Missing data estimation method and system for UPFC sub-module state parameter monitoring

Abstract

The invention discloses a missing data estimation method and system for UPFC sub-module state parameter monitoring, and the method specifically comprises the steps: obtaining the operation data and historical data of a UPFC sub-module, converting the data from different sources into a unified format, and further carrying out the preprocessing of the data, and forming an operation data set and a historical data set; the operation data set and the historical data set are detected, missing data points are searched and positioned, data are perfected and supplemented, and a repaired operation data set is formed; a support vector machine method is further combined to calibrate and update data in the repaired operation data set to form an operation data set with higher estimation precision; and integrating the operation data set with higher estimation precision with the historical data set, and updating the historical data set. The method can form operation data with higher estimation precision, and has the advantages of high data recovery reliability, high data integrity and the like.

Description

Technical Field

[0001] The present invention relates to the field of power electronics technology, and in particular to a missing data estimation method and system for monitoring state parameters of a UPFC submodule. Background Art

[0002] As the third generation of flexible AC transmission device, Unified Power Flow Controller (UPFC) can realize voltage regulation, phase angle regulation, line impedance compensation, dynamic power flow control, reactive power compensation and other functions. It is an important basic equipment for building future new energy power systems. However, the huge volume of sub-modules has brought great challenges to module status monitoring and data analysis, and the "black module" problem found during operation and maintenance has led to communication interruption and sub-module data loss. Therefore, how to ensure the reliability and integrity of the information data of UPFC sub-modules is crucial to improving the operation and maintenance and analysis capabilities of the UPFC system.

[0003] Existing technologies for data repair include: The invention with publication number CN116578551A proposes a power grid data repair method based on GRU-GAN, which includes the following specific steps: Step 1: data acquisition and preprocessing, based on the voltage history data collected by TTU, find a complete data set X that has been cleaned and outliers removed, and normalize the collected data set to make the normalized data set conform to the normal distribution; Step 2: Construct a generative adversarial network GAN based on the gated recurrent neural unit GRU; Step 3: GRU-GAN neural network training, so that the data distribution generated by random noise is close to the distribution of real data; Step 4: Repair of missing data; Step 5: Missing data error analysis, using the root mean square error RMSE indicator to evaluate the performance of the model.

[0004] In addition, a method for repairing electric power data is proposed in the patent document with authorization announcement number CN113485986B, which uses a SOM neural network to classify and process the electric power data in the historical electric power data set; uses the Pearson correlation coefficient theory to obtain the influencing factors of the electric power data type that meets the correlation threshold; inputs the influencing factors of the missing data into the trained LSTM neural network to obtain the electric power data type of the missing data; and uses different methods to repair the data according to the electric power data type of the missing data.

[0005] However, the missing data estimation methods in the above literature are slightly complicated in processing specific application scenarios, and the repair accuracy is difficult to guarantee when computing power is limited. On the other hand, the above methods are not universal and are difficult to promote and apply in the face of the actual needs of UPFC submodule status monitoring. Summary of the invention

[0006] In order to solve the technical problems mentioned in the above background technology, the present invention provides a missing data estimation method and system for UPFC sub-module state parameter monitoring, which repairs the UPFC operation data set and historical data set through multiple Bayesian interpolation and support vector machine methods, and provides an effective solution to the UPFC data missing problem caused by sampling errors, communication failures and black modules, improves data integrity and reliability, and has good practical value.

[0007] The present invention adopts the following technical solution.

[0008] The first aspect of the present invention provides a missing data estimation method for monitoring state parameters of a UPFC submodule, wherein the UPFC submodule includes a submodule positive terminal SM+, a submodule negative terminal SM-, a submodule capacitor C SM , the first switch tube S1, the second switch tube S2, the submodule positive terminal SM+ is connected to the emitter of the first switch tube S1 and the collector of the second switch tube S2, the collector of the first switch tube S1 and the submodule capacitor C SM One end of the submodule capacitor C SM The other end is connected to the emitter of the second switch tube S2 and the negative terminal SM- of the submodule; the missing data estimation method comprises the following steps: Step S1: obtaining the current operating data and historical data of the state parameters of the UPFC submodules, converting the data of different state parameters into a unified format, and further preprocessing the data to form an operating data set and a historical data set; Step S2: Detect the operating data set and the historical data set, find and locate the missing data points, and complete and supplement the data to form a repaired operating data set; Step S3: Based on the support vector machine, calibrate and update the data in the repaired running data set; Step S4: Integrate the calibrated and updated operating data set with the historical data set to update the historical data set.

[0009] Preferably, in step S1, the data of different state parameters include submodule capacitance C SM Voltage u c , the current of the positive terminal SM+ of the submodule i SM , submodule temperature T SM ; Converting data of different state parameters into a unified format means aligning data of different parameters according to time series and associating the aligned data according to the data source.

[0010] Preferably, in step S1, the data is preprocessed as follows: a filter is used to remove measurement noise and system noise, and after the noise is removed, the average value of the historical data is calculated, and it is determined whether the data exceeds a threshold value set by a multiple of the average value. If so, it is an outlier, otherwise it is not an outlier, all outliers are removed, and multiple sensors are used to sample the same data, and verification is performed by checking whether the data are the same.

[0011] Preferably, in step S2, the operating data set and the historical data set are tested, missing data points are found and located, and the data are supplemented to form a repaired operating data set. Specifically, for the state parameter, after excluding abnormal values, the historical data and the operating data are divided according to the equal periodic intervals of the parameter, and there are n data points; all complete data sets divided into equal periodic intervals according to this parameter are recorded as a set X , X for n × q matrix, q is the number of complete data sets; any data set with missing records divided by equal periodic intervals according to this parameter is selected as Y , Y A historical data set or a current running data set with missing data for a certain period of time; Y and X There is a linear relationship between Y There are n One non-missing observation is recorded as Y 1, which corresponds to X The part is recorded as X 1; Y middle n 0 missing values ​​are recorded as Y m , its corresponding X Some records are X m ; Based on the part of the data that is not missing, Y and X The parameter relationship between them is estimated and the posterior distribution of the parameters is constructed; then for each missing data, a sample is drawn from the posterior distribution to obtain an estimate of the missing value.

[0012] Preferably, the equal-periodic interval length is set to one or more period lengths corresponding to the state parameter, and is set according to the degree of data integrity.

[0013] Preferably, Y and X The parameter correspondence is set to have a normal linear correlation, that is, Y ~N ( Xβ , s 2 ), β is the regression coefficient, which is a q dimensional vector, s 2 is the residual, for The normal distribution function of .

[0014] Preferably, the Y and X The parameters of the linear relationship between are estimated and the posterior distribution of the parameters is constructed. The specific process is: Y There are n One non-missing observation is recorded as Y 1, which corresponds to X The part is recorded as X 1; Y middle n 0 missing values ​​are recorded as Y m , its corresponding X Some records are X m ; Based on the part with no missing data, it is estimated by the following formula: , In the above formula, is the estimated value of the regression coefficient, is the estimated value of the residual; for X The transpose of 1, T is the transpose symbol; Build β and s 2 The posterior distribution of s 2 The posterior distribution exhibits chi-squared distribution characteristics: , in, n is the number of state parameter observations within the same period, is the chi-square distribution function; β The posterior distribution of s 2 , is the conditional posterior distribution, and follows a normal distribution: , In the above formula, β is the regression coefficient, which is a q dimensional vector, s 2 is the residual, is the estimated value of the regression coefficient, is the estimated value of the residual.

[0015] Preferably, for each missing data, sampling is performed from the posterior distribution to obtain an estimate of the missing value, and the specific process is: generating any multiple interpolation value sets according to the following interpolation model: , In the above formula, For any set of interpolation values, s * is any random draw from the posterior distribution, and β * for Any random draw from the conditional posterior distribution when Z 1 Yes n 0 random standard normal distribution values, that is: , In the above formula, g Originated from x 2 ( n 1- q ) distribution, Z 2 Yes q random standard normal distribution values; Take the arithmetic average of all interpolated data sets to get the repaired data set Y’ .

[0016] Preferably, in step S3, the data in the repaired running data set is calibrated and updated, specifically: the repaired running data set is recorded as Y’ , given a set of data ( x i , y i ), i =1,2,…, m , x i ∈ Y ', y i ∈{-1, 1}, where m is the total number of a given set of data. When the samples are linearly separable, an optimal hyperplane can be found to separate them, that is:

[0017] In the above formula, and Represent the weight vector and the optimal hyperplane offset respectively, yes The transpose of is the independent variable of the hyperplane; The support vector machine method maps samples from the original space to a high-dimensional feature space through soft margin optimization and kernel techniques. The support vector machine method finds a linear hyperplane with the largest margin in this high-dimensional space to classify the data. The linear hyperplane at this time is ,in is a function that maps high-dimensional space; for x The feature vector after being mapped to the high-dimensional space; solve the following optimization problem: , In the above formula, C >0 is the penalty factor, x i is the slack variable, for x i The eigenvector after being mapped to the high-dimensional space; in the calculation process of the above formula, it is necessary to solve 0, and the kernel function is used to solve it. The kernel function is defined as follows: , in, is one of the independent variables of the kernel function, and we need to solve hour , yes The transpose of

[0018] In the support vector machine method, there are four kernel functions, which are as follows: , In the above formula, r , c and d These are kernel function parameters; for The transpose of , , , are the names of four kernel functions; The restored dataset calibrated by the support vector machine method is added to the historical dataset to form an updated dataset.

[0019] The second aspect of the present invention proposes a missing data estimation system for monitoring state parameters of UPFC submodules, using the method described in the first aspect of the present invention, including a data acquisition module, a data preprocessing module, a repair operation data set module, a repair data calibration module, and a data update module, specifically: Data acquisition module, which obtains the operation data and historical data of UPFC submodules and converts them into a unified format; The data preprocessing module preprocesses the data, removes noise in specific frequency bands, monitors outliers in the data, and cross-validates the data by comparing historical data with other sensor data to optimize the data filtering effect and form operating data sets and historical data sets; Repair the running data set module, find and locate missing data points, and supplement the missing data; The repair data calibration module, combined with the support vector machine method, calibrates and updates the repaired data to form a running data set with higher estimation accuracy; The data update module integrates the operating data set with higher estimation accuracy with the historical data set and updates the historical data set.

[0020] The beneficial effect of the present invention is that, compared with the prior art, the present invention adopts multiple Bayesian interpolation and support vector machine methods to process and optimize the UPFC submodule status information, unifies and preprocesses the submodule operation data and historical data, then locates the missing data points of the historical data and the operation data, and uses the Bayesian method to multiple interpolate the missing data, which is universal and meets the actual needs of UPFC submodule status monitoring, and then further combines the repaired data set with the support vector machine method to locate abnormal interpolation data, update the data, and form a higher estimation accuracy operation data set; the algorithm has low computational complexity and improves the repair accuracy, greatly improves the reliability of data recovery, achieves data integrity, and has stronger use value. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic flow chart of a missing data estimation method for monitoring state parameters of a UPFC submodule according to the present invention; Figure 2 It is a topological diagram of the UPFC submodule disclosed in the embodiment of the present invention; Figure 3 A schematic diagram of the optimal hyperplane when the support vector machine disclosed in the embodiment of the present invention is linearly separable; Figure 4 This is the model result under the condition of 50% missing data disclosed in the embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only embodiments of a part of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0023] like Figure 1As shown, Embodiment 1 of the present invention provides a missing data estimation method for monitoring state parameters of UPFC submodules, such as Figure 2 As shown, the UPFC submodule includes a submodule positive terminal SM+, a submodule negative terminal SM-, and a submodule capacitor C SM , the first switch tube S1, the second switch tube S2, the submodule positive terminal SM+ is connected to the emitter of the first switch tube S1 and the collector of the second switch tube S2, the collector of the first switch tube S1 and the submodule capacitor C SM One end of the submodule capacitor C SM The other end is connected to the emitter of the second switch tube S2 and the negative terminal SM- of the submodule; the missing data estimation method comprises the following steps: Step S1: Acquire the current operation data and historical data of the state parameters of the UPFC submodules, convert the data of different state parameters into a unified format, and further pre-process the data to form an operation data set and a historical data set.

[0024] Preferably, in step S1, the data of different state parameters include submodule capacitance C SM Voltage u c , the current of the positive terminal SM+ of the submodule i SM , submodule temperature T SM ; Converting data of different state parameters into a unified format means aligning data of different parameters according to time series and associating the aligned data according to the data source.

[0025] Preferably, in step S1, the data is preprocessed as follows: a filter is used to remove measurement noise and system noise, and after the noise is removed, the average value of the historical data is calculated, and it is determined whether the data exceeds a threshold value set by a multiple of the average value. If so, it is an outlier, otherwise it is not an outlier, all outliers are removed, and multiple sensors are used to sample the same data, and verification is performed by checking whether the data are the same.

[0026] Preferably, the filter is an adaptive Kalman filter, which is expressed as: , In the above formula, for k Estimate the state vector at each moment, for k-1 Estimate the state vector at each moment, for k The one-step predicted state vector at time , G kis the adaptive Kalman filter gain, z k for k The time measurement vector, H k for k The moment measurement matrix, F k-1 yes k -1 time state transfer matrix, P k|k-1 for k The one-step forecast state error covariance matrix at time , P k for k The estimated state error covariance matrix at time instant, β k for k Time adaptive factor, R k and Q k-1 They are k The covariance matrix of the estimated measurement noise at each moment and k -1 time to estimate the covariance matrix of the system noise, I is the identity matrix; The abnormal value is a parameter value whose standard deviation exceeds a set multiple of the average value, and the number of selected parameter data points is set according to the parameter time scale and the change trend.

[0027] Step S2: Detect the operating data set and the historical data set, find and locate the missing data points, and complete and supplement the data to form a repaired operating data set.

[0028] Preferably, in step S2, the operating data set and the historical data set are tested, missing data points are found and located, and the data are supplemented to form a repaired operating data set. Specifically, for the state parameter, after excluding abnormal values, the historical data and the operating data are divided according to the equal periodic intervals of the parameter, and there are n data points; all complete data sets divided into equal periodic intervals according to this parameter are recorded as a set X , X for n × q matrix, q is the number of complete data sets; any data set with missing records divided by equal periodic intervals according to this parameter is selected as Y , Y A historical data set or a current running data set with missing data for a certain period of time; Y and X There is a linear relationship between Y There aren One non-missing observation is recorded as Y 1, which corresponds to X The part is recorded as X 1; Y middle n 0 missing values ​​are recorded as Y m , its corresponding X Some records are X m ; Based on the part of the data that is not missing, Y and X The parameter relationship between them is estimated and the posterior distribution of the parameters is constructed; then for each missing data, a sample is drawn from the posterior distribution to obtain an estimate of the missing value.

[0029] Preferably, the equal-periodic interval length is set to one or more period lengths corresponding to the state parameter, and is set according to the degree of data integrity.

[0030] Preferably, Y and X The parameter correspondence is set to have a normal linear correlation, that is, Y ~ N ( Xβ , s 2 ), β is the regression coefficient, which is a q dimensional vector, s 2 is the residual, for The normal distribution function of .

[0031] Preferably, the Y and X The parameters of the linear relationship between are estimated and the posterior distribution of the parameters is constructed. The specific process is: Y There are n One non-missing observation is recorded as Y 1, which corresponds to X The part is recorded as X 1; Y middle n 0 missing values ​​are recorded as Y m , its corresponding X Some records are X m ; Based on the part with no missing data, it is estimated by the following formula: , In the above formula, is the estimated value of the regression coefficient, is the estimated value of the residual; for X The transpose of 1, T is the transpose symbol; Build β and s 2 The posterior distribution of s 2 The posterior distribution exhibits chi-squared distribution characteristics: , in, n is the number of state parameter observations within the same period, is the chi-square distribution function; β The posterior distribution of s 2 , is the conditional posterior distribution, and follows a normal distribution: , In the above formula, β is the regression coefficient, which is a q dimensional vector, s 2 is the residual, is the estimated value of the regression coefficient, is the estimated value of the residual.

[0032] Preferably, for each missing data, sampling is performed from the posterior distribution to obtain an estimate of the missing value, and the specific process is: generating any multiple interpolation value sets according to the following interpolation model: , In the above formula, For any set of interpolation values, s * is any random draw from the posterior distribution, and β * for Any random draw from the conditional posterior distribution when Z 1 Yes n 0 random standard normal distribution values, that is: , In the above formula, g Originated from x 2 ( n 1- q ) distribution, Z 2 Yes q random standard normal distribution values; Take the arithmetic average of all interpolated data sets to get the repaired data set Y’ .

[0033] Step S3: Based on the support vector machine, calibrate and update the data in the repaired running data set; Preferably, in step S3, the data in the repaired running data set is calibrated and updated, specifically: the repaired running data set is recorded as Y’ , given a set of data ( x i , y i ), i =1,2,…, m , x i ∈ Y ', y i ∈{-1, 1}, where m is the total number of a given set of data. When the samples are linearly separable, an optimal hyperplane can be found to separate them, that is:

[0034] In the above formula, and Represent the weight vector and the optimal hyperplane offset respectively, yes The transpose of is the independent variable of the hyperplane; The support vector machine method maps samples from the original space to a high-dimensional feature space through soft margin optimization and kernel techniques. The support vector machine method finds a linear hyperplane with the largest margin in this high-dimensional space to classify the data. The linear hyperplane at this time is ,in is a function that maps high-dimensional space; for x The feature vector after being mapped to the high-dimensional space; solve the following optimization problem: , In the above formula, C >0 is the penalty factor, x i is the slack variable, for x i The eigenvector after being mapped to the high-dimensional space; in the calculation process of the above formula, it is necessary to solve ,in is a function that maps high-dimensional space; for x The feature vector after being mapped to the high-dimensional space; solve the following optimization problem: , in, is one of the independent variables of the kernel function, and we need to solve hour , yes The transpose of In the support vector machine method, there are four kernel functions, which are as follows: , In the above formula, r , c and d are kernel function parameters, for The transpose of , , , are the names of four kernel functions; The restored dataset calibrated by the support vector machine method is added to the historical dataset to form an updated dataset.

[0035] Step S4: Integrate the calibrated and updated operating data set with the historical data set to update the historical data set.

[0036] Specifically, three types of symbols are set to represent submodule voltages: u c , submodule current i SM , submodule temperature T SM , set the minimum and maximum iterations to 50 and 1000 respectively, the termination tolerance to 1e-9, and the missing ratio of each group of data to 50%. Both historical data and operating data are generated randomly, and the simulated data missing meets the assumption of random missing (MAR). The model results under the condition of 50% missing data disclosed in the embodiment of the present invention are as follows Figure 4 As shown, it can be seen that after the Bayesian multiple interpolation and support vector machine method provided by the present invention, the missing proportion of the three symbol data is controlled at 9.4%, and the confidence intervals are 63.5%, 24.4%, and 12.1% respectively.

[0037] Embodiment 2 of the present invention provides a missing data estimation system for monitoring state parameters of UPFC submodules, using the method described in Embodiment 1 of the present invention, including a data acquisition module, a data preprocessing module, a repair operation data set module, a repair data calibration module, and a data update module, including the following modules: Data acquisition module, which obtains the operation data and historical data of UPFC submodules and converts them into a unified format; The data preprocessing module preprocesses the data, removes noise in specific frequency bands, monitors outliers in the data, and cross-validates the data by comparing historical data with other sensor data to optimize the data filtering effect and form operating data sets and historical data sets; Repair the running data set module, find and locate missing data points, and supplement the missing data; The repair data calibration module, combined with the support vector machine method, calibrates and updates the repaired data to form a running data set with higher estimation accuracy; The data update module integrates the operating data set with higher estimation accuracy with the historical data set and updates the historical data set.

[0038] The beneficial effect of the present invention is that, compared with the prior art, the present invention adopts multiple Bayesian interpolation and support vector machine methods to process and optimize the UPFC submodule status information, unifies and preprocesses the submodule operation data and historical data, then locates the missing data points of the historical data and the operation data, and uses the Bayesian method to multiple interpolate the missing data, which is universal and meets the actual needs of UPFC submodule status monitoring, and then further combines the repaired data set with the support vector machine method to locate abnormal interpolation data, update the data, and form a higher estimation accuracy operation data set; the algorithm has low computational complexity and improves the repair accuracy, greatly improves the reliability of data recovery, achieves data integrity, and has stronger use value.

[0039] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0040] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0041] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0042] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A missing data estimation method for monitoring state parameters of a UPFC submodule, wherein the UPFC submodule comprises a submodule positive terminal SM+, a submodule negative terminal SM-, a submodule capacitor C SM , the first switch tube S1, the second switch tube S2, the submodule positive terminal SM+ is connected to the emitter of the first switch tube S1 and the collector of the second switch tube S2, the collector of the first switch tube S1 and the submodule capacitor C SM One end of the submodule capacitor C SM The other end is connected to the emitter of the second switch tube S2 and the negative terminal SM- of the submodule; characterized in that, The missing data estimation method comprises the following steps: Step S1: obtaining the current operating data and historical data of the state parameters of the UPFC submodules, converting the data of different state parameters into a unified format, and further preprocessing the data to form an operating data set and a historical data set; Step S2: Detect the operating data set and the historical data set, find and locate the missing data points, and complete and supplement the data to form a repaired operating data set; Step S3: Based on the support vector machine, calibrate and update the data in the repaired running data set; Step S4: Integrate the calibrated and updated operating data set with the historical data set to update the historical data set.

2. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 1, characterized in that: In step S1, the data of different state parameters include the submodule capacitance C SM Voltage u c , the current of the positive terminal SM+ of the submodule i SM , submodule temperature T SM ; Converting data of different state parameters into a unified format means aligning data of different parameters according to time series and associating the aligned data according to the data source.

3. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 1 or 2, characterized in that: In step S1, the data is preprocessed specifically as follows: a filter is used to remove measurement noise and system noise, the average value of historical data is calculated after the noise is removed, and it is determined whether the data exceeds a threshold value set by a multiple of the average value. If so, it is an outlier, otherwise it is not an outlier, all outliers are removed, and multiple sensors are used to sample the same data, and verification is performed by checking whether the data are the same.

4. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 1, characterized in that: In step S2, the operation data set and the historical data set are tested, missing data points are found and located, and the data are supplemented to form a repaired operation data set. Specifically, for the state parameter, after excluding abnormal values, the historical data and the operation data are divided according to the equal periodic intervals of the parameter. n data points; all complete data sets divided into equal periodic intervals according to this parameter are recorded as a set X , X for n × q matrix, q is the number of complete data sets; Randomly select a missing data set divided into equal periodic intervals according to the parameter as Y , Y A historical data set or a current running data set with missing data for a certain period of time; Y and X There is a linear relationship between Y There are n One non-missing observation is recorded as Y 1, which corresponds to X The part is recorded as X 1; Y middle n 0 missing values ​​are recorded as Y m , its corresponding X Some records are X m ; Based on the part of the data that is not missing, Y and X The parameter relationship between them is estimated and the posterior distribution of the parameters is constructed; then for each missing data, a sample is drawn from the posterior distribution to obtain an estimate of the missing value.

5. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 4, characterized in that: The equal periodic interval length is set to one or more period lengths corresponding to the state parameter, and is set according to the degree of data integrity.

6. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 4, characterized in that: Will Y and X The parameter correspondence is set to have a normal linear correlation, that is, Y ~ N ( Xβ , σ 2 ), β is the regression coefficient, which is a q dimensional vector, σ 2 is the residual, for The normal distribution function of .

7. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 6, characterized in that: The pair Y and X The parameters of the linear relationship between are estimated and the posterior distribution of the parameters is constructed. The specific process is: Y There are n One non-missing observation is recorded as Y 1, which corresponds to X The part is recorded as X 1; Y middle n 0 missing values ​​are recorded as Y m , its corresponding X Some records are X m ; Based on the part with no missing data, it is estimated by the following formula: ， In the above formula, is the estimated value of the regression coefficient, is the estimated value of the residual; for X The transpose of 1, T is the transpose symbol; Build β and σ 2 The posterior distribution of σ 2 The posterior distribution exhibits chi-squared distribution characteristics: ， in, n is the number of state parameter observations within the same period, is the chi-square distribution function; β The posterior distribution of σ 2 , is the conditional posterior distribution, and follows a normal distribution: ， In the above formula, β is the regression coefficient, which is a q dimensional vector, σ 2 is the residual, is the estimated value of the regression coefficient, is the estimated value of the residual.

8. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 7, characterized in that: For each missing data, sampling is performed from the posterior distribution to obtain an estimate of the missing value. The specific process is: generating any number of interpolation value sets according to the following interpolation model: ， In the above formula, For any set of interpolation values, σ * is any random draw from the posterior distribution, and β * for Any random draw from the conditional posterior distribution when Z 1 Yes n 0 random standard normal distribution values, that is: ， In the above formula, g Originated from χ 2 ( n 1 - q ) distribution, Z 2 Yes q random standard normal distribution values; Take the arithmetic average of all interpolated data sets to get the repaired data set Y’ .

9. A missing data estimation method for monitoring state parameters of a UPFC submodule according to claim 1, characterized in that: In step S3, the data in the repaired running data set is calibrated and updated, specifically: the repaired running data set is recorded as Y’ , given a set of data ( x i , y i ), i =1,2,…, m , x i ∈ Y ', y i ∈{-1, 1}, where m is the total number of a given set of data. When the samples are linearly separable, an optimal hyperplane can be found to separate them, that is: ， In the above formula, and Represent the weight vector and the optimal hyperplane offset respectively, yes The transpose of is the independent variable of the hyperplane; The support vector machine method maps samples from the original space to a high-dimensional feature space through soft margin optimization and kernel techniques. The support vector machine method finds a linear hyperplane with the largest margin in this high-dimensional space to classify the data. The linear hyperplane at this time is ,in is a function that maps high-dimensional space; for x The feature vector after being mapped to the high-dimensional space; solve the following optimization problem: ， In the above formula, C >0 is the penalty factor, ξ i is the slack variable, is the eigenvector after being mapped to the high-dimensional space; in the calculation process of the above formula, it is necessary to solve , use the kernel function to solve, and define the kernel function as follows: ， in, is one of the independent variables of the kernel function, and we need to solve hour , yes The transpose of In the support vector machine method, there are four kernel functions, which are as follows: ， In the above formula, r , γ and d are kernel function parameters, for The transpose of , , , are the names of four kernel functions; The restored dataset calibrated by the support vector machine method is added to the historical dataset to form an updated dataset.

10. A missing data estimation system for monitoring state parameters of UPFC submodules, using the method according to any one of claims 1 to 9, comprising a data acquisition module, a data preprocessing module, a repair operation data set module, a repair data calibration module, and a data update module, characterized in that: Data acquisition module, which obtains the operation data and historical data of UPFC submodules and converts them into a unified format; The data preprocessing module preprocesses the data, removes noise in specific frequency bands, monitors outliers in the data, and cross-validates the data by comparing historical data with other sensor data to optimize the data filtering effect and form operating data sets and historical data sets; Repair the running data set module, find and locate missing data points, and supplement the missing data; The repair data calibration module, combined with the support vector machine method, calibrates and updates the repaired data to form a running data set with higher estimation accuracy; The data update module integrates the operating data set with higher estimation accuracy with the historical data set and updates the historical data set.

Images