Active power distribution network robust three-phase state estimation method based on multi-source mixed data

Abstract

The invention discloses an active power distribution network robust three-phase state estimation method based on multi-source mixed data. The method comprises the following steps: a, establishing three-phase measurement-state models of multi-source SCADA, PMU and AMI data; b, performing measurement section normalization processing on the PMU data, the SCADA data and the AMI data by calculating the optimal buffer length of the PMU data and the SCADA data; and c, establishing an active power distribution network robust state estimation model based on an exponential weight robust state estimation method, and proposing an active power distribution network robust three-phase state estimation method. The method has the advantages that time delay processing is performed on PMU and SCADA data of a waiting area through a data buffering method, so that section measurement normalization processing is performed on the PMU, SCADA and AMI data, and after the AMI data arrives, the power distribution network robust three-phase state estimation model is established. And then SE calculation of the section can be carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of a three-phase state estimation method for a distribution network, and more specifically to the technical field of a robust three-phase state estimation method for an active distribution network. Background Art

[0002] The active distribution network (ADN) contains a large amount of distributed renewable energy. It can use advanced communication technology to independently plan, manage and coordinate new energy and energy storage devices to ensure the safe and reliable operation of the power grid while actively absorbing distributed energy.

[0003] At present, the smart distribution network is developing rapidly, and smart meters with two-way communication capabilities have been widely deployed in the Advanced Metering Infrastructure (AMI). The data types collected by smart meters include data such as power and voltage of each phase, and the default is generally to collect data once every fifteen minutes. In contrast, the phasor measurement unit (PMU) can collect phasor information at a very high frequency, generally collecting data dozens of times within 1 second, and with high accuracy.

[0004] However, due to the gradual configuration of the acquisition system, the Supervisory Control And Data Acquisition (SCADA), PMU and AMI in the active distribution network will coexist for a long time in the future. In state estimation (SE), purely using only a certain type of data may not meet the requirements of observability, and the measurement redundancy is low. Therefore, in the active distribution network, the integration of SCADA, AMI and PMU data for state estimation can significantly increase the measurement redundancy of SE and improve the accuracy of SE results. Since SCADA data, AMI smart meter data and PMU data are non-homologous data, there are time deviation and data delay problems. In order to effectively combine various types of non-homologous data, it is necessary to unify various types of measurements under the same time section.

[0005] The structure of the distribution network is a few-ring, and generally radial, with a high degree of three-phase asymmetry and incomplete measurement configuration. Some existing methods use branch power or branch current as state quantities and establish SE models based on weighted least squares (WLS). Although these methods can achieve state estimation of the distribution network, they generally do not have robustness. In order to increase measurement redundancy and improve estimation accuracy, meet the ADN's requirements for accurate state perception, integrate multi-source data, and improve the measurement redundancy of distribution network estimation and the robustness of the algorithm, it has been a research hotspot in recent years.

[0006] As a state perception technology of smart grid, state estimation has become mature in transmission network, but it has not yet been fully implemented in distribution network. State estimation can use multi-source data for filtering and restore the most realistic state of the system. Therefore, it is an urgent problem to be solved to use it as a function of active distribution network state perception, reasonably coordinate the data of SCADA, PMU and AMI systems, normalize the state data with different sampling periods to the same section, and jointly perform state estimation to obtain the most accurate current state of the system. Summary of the invention

[0007] The present invention provides a robust three-phase state estimation method for active distribution network based on multi-source mixed data to solve the above-mentioned technical problems.

[0008] The technical solutions adopted by the present invention to solve the above technical problems are as follows:

[0009] A robust three-phase state estimation method for active distribution network based on multi-source mixed data, the method steps are as follows:

[0010] a. Establish three-phase measurement-state models for multi-source SCADA, PMU and AMI data;

[0011] b. By calculating the optimal buffer length of PMU and SCADA data, the PMU, SCADA and AMI data are normalized for measurement sections;

[0012] c. Based on the exponential weighted robust state estimation method, a robust state estimation model for active distribution network is established, and a robust three-phase state estimation method for active distribution network is proposed.

[0013] Preferably, the step a is specifically as follows:

[0014] a1. AMI measurement-state model

[0015] AMI smart meters are configured at the distributed power supply endpoints and load endpoints in the active distribution network. The smart meters are set to collect data every 15 minutes to collect the three-phase voltage module values ​​at the endpoints. Injection current magnitude Injected active power size P i d , Injection reactive power

[0016] According to the injection P i d , The current injection equation can be established as follows:

[0017]

[0018] Expanding the above formula, we can get the real part and imaginary part equations respectively:

[0019]

[0020] Where: is the power injected into endpoint i; P i d is the active power injected into endpoint i; Reactive power for injection endpoint i; f i d are the real and imaginary parts of the voltage at the terminal i in phase d; is the mutual conductance and mutual admittance of the endpoints,

[0021] According to the voltage modulus The voltage modulus equation can be established as follows:

[0022]

[0023] According to the injection current modulus The current modulus equation can be established as follows:

[0024]

[0025] a2. PMU measurement-state model

[0026] At the endpoint of the active distribution network, a PMU measurement device is configured to obtain the relevant voltage and current phasors at this endpoint through the PMU. The measurement types mainly include: voltage phasors of each phase Each branch current phasor associated with the PMU endpoint (t∈B p , and j≠i),

[0027] Voltage phasor for PMU t∈B p , the voltage equation can be established as follows:

[0028]

[0029] Expanding the above formula, we can get the real part and imaginary part equations:

[0030]

[0031] Where: for The real and imaginary parts of

[0032] For each branch current phasor associated with the PMU terminal t∈B1, the branch current phasor equation can be established as follows:

[0033]

[0034] Expanding the above formula, we can get the real part and imaginary part equations:

[0035]

[0036] In the formula: real and imag represent the real and imaginary parts respectively;

[0037] a3. SCADA measurement-state model

[0038] The RTU device in the SCADA system can measure the injected power P i d , Branch power and voltage modulus The SCADA system measures the entire quantity.

[0039] For voltage modulus The voltage modulus equation can be established as follows:

[0040]

[0041] For the injection power P i d , The current injection equation can be established as follows:

[0042]

[0043] Expanding the above formula, we can get the real part and imaginary part equations respectively:

[0044]

[0045] For branch power The branch power equation can be established as follows:

[0046]

[0047] Expanding the above formula, we can get the real part and imaginary part equations:

[0048]

[0049] a4. Contact endpoint virtual measurement-state model

[0050] In the active distribution network, there are zero injection endpoints that are neither directly connected to the generator nor directly connected to the load and only serve as a connection. The virtual measurement equation can be established as follows:

[0051]

[0052] Preferably, the step b is as follows:

[0053] b1. Calculation of the optimal buffer length of PMU

[0054] 30 seconds before the arrival of AMI, the PMU has sampled 750 measurement data. The method to calculate the optimal buffer length is to assume that the PMU sampling frequency is n r sample / s. Every time the next AMI data arrives, the system updates the status. 30s before the AMI arrives:

[0055] (1) The number of PMU sampling measurements is N = 30*n r , divide N sample measurements into n subset There are measurement subsets, usually n subset =30, so the number of each measurement subset is n meas =n r , n subset The subsets are represented as: Z(1), Z(2), ..., Z(n subset ),

[0056] (2) For the latest PMU measurement subset Z(n subset ) to perform change point detection:

[0057] ① Calculate Z(n subset ) and compare the calculated standard deviation with the threshold value, the amplitude threshold value is set to 10 -4 pu, the phase angle threshold is set to 10 -3 If the standard deviation is smaller than the threshold value, it means that Z(n subset ) has no change point, Z(n subset ) is all contained in the buffer, go to (3);

[0058] ②If the standard deviation is larger than the threshold value, Z(n subset ) has a change point, then take Z(n subset ) is used as the measurement value of the equivalent section, and the calculation is completed.

[0059] (3) Based on the internal function ttest2 in Matlab, Z(n subset ) and Z(n subset -1), ..., Z(3) and Z(2), Z(2) and Z(1) are tested with a default probability of 95%, and the results are expressed as h i (i=1,2,...,n subset -1) is expressed as:

[0060]

[0061] Where: N op_bf is the optimal buffer length of PMU; the overline is the "not" operator symbol,

[0062] Therefore, based on the above steps, the optimal buffer length can be calculated for each PMU measurement point, and the mean and variance of the PMU statistical data in the buffer zone can be calculated to normalize the cross section.

[0063] b2. SCADA optimal buffer length calculation

[0064] In 15 minutes, the SCADA sampling data series has 450 measurement data, and the method for calculating the optimal buffer length is similar to that of PMU.

[0065] After normalizing the measurement sections of PMU and SCADA, all processed data and incoming AMI data can be regarded as data within the same section, and thus combined together for SE calculation.

[0066] Preferably, the step c is as follows:

[0067] After the above cross-section normalization processing, the three types of data, AMI, PMU and SCADA, are combined, and the active distribution network robust estimation model is established based on the exponential weight robust estimation method. The quantity measurement of the active distribution network system is:

[0068]

[0069] Where: z represents the full measurement of the active distribution network; z AMI 、z PMU and z SCADA Respectively represent the three types of measurements: AMI, PMU and SCADA.

[0070] Based on the normalized full measurement above, the index-weighted robust estimation model of the main distribution network is established and solved by Newton method. The index-weighted robust three-phase state estimation method of the active distribution network is proposed.

[0071] c1. Exponentially weighted robust state estimation model for active distribution network

[0072] In the exponentially weighted robust state estimation model, the objective function is to minimize the sum of the exponentially weighted residuals:

[0073]

[0074] stc(x)=0 (23)

[0075] Where: represents measurement; h(x) represents measurement function; c(x) represents equality constraint; represents the measurement residual; x represents the state quantity; W represents the diagonal matrix of the weight function, W ii =w i * ,

[0076] w i * Calculate the expression:

[0077]

[0078] Where: represents fixed weight; δ represents standard deviation, and the initial value calculation formula is δ 2 =r T Pr / (mn), P represents the prior diagonal matrix, the diagonal elements m represents the number of measurements, and n represents the number of states;

[0079] c2. Exponentially weighted robust state estimation algorithm for active distribution network

[0080] Based on the Lagrange multiplier λ, the equality constraint is transformed to obtain the unconstrained objective function, which is then optimized and solved. The calculation formula is as follows:

[0081]

[0082] Where: L(x,λ) represents the augmented Lagrange function,

[0083] Taylor expansion is performed on h(x) and c(x) and linearization of nonlinear equations is performed:

[0084]

[0085] Where: H(x0) and C(x0) represent Jocabian matrices, which are and (abbreviated as H and C),

[0086] Find the first-order partial derivative of formula (25):

[0087]

[0088] Based on the Newton method iterative solution, the incremental update matrix form is:

[0089]

[0090] Where: k represents the number of iterations; matrix Δx (k) =x (k+1) -x (k) ,

[0091] The calculation is based on the exponential weight robust state estimation algorithm. The exponential weight can automatically correct the measurement weight and suppress the influence of bad data.

[0092] The detailed calculation steps of the algorithm are as follows:

[0093] ① Initialization: set the initial values ​​of the real and imaginary parts of the three-phase voltages of the distribution network balance, load and interconnection endpoints, take the phase A of the balance endpoint as the phase reference, the earth as the potential reference, set the fixed weight matrix to R-1, set the maximum number of iterations to Tmax, set the convergence threshold to ε, and set time = 1 for the initial iteration;

[0094] ② Calculate the endpoint admittance matrix and Jocabian matrix, calculate the imbalance Δz and Δc, and calculate the diagonal element value of the weight matrix W to form the Jocabian matrix H and C, and calculate H T WH and H T WΔz;

[0095] ③Calculate the state deviation Δx (k) , and update the state value: x (k+1) =x (k) +Δx (k) , time = time + 1;

[0096] ④ Convergence judgment, if max(Δx (k) )<ε, the calculation converges and the result is output; if max(Δx (k) )>ε and time>Tmax, the iteration stops, "Not converged!". Otherwise, go to step ⑤;

[0097] ⑤ Modify the dynamic scale parameter, if (δ (k) ) 2 ≥0.01, (δ (k+1) ) 2 =0.001(δ (k) ) 2 ; If (δ (k) ) 2 <0.01, (δ (k+1) ) 2 =(δ (k) ) 2 , go to step ②.

[0098] The beneficial effects that can be achieved by the present invention using the above technical solution are:

[0099] 1. For the multi-source SCADA, PMU and AMI data in the active distribution network, three-phase measurement-state models were established respectively, so as to establish a three-phase state estimation model of the active distribution network;

[0100] 2. Through the data buffering method, the PMU and SCADA data in the waiting area are delayed, so that the PMU, SCADA and AMI data are normalized for cross-section measurement. After the AMI data arrives, the SE calculation of the cross-section can be performed;

[0101] 3. Based on the constructed IEEE-13 node distribution system, an active distribution network with distributed power sources was constructed, and SCADA, PMU and AMI measurements were configured respectively. By simulating this method in the constructed ADN network, the superiority of the three-phase static state estimation method of the active distribution network based on multi-source mixed data proposed in this method in terms of the effectiveness and robustness of SE calculation was verified. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 It is the branch equivalent structure model in the present invention;

[0103] Figure 2 It is a flow chart of the exponential weighted robust state estimation algorithm in the present invention;

[0104] Figure 3 This is the IEEE13 node correction system diagram in the present invention. DETAILED DESCRIPTION

[0105] A robust three-phase state estimation method for active distribution network based on multi-source mixed data, the method steps are as follows:

[0106] a. Establish three-phase measurement-state models for multi-source SCADA, PMU and AMI data;

[0107] Branch Model

[0108] In the active distribution network, the equivalent model of the network structure between the two terminals is as follows: Figure 1 As shown,

[0109] In the figure: i and j are endpoint numbers; and is the current and voltage at the endpoints i and j in phase t (t∈B p , B p ={a,b,c}, which is the set of a, b, c); is the current of branch i and j in phase t; is the self-impedance and mutual impedance of branch i and j (x∈B p , t≠x); and is the admittance of phase t and phase x at endpoints i and j.

[0110] The current injection equation at terminal i is:

[0111]

[0112] Where: is the injected d-phase current, d∈B P ; is the set of endpoints directly connected to endpoint i, and includes i; is the endpoint admittance; is the voltage at terminal k at phase t.

[0113] a1. AMI measurement-state model

[0114] AMI smart meters are configured at the distributed power supply endpoints and load endpoints in the active distribution network. The smart meters are set to collect data every 15 minutes to collect the three-phase voltage module values ​​at the endpoints. Injection current magnitude Injected active power size P i d , Injection reactive power

[0115] According to the injection P i d , The current injection equation can be established as follows:

[0116]

[0117] Expanding the above formula, we can get the real part and imaginary part equations respectively:

[0118]

[0119] Where: is the power injected into endpoint i; P i d is the active power injected into endpoint i; Reactive power for injection endpoint i; f i d are the real and imaginary parts of the voltage at the terminal i in phase d; is the mutual conductance and mutual admittance of the endpoints,

[0120] According to the voltage modulus The voltage modulus equation can be established as follows:

[0121]

[0122] According to the injection current modulus The current modulus equation can be established as follows:

[0123]

[0124] a2. PMU measurement-state model

[0125] At the endpoint of the active distribution network, a PMU measurement device is configured to obtain the relevant voltage and current phasors at this endpoint through the PMU. The measurement types mainly include: voltage phasors of each phase (t∈B p );Each branch current phasor associated with the PMU endpoint (t∈B p , and j≠i),

[0126] Voltage phasor for PMU t∈B p , the voltage equation can be established as follows:

[0127]

[0128] Expanding the above formula, we can get the real part and imaginary part equations:

[0129]

[0130] Where: for The real and imaginary parts of

[0131] For each branch current phasor associated with the PMU terminal t∈B1, the branch current phasor equation can be established as follows:

[0132]

[0133] Expanding the above formula, we can get the real part and imaginary part equations:

[0134]

[0135] In the formula: real and imag represent the real and imaginary parts respectively;

[0136] a3. SCADA measurement-state model

[0137] The RTU device in the SCADA system can measure the injected power P i d , Branch power and voltage modulus The SCADA system measures the entire quantity.

[0138] For voltage modulus The voltage modulus equation can be established as follows:

[0139]

[0140] For the injection power P i d , The current injection equation can be established as follows:

[0141]

[0142] Expanding the above formula, we can get the real part and imaginary part equations respectively:

[0143]

[0144] For branch power The branch power equation can be established as follows:

[0145]

[0146] Expanding the above formula, we can get the real part and imaginary part equations:

[0147]

[0148] a4. Contact endpoint virtual measurement-state model

[0149] In the active distribution network, there are zero injection endpoints that are neither directly connected to the generator nor directly connected to the load and only serve as a connection. The virtual measurement equation can be established as follows:

[0150]

[0151] b. By calculating the optimal buffer length of PMU and SCADA data, the PMU, SCADA and AMI data are normalized for measurement sections;

[0152] Since the sampling periods of AMI measurement, SCADA measurement and PMU measurement are quite different, they cannot be directly combined with SCADA measurement and PMU measurement due to the section time difference problem. They must be section normalized first.

[0153] In the actual operation of active distribution network, the shortest PMU sampling period is about 40ms, with the highest frequency; the SCADA sampling period is about 2s; the longest AMI sampling period is 15min, with the lowest sampling frequency. Therefore, this section combines AMI measurement with SCADA and PMU measurement in the same section. Before the AMI sampling arrives, the optimal buffer length of SCADA and PMU measurement is calculated respectively, and the equivalent section data is calculated, so that the three data can be combined for state estimation calculation.

[0154] b1. Calculation of the optimal buffer length of PMU

[0155] Since the PMU performs sampling approximately every 40 ms, a large amount of PMU data has been collected and transmitted to the data center between two AMI data. Therefore, the optimal PMU buffer length is calculated so that the equivalent section data and AMI data can be combined for application.

[0156] Since the PMU sampling data series is too long within 15 minutes, the algorithm in this paper sets the PMU data 30 seconds before the arrival of AMI as the standby data in the buffer, and uses this data length to determine the optimal buffer length. By calculating the mean and variance of the PMU statistical data in the optimal buffer length area, the normalization of various data sections is performed so that they can be used together for state estimation.

[0157] 30 seconds before the arrival of AMI, the PMU has sampled 750 measurement data. The method to calculate the optimal buffer length is to assume that the PMU sampling frequency is n r sample / s. Every time the next AMI data arrives, the system updates the status. 30s before the AMI arrives:

[0158] (1) The number of PMU sampling measurements is N = 30*n r , divide N sample measurements into n subset There are measurement subsets, usually n subset =30, so the number of each measurement subset is n meas =n r , n subset The subsets are represented as: Z(1), Z(2), ..., Z(n subset ),

[0159] (2) For the latest PMU measurement subset Z(n subset ) to perform change point detection:

[0160] ① Calculate Z(n subset ) and compare the calculated standard deviation with the threshold value, the amplitude threshold value is set to 10 -4 pu, the phase angle threshold is set to 10 -3If the standard deviation is smaller than the threshold value, it means that Z(n subset ) has no change point, Z(n subset ) is all contained in the buffer, go to (3);

[0161] ②If the standard deviation is larger than the threshold value, Z(n subset ) has a change point, then take Z(n subset ) is used as the measurement value of the equivalent section, and the calculation is completed.

[0162] (3) Based on the internal function ttest2 in Matlab, Z(n subset ) and Z(n subset -1), ..., Z(3) and Z(2), Z(2) and Z(1) are tested with a default probability of 95%, and the results are expressed as h i (i=1,2,...,n subset -1) is expressed as:

[0163]

[0164] Where: N op_bf is the optimal buffer length of PMU; the overline is the "not" operator symbol,

[0165] Therefore, based on the above steps, the optimal buffer length can be calculated for each PMU measurement point, and the mean and variance of the PMU statistical data in the buffer zone can be calculated to normalize the cross section.

[0166] b2. SCADA optimal buffer length calculation

[0167] Similarly, for SCADA measurements, since it samples approximately every 2 seconds, a large amount of SCADA data has been collected and transmitted to the data center between two AMI data. Therefore, it is also necessary to calculate the optimal SCADA buffer length so that the equivalent section data and AMI data can be combined for application.

[0168] In 15 minutes, the SCADA sampling data series has 450 measurement data, and the method for calculating the optimal buffer length is similar to that of PMU.

[0169] After normalizing the measurement sections of PMU and SCADA, all processed data and incoming AMI data can be regarded as data within the same section, and thus combined together for SE calculation.

[0170] c. Based on the exponential weighted robust state estimation method, a robust state estimation model for active distribution network is established, and a robust three-phase state estimation method for active distribution network is proposed.

[0171] After the above cross-section normalization processing, the three types of data, AMI, PMU and SCADA, are combined, and the active distribution network robust estimation model is established based on the exponential weight robust estimation method. The quantity measurement of the active distribution network system is:

[0172]

[0173] Where: z represents the full measurement of the active distribution network; z AMI 、z PMU and z SCADA Respectively represent the three types of measurements: AMI, PMU and SCADA.

[0174] Based on the normalized full measurement above, the index-weighted robust estimation model of the main distribution network is established and solved by Newton method. The index-weighted robust three-phase state estimation method of the active distribution network is proposed.

[0175] c1. Exponentially weighted robust state estimation model for active distribution network

[0176] In the exponentially weighted robust state estimation model, the objective function is to minimize the sum of the exponentially weighted residuals:

[0177]

[0178] stc(x)=0 (23)

[0179] Where: represents measurement; h(x) represents measurement function; c(x) represents equality constraint; represents the measurement residual; x represents the state quantity; W represents the diagonal matrix of the weight function, W ii =w i * ,

[0180] w i * Calculate the expression:

[0181]

[0182] Where: R i -1 represents fixed weight; δ represents standard deviation, and the initial value calculation formula is δ 2 =r T Pr / (mn), P represents the prior diagonal matrix, the diagonal elements m represents the number of measurements, and n represents the number of states;

[0183] c2. Exponentially weighted robust state estimation algorithm for active distribution network

[0184] Based on the Lagrange multiplier λ, the equality constraint is transformed to obtain the unconstrained objective function, which is then optimized and solved. The calculation formula is as follows:

[0185]

[0186] Where: L(x,λ) represents the augmented Lagrange function,

[0187] Taylor expansion is performed on h(x) and c(x) and linearization of nonlinear equations is performed:

[0188]

[0189] Where: H(x0) and C(x0) represent Jocabian matrices, which are and (abbreviated as H and C),

[0190] Find the first-order partial derivative of formula (25):

[0191]

[0192] Based on the Newton method iterative solution, the incremental update matrix form is:

[0193]

[0194] Where: k represents the number of iterations; matrix Δx (k) =x (k+1) -x (k) ,

[0195] The calculation is based on the exponential weight robust state estimation algorithm. The exponential weight can automatically correct the measurement weight and suppress the influence of bad data. The calculation process is as follows Figure 2 shown.

[0196] The detailed calculation steps of the algorithm are as follows:

[0197] ① Initialization: set the initial values ​​of the real and imaginary parts of the three-phase voltages of the distribution network balance, load and interconnection endpoints, take the phase A of the balance endpoint as the phase reference, the earth as the potential reference, set the fixed weight matrix to R-1, set the maximum number of iterations to Tmax, set the convergence threshold to ε, and set time = 1 for the initial iteration;

[0198] ② Calculate the endpoint admittance matrix and Jocabian matrix, calculate the imbalance Δz and Δc, and calculate the diagonal element value of the weight matrix W to form the Jocabian matrix H and C, and calculate H T WH and H T WΔz;

[0199] ③Calculate the state deviation Δx(k) , and update the state value: x (k+1) =x (k) +Δx (k) , time = time + 1;

[0200] ④ Convergence judgment, if max(Δx (k) )<ε, the calculation converges and the result is output; if max(Δx (k) )>ε and time>Tmax, the iteration stops, "Not converged!". Otherwise, go to step ⑤;

[0201] ⑤ Modify the dynamic scale parameter, if (δ (k) ) 2 ≥0.01, (δ (k+1) ) 2 =0.001(δ (k) ) 2 ; If (δ (k) ) 2 <0.01, (δ (k+1) ) 2 =(δ (k) ) 2 , go to step ②.

[0202] Example simulation and analysis

[0203] Basic data and simulation conditions

[0204] In order to verify the effectiveness and robustness of the proposed robust three-phase state estimation method for active distribution network based on multi-source mixed data, a simulation experiment is carried out based on the constructed IEEE-13 node ADN network. Figure 3 shown.

[0205] In order to adapt to the research of active distribution network of this method, the standard system is modified as follows:

[0206] ① The original system has one branch model each, all of which are unified as 501;

[0207] ② The original system node 5 is connected to the photovoltaic power supply, and the three-phase output of the node power supply is set to 40kW and the power factor is 0.98;

[0208] ③SCADA measurement is arranged throughout the system. Node 1 is located on the low-voltage side of the distribution transformer as the system balancing node. AMI smart meters are configured on the low-voltage side of the distribution transformer and each load terminal. PMUs are arranged at nodes 2, 6, 9 and 11.

[0209] In order to simulate actual measurement, in the simulation experiment, the improved system is calculated by the Newton method to obtain the power flow results as the true values ​​of the physical quantities of the system. On this basis, the random distribution errors are superimposed to simulate the actual measurement. Among them, for SCADA data, the measurement error standard deviation is set to 0.02, the PMU measurement error standard deviation is set to 0.002, and the AMI measurement error standard deviation is set to 0.01.

[0210] Based on the above improved system and simulation measurements, the following two simulation cases are studied:

[0211] Example 1: There is no bad data in the system, and all simulation measurements are generated by superimposing random distribution errors on the true value;

[0212] Example 2: The system has bad data. Assume that the bad data is generated by the active power of phase a at node 2 and the voltage modulus of phase C at node 6. Set the active power error to 30% of the original value and the modulus to 55% of the original value to simulate the bad data collected due to external reasons.

[0213] In the above example, this method is compared with the WLS method.

[0214] Simulation Results and Analysis

[0215] (1) Validity analysis of ADN estimation results

[0216] In order to verify the effectiveness of this method, the following two indicators are defined for evaluation:

[0217]

[0218] Where: S1 represents the system error; S2 represents the maximum error; x i represents the truth value; represents the estimated value; n represents the number of state quantities.

[0219] The SE method and WLS method of the multi-source hybrid data active distribution network proposed in this method are simulated in Example 1 and Example 2 respectively. The simulation results are shown in Table 1.

[0220] Table 1 Comparison of active distribution network state estimation errors

[0221]

[0222] As shown in Table 1, in Example 1, for the voltage modulus, the system error and maximum error of the SE method for multi-source hybrid data active distribution network are 1.1322×10 -3 5.5124×10 -4 The system error and maximum error of WLS are 7.9231×10 -35.7981×10 -4 ; For the voltage phase angle, the system error and maximum error of the SE method of multi-source hybrid data active distribution network are 0.04913 and 1.6824×10 -3 The system error and maximum error of WLS are 0.05537 and 1.9721×10 -3 Obviously, in the case of no bad data in Example 1, the system error and maximum error of the SE method and WLS method of multi-source hybrid data active distribution network of this method are smaller, but the error of this method is smaller and the estimation result is better.

[0223] In Example 2, for the voltage modulus, the system error and maximum error of the SE method for active distribution network with multi-source hybrid data are 3.1765×10 -3 ,8.6133×10 -4 The system error and maximum error of WLS are 0.1327 and 5.6521×10 -3 ; For the voltage phase angle, the system error and maximum error of the SE method of multi-source hybrid data active distribution network are 0.0944 and 1.2563×10 -2 , the system error and maximum error of WLS are 1.7514 and 0.1124 respectively. Obviously, in the case of bad data in Example 2, the system error and maximum error of the SE method and WLS method of multi-source hybrid data active distribution network of this method are increased compared with Example 1, but the error increase of this method is small, and it has excellent robust performance.

[0224] (2) Analysis of the accuracy of ADN estimation results

[0225] This method conducts three-phase SE of active distribution network based on multi-source mixed data and can calculate the three-phase state. Based on Example 1 and Example 2, the following two simulation scenarios are set up and compared with the true value to verify the accuracy of the calculation results of three-phase SE of active distribution network based on multi-source mixed data:

[0226] (1) The phasors of the active distribution network balancing endpoints are set to be symmetrical;

[0227] (2) The phasors of the active distribution network balancing endpoints are set to be asymmetric.

[0228] Tables 2 and 3 are the comparison results of the three-phase voltage SE values ​​and true values ​​of the three endpoints 1, 2 and 5 of the active distribution network in the two simulation scenarios of Example 1 and Example 2 respectively.

[0229] As can be seen from Table 2, based on Example 1, regardless of whether the balance endpoints are set symmetrically or not, the results obtained by the three-phase SE method of the active distribution network with multi-source hybrid data proposed in this method are consistent with the true value, and the three-phase SE of the active distribution network can be accurately performed. As can be seen from Table 3, based on Example 2, regardless of whether the balance endpoints are set symmetrically or not, the results obtained by the three-phase SE method of the active distribution network with multi-source hybrid data proposed in this method are also very close to the true value and basically consistent. The main reason is that Example 2 sets some bad data based on Example 1. The multi-source hybrid data robust SE method proposed in this method can effectively filter out the bad data and obtain the best estimate closest to the true value, but there will be some deviations compared to the case without bad data, but the deviation is still very small, and the three-phase SE of the active distribution network can be accurately performed.

[0230] Table 2 Three-phase SE results of active distribution network example 1

[0231]

[0232] Table 3 Three-phase SE results of active distribution network example 2

[0233]

[0234] (3) Convergence analysis of active distribution network estimation results

[0235] In order to verify the convergence of the proposed multi-source hybrid data active distribution network SE method with the increase of distributed generation penetration, the following three situations are set:

[0236] (1) Connecting a distributed power source: Connecting a photovoltaic power source to the original system node 5, setting the node power three-phase output to 40 kW and the power factor to 0.98;

[0237] (2) Connecting two distributed power sources: Connecting photovoltaic power sources to nodes 5 and 7 of the original system, setting the three-phase output of the node power supply to 40 kW and the power factor to 0.98;

[0238] (3) Connecting three distributed power sources: Connecting photovoltaic power sources to nodes 5, 7, and 9 of the original system, setting the three-phase output of the node power supply to 40 kW and the power factor to 0.98;

[0239] In the above three situations, based on Example 1 and Example 2 set up by this method, the multi-source mixed data robust SE method and WLS method proposed by this method are used to perform SE calculation. The convergence times and the estimation errors in the corresponding situations are shown in Table 4 and Table 5, respectively.

[0240] Table 4. The number of iterations under different distributed generation access conditions in active distribution network

[0241]

[0242] As shown in Table 4, based on Example 1, in the three distributed power access scenarios, the iteration numbers of the multi-source mixed data robust SE method proposed by this method are 5, 6 and 6 respectively, while the iteration numbers of the WLS method are 4, 5 and 6 respectively. The iteration number of the WLS method is relatively less than that of the multi-source mixed data robust SE method proposed in the method paper, but they are almost close. The reason is that there is no bad data in Example 1. Both methods can achieve reliable calculation and converge well. Based on Example 2, in the three distributed power access scenarios, the iteration numbers of the multi-source mixed data robust SE method proposed by this method are 6, 7 and 7 respectively, while the iteration numbers of the WLS method are 8, 9 and 11 respectively. The iteration number of the WLS method is significantly more than that of the multi-source mixed data robust SE method proposed by this method. The reason is that bad data is set in Example 1. With the increase of distributed photovoltaic power sources, WLS cannot effectively process bad data, and the iteration number increases.

[0243] Table 5 Estimation error of active distribution network under different distributed generation access situations

[0244]

[0245] It can be seen from Table 5 that in Example 1, for the voltage modulus and phase angle, the system errors of the multi-source hybrid data active distribution network SE method in the three cases are 1.1322×10 -3 , 1.1431×10 -3 , 1.1437×10 -3 and 0.04913, 0.05011, 0.05024, and the systematic errors of WLS are 7.9231×10 -3 7.9351×10 -3 7.9377×10 -3 and 0.05537, 0.05734, and 0.05759. In Example 2, for the voltage modulus and phase angle, the system errors of the multi-source hybrid data active distribution network SE method in the three cases are 3.1765×10 -3 ,3.1861×10 -3 3.1951×10 -3 and 0.0944, 0.0954, 0.0962, and the systematic errors of WLS are 0.1327, 0.1433, 0.1459 and 1.7514, 1.7667, 1.7721, respectively.

[0246] Obviously, in the case of no bad data in Example 1, the multi-source hybrid data active distribution network SE method and WLS method of this method have lower system errors and better convergence. However, after setting bad data in Example 2, with the increase of distributed power sources, WLS cannot effectively handle bad data, not only the number of iterations increases, but also the SE results are poor. The multi-source hybrid data robust SE method proposed in this method can achieve reliable calculation, good convergence, and can maintain good accuracy.

Claims

1. A robust three-phase state estimation method for active distribution network based on multi-source mixed data, characterized in that: The steps are as follows: a. Establish three-phase measurement-state models for multi-source SCADA, PMU and AMI data; b. By calculating the optimal buffer length of PMU and SCADA data, the PMU, SCADA and AMI data are normalized for measurement sections; c. Based on the exponential weighted robust state estimation method, a robust state estimation model for active distribution network is established, and a robust three-phase state estimation method for active distribution network is proposed.

2. The method for robust three-phase state estimation of active distribution network based on multi-source mixed data according to claim 1 is characterized in that: The step a is specifically as follows: a1. AMI measurement-state model AMI smart meters are configured at the distributed power supply endpoints and load endpoints in the active distribution network. The smart meters are set to collect data every 15 minutes to collect the three-phase voltage module values ​​at the endpoints. Injection current magnitude Injected active power Injected reactive power According to injection The current injection equation can be established as follows: Expanding the above formula, we can get the real part and imaginary part equations respectively: Where: is the power injected into endpoint i; is the active power injected into endpoint i; Reactive power for injection endpoint i; are the real and imaginary parts of the voltage at the terminal i in phase d; is the mutual conductance and mutual admittance of the endpoints, According to the voltage modulus The voltage modulus equation can be established as follows: According to the injection current modulus The current modulus equation can be established as follows: a2. PMU measurement-state model At the endpoint of the active distribution network, a PMU measurement device is configured to obtain the relevant voltage and current phasors at this endpoint through the PMU. The measurement types mainly include: voltage phasors of each phase Each branch current phasor associated with the PMU endpoint Voltage phasor for PMU The voltage equation can be established as follows: Expanding the above formula, we can get the real part and imaginary part equations: Where: for The real and imaginary parts of For each branch current phasor associated with the PMU terminal t∈B1, the branch current phasor equation can be established as follows: Expanding the above formula, we can get the real part and imaginary part equations: In the formula: real and imag represent the real and imaginary parts respectively; a3. SCADA measurement-state model RTU device in SCADA system, can measure the injected power Branch power and voltage modulus The SCADA system measures the entire quantity. For voltage modulus The voltage modulus equation can be established as follows: For the injected power The current injection equation can be established as follows: Expanding the above formula, we can get the real part and imaginary part equations respectively: For branch power The branch power equation can be established as follows: Expanding the above formula, we can get the real part and imaginary part equations: a4. Contact endpoint virtual measurement-state model In the active distribution network, there are zero injection endpoints that are neither directly connected to the generator nor directly connected to the load and only serve as a connection. The virtual measurement equation can be established as follows:

3. The method for robust three-phase state estimation of active distribution network based on multi-source mixed data according to claim 1 is characterized in that: The step b is specifically as follows: b1. Calculation of the optimal buffer length of PMU 30 seconds before the arrival of AMI, the PMU has sampled 750 measurement data. The method to calculate the optimal buffer length is to assume that the PMU sampling frequency is n r sample / s. Every time the next AMI data arrives, the system updates the status. 30s before the AMI arrives: (1) The number of PMU sampling measurements is N = 30*n r , divide N sample measurements into n subset There are measurement subsets, usually n subset =30, so the number of each measurement subset is n meas =n r , n subset The subsets are represented as: Z(1), Z(2), ..., Z(n subset ), (2) For the latest PMU measurement subset Z(n subset ) to perform change point detection: ① Calculate Z(n subset ) and compare the calculated standard deviation with the threshold value, the amplitude threshold value is set to 10 -4 pu, the phase angle threshold is set to 10 -3 If the standard deviation is smaller than the threshold value, it means that Z(n subset ) has no change point, Z(n subset ) is all contained in the buffer, go to (3); ②If the standard deviation is larger than the threshold value, Z(n subset ) has a change point, then take Z(n subset ) is used as the measurement value of the equivalent section, and the calculation is completed. (3) Based on the internal function ttest2 in Matlab, Z(n subset ) and Z(n subset -1), ..., Z(3) and Z(2), Z(2) and Z(1) are tested with a default probability of 95%, and the results are expressed as h i (i=1,2,...,n subset -1) is expressed as: Where: N op_bf is the optimal buffer length of PMU; the overline is the "not" operator symbol, Therefore, based on the above steps, the optimal buffer length can be calculated for each PMU measurement point, and the mean and variance of the PMU statistical data in the buffer zone can be calculated to normalize the cross section. b2. SCADA optimal buffer length calculation In 15 minutes, the SCADA sampling data series has 450 measurement data. The method for calculating the optimal buffer length is similar to that of PMU. After normalizing the measurement sections of PMU and SCADA, all processed data and incoming AMI data can be regarded as data within the same section, and thus combined together for SE calculation.

4. The method for robust three-phase state estimation of active distribution network based on multi-source mixed data according to claim 1 is characterized in that: The step c is specifically as follows: After the above cross-section normalization processing, the three types of data, AMI, PMU and SCADA, are combined, and the active distribution network robust estimation model is established based on the exponential weight robust estimation method. The quantity measurement of the active distribution network system is: Where: z represents the full measurement of the active distribution network; z AMI 、z PMU and z SCADA Respectively represent the three types of measurements: AMI, PMU and SCADA. Based on the normalized full measurement above, the index-weighted robust estimation model of the main distribution network is established and solved by Newton method. The index-weighted robust three-phase state estimation method of the active distribution network is proposed. c1. Exponentially weighted robust state estimation model for active distribution network In the exponentially weighted robust state estimation model, the objective function is to minimize the sum of the exponentially weighted residuals: Where: represents measurement; h(x) represents measurement function; c(x) represents equality constraint; represents the measurement residual; x represents the state quantity; W represents the diagonal matrix of the weight function, Calculate the expression: Where: represents fixed weight; δ represents standard deviation, and the initial value calculation formula is δ 2 =r T Pr / (mn), P represents the prior diagonal matrix, the diagonal elements m represents the number of measurements, and n represents the number of states; c2. Exponentially weighted robust state estimation algorithm for active distribution network Based on the Lagrange multiplier λ, the equality constraint is transformed to obtain the unconstrained objective function, which is then optimized and solved. The calculation formula is as follows: Where: L(x,λ) represents the augmented Lagrange function, Taylor expansion is performed on h(x) and c(x) and linearization of nonlinear equations is performed: Where: H(x0) and C(x0) represent Jocabian matrices, which are and (abbreviated as H and C), Find the first-order partial derivative of formula (25): Based on the Newton method iterative solution, the incremental update matrix form is: Where: k represents the number of iterations; matrix Δx (k) =x (k+1) -x (k) , The calculation is based on the exponential weight robust state estimation algorithm. The exponential weight can automatically correct the measurement weight and suppress the impact of bad data. The detailed calculation steps of the algorithm are as follows: ① Initialization: set the initial values ​​of the real and imaginary parts of the three-phase voltages of the distribution network balance, load and interconnection endpoints, take the phase A of the balance endpoint as the phase reference, the earth as the potential reference, set the fixed weight matrix to R-1, set the maximum number of iterations to Tmax, set the convergence threshold to ε, and set time = 1 for the initial iteration; ② Calculate the endpoint admittance matrix and Jocabian matrix, calculate the imbalance Δz and Δc, and calculate the diagonal element value of the weight matrix W to form the Jocabian matrix H and C, and calculate H T WH and H T WΔz; ③Calculate the state deviation Δx (k) , and update the state value: x (k+1) =x (k) +Δx (k) , time = time + 1; ④ Convergence judgment, if max(Δx (k) )<ε, the calculation converges and the result is output; if max(Δx (k) )>ε and time>Tmax, the iteration stops, "Not converged!", otherwise, go to step ⑤; ⑤ Modify the dynamic scale parameter, if (δ (k) ) 2 ≥0.01, (δ (k+1) ) 2 =0.001(δ (k) ) 2 ; If (δ (k) ) 2 <0.01, (δ (k+1) ) 2 =(δ (k) ) 2 , go to step ②.

Images