Low-voltage transformer area power grid three-phase dynamic state estimation method based on data fusion technology

Abstract

The invention discloses a low-voltage transformer area power grid three-phase dynamic state estimation method based on a data fusion technology. The method comprises the following steps: a, establishing a low-voltage transformer area power grid measurement equation; b, unscented transformation and a dynamic prediction method; the low-voltage transformer area power grid three-phase DSE method based on the DF technology has the advantages that the operation state of the low-voltage transformer area power grid is accurately estimated in real time by combining the mixed measurement data of the intelligent electric meter and the synchronous phasor measurement unit (PMU) and utilizing the data fusion technology, powerful support is provided for safe and economical operation of the power grid, and the reliability of the low-voltage transformer area power grid is improved. The method also has relatively high adaptability and expandability, and can meet the state estimation requirements of different low-voltage transformer area power grids.

Description

Technical Field

[0001] The present invention relates to the technical field of a three-phase dynamic state estimation method for a power grid, and more specifically to a three-phase dynamic state estimation method for a low-voltage power grid. Background Art

[0002] With the rapid development of smart grids, state estimation (SE) of low-voltage distribution networks (LDNs) has become an important part of power system analysis and management. However, due to the complexity and uncertainty of low-voltage distribution networks, measurement data often have problems such as noise and outliers, which require effective preprocessing and filtering. At the same time, traditional state estimation methods often fail to make full use of multi-source information, resulting in limited accuracy and robustness of estimation results.

[0003] With the development of Internet of Things technology, more and more smart meters (SM) and phasor measurement units (PMU) are deployed in low-voltage power grids, realizing real-time measurement of the power grid operation status. The key technologies of SE in LDN include preprocessing technology of measurement data, optimization technology of SE algorithm, and identification technology of model parameters. The preprocessing technology of measurement data includes data cleaning, filtering, interpolation, etc., which aims to eliminate noise and outliers and improve the accuracy and reliability of data. The optimization technology of SE algorithm focuses on how to improve the convergence speed of the algorithm, reduce the computational complexity, and improve the accuracy and robustness of the estimation results. The identification technology of model parameters accurately estimates the structure and parameters of the power grid and provides a reliable model basis for SE.

[0004] The structure of LDN is complex, and the amount of measurement data is large and scattered. How to effectively integrate and utilize this data is a difficult problem. With the rapid development of information technology, data fusion (DF) technology has gradually become a major research hotspot in the current information technology field. In short, data fusion is to obtain more comprehensive, accurate and valuable information by collecting and processing data from different information sources, thereby providing strong support for decision-making. The power system SE is based on the underlying collection of perception data from multiple sources to filter out data harmonics and restore the most realistic state of the system. The advantage of DF technology is that it can make full use of the advantages of various information sources, make up for the shortcomings of a single information source, and improve the reliability and effectiveness of information. At the same time, data fusion technology can also model and process uncertainties in complex environments, and improve the robustness and adaptability of decision-making. Summary of the invention

[0005] The present invention provides a three-phase dynamic state estimation (DSE) method for low-voltage power grid based on data fusion technology. The method combines the mixed measurement data of smart meters and PMUs and uses data fusion technology for processing, so as to achieve accurate and real-time estimation of the operating status of low-voltage power grids. The method can not only improve the accuracy and robustness of the estimation results, but also make full use of the advantages of various information sources to improve the reliability and effectiveness of information. In addition, the method also has strong adaptability and scalability, and can meet the state estimation requirements of different low-voltage power grids.

[0006] The technical solutions adopted by the present invention to solve the above technical problems are as follows:

[0007] The three-phase dynamic state estimation method of low-voltage power grid based on data fusion technology has the following steps:

[0008] a. Establish the low voltage area power grid measurement equation

[0009] a1. Smart meter measurement equation

[0010] SM measurement includes three-phase injected active power P i d , inject reactive power Voltage modulus Injection current magnitude where d∈B p , B p ={a,b,c} is a three-phase node set,

[0011] Based on three-phase active injection P i d 、Reactive Calculate the three-phase equivalent injection current and take the real and imaginary parts respectively, and we can get:

[0012]

[0013] Where: P i d and Inject active and reactive power into three phases respectively; f i d and f i n are the real and imaginary parts of the d-phase and neutral point n voltages at node i, respectively; are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes,

[0014] Based on the three-phase equivalent injection current, the real and imaginary parts of the injected neutral point current are as follows:

[0015]

[0016] Where: P i d and Inject active and reactive power into three phases respectively; f i d and f i n are the real and imaginary parts of the d-phase and neutral point n voltages at node i, respectively; are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes; B p ={a,b,c} is a three-phase node set,

[0017] Three-phase voltage modulus The equation is expressed as follows:

[0018]

[0019] Where: f i d and f i n are the real and imaginary parts of the voltage of phase d and neutral point n at node i, respectively; B p ={a,b,c} is a set of three-phase nodes; is the d-phase voltage modulus at node i, and the three-phase injection current modulus The equation is expressed as follows:

[0020]

[0021] Where: are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes; B p ={a,b,c} is a set of three-phase nodes; The magnitude of the d-phase current at node i,

[0022] a2. Synchronous phasor measurement equation

[0023] PMU measurements include phase voltages on nodes (t∈B1) and the associated branch current (t∈B1, and j≠i),

[0024] Voltage phasors of each node phase The partial and imaginary part equations of t∈B1 are expressed as follows:

[0025]

[0026] Where: The measurement Real part, imaginary part; f i t The real and imaginary parts of the t-phase voltage at node i, respectively,

[0027] Based on the associated branch current phasor t∈B1, and taking its real and imaginary parts, we get:

[0028]

[0029] In the formula: real and imag are operations to preserve the real and imaginary parts respectively; f i x and are the real and imaginary parts of the x-phase voltage at nodes i and j, respectively; and They are mutual guidance and mutual acceptance; is the phasor value of the branch current; B1={a,b,c,n} is the set of three-phase and neutral point nodes;

[0030] b. Untraceable transformation and dynamic prediction method

[0031] b1. Dynamic state model

[0032] The dynamic model and measurement equation of the state quantity are as follows:

[0033]

[0034] Where: X k , X k+1 are the state quantities at time k and k+1 respectively; Z k is the quantity measured at time k; f is the state prediction function; h is the measurement equation; ν k is the measured noise; ωk is the state noise,

[0035] b2. Dynamic prediction model

[0036] State prediction function for:

[0037]

[0038] Where: They are the predicted states at time k and k+1 respectively; is the estimated state at time k; a k 、a k-1 are the horizontal prediction components at time k and k-1 respectively; b k 、b k-1 are the tilt prediction components at time k and k-1 respectively; α and β are prediction coefficients,

[0039] b3. Traceless transformation method

[0040] The estimated value of the n-dimensional state under known initial conditions and the n×n dimensional variance matrix P x,0 , the following Sigma point set can be constructed

[0041]

[0042] Where: is the estimated state at time k; is the Sigma point at time k; n is the dimension of the state quantity; P x,k is the variance matrix of the state quantity at time k,

[0043] Based on the state prediction function f(·), Perform state prediction at time k+1:

[0044]

[0045] Where: is the Sigma point at time k; is the Sigma point at time k+1 after prediction; f(·) is the state prediction function,

[0046] Calculate predicted status The mean of :

[0047]

[0048] Where: i =1 / 2n is the weight; is the predicted value of the state at time k+1; is the Sigma point at time k+1 after prediction,

[0049] The state prediction variance is:

[0050]

[0051] Where: is the state prediction variance at time k+1; ω i =1 / 2n is the weight; is the predicted value of the state at time k+1; is the Sigma point at time k+1 after prediction; T is the transposition operation,

[0052] State prediction value As a virtual measurement, the measurement equation can be obtained as:

[0053]

[0054] Where: x k+1 is the state truth value; e 1,k+1 is the predicted state error; is the predicted value of the state at time k+1;

[0055] c. Three-phase DSE method for low-voltage power grid based on DF technology

[0056] c1. Parallel filter estimation based on PMU measurement and SM measurement

[0057] In order to make full use of the system data and increase the LDN measurement redundancy, the SM measurement Z sm and PMU measurement Z pmu Respectively with the predicted state Combined, parallel filtering estimation is performed. When time k+1 arrives, let represents the combined measurement of the smart meter and the predicted state, let represents the combined measurement of PMU and predicted state,

[0058] Combined measurement of SM measurement and predicted state The filtering model is established as follows. The measurement equation is:

[0059]

[0060] Where: is the measurement implicit function; is the measurement error, The measurement error variance Σ 1,k+1 for:

[0061]

[0062] Where: R 1,k+1 v1,k+1 The variance of

[0063] The optimization goal is to minimize the sum of residual variances. The objective function J1(x k+1 )for:

[0064]

[0065] stc1(x k+1 )=0 (20)

[0066] Where: is the residual; W is the weight; c1(x k+1 ) is the constraint equation, for the objective function J1(x k+1 ), based on the Lagrange multiplier method and Newton method, the state correction equation can be obtained:

[0067]

[0068] Where: superscript j is the jth iteration; is the state quantity increment; is the Lagrange multiplier; H1 and C1 are the measurement equations and the constraint equation c1(x k+1 ),

[0069] The status is modified to:

[0070]

[0071] The convergence threshold is:

[0072]

[0073] SE error variance Γ 1,k+1 for:

[0074]

[0075] Thus, the k+1 time measurement can be obtained The filtered state estimate and variance Γ 1,k+1 Similarly, the combined measurement of PMU and predicted state By building a filtering model and solving it, we can obtain the estimated state value after filtering at time k+1. and variance Γ 2,k+1 ;

[0076] C2. Parallel filtering DSE method based on DF technology

[0077] At time k+1, the measurement The estimated state after filtering and variance Γ 1,k+1 , Measurement The estimated state after filtering and variance Γ 2,k+1 , through information fusion processing, the final estimated state is obtained. Estimated state and The fusion expression is as follows:

[0078]

[0079] Where: is the final estimated state at time k+1.

[0080] The beneficial effects that can be achieved by the present invention using the above technical solution are:

[0081] This method combines the hybrid measurement data of smart meters and synchronized phasor measurement units (PMUs) and uses data fusion technology to achieve accurate and real-time estimation of the operating status of the low-voltage power grid, providing strong support for the safe and economical operation of the power grid.

[0082] This method can not only improve the accuracy and robustness of the estimation results, but also make full use of the advantages of various information sources to improve the reliability and effectiveness of the information.

[0083] This method also has strong adaptability and scalability, and can meet the state estimation needs of power grids in different low-voltage areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 This is a diagram of the LDN line model in the present invention;

[0085] Figure 2 This is a schematic diagram of SM wiring in the present invention;

[0086] Figure 3 It is a flow chart of LDN parallel filtering DSE based on DF technology in the present invention;

[0087] Figure 4 This is a measurement configuration diagram of the smart meter and PMU in the system of the present invention. DETAILED DESCRIPTION

[0088] The three-phase dynamic state estimation method of low-voltage power grid based on data fusion technology has the following steps:

[0089] a. Establish the low voltage area power grid measurement equation

[0090] LDN line model diagram Figure 1 As shown,

[0091] a1. Smart meter measurement equation

[0092] SM measurement includes three-phase injected active power P i d , inject reactive power Voltage modulus Injection current magnitude where d∈B p , B p ={a,b,c} is a three-phase node set, Figure 2 The following is the SM wiring diagram.

[0093] Based on three-phase active injection P i d 、Reactive Calculate the three-phase equivalent injection current and take the real and imaginary parts respectively, and we can get:

[0094]

[0095] Where: P i d and Inject active and reactive power into three phases respectively; f i d and f i n are the real and imaginary parts of the d-phase and neutral point n voltages at node i, respectively; are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes,

[0096] Based on the three-phase equivalent injection current, the real and imaginary parts of the injected neutral point current are as follows:

[0097]

[0098]

[0099] Where: P i d and Inject active and reactive power into three phases respectively; f i d and f i n are the real and imaginary parts of the d-phase and neutral point n voltages at node i, respectively; are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes; B p ={a,b,c} is a three-phase node set,

[0100] Three-phase voltage modulus The equation is expressed as follows:

[0101]

[0102] Where: f i d and f i n are the real and imaginary parts of the voltage of phase d and neutral point n at node i, respectively; B p ={a,b,c} is a set of three-phase nodes; is the d-phase voltage modulus at node i, and the three-phase injection current modulus The equation is expressed as follows:

[0103]

[0104] Where: are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes; B p ={a,b,c} is a set of three-phase nodes; The magnitude of the d-phase current at node i,

[0105] a2. Synchronous phasor measurement equation

[0106] PMU measurements include phase voltages on nodes (t∈B1) and the associated branch current (t∈B1, and j≠i),

[0107] Voltage phasors of each node phase The partial and imaginary part equations of t∈B1 are expressed as follows:

[0108]

[0109] Where: The measurement Real part, imaginary part; f i tThe real and imaginary parts of the t-phase voltage at node i, respectively,

[0110] Based on the associated branch current phasor t∈B1, and taking its real and imaginary parts, we get:

[0111]

[0112] In the formula: real and imag are operations to preserve the real and imaginary parts respectively; f i x and are the real and imaginary parts of the x-phase voltage at nodes i and j, respectively; and They are mutual guidance and mutual acceptance; is the phasor value of the branch current; B1={a,b,c,n} is the set of three-phase and neutral point nodes;

[0113] b. Untraceable transformation and dynamic prediction method

[0114] b1. Dynamic state model

[0115] The dynamic model and measurement equation of the state quantity are as follows:

[0116]

[0117] Where: X k , X k+1 are the state quantities at time k and k+1 respectively; Z k is the quantity measured at time k; f is the state prediction function; h is the measurement equation; ν k is the measured noise; ω k is the state noise,

[0118] b2. Dynamic prediction model

[0119] State prediction function for:

[0120]

[0121] Where: They are the predicted states at time k and k+1 respectively; is the estimated state at time k; a k 、a k-1 are the horizontal prediction components at time k and k-1 respectively; b k 、b k-1 are the tilt prediction components at time k and k-1 respectively; α and β are prediction coefficients,

[0122] b3. Traceless transformation method

[0123] The estimated value of the n-dimensional state under known initial conditions and the n×n dimensional variance matrix P x,0 , the following Sigma point set can be constructed

[0124]

[0125] Where: is the estimated state at time k; is the Sigma point at time k; n is the dimension of the state quantity; P x,k is the variance matrix of the state quantity at time k,

[0126] Based on the state prediction function f(·), Perform state prediction at time k+1:

[0127]

[0128] Where: is the Sigma point at time k; is the Sigma point at time k+1 after prediction; f(·) is the state prediction function,

[0129] Calculate predicted status The mean of :

[0130]

[0131] Where: i =1 / 2n is the weight; is the predicted value of the state at time k+1; is the Sigma point at time k+1 after prediction,

[0132] The state prediction variance is:

[0133]

[0134] Where: is the state prediction variance at time k+1; ω i =1 / 2n is the weight; is the predicted value of the state at time k+1; is the Sigma point at time k+1 after prediction; T is the transposition operation,

[0135] State prediction value As a virtual measurement, the measurement equation can be obtained as:

[0136]

[0137] Where: x k+1 is the state truth value; e 1,k+1is the predicted state error; is the predicted value of the state at time k+1;

[0138] c. Three-phase DSE method for low-voltage power grid based on DF technology

[0139] c1. Parallel filter estimation based on PMU measurement and SM measurement

[0140] In order to make full use of the system data and increase the LDN measurement redundancy, the SM measurement Z sm and PMU measurement Z pmu Respectively with the predicted state Combined, parallel filtering estimation is performed. When time k+1 arrives, let represents the combined measurement of the smart meter and the predicted state, let represents the combined measurement of PMU and predicted state,

[0141] Combined measurement of SM measurement and predicted state The filtering model is established as follows. The measurement equation is:

[0142]

[0143] Where: is the measurement implicit function; is the measurement error, The measurement error variance Σ 1,k+1 for:

[0144]

[0145] Where: R 1,k+1 v 1,k+1 The variance of

[0146] The optimization goal is to minimize the sum of residual variances. The objective function J1(x k+1 )for:

[0147]

[0148] stc1(x k+1 )=0 (20)

[0149] Where: is the residual; W is the weight; c1(x k+1 ) is the constraint equation, for the objective function J1(x k+1 ), based on the Lagrange multiplier method and Newton method, the state correction equation can be obtained:

[0150]

[0151] Where: superscript j is the jth iteration; is the state quantity increment; is the Lagrange multiplier; H1 and C1 are the measurement equations and the constraint equation c1(x k+1 ),

[0152] The status is modified to:

[0153]

[0154] The convergence threshold is:

[0155]

[0156] SE error variance Γ 1,k+1 for:

[0157] Γ 1,k+1 =[H1 T WH1] -1 H1 T WΣ 1,k+1 WH1([H1 T WH1] -1 ) T (twenty four)

[0158] Thus, the k+1 time measurement can be obtained The filtered state estimate

[0159] and variance Γ 1,k+1 Similarly, the combined measurement of PMU and predicted state By building a filtering model and solving it, we can obtain the estimated state value after filtering at time k+1. and variance Γ 2,k+1 ;

[0160] c2. Parallel filtering DSE method based on DF technology

[0161] At time k+1, the measurement The estimated state after filtering and variance Γ 1,k+1 , Measurement The estimated state after filtering and variance Γ 2,k+1 , through information fusion processing, the final estimated state is obtained. Estimated state and The fusion expression is as follows:

[0162]

[0163] Where: is the final estimated state at time k+1.

[0164] Figure 3 Shown is the calculation process of the parallel filtering DSE method based on DF technology in LDN.

[0165] Simulation cases and conditions

[0166] In order to verify the effectiveness of the three-phase dynamic state estimation method of low-voltage power grid based on data fusion technology proposed in this method, a simulation analysis is carried out based on the IEEE-13 node distribution correction system. Figure 4 Shown is the system smart meter and PMU measurement configuration diagram.

[0167] The daily load fluctuation law of the power grid in the substation area can be found in the literature [Liu Xianglin, Xiang Xiaomin, Yang Zhe, et al. Two-stage dynamic state estimation method of power system considering data delay [J]. Smart Power, 2021, 49(04):75-81. ], the measurement error of the voltage modulus and current modulus of the smart meter is set to 0.01, and the measurement error of active and reactive power is set to 0.008; the measurement error of the PMU phase measurement is set to 0.001, and the simulation of the measurement data is the same as the literature [Chen Xiaoyi, Wang Dong, Cao Zhi, et al. Three-phase state estimation of low-voltage distribution network based on hybrid measurement of smart meter and PMU [J]. Power System and Clean Energy, 2019, 35(03):14-20. ].

[0168] The following three schemes are constructed for simulation comparison:

[0169] 1) Solution 1: LDN three-phase DSE method based on hybrid SM measurement and PMU measurement;

[0170] 2) Solution 2: LDN two-stage DSE method based on SM measurement and PMU measurement;

[0171] 3) Scheme 3: The LDN three-phase DSE method based on DF technology proposed in this method.

[0172] For the simulated measurement data, the following two simulation conditions are set:

[0173] 1) No bad data in analog measurement;

[0174] 2) The simulated measurement data contains bad data, and the power measured by the smart meter on node 4 is changed to 150% of the original power.

[0175] Simulation Results and Analysis

[0176] (1) Estimation accuracy analysis

[0177] In order to evaluate the accuracy of DSE results under different policy schemes and conditions, the root mean square error E of the LDN three-phase DSE is defined as RMS for:

[0178]

[0179] E RMS,max =max(E RMS,k (x)) (25)

[0180]

[0181] Where: is the estimated value of the i-th state at time k; x k,i is the true value of the i-th state at time k; n is the number of nodes; E RMS,k (x) is the three-phase DSE root mean square error at time k; E RMS,max is the maximum RMS error of three-phase DSE; E RMS,mean is the average RMS error of the three-phase DSE.

[0182] The three constructed simulation schemes are compared under two simulation conditions. Tables 1 and 2 show the root mean square error E of the LDN three-phase DSE with and without bad data. RMS result.

[0183] Table 1. RMS error results of LDN three-phase DSE under simulation conditions without bad data

[0184]

[0185] Table 2. LDN three-phase DSE RMS error results under bad data simulation conditions

[0186]

[0187] In the LDN three-phase DSE RMS error results under the simulation condition without bad data in Table 1, the E RMS,max (V) and E RMS,mean (V) and E RMS,max (θ) and E RMS,mean (θ) are 0.0243pu, 0.0078pu, 0.0222rad, and 0.0035rad respectively; E of Scheme 2 RMS,max (V) and E RMS,mean (V) and E RMS,max (θ) and E RMS,mean (θ) are 0.0218pu, 0.0072pu, 0.0197rad, and 0.0032rad respectively; E of the proposed method in Scheme 3 RMS,max (V) and E RMS,mean (V) and E RMS,max (θ) and E RMS,mean(θ) are 0.0196pu, 0.0067pu, 0.0184rad, and 0.0028rad respectively. Comparing the simulation results of the three schemes without bad data, it can be seen that in the LDN three-phase DSE, the root mean square error E of the LDN three-phase DSE proposed in this paper is RMS Both are minimum.

[0188] In the LDN three-phase DSE RMS error results under the simulation condition without bad data in Table 2, the E RMS,max (V) and E RMS,mean (V) and E RMS,max (θ) and E RMS,mean (θ) are 0.1027pu, 0.0131pu, 0.0572rad, and 0.0093rad respectively; E of Scheme 2 RMS,max (V) and E RMS,mean (V) and E RMS,max (θ) and E RMS,mean (θ) are 0.0579pu, 0.0102pu, 0.0495rad, and 0.0074rad respectively; the E of the proposed method in scheme 3 RMS,max (V) and E RMS,mean (V) and E RMS,max (θ) and E RMS,mean (θ) are 0.0329pu, 0.0089pu, 0.0354rad, and 0.0049rad respectively. Comparing the simulation results of the three schemes with bad data, it can be seen that in the LDN three-phase DSE, the root mean square error E of the LDN three-phase DSE proposed in this paper is RMS Still minimal.

[0189] From the results in Tables 1 and 2 above, it can be seen that the accuracy of the method proposed in Scheme 3 is the highest regardless of whether there is bad data or not. The main reason is that Scheme 1 uses the SM measurement and PMU measurement in LDN to mix and filter directly. The SM measurement accuracy is low and the number of measurements is large. The PMU measurement with higher measurement accuracy will be polluted during the mixed data filtering process; Scheme 2 uses a two-stage estimation method of staged filtering of SM measurement and PMU measurement. Compared with mixed data filtering, it can effectively avoid the data pollution problem of SM measurement on PMU and increase the accuracy of three-phase estimation; The method proposed in Scheme 3 uses data fusion technology to combine SM measurement and PMU measurement with predicted values ​​for parallel filtering. Compared with the first two schemes, it can not only effectively avoid the data pollution problem of SM measurement on PMU, but also DF technology can further make full use of multi-source SM and PMU information to improve the robustness and accuracy of the estimation results.

[0190] (2) Computational efficiency analysis

[0191] Table 3 shows the calculation time used by the LDN three-phase DSE for the three constructed simulation schemes under two simulation conditions.

[0192] Table 3 LDN three-phase DSE calculation time

[0193]

[0194] In Table 3, under the simulation condition without bad data, the calculation time of LDN three-phase DSE of Scheme 1, Scheme 2 and Scheme 3 is 177.6921s, 196.4836s and 139.8721s respectively; under the simulation condition with bad data, the calculation time of LDN three-phase DSE of Scheme 1, Scheme 2 and Scheme 3 is 234.3375s, 279.3321s and 187.4962s respectively.

[0195] It can be seen that in the LDN three-phase DSE, Scheme 2 takes the longest time and has the lowest computational efficiency, while the method proposed in Scheme 3 has the shortest computational time and the highest computational efficiency. The main reason is that in Scheme 2, SM measurement and PMU measurement are estimated separately in two stages in the LDN three-phase DSE. Compared with Scheme 3 and Scheme 1, a filtering process is added, which greatly increases the overall time. In Scheme 3, the LDN three-phase DSE method based on DF technology proposed in this paper performs parallel filtering estimation on SM measurement and PMU measurement separately, which can greatly reduce the computational time, so it has the highest computational efficiency.

Claims

1. A method for estimating the three-phase dynamic state of a low-voltage power grid based on data fusion technology, characterized in that: Here are the steps: a. Establish the measurement equation of low voltage power grid a1. Smart meter measurement equation SM measurement includes three-phase injected active Injection reactive power Voltage modulus Injection current magnitude where d∈B p , B p ={a,b,c} is a three-phase node set, Based on three-phase active injection Reactive Calculate the three-phase equivalent injection current and take the real and imaginary parts respectively, and we can get: Where: and Inject active and reactive power into three phases respectively; and are the real and imaginary parts of the d-phase and neutral point n voltages at node i, respectively; are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes, Based on the three-phase equivalent injection current, the real and imaginary parts of the injected neutral point current are as follows: Where: and Inject active and reactive power into three phases respectively; and are the real and imaginary parts of the d-phase and neutral point n voltages at node i, respectively; are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes; B p ={a,b,c} is a three-phase node set, Three-phase voltage modulus The equation is expressed as follows: Where: and are the real and imaginary parts of the voltage of phase d and neutral point n at node i, respectively; B p ={a,b,c} is a set of three-phase nodes; is the d-phase voltage modulus at node i, Three-phase injection current modulus The equation is expressed as follows: Where: are the real and imaginary parts of phase t at node k respectively; and They are mutual guidance and mutual acceptance; is the set of nodes connected to node i (including node i); B1 = {a, b, c, n} is the set of three-phase and neutral point nodes; B p ={a,b,c} is a set of three-phase nodes; The d-phase current modulus of node i; a2. Synchronous phasor measurement equation PMU measurements include phase voltages on nodes and the associated branch currents The phasor value of Voltage phasors of each node phase The partial and imaginary part equations of t∈B1 are expressed as follows: Where: The measurement Real part, imaginary part; The real and imaginary parts of the t-phase voltage at node i, respectively, Based on the associated branch current phasor t∈B1, and taking its real and imaginary parts, we get: In the formula: real and imag are operations to preserve the real and imaginary parts respectively; and are the real and imaginary parts of the x-phase voltage at nodes i and j, respectively; and They are mutual guidance and mutual acceptance; is the phasor value of the branch current; B1={a,b,c,n} is the set of three-phase and neutral point nodes; b. Untraceable transformation and dynamic prediction method b1. Dynamic state model The dynamic model and measurement equation of the state quantity are as follows: Where: X k , X k+1 are the state quantities at time k and k+1 respectively; Z k is the quantity measured at time k; f is the state prediction function; h is the measurement equation; ν k is the measured noise; k is the state noise; b2. Dynamic prediction model State prediction function for: Where: They are the predicted states at time k and k+1 respectively; is the estimated state at time k; a k 、a k-1 are the horizontal prediction components at time k and k-1 respectively; b k , b k-1 are the tilt prediction components at time k and k-1 respectively; α and β are the prediction coefficients; b3. Traceless transformation method The estimated value of the n-dimensional state under known initial conditions and the n×n dimensional variance matrix P x,0 , the following Sigma point set can be constructed Where: is the estimated state at time k; is the Sigma point at time k; n is the dimension of the state quantity; P x,k is the variance matrix of the state quantity at time k, Based on the state prediction function f(·), Perform state prediction at time k+1: Where: is the Sigma point at time k; is the Sigma point at time k+1 after prediction; f(·) is the state prediction function, Calculate predicted status The mean of : Where: i =1 / 2n is the weight; is the predicted value of the state at time k+1; is the Sigma point at time k+1 after prediction, The state prediction variance is: Where: is the state prediction variance at time k+1; ω i =1 / 2n is the weight; is the predicted value of the state at time k+1; is the Sigma point at time k+1 after prediction; T is the transpose operation, State prediction value As a virtual measurement, the measurement equation can be obtained as: Where: x k+1 is the state truth value; e 1,k+1 is the predicted state error; is the predicted value of the state at time k+1; c. Three-phase DSE method for low-voltage power grid based on DF technology c1. Parallel filter estimation based on PMU measurement and SM measurement When the k+1 moment arrives, let represents the combined measurement of the smart meter and the predicted state, let represents the combined measurement of PMU and predicted state, Combined measurement of SM measurement and predicted state The filter model is established as follows: The measurement equation is: Where: is the measurement implicit function; is the measurement error, The measurement error variance Σ 1,k+1 for: Where: R 1,k+1 v 1,k+1 The variance of The optimization goal is to minimize the sum of residual variances. The objective function J1(x k+1 )for: Where: is the residual; W is the weight; c1(x k+1 ) is the constraint equation, for the objective function J1(x k+1 ), based on the Lagrange multiplier method and Newton method, the state correction equation can be obtained: Where: superscript j is the jth iteration; is the state quantity increment; is the Lagrange multiplier; H1 and C1 are the measurement equations and the constraint equation c1(x k+1 ), The status is modified to: The convergence threshold is: SE error variance Γ 1,k+1 for: Thus, the k+1 time measurement can be obtained The filtered state estimate and variance Γ 1,k+1 Similarly, the combined measurement of PMU and predicted state By building a filtering model and solving it, we can obtain the estimated state value after filtering at time k+1. and variance Γ 2,k+1 ; c2. Parallel filtering DSE method based on DF technology At time k+1, the measurement The estimated state after filtering and variance Γ 1,k+1 , Measurement The estimated state after filtering and variance Γ 2,k+1 , through information fusion processing, the final estimated state is obtained, the estimated state and The fusion expression is as follows: Where: is the final estimated state at time k+1.

Images