponding to the missing data constitute a monitoring cycle.

[0169] The first matrix generating submodule is used to generate a distribution network measurement data matrix based on a plurality of distribution network measurement data sequences of different monitoring periods.

[0170] The first tensor generation submodule is used to generate a distribution network measurement data tensor based on a plurality of distribution network measurement data matrices of different data types, wherein the distribution network measurement data tensor includes data elements corresponding to missing data.

[0171] According to an embodiment of the present invention, the second data determination module 540 includes a tensor decomposition submodule and a data repair submodule.

[0172] The tensor decomposition submodule is used to decompose the distribution network measurement data tensor according to the data type dimension, monitoring cycle dimension and monitoring time dimension to obtain multiple decomposition factor matrices.

[0173] The data repair submodule is used to repair the missing data by taking the minimum sum of the ranks of multiple decomposition factor matrices as the objective function to obtain second initial repair data that matches the missing data.

[0174] According to an embodiment of the present invention, the data repair submodule includes a first function transformation unit, a second function transformation unit and a data repair unit.

[0175] The first function transformation unit is used to transform the objective function by using the augmented Lagrangian function to obtain a relaxed expression function.

[0176] The second function transformation unit is used to transform the relaxation expression function based on the auxiliary distribution network measurement parameter matrix to obtain a simplified function.

[0177] The data repair unit is used to repair the missing data based on the simplified function to obtain second initial repair data.

[0178] According to an embodiment of the present invention, the data repair unit includes a function transformation subunit, a function iteration subunit and a data determination subunit.

[0179] The function transformation subunit is used to transform the simplified function based on the Lagrange multiplier matrix to obtain a multi-parameter function, wherein the multi-parameter function includes a sub-function corresponding to the auxiliary distribution network measurement parameter matrix, a sub-function corresponding to the Lagrange multiplier matrix, a sub-function corresponding to the distribution network measurement data tensor and a sub-function corresponding to multiple decomposition factor matrices.

[0180] The function iteration subunit is used to fix the variables of other sub-functions in the multi-parameter function except the current sub-function, and iterate the current sub-function according to the iteration method matching the current sub-function type.

[0181] The data determination subunit is used to repeatedly execute the above iterative process until a predetermined condition is met, and obtain the second initial repair data based on the current round of distribution network measurement data tensor that meets the predetermined condition.

[0182] According to an embodiment of the present invention, the distribution network measurement data repairing device 500 further includes a second tensor generating module, a third data determining module and a second data repairing module.

[0183] The second tensor generation module is used to generate an object distribution network measurement data tensor based on the distribution network measurement data set when it is determined that the data types of multiple measurement data in the distribution network measurement data set are the same and there are measurement data of different multiple usage objects in the distribution network measurement data set, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data.

[0184] The third data determination module is used to perform data repair by performing tensor decomposition on the object distribution network measurement data tensor to obtain third initial repair data matching the missing data.

[0185] The second data repair module is used to obtain repair data matching the missing data based on the third initial repair data.

[0186] According to an embodiment of the present invention, the second tensor generation module includes a second sequence determination submodule, a second matrix generation submodule and a second tensor generation submodule.

[0187] The second sequence determination submodule is used to determine, from the distribution network measurement data set, an object distribution network measurement data sequence of a single user object including the monitoring moment corresponding to the missing data, wherein the object distribution network measurement data sequence also includes the measurement data of multiple other monitoring moments, and the multiple other monitoring moments and the monitoring moment corresponding to the missing data constitute a monitoring cycle.

[0188] The second matrix generating submodule is used to generate a target distribution network measurement data matrix based on a plurality of target distribution network measurement data sequences of different monitoring periods.

[0189] The second tensor generation submodule is used to generate an object distribution network measurement data tensor based on multiple distribution network measurement data matrices of different usage objects.

[0190] According to an embodiment of the present invention, the first data repair module 550 includes a difference determination submodule and a data summation submodule.

[0191] The difference determination submodule is used to determine the data difference between the first initial repair data and the second initial repair data.

[0192] The data summing submodule is used to perform weighted summation on the first initial repair data and the second initial repair data to obtain the repair data when it is determined that the data difference is less than or equal to a preset threshold.

[0193] According to an embodiment of the present invention, any multiple modules of the type determination module 510, the first data determination module 520, the first tensor generation module 530, the second data determination module 540 and the first data repair module 550 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the type determination module 510, the first data determination module 520, the first tensor generation module 530, the second data determination module 540 and the first data repair module 550 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or by any one of the three implementation methods of software, hardware and firmware or by a suitable combination of any of them. Alternatively, at least one of the type determination module 510, the first data determination module 520, the first tensor generation module 530, the second data determination module 540 and the first data repair module 550 can be at least partially implemented as a computer program module, which can perform the corresponding function when the computer program module is executed.

[0194] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0195] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present invention may be combined and / or combined in various ways, even if such combinations and / or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments and / or claims of the present invention may be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All of these combinations and / or combinations fall within the scope of the present invention.

[0196] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination. The scope of the present invention is defined by the attached claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A method for repairing distribution network measurement data, characterized in that: The method comprises: In the case of determining that there is missing data in the distribution network measurement data set, determining the data type of the missing data and the data type of each of the plurality of measurement data in the distribution network measurement data set; When it is determined that the data types of the plurality of measurement data are different and there is only one missing data at the same monitoring time, determining first initial repair data of the missing data based on the plurality of measurement data at the same monitoring time as the missing data and the electrical correlation function; Based on the distribution network measurement data set, generating a distribution network measurement data tensor, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data; Performing data repair by tensor decomposition on the distribution network measurement data tensor to obtain second initial repair data matching the missing data; and Based on the first initial repair data and the second initial repair data, repair data matching the missing data is obtained.

2. The method according to claim 1, characterized in that The generating a distribution network measurement data tensor based on the distribution network measurement data set includes: Determine a distribution network measurement data sequence including the monitoring time corresponding to the missing data from the distribution network measurement data set, wherein the distribution network measurement data sequence also includes measurement data of multiple other monitoring times, and the multiple other monitoring times and the monitoring time corresponding to the missing data constitute a monitoring cycle; generating a distribution network measurement data matrix based on a plurality of distribution network measurement data sequences of different monitoring periods; and Based on a plurality of the distribution network measurement data matrices of different data types, the distribution network measurement data tensor is generated, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data.

3. The method according to claim 1 or 2, characterized in that: The data repairing of performing tensor decomposition on the distribution network measurement data tensor to obtain second initial repair data matching the missing data includes: Decomposing the distribution network measurement data tensor according to the data type dimension, the monitoring cycle dimension and the monitoring time dimension respectively to obtain a plurality of decomposition factor matrices; and The missing data is repaired by taking the minimization of the sum of the ranks of the plurality of decomposition factor matrices as the objective function to obtain the second initial repaired data matching the missing data.

4. The method according to claim 3, characterized in that: The method of repairing the missing data by taking the sum of the ranks of the plurality of decomposition factor matrices as the minimum as the objective function to obtain the second initial repaired data matching the missing data includes: The objective function is transformed by using an augmented Lagrangian function to obtain a relaxed expression function; Based on the auxiliary distribution network measurement parameter matrix, transforming the relaxation expression function to obtain a simplified function; and The missing data is repaired based on the simplified function to obtain the second initial repair data.

5. The method according to claim 4, characterized in that The repairing the missing data based on the simplified function to obtain the second initial repair data includes: Based on the Lagrange multiplier matrix, the simplified function is transformed to obtain a multi-parameter function, wherein the multi-parameter function includes a sub-function corresponding to the auxiliary distribution network measurement parameter matrix, a sub-function corresponding to the Lagrange multiplier matrix, a sub-function corresponding to the distribution network measurement data tensor, and a sub-function corresponding to a plurality of the decomposition factor matrices; and Fix the variables of other sub-functions in the multi-parameter function except the current sub-function, and iterate the current sub-function in an iteration mode matching the type of the current sub-function; The iterative process is repeatedly performed until a predetermined condition is met, and the second initial repair data is obtained based on the current round of distribution network measurement data tensor that meets the predetermined condition.

6. The method according to claim 5, characterized in that The predetermined condition includes: the ratio of the norm of the difference between the current round of distribution network measurement data tensor and the previous round of distribution network measurement data tensor to the norm of the distribution network measurement data is less than or equal to a preset norm threshold.

7. The method according to claim 1, characterized in that The method further comprises: In a case where it is determined that the data types of the plurality of measurement data in the distribution network measurement data set are the same and the distribution network measurement data set contains measurement data of different plurality of user objects, generating an object distribution network measurement data tensor based on the distribution network measurement data set, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data; Performing data repair by tensor decomposition on the object distribution network measurement data tensor to obtain third initial repair data matching the missing data; and Based on the third initial repair data, repair data matching the missing data is obtained.

8. The method according to claim 7, characterized in that The step of generating a target distribution network measurement data tensor based on the distribution network measurement data set includes: Determine, from the distribution network measurement data set, a target distribution network measurement data sequence of a single user object including the monitoring time corresponding to the missing data, wherein the target distribution network measurement data sequence also includes measurement data of multiple other monitoring times, and the multiple other monitoring times and the monitoring time corresponding to the missing data constitute a monitoring cycle; generating a target distribution network measurement data matrix based on a plurality of target distribution network measurement data sequences of different monitoring periods; and Based on a plurality of the power distribution network measurement data matrices of different usage objects, the object power distribution network measurement data tensor is generated.

9. The method according to claim 1, characterized in that: The obtaining, based on the first initial repair data and the second initial repair data, repair data matching the missing data comprises: determining a data difference between the first initial repair data and the second initial repair data; and When it is determined that the data difference is less than or equal to a preset threshold, a weighted sum is performed on the first initial repair data and the second initial repair data to obtain the repair data.

10. A distribution network measurement data repair device, characterized in that: The device comprises: A type determination module, for determining the data type of the missing data and the data type of each of the plurality of measurement data in the distribution network measurement data set when it is determined that there is missing data in the distribution network measurement data set; A first data determination module is used to determine, when it is determined that the data types of the plurality of measurement data are different and there is only one missing data at the same monitoring time, first initial repair data of the missing data based on the plurality of measurement data at the same monitoring time as the missing data and the electrical correlation function; A first tensor generation module, configured to generate a distribution network measurement data tensor based on the distribution network measurement data set, wherein the distribution network measurement data tensor includes data elements corresponding to the missing data; A second data determination module is used to perform data repair of tensor decomposition on the distribution network measurement data tensor to obtain second initial repair data matching the missing data; and The first data repair module is used to obtain repair data matching the missing data based on the first initial repair data and the second initial repair data.

Images