Data alignment method and system
By constructing multi-dimensional time series and energy interaction operator similarity calculation, the problem of inconsistency in power system data caused by ignoring important variables in the prior art is solved, and more accurate and reliable data alignment is achieved, ensuring the global consistency of power grid data.
Patent Information
- Application Number
- CN202410944087.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-15
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art performs correlation analysis on one-dimensional measurement data in power systems, ignoring other potential important variables or factors, resulting in incomplete and inconsistent global data, especially when synchronizing PMU and RTU data, there are errors and Problem of insufficient accuracy.
By constructing a multidimensional time series, using preset angle constraint thresholds and ABR algorithms to extract important time points, combining energy interaction operators to perform similarity calculations, obtain the multidimensional time series with the greatest correlation, and determine the time of RTU data using PMU data as the benchmark. to achieve data alignment.
It improves the accuracy and reliability of data alignment, ensures the global consistency of power grid data, reduces the impact of complex multi-dimensional data structures on data consistency, and provides more comprehensive information and more efficient computing efficiency.
Smart Images

Figure CN118940049A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power system dispatching automation, and in particular to a data alignment method and system. Background Art
[0002] With the development of science and technology, the measurement devices of power systems are constantly upgraded, and the quantitative analysis of power systems can be completed more accurately and quickly. Among them, PMU technology has been widely used in power systems at home and abroad, but PMU measurement can no longer meet the needs of power systems, and RTU measurement is still the main source of information. Therefore, in practical applications, it is necessary to synchronize the data of the two.
[0003] At present, most of the existing technologies use advanced monitoring data collection systems (SCADA), which can collect and analyze various types of quantitative analysis, such as the active and reactive power provided by the line nodes, the active and reactive power transmitted by the branches, and the changes in the bus current. The data obtained by the system is transmitted through a remote transmission device. The errors of the transmission device and the errors generated in each link during the transmission process make the real-time accuracy of the voltage phase angle poor and the precision low, and the SCADA device cannot measure the phase angle of the line node; and the existing technology is based on the GPS system to upgrade the measurement device to form a phase angle measurement device (PMU). PMU can measure synchronous phasors in real time under synchronous clock signals and maintain the signal punctuality in the out-of-synchronization state. However, compared with SCADA equipment, PMU equipment is more expensive and more difficult to maintain. Therefore, the domestic power system is insufficiently deployed, and it is impossible to ensure that each node has PMU measurement, but domestic SCADA devices are almost everywhere in all measurement nodes. Therefore, some existing technologies use data correlation theory, take the maximum correlation coefficient as the objective function, add time stamps to SCADA data, and thus achieve time section consistency; some existing technologies find a set of data within the SCADA measurement cycle that minimizes the difference with the PMU data, and then the time stamp corresponding to the SCADA data at this time is the PMU time stamp at that moment, thereby adding time stamps to the SCADA data to achieve synchronization. However, existing technologies generally perform correlation analysis on one-dimensional measurement data, which may ignore the influence of other potential important variables or factors, resulting in problems such as incomplete global data and inconsistent global data. Summary of the invention
[0004] Based on the above problems, the present invention proposes a data alignment method and system to solve the problem that the prior art generally performs correlation analysis on one-dimensional measurement data and ignores the influence of other potential important variables or factors, resulting in global data inconsistency.
[0005] To achieve the above object, an embodiment of the present invention provides a data alignment method, comprising:
[0006] Acquire historical PMU measurement data and historical RTU measurement data, and construct a first PMU multidimensional time series and a first RTU multidimensional time series;
[0007] Based on a preset angle constraint threshold, extracting important time points from the first PMU multidimensional time series and the first RTU multidimensional time series through a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series;
[0008] Based on a preset energy interaction operator, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series by using a preset similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients;
[0009] Based on the PMU correlation coefficient and the RTU correlation coefficient, acquiring the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation, and obtaining a third PMU multidimensional time series and a third RTU multidimensional time series;
[0010] Based on the data of the third PMU multidimensional time series, the data time stamp of the third RTU multidimensional time series is determined to obtain a data alignment result.
[0011] The embodiment of the present invention proposes a data alignment method, which constructs a multidimensional time series for subsequent data alignment steps by acquiring relevant data of historical moments, and can reveal the complex relationship between multiple variables through analysis of multidimensional data, and provide more comprehensive information; due to the influence of factors such as random fluctuations and measurement errors in the power grid, it is easy to produce errors when mining multidimensional measurement data directly. By selecting important time points, redundant data can be removed, the efficiency, accuracy and reliability of the algorithm can be improved, and the accuracy of data alignment data can be guaranteed; an energy interaction operator is introduced to calculate the similarity of the time series and obtain the multidimensional time series with the maximum correlation. The energy interaction operator is used to characterize the relationship between a certain time point and its previous and subsequent moments, and the reflection of the global consistency of the data is reduced due to the unpredictable complex structure of the multidimensional data; finally, the time stamp of the RTU data is determined based on the PMU data, and the data alignment is completed to ensure the consistency of the global data.
[0012] Further, the acquisition of historical PMU measurement data and historical RTU measurement data to construct a first PMU multidimensional time series and a first RTU multidimensional time series is specifically as follows:
[0013] Obtain historical PMU measurement data and historical RTU measurement data;
[0014] Based on a preset time length, the PMU measurement data at the historical moment and the RTU measurement data at the historical moment are time-sequenced to obtain a first PMU multi-dimensional time series and an initial RTU multi-dimensional time series;
[0015] Based on a preset time step, the initial RTU multi-dimensional time series is subjected to an offset time RTU measurement acquisition to obtain a first RTU multi-dimensional time series.
[0016] A data alignment method proposed in an embodiment of the present invention constructs multidimensional measurement time series data. It is necessary to organize the status data of historical moments into multidimensional time series according to a certain time length, make full use of power grid measurement information, reveal the complex relationship between multidimensional data, and lay a solid foundation for subsequent data consistency analysis.
[0017] Further, based on the preset angle constraint threshold, the first PMU multidimensional time series and the first RTU multidimensional time series are extracted with important time points by a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series, specifically:
[0018] Acquire a first PMU intermediate data point and a first RTU intermediate data point at which an angle is formed by three consecutive data points in the first PMU multidimensional time series and the first RTU multidimensional time series;
[0019] Based on a preset angle constraint threshold, an ABR algorithm is used to respectively judge the important data points of the first PMU intermediate data point and the first RTU intermediate data point, to obtain a PMU data point importance judgment result and an RTU data point importance judgment result;
[0020] The PMU data point importance judgment result and the RTU data point importance judgment result are respectively subjected to important time point extraction and time series reconstruction to obtain a second PMU multi-dimensional time series and a second RTU multi-dimensional time series.
[0021] Further, based on the preset angle constraint threshold, the ABR algorithm is used to respectively judge the important data points of the first PMU intermediate data point and the first RTU intermediate data point, and obtain the PMU data point importance judgment result and the RTU data point importance judgment result, which are specifically:
[0022] Based on the ABR algorithm, when it is determined that the angle between the first PMU intermediate data point and the adjacent data point is greater than the preset angle constraint threshold, the first PMU intermediate data point is judged to be an important time point;
[0023] Based on the ABR algorithm, when it is determined that the angle between the first PMU intermediate data point and the adjacent data point is less than or equal to the preset angle constraint threshold, the first PMU intermediate data point is judged to be a non-important time point.
[0024] Further, based on the preset angle constraint threshold, the important data point judgment is performed on the first PMU intermediate data point and the first RTU intermediate data point respectively through the ABR algorithm to obtain the PMU data point importance judgment result and the RTU data point importance judgment result, and also includes:
[0025] Based on the ABR algorithm, when it is determined that the angle between the first RTU intermediate data point and the adjacent data point is greater than the preset angle constraint threshold, the first RTU intermediate data point is judged to be an important time point;
[0026] Based on the ABR algorithm, when it is determined that the angle between the first RTU intermediate data point and the adjacent data point is less than or equal to the preset angle constraint threshold, the first RTU intermediate data point is judged to be a non-important time point.
[0027] A data alignment method proposed in an embodiment of the present invention adopts an ABR algorithm to select a part of points from multidimensional measurement sequence data as important nodes to fit the original sequence, thereby improving the calculation efficiency by compressing the multidimensional measurement time series data while retaining the sequence information as much as possible; and selecting and retaining important points to improve the overall reliability of the data.
[0028] Further, the PMU data point importance judgment result and the RTU data point importance judgment result are respectively subjected to important time point extraction and time series reconstruction in sequence to obtain a second PMU multidimensional time series and a second RTU multidimensional time series, specifically:
[0029] Based on the PMU data point importance judgment result and the RTU data point importance judgment result, important time points are retained and unimportant time points are eliminated to obtain a PMU important time point set and an RTU important time point set;
[0030] Based on the PMU important time point set and the RTU important time point set, a second PMU multidimensional time series and a second RTU multidimensional time series are constructed.
[0031] Furthermore, the method further comprises the following steps:
[0032] Based on the preset energy interaction operator, energy interaction is defined for the second PMU multidimensional time series and the second RTU multidimensional time series respectively, to obtain PMU energy definition and RTU energy definition and to construct corresponding PMU energy matrix and RTU energy matrix.
[0033] Further, based on the preset energy interaction operator, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series through a preset similarity measurement algorithm to obtain the corresponding PMU correlation coefficient and RTU correlation coefficient, which are specifically:
[0034] By performing singular value decomposition on the PMU energy matrix and the RTU energy matrix, a corresponding PMU singular value vector set and RTU singular value vector set are obtained;
[0035] Summing the PMU singular value vector set and the RTU singular value vector set respectively to obtain a weight vector of a corresponding PMU singular value and a weight vector of a corresponding RTU singular value;
[0036] Based on the weight vector of the PMU singular value and the weight vector of the RTU singular value, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series through a multidimensional time series similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients.
[0037] Further, based on the PMU correlation coefficient and the RTU correlation coefficient, the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation are obtained to obtain the third PMU multidimensional time series and the third RTU multidimensional time series, specifically:
[0038] Update the offset parameters and record the PMU correlation coefficients and RTU correlation coefficients calculated corresponding to each parameter until the preset requirements are met, and obtain the PMU correlation coefficient matrix and the RTU correlation coefficient matrix;
[0039] Based on the PMU correlation coefficient matrix and the RTU correlation coefficient matrix, the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation are obtained by comparing the correlation coefficients, so as to obtain a third PMU multidimensional time series and a third RTU multidimensional time series.
[0040] A data alignment method proposed in an embodiment of the present invention quantifies the interactivity of two multidimensional time series by introducing a nonlinear measure called an energy interaction operator, and then obtains the energy principal components by constructing an energy matrix and performing singular value decomposition. The similarity of the two sequences is measured by comparing the differences between the principal components of the energy matrices of the multidimensional time series. Based on the similarity analysis, the relative changes of the multidimensional data are obtained to ensure the consistency of the reflection of the global data.
[0041] Further, based on the data of the third PMU multidimensional time series, the data time stamp of the third RTU multidimensional time series is determined to obtain the data alignment result, specifically:
[0042] Based on the data of the third PMU multidimensional time series, synchronize the data of the third RTU multidimensional time series to obtain a synchronization result of the third RTU multidimensional time series;
[0043] A data alignment result is obtained by setting a data time stamp for the third RTU multi-dimensional time series synchronization result.
[0044] The embodiment of the present invention further provides a data alignment system, comprising:
[0045] Time series construction module, important time point extraction module, similarity calculation module, maximum correlation acquisition module and data alignment module;
[0046] The time series construction module is used to obtain PMU measurement data at historical moments and RTU measurement data at historical moments, and to construct a first PMU multidimensional time series and a first RTU multidimensional time series;
[0047] The important time point extraction module is used to extract important time points from the first PMU multidimensional time series and the first RTU multidimensional time series based on a preset angle constraint threshold through a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series;
[0048] The similarity calculation module is used to perform similarity calculation on the second PMU multidimensional time series and the second RTU multidimensional time series based on a preset energy interaction operator and through a preset similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients;
[0049] The maximum correlation acquisition module is used to acquire the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation based on the PMU correlation coefficient and the RTU correlation coefficient, and obtain a third PMU multidimensional time series and a third RTU multidimensional time series;
[0050] The data alignment module is used to determine the data time stamp of the third RTU multidimensional time series based on the data of the third PMU multidimensional time series, and obtain a data alignment result.
[0051] The embodiment of the present invention proposes a data alignment system, which obtains relevant data of historical moments through a time series construction module to construct a multidimensional time series for subsequent data alignment steps, and can reveal the complex relationship between multiple variables through multidimensional data analysis to provide more comprehensive information; through the important time point extraction module, important time points are selected to remove redundant data, improve the efficiency, accuracy and reliability of the algorithm, and ensure the accuracy of data alignment data; through the similarity calculation module and the maximum correlation acquisition module, an energy interaction operator is introduced to perform similarity calculation on the time series and obtain a multidimensional time series with maximum correlation, and the relationship between a certain time point and its previous and subsequent moments is characterized through the energy interaction operator, reducing the reflection of the global consistency of the data due to the unpredictable complex structure of the multidimensional data; finally, the data alignment module uses the PMU data as a benchmark to determine the time stamp of the RTU data, completes the data alignment, and ensures the consistency of the global data. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 A schematic diagram of a data alignment method according to an embodiment of the present invention;
[0053] Figure 2 A schematic diagram of a process flow of a data alignment method provided by an embodiment of the present invention;
[0054] Figure 3 A schematic diagram of the result when the RTU data delay update is 360ms according to a data alignment method provided by an embodiment of the present invention;
[0055] Figure 4 A schematic diagram of the result when the RTU delay is updated for 1s in a data alignment method provided by an embodiment of the present invention;
[0056] Figure 5 A schematic diagram of the result when the RTU is updated 640ms in advance according to a data alignment method provided by an embodiment of the present invention;
[0057] Figure 6 A schematic diagram of the module structure of a data alignment system provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0059] Example 1
[0060] See also Figure 1 , Figure 1 A schematic diagram of a data alignment method according to an embodiment of the present invention. Figure 1 As shown, the embodiment of the present invention proposes a data alignment method, including steps 101 to 105, each step is specifically as follows:
[0061] Step 101, obtaining historical PMU measurement data and historical RTU measurement data, and constructing a first PMU multidimensional time series and a first RTU multidimensional time series;
[0062] As an example of this embodiment, PMU measurement data at historical moments and RTU measurement data at historical moments are obtained; based on a preset time length, the PMU measurement data at historical moments and the RTU measurement data at historical moments are sequentially organized to obtain a first PMU multidimensional time series and an initial RTU multidimensional time series; based on a preset time step, the initial RTU multidimensional time series is subjected to an offset time RTU measurement acquisition to obtain a first RTU multidimensional time series. Specifically, the power grid data comes from the WAMS system and the SCADA system of the power grid monitoring point, which measure and upload the monitoring data in real time. The state data of the power grid consists of directly measured voltage and indirectly calculated power, etc. For nodes in the power grid that are simultaneously equipped with PMU and RTU monitoring, each measurement data of a certain time length is taken to form the initial multidimensional time series of PMU and RTU respectively. According to a certain time step, the RTU data measurement values corresponding to different offset times are taken to form multiple multidimensional time series and the PMU multidimensional time series X (equivalent to the first PMU multidimensional time series and the first RTU multidimensional time series) for correlation analysis in subsequent steps.
[0063] X PMU (t)=(x1(t),...,x i (t),...,x n (t))
[0064] X RTU (t'+t shift )=(x1(t'+t shift ),...,x i (t'+t shift ),...,x n (t'+t shift ))
[0065] t shift =k*delta_t
[0066] Where t and t' represent the time points of observation, t is known and t' is unknown, i represents a variable of the observed value, n represents the number of variables, tshift It represents the RTU measurement time offset, delta_t represents the translation time step, and k can take any integer value.
[0067] Whether it is RTU or PMU measurement data, their time series variables are composed of voltage amplitude, node injected active and reactive power, and associated branch active and reactive power.
[0068] Step 102: Based on a preset angle constraint threshold, extract important time points from the first PMU multidimensional time series and the first RTU multidimensional time series by using a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series;
[0069] As an example of this embodiment, the first PMU intermediate data point and the first RTU intermediate data point where three consecutive data points form an angle in the first PMU multidimensional time series and the first RTU multidimensional time series are obtained; based on the preset angle constraint threshold, the first PMU intermediate data point and the first RTU intermediate data point are respectively judged as important data points by the ABR algorithm to obtain the PMU data point importance judgment result and the RTU data point importance judgment result; the PMU data point importance judgment result is successively subjected to important time point extraction and time series reconstruction to obtain the second PMU multidimensional time series; the RTU data point importance judgment result is successively subjected to important time point extraction and time series reconstruction to obtain the second RTU multidimensional time series. Based on the ABR algorithm, when it is determined that the angle between the first PMU intermediate data point and the adjacent data point is greater than the preset angle constraint threshold, the first PMU intermediate data point is judged to be an important time point; based on the ABR algorithm, when it is determined that the angle between the first PMU intermediate data point and the adjacent data point is less than or equal to the preset angle constraint threshold, the first PMU intermediate data point is judged to be a non-important time point. Based on the ABR algorithm, when it is determined that the angle between the first RTU intermediate data point and the adjacent data point is greater than the preset angle constraint threshold, the first RTU intermediate data point is judged to be an important time point; based on the ABR algorithm, when it is determined that the angle between the first RTU intermediate data point and the adjacent data point is less than or equal to the preset angle constraint threshold, the first RTU intermediate data point is judged to be a non-important time point. Based on the PMU data point importance judgment result and the RTU data point importance judgment result, the important time points are retained and the non-important time points are eliminated to obtain the PMU important time point set and the RTU important time point set; based on the PMU important time point set and the RTU important time point set, the second PMU multidimensional time series and the second RTU multidimensional time series are constructed. Specifically, due to the influence of factors such as random fluctuations and measurement errors in the power grid, directly mining the multidimensional measurement data is prone to errors, affecting the accuracy and reliability of the algorithm. On the other hand, the data volume is large and the algorithm efficiency is low. Therefore, the ABR algorithm is used to select a part of points from the multidimensional measurement sequence data as important nodes to fit the original sequence. The ABR algorithm is an idea that utilizes the human ability to extract important points during waveform perception. It regards a multivariate time series as a waveform in a multidimensional space, and scans the time series data set once. It uses the given angle threshold C constraint to determine whether the middle data point is an important point in the time series based on the angle formed by three consecutive data points. If so, it is retained; otherwise, the corresponding data point is filtered out. Finally, the original sequence is represented by an important point sequence (equivalent to obtaining the second PMU multidimensional time series and the second RTU multidimensional time series). For example, if a threshold C is given, when the multidimensional time series X=(x1,x2,...,xn ) in x i ( i When the angle θ at the observation time point is greater than C, it is called x i is an important point in the sequence, keep it.
[0070] Step 103: Based on a preset energy interaction operator, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series by using a preset similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients;
[0071] As an example of this embodiment, based on a preset energy interaction operator, energy interaction definitions are performed on the second PMU multidimensional time series and the second RTU multidimensional time series respectively to obtain PMU energy definitions and RTU energy definitions and to construct corresponding PMU energy matrices and RTU energy matrices; by performing singular value decomposition on the PMU energy matrix and the RTU energy matrix, corresponding PMU singular value vector sets and RTU singular value vector sets are obtained; the PMU singular value vector sets and the RTU singular value vector sets are summed respectively to obtain corresponding PMU singular value weight vectors and RTU singular value weight vectors; based on the PMU singular value weight vectors, the RTU singular value weight vectors and preset offset parameters, similarity calculations are performed on the second PMU multidimensional time series and the second RTU multidimensional time series by a multidimensional time series similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients. Specifically, traditional similarity measures have achieved good results in the research and application of one-dimensional time series data, but in the case of multidimensional time series data, since the distribution of data often has an unpredictable complex structure, it cannot reflect the global consistency of multidimensional data. Therefore, the present invention introduces the nonlinear measure of energy interaction operator Ψ to quantify the interactivity of two multidimensional time series. From the perspective of energy distribution, the energy matrix of the multidimensional time series is calculated and the singular value decomposition is performed to obtain the energy principal components. The similarity of the two sequences is measured by comparing the differences between the principal components of the energy matrix of the multidimensional time series. Since the energy interaction operator describes the relationship between a certain time point and the moments before and after it, the energy matrix of the multidimensional time series contains the temporal information of the multidimensional time series and its relative changes. More specifically, suppose the multidimensional time series composed of RTU measurement data is X(t)=(x1(t),...,x n (t)) T , the multidimensional time series composed of PMU measurement data is Y(t)=(y1(t),...,y n (t)) T The energy interaction between them at time t is defined as:
[0072]
[0073] Where T stands for transpose.
[0074] When MB When (X,Y)=0, there is no energy interaction between X(t) and Y(t), and X(t) and Y(t) are said to be uncorrelated.
[0075] Similarly, the energy of X(t) at time t is defined as:
[0076]
[0077] Therefore, the energy matrix of X(t) is defined as:
[0078]
[0079] Similarly, the energy matrix of Y(t) is defined as:
[0080]
[0081] Singular value decomposition of the energy matrix:
[0082] [U,S,V]=SVD(Ψ(X))
[0083] [U,S,V]=SVD(Ψ(Y))
[0084] As an explanation of this embodiment, in the singular value decomposition (SVD), U, S and V represent different contents respectively: U is a left singular vector matrix, and its column vector contains the left singular vectors of the original matrix. These vectors describe the projection of the original matrix under a new orthogonal basis vector. S is a singular value matrix, which is a diagonal matrix, and the elements on the diagonal are singular values. The singular values represent the importance of the original matrix in each direction and are a way to measure the "importance" of the matrix. V is a right singular vector matrix, and its column vector contains the right singular vectors of the original matrix. These vectors describe the projection of the original matrix under another new orthogonal basis vector. In general, U, S and V together describe the structure and characteristics of the original matrix, through which we can perform operations such as dimensionality reduction and feature extraction on the matrix, so as to better understand and utilize the data.
[0085] Calculate the energy matrix M for X and Y respectively i , calculate M i The singular value vector of Then, sum all the singular value vectors to get
[0086]
[0087] The weight vector is calculated as follows:
[0088]
[0089] Where i represents the number of multidimensional time series.
[0090] The method used to measure the synchronization relationship of different measurement signals in the time domain is to calculate the correlation coefficient of PMU measurement data X and RTU measurement data Y of a multidimensional time series based on the multidimensional time series similarity measurement algorithm of the energy interaction operator:
[0091]
[0092] t shift =k*Δt
[0093] Where Erui is the correlation coefficient between RTU measurement X and PMU measurement Y; t1 is the time of the current RTU measurement. Since the RTU measurements are not consistent with each other and there is no time scale, the value of t1 is unknown; t2 is the end time of the current PMU measurement sequence; t shift is the RTU measurement time offset, Δt is the translation time step, k∈Z; i ,b i > is a i With b i The inner product operation, a i With b i are the right singular vectors of the energy matrices of X and Y respectively.
[0094] Step 104: based on the PMU correlation coefficient and the RTU correlation coefficient, obtain the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation, and obtain a third PMU multidimensional time series and a third RTU multidimensional time series;
[0095] As an example of this embodiment, the offset parameter is updated until a preset threshold is reached, the PMU correlation coefficient and the RTU correlation coefficient calculated based on each offset parameter are obtained, and a PMU correlation coefficient matrix and an RTU correlation coefficient matrix are constructed; based on the PMU correlation coefficient matrix and the RTU correlation coefficient matrix, the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation are obtained by comparing the correlation coefficients, and the third PMU multidimensional time series and the third RTU multidimensional time series are obtained. Specifically, the similarity between the RTU measurement data and the PMU measurement data is calculated according to the similarity Erui definition formula, and the k value is changed, t shift As it changes, that is, the translation amount of the RTU changes, and steps S101-S103 are repeated until k reaches the set value, and the calculation is stopped and recorded and saved. Finally, Erui mn (t1,t2+tshift ,w) is a correlation coefficient matrix of a series of RTU measurements and PMU measurements arranged in chronological order. By analyzing the correlation coefficient matrix of RTU and PMU measurements, the correlation coefficients corresponding to different RTU measurement shift times are compared, and the shift time corresponding to the maximum correlation coefficient and the RTU measurement value after shifting are found (equivalent to the third PMU multidimensional time series and the third RTU multidimensional time series).
[0096] Step 105: Based on the data of the third PMU multidimensional time series, determine the data time stamp of the third RTU multidimensional time series to obtain a data alignment result.
[0097] As an example of this embodiment, based on the data of the third PMU multidimensional time series, the third RTU multidimensional time series is synchronized to obtain the third RTU multidimensional time series synchronization result; by setting the data time stamp for the third RTU multidimensional time series synchronization result, the data alignment result is obtained. Specifically, when the data similarity between the two reaches the maximum value, the data is close to synchronization, thereby obtaining the reference measurement moment of the PMU / RTU hybrid measurement system, realizing the synchronization of RTU data using the precise data generated by the PMU, and adding the time coordinate to it to complete the data alignment.
[0098] Erui(t1,t2+t shift ,w)=maxErui mn (t1,t2+t shift ,w)
[0099] t st =t2+t shift
[0100] Where: t st It is the reference measurement time of the hybrid measurement system.
[0101] The embodiment of the present invention proposes a data alignment method, which constructs a multidimensional time series for subsequent data alignment by acquiring relevant data at historical moments, and can reveal the complex relationship between multiple variables through multidimensional data analysis, providing more comprehensive information; the influence of random fluctuations and measurement errors in the power grid, directly mining multidimensional measurement data is prone to errors, and by selecting important time points, redundant data can be removed, the efficiency, accuracy and reliability of the algorithm can be improved, and the accuracy of data alignment data can be guaranteed; the energy interaction operator is introduced to calculate the similarity of the time series and obtain the multidimensional time series with the maximum correlation, and the relationship between a certain time point and its previous and subsequent moments is characterized by the energy interaction operator, reducing the reflection of the global consistency of the data due to the unpredictable complex structure of the multidimensional data; finally, the time scale of the RTU data is determined based on the PMU data, and the data alignment is completed to ensure the consistency of the global data. By constructing multidimensional measurement time series data, it is necessary to form a multidimensional time series with the state data of the historical moments according to a certain time length, make full use of the power grid measurement information, reveal the complex relationship between the multidimensional data, and lay a solid foundation for the subsequent analysis of data consistency. The ABR algorithm is used to select a part of points from the multidimensional measurement sequence data as important nodes to fit the original sequence. By compressing the multidimensional measurement time series data, the information of the sequence is retained as much as possible, which improves the calculation efficiency; the important points are screened and retained, so that the overall reliability of the data is improved. The energy interaction operator is introduced as a nonlinear measure to quantify the interactivity of the two multidimensional time series, and then the energy matrix is constructed and the energy principal component is obtained by singular value decomposition. The similarity of the two sequences is measured by comparing the differences between the principal components of the energy matrix of the multidimensional time series. Based on the analysis of similarity, the relative changes of the multidimensional data are obtained to ensure the consistency of the reflection of the global data.
[0102] Example 2
[0103] See also Figure 2 , Figure 2 A flowchart of a data alignment method according to an embodiment of the present invention; Figure 2As shown, a data alignment method proposed in an embodiment of the present invention is one of the feasible explanations, and is expressed in the program as follows: run the program, construct PMU measurement data and RTU measurement data at historical moments through power flow calculation, initialize the RTU measurement time offset, translation time step and integer k, set the maximum value of k to 10, and enter the similarity calculation loop: construct RTU translation time, construct the RTU multidimensional time series and PMU multidimensional time series after translation (equivalent to the first PMU multidimensional time series and the first RTU multidimensional time series) after translation, perform preprocessing such as importance judgment and energy interaction definition on the data, and then perform energy matrix calculation, singular value decomposition and weight vector calculation to obtain correlation coefficients and save them until k is greater than or equal to 10, and the loop ends; compare the sizes of all correlation coefficients, and find the translation time corresponding to the maximum correlation coefficient and the RTU measurement value after translation, align the data, and the program ends.
[0104] More specifically, based on a data alignment method proposed in an embodiment of the present invention and the application of the above-mentioned program flow, the IEEE9 node system is used for simulation verification, and the program flow is implemented and run in MATLAB 2021b. The experimental data consists of the two data types mentioned above, namely PMU measurement and RTU measurement. The PMU measurement data includes the voltage amplitude, node injection power, and the active and reactive power of the associated branches. The RTU measurement data includes the voltage amplitude, node injection power, and the active and reactive power of the associated branches, thereby forming two multidimensional time series. The open circuit and short circuit fault conditions are simulated respectively, and the simulation time is set to 51s. The simulated PMU measurement step is 10ms, and the RTU measurement step is 1s. The simulation shift time unit is set to 20ms, and simulations are performed for different asynchronous errors and different measurement errors respectively. The simulation results are shown below. See Figure 3 , Figure 4 and Figure 5 , Figure 3 A schematic diagram of the result when the RTU data delay update is 360ms according to a data alignment method provided by an embodiment of the present invention; Figure 4 A schematic diagram of the result when the RTU delay is updated for 1s in a data alignment method provided by an embodiment of the present invention; Figure 5 A schematic diagram of the result when the RTU is updated 640ms in advance according to a data alignment method provided by an embodiment of the present invention; wherein, Figure 3 , Figure 4 and Figure 5(a) and (c) show the similarity curves of individual nodes obtained by similarity calculation in the data alignment method proposed by the present invention. It can be seen that the correlation between the RTU simulation value and the PMU simulation value reaches the maximum after the RTU simulation value is updated in advance or delayed for a certain time; at the same time, (b) and (d) show the voltage amplitude curve comparison diagrams of the third RTU multidimensional time series (RTU after synchronization), the first RTU multidimensional time series (RTU before synchronization) and the first PMU multidimensional time series. The PMU measurement value follows a normal distribution with a standard deviation of 0.002 and an error mean of 0, and the RTU measurement value follows a normal distribution with a standard deviation of 0.005 and an error mean of 0.
[0105] Table 1 Relative error between PMU and RTU data before and after synchronization
[0106]
[0107]
[0108] The PMU measurement values follow a normal distribution with a standard deviation of 0.002 and an error mean of 0, and the RTU measurement values follow a normal distribution with a standard deviation of 0.01 and an error mean of 0.
[0109] Table 2 Relative error between PMU and RTU data before and after synchronization
[0110]
[0111]
[0112]
[0113] It can be seen from Table 1 and Table 2 that after the RTU simulation value is updated in advance or delayed in time, the Erui similarity between the RTU measurement and the PMU measurement is at its maximum value on most nodes. Therefore, the update delay of the PMU and RTU measurements can be obtained based on the maximum value of the Erui similarity between the two, so that the RTU data can be shifted in time to synchronize the two data, thereby better performing data fusion. At the same time, the greater the difference in measurement error between the PMU and RTU data, the greater the data error after synchronization.
[0114] It can be concluded from the simulation case proposed in the embodiment of the present invention that the proposed data alignment method, based on the data correlation analysis theory, maximizes the correlation between PMU data and RTU data, adds a time stamp to the RTU data, thereby achieving the purpose of data synchronization between the two, so as to improve the consistency of data sections and provide a data basis guarantee for subsequent applications in the system. The proposed similarity measurement method can better characterize the essential characteristics of multidimensional time series than the traditional similarity definition due to the introduction of temporal information and relative changes in the sequence, and has obvious advantages in accuracy. Secondly, the length m of the multidimensional time series composed of measurement data is usually much larger than the number of variables n. The energy matrix is only an n×n-order matrix, which greatly reduces the scale of the data and greatly improves the calculation speed. By decomposing the singular value of the multidimensional time series, the influence of non-main factors can be effectively suppressed, thereby extracting key information, and it has good noise reduction performance.
[0115] Example 3
[0116] See also Figure 6 , Figure 6 A schematic diagram of the module structure of a data alignment system provided by an embodiment of the present invention. Figure 6 As shown, an embodiment of the present invention provides a data alignment system, including:
[0117] A time series construction module 601, an important time point extraction module 602, a similarity calculation module 603, a maximum correlation acquisition module 604 and a data alignment module 605;
[0118] The time series construction module 601 is used to obtain PMU measurement data at historical moments and RTU measurement data at historical moments, and construct a first PMU multidimensional time series and a first RTU multidimensional time series;
[0119] The important time point extraction module 602 is used to extract important time points from the first PMU multidimensional time series and the first RTU multidimensional time series based on a preset angle constraint threshold through a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series;
[0120] The similarity calculation module 603 is used to perform similarity calculation on the second PMU multidimensional time series and the second RTU multidimensional time series based on a preset energy interaction operator and through a preset similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients;
[0121] The maximum correlation acquisition module 604 is used to acquire the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation based on the PMU correlation coefficient and the RTU correlation coefficient, and obtain a third PMU multidimensional time series and a third RTU multidimensional time series;
[0122] The data alignment module 605 is used to determine the data time stamp of the third RTU multi-dimensional time series based on the data of the third PMU multi-dimensional time series, and obtain a data alignment result.
[0123] The embodiment of the present invention proposes a data alignment system, which obtains relevant data of historical moments through a time series construction module to construct a multidimensional time series for subsequent data alignment steps, and can reveal the complex relationship between multiple variables through multidimensional data analysis to provide more comprehensive information; through the important time point extraction module, important time points are selected to remove redundant data, improve the efficiency, accuracy and reliability of the algorithm, and ensure the accuracy of data alignment data; through the similarity calculation module and the maximum correlation acquisition module, an energy interaction operator is introduced to perform similarity calculation on the time series and obtain a multidimensional time series with maximum correlation, and the relationship between a certain time point and its previous and subsequent moments is characterized through the energy interaction operator, reducing the reflection of the global consistency of the data due to the unpredictable complex structure of the multidimensional data; finally, the data alignment module uses the PMU data as a benchmark to determine the time stamp of the RTU data, completes the data alignment, and ensures the consistency of the global data.
[0124] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
[0125] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.
[0126] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
Claims
1. A data alignment method, characterized in that: The following steps are involved: Acquire historical PMU measurement data and historical RTU measurement data, and construct a first PMU multidimensional time series and a first RTU multidimensional time series; Based on a preset angle constraint threshold, extracting important time points from the first PMU multidimensional time series and the first RTU multidimensional time series through a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series; Based on a preset energy interaction operator, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series by using a preset similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients; Based on the PMU correlation coefficient and the RTU correlation coefficient, acquiring the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation, and obtaining a third PMU multidimensional time series and a third RTU multidimensional time series; Based on the data of the third PMU multidimensional time series, the data time stamp of the third RTU multidimensional time series is determined to obtain a data alignment result.
2. A data alignment method as claimed in claim 1, characterized in that: The obtaining of historical PMU measurement data and historical RTU measurement data, and the construction of a first PMU multi-dimensional time series and a first RTU multi-dimensional time series, is specifically as follows: Obtain historical PMU measurement data and historical RTU measurement data; Based on a preset time length, the PMU measurement data at the historical moment and the RTU measurement data at the historical moment are time-sequenced to obtain a first PMU multi-dimensional time series and an initial RTU multi-dimensional time series; Based on a preset time step, the initial RTU multi-dimensional time series is subjected to an offset time RTU measurement acquisition to obtain a first RTU multi-dimensional time series.
3. A data alignment method as claimed in claim 1, characterized in that: Based on the preset angle constraint threshold, the first PMU multidimensional time series and the first RTU multidimensional time series are extracted with important time points by a preset data processing algorithm to obtain the second PMU multidimensional time series and the second RTU multidimensional time series, specifically: Acquire a first PMU intermediate data point and a first RTU intermediate data point at which an angle is formed by three consecutive data points in the first PMU multidimensional time series and the first RTU multidimensional time series; Based on a preset angle constraint threshold, an ABR algorithm is used to respectively judge the important data points of the first PMU intermediate data point and the first RTU intermediate data point, to obtain a PMU data point importance judgment result and an RTU data point importance judgment result; Extracting important time points and reconstructing time series are performed on the importance judgment results of the PMU data points in sequence to obtain a second PMU multidimensional time series; The importance judgment results of the RTU data points are sequentially subjected to important time point extraction and time series reconstruction to obtain a second RTU multi-dimensional time series.
4. A data alignment method as claimed in claim 3, characterized in that: Based on the preset angle constraint threshold, the ABR algorithm is used to respectively judge the important data points of the first PMU intermediate data point and the first RTU intermediate data point, and obtain the PMU data point importance judgment result and the RTU data point importance judgment result, which are specifically: Based on the ABR algorithm, when it is determined that the angle between the first PMU intermediate data point and the adjacent data point is greater than the preset angle constraint threshold, the first PMU intermediate data point is judged to be an important time point; Based on the ABR algorithm, when it is determined that the angle between the first PMU intermediate data point and the adjacent data point is less than or equal to the preset angle constraint threshold, the first PMU intermediate data point is judged to be a non-important time point.
5. A data alignment method as claimed in claim 4, characterized in that: The method further comprises: performing important data point judgment on the first PMU intermediate data point and the first RTU intermediate data point respectively based on the preset angle constraint threshold by using the ABR algorithm to obtain a PMU data point importance judgment result and an RTU data point importance judgment result. Based on the ABR algorithm, when it is determined that the angle between the first RTU intermediate data point and the adjacent data point is greater than the preset angle constraint threshold, the first RTU intermediate data point is judged to be an important time point; Based on the ABR algorithm, when it is determined that the angle between the first RTU intermediate data point and the adjacent data point is less than or equal to the preset angle constraint threshold, the first RTU intermediate data point is judged to be a non-important time point.
6. A data alignment method as claimed in claim 5, characterized in that: The important time points are extracted and the time series are reconstructed respectively on the PMU data point importance judgment result and the RTU data point importance judgment result, to obtain a second PMU multi-dimensional time series and a second RTU multi-dimensional time series, specifically: Based on the PMU data point importance judgment result and the RTU data point importance judgment result, important time points are retained and unimportant time points are eliminated to obtain a PMU important time point set and an RTU important time point set; Based on the PMU important time point set and the RTU important time point set, a second PMU multidimensional time series and a second RTU multidimensional time series are constructed.
7. A data alignment method as claimed in claim 1, characterized in that: Based on the preset energy interaction operator, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series through a preset similarity measurement algorithm to obtain the corresponding PMU correlation coefficient and RTU correlation coefficient, which are specifically: Based on a preset energy interaction operator, respectively perform energy interaction definition on the second PMU multidimensional time series and the second RTU multidimensional time series, obtain a PMU energy definition and an RTU energy definition, and construct corresponding PMU energy matrices and RTU energy matrices; By performing singular value decomposition on the PMU energy matrix and the RTU energy matrix, a corresponding PMU singular value vector set and RTU singular value vector set are obtained; Summing the PMU singular value vector set and the RTU singular value vector set respectively to obtain a weight vector of a corresponding PMU singular value and a weight vector of a corresponding RTU singular value; Based on the weight vector of the PMU singular value, the weight vector of the RTU singular value and the preset offset parameter, similarity calculation is performed on the second PMU multidimensional time series and the second RTU multidimensional time series through a multidimensional time series similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients.
8. A data alignment method as claimed in claim 7, characterized in that: The acquiring, based on the PMU correlation coefficient and the RTU correlation coefficient, the second PMU multidimensional time series and the second RTU multidimensional time series having the maximum correlation, and obtaining the third PMU multidimensional time series and the third RTU multidimensional time series, is specifically: The offset parameter is updated until a preset threshold is reached, a PMU correlation coefficient and an RTU correlation coefficient calculated based on each offset parameter are obtained, and a PMU correlation coefficient matrix and an RTU correlation coefficient matrix are constructed; Based on the PMU correlation coefficient matrix and the RTU correlation coefficient matrix, the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation are obtained by comparing the correlation coefficients, so as to obtain a third PMU multidimensional time series and a third RTU multidimensional time series.
9. A data alignment method as claimed in claim 1, characterized in that: The data time stamp of the third RTU multidimensional time series is determined based on the data of the third PMU multidimensional time series to obtain the data alignment result, specifically: Based on the data of the third PMU multidimensional time series, synchronize the data of the third RTU multidimensional time series to obtain a synchronization result of the third RTU multidimensional time series; A data alignment result is obtained by setting a data time stamp for the third RTU multi-dimensional time series synchronization result.
10. A data alignment system, characterized in that: include: Time series construction module, important time point extraction module, similarity calculation module, maximum correlation acquisition module and data alignment module; The time series construction module is used to obtain PMU measurement data at historical moments and RTU measurement data at historical moments, and to construct a first PMU multidimensional time series and a first RTU multidimensional time series; The important time point extraction module is used to extract important time points from the first PMU multidimensional time series and the first RTU multidimensional time series based on a preset angle constraint threshold through a preset data processing algorithm to obtain a second PMU multidimensional time series and a second RTU multidimensional time series; The similarity calculation module is used to perform similarity calculation on the second PMU multidimensional time series and the second RTU multidimensional time series based on a preset energy interaction operator and through a preset similarity measurement algorithm to obtain corresponding PMU correlation coefficients and RTU correlation coefficients; The maximum correlation acquisition module is used to acquire the second PMU multidimensional time series and the second RTU multidimensional time series with the maximum correlation based on the PMU correlation coefficient and the RTU correlation coefficient, and obtain a third PMU multidimensional time series and a third RTU multidimensional time series; The data alignment module is used to determine the data time stamp of the third RTU multidimensional time series based on the data of the third PMU multidimensional time series, and obtain a data alignment result.
Citation Information
Patent Citations
Method for aligning section raw data on basis of double time scale delay evaluation
CN102361353A
Power system state estimation method combining particle filtering and convolutional neural network
CN111797974A
Method and apparatus for checking power grid measurement data, device, storagemedium and program product
US20230369854A1
