Robust three-phase state estimation method for active distribution network based on multi-source mixed data

Abstract

The application discloses a kind of active distribution network robust three-phase state estimation method based on multi-source mixed data, and method steps are as follows: a, the three-phase measurement-state model of each of multi-source SCADA, PMU and AMI data is established;B, by calculating the optimal buffer length of PMU and SCADA data, PMU, SCADA and AMI data are normalized processing measurement section;C, based on exponential weight robust state estimation method, establish active distribution network robust state estimation model, propose active distribution network robust three-phase state estimation method, the advantage of the application is, by the method of data buffering, the time delay processing of PMU and SCADA data in waiting area, so that PMU, SCADA and AMI data are section measurement normalization processing, after AMI data arrives, SE calculation of the section can be carried out.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-phase state estimation method of power distribution network, and more particularly to the technical field of robust three-phase state estimation method of active distribution network. BACKGROUND

[0002] Active distribution network (ADN) contains a large number of distributed renewable energy sources, and can use advanced communication technology to autonomously plan, manage and coordinate new energy and energy storage devices, to ensure safe and reliable operation of the power grid on the basis of actively consuming distributed energy.

[0003] At present, the smart distribution network is developing rapidly, and the smart meters with bidirectional communication capability in the advanced metering infrastructure (AMI) have been widely deployed. The data types collected by the smart meters include phase power voltage and other data, which are generally collected once every fifteen minutes. In contrast, the phasor measurement unit (PMU) can collect phasor information, and the collection frequency is very high, generally collecting tens of times of data within 1 second, and the accuracy is relatively high.

[0004] However, due to the gradual configuration of the collection system, in the future for a long period of time, the supervisory control and data acquisition system (SCADA), PMU and AMI in the active distribution network will coexist for a long time. In state estimation (SE), pure use of only one type of data may not meet the observability requirement, and the measurement redundancy is low. Therefore, in the active distribution network, the SCADA, AMI and PMU data are fused together for state estimation, which can significantly increase the measurement redundancy of SE and improve the accuracy of SE results. Since the SCADA data, AMI smart meter data and PMU data are non-homogeneous data, there are problems of time deviation and data time delay. In order to effectively combine various types of non-homogeneous data, it is necessary to unify each type of measurement to the same time section.

[0005] The structure of the distribution network is few ring and generally presents a radiation type, the three-phase asymmetry degree is high, and the measurement configuration is not complete. Some existing methods take branch power or branch current as state quantity, and establish SE model based on weighted least square (WLS). These methods can realize state estimation of the distribution network, but generally do not have robustness. In order to increase the measurement redundancy and improve the estimation accuracy, meet the accurate perception requirements of ADN on the state, fuse multi-source data, and improve the measurement redundancy of the distribution network estimation and the robustness of the algorithm, it is a research hotspot in recent years.

[0006] State estimation as a state perception technology of smart grid has been mature in transmission network, but has not been completely landed in distribution network. State estimation can utilize multi-source data for filtering, and restore the most real state of the system, so as to be used as a function of active distribution network state perception, reasonably coordinate SCADA, PMU and AMI three system data, unify the state data with different sampling periods to the same section, and jointly perform state estimation to obtain the most accurate state of the system, which is an urgent problem to be solved. SUMMARY

[0007] The application provides a robust three-phase state estimation method for active distribution network based on multi-source hybrid data, which solves the above technical problems.

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

[0009] The robust three-phase state estimation method for active distribution network based on multi-source hybrid data has the following method steps:

[0010] a. Establishing a three-phase measurement-state model of multi-source SCADA, PMU and AMI data;

[0011] b. Through calculation of the optimal buffer length of PMU and SCADA data, the measurement section normalization processing is performed on PMU, SCADA and AMI data;

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

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

[0014] a1. AMI measurement-state model

[0015] The AMI intelligent electric meter is configured at the distributed power source end point and the load end point connected with the load in the active distribution network, the intelligent electric meter is set to collect data every 15 minutes, and the three-phase voltage module value and the injected current module value of the end point are collected . , the active power injection size , the reactive power injection size ,

[0016] According to the injection , , the current injection equation is established as follows:

[0017] (2)

[0018] Expanding the above equation, the real part and imaginary part equations are obtained respectively:

[0019] (3)

[0020] (4)

[0021] In the formula: is the active power at the injection end point ; is the reactive power at the injection end point ; is the voltage at the injection end point ; , is the real part and imaginary part of the voltage at the end point ; , , is the mutual conductance and mutual susceptance at the end point

[0022] According to the voltage modulus value , the voltage modulus value equation is established as follows:

[0023] (5)

[0024] According to the injection current modulus value , the current modulus value equation is established as follows:

[0025] ; (6)

[0026] a2, PMU measurement-state model

[0027] At the end point of the active distribution network, a PMU measurement device is configured, and the relevant voltage and current phasors at the end point are obtained through the PMU. The types of measurements mainly include: each phase voltage phasor , ; and each branch current phasor associated with the PMU end point , , and ,

[0028] The voltage phasor of the PMU The voltage equation is established as follows:

[0029] (7)

[0030] Expanding the above equation, the real and imaginary parts equations are:

[0031] (8)

[0032] (9)

[0033] where: are the real and imaginary parts of

[0034] For each branch current phasor associated with the PMU end point The branch current phasor equation is established as follows:

[0035] (10)

[0036] Expanding the above equation, the real and imaginary parts equations are:

[0037] (11)

[0038] (12)

[0039] where: denote taking the real and imaginary parts, respectively;

[0040] a3, SCADA measurement-state model

[0041] At the RTU device in the SCADA system, the injected power , the branch power and the voltage modulus are measured, and the SCADA system measurement is full measurement,

[0042] For the voltage modulus The voltage modulus equation is established as follows:

[0043] (13)

[0044] For the injected power The current injection equation is established as follows:

[0045] ​​​​​​​​(14)

[0046] Unfold the above equation, get real part, imaginary part equation:

[0047] (15)

[0048] (16)

[0049] For branch power , , its branch power equation is established as follows:

[0050] (17)

[0051] Unfold the above equation, get real part, imaginary part equation:

[0052] (18)

[0053] ;(19)

[0054] a4, contact end virtual measurement-state model

[0055] In active distribution network, there is a zero injection end which is neither directly connected with generator nor directly connected with load, only for contact. The virtual measurement equation is established as follows:

[0056] (20).

[0057] Preferably, the step b is specifically as follows,

[0058] b1, PMU optimal buffer length calculation

[0059] Before 30s of AMI, PMU sampling has 750 measurement data, the method of calculating optimal buffer length is to assume that PMU sampling frequency is sample / s, the system is updated once per next AMI data, before 30s of AMI:

[0060] (1) PMU sampling measurement number is , divide sampling measurements into measurement subsets, generally take , so that each measurement subset number is , subsets are expressed as: ,

[0061] (2) Perform change point detection on the latest PMU measurement subset ​

[0062] ①Calculate the standard deviation of the measurement data in the buffer, and compare the calculated standard deviation with a threshold value, the amplitude threshold value is 10 -4 , and the phase angle threshold value is 10 -3 . If the standard deviation is smaller than the threshold value, it indicates that there is no change point in the buffer, and the entire buffer is included in the buffer, go to step (3);

[0063] ②If the standard deviation is larger than the threshold value, there is a change point in the buffer, take the latest PMU in the buffer as the measurement value of the equivalent section, and the calculation is completed,

[0064] (3) Based on the internal function ttest2 in Matlab, perform hypothesis testing on and , , and , and with the default value of 95% probability, and the result is represented as :

[0065] (21)

[0066] In the formula: is the optimal buffer length of the PMU; the overline is the "not" operator,

[0067] Therefore, based on the above steps, the optimal buffer length of each PMU measurement point is calculated, and the mean and variance of the PMU statistical data in the buffer are calculated, so as to normalize the section,

[0068] b2, calculation of the optimal buffer length of SCADA

[0069] Within 15 minutes, there are 450 measurement data in the SCADA sampling data series, and the method for calculating the optimal buffer length is similar to that of PMU.

[0070] After the measurement section normalization processing of PMU and SCADA, all the processed data and the incoming AMI data are regarded as data in the same section, so as to be combined together and used for SE calculation.

[0071] Preferably, step c is specifically as follows,

[0072] Combine AMI, PMU and SCADA data, establish a robust estimation model of active distribution network based on exponential weight robust estimation method, and the measurement of active distribution network system is: ​

[0073] (22)

[0074] wherein: represents full measurement of active distribution network; 、 and respectively represent AMI, PMU and SCADA three types of measurement,

[0075] Based on the above normalized full measurement, an index weight robust estimation model of active distribution network is established, and the Newton method is used for solution, and an index weight robust three-phase state estimation method of active distribution network is proposed,

[0076] c1, index weight robust state estimation model of active distribution network

[0077] In the index weight robust state estimation model, the objective function is the minimum sum of index weight residual:

[0078] (23)

[0079] wherein: represents measurement; represents measurement function; c(x) represents equality constraint; represents measurement residual; x represents state quantity; represents diagonal matrix of weight function, ,

[0080] Calculation expression:

[0081] (24)

[0082] wherein: represents fixed weight; represents standard deviation, and the initial value calculation formula is , P represents prior diagonal matrix, and the diagonal element , m represents measurement number, and n represents state number;

[0083] c2, index weight robust state estimation algorithm of active distribution network

[0084] Based on Lagrange multiplier , equality constraint is transformed and processed to obtain an unconstrained objective function, and optimization solution is carried out, and the calculation formula is as follows:

[0085] (25)

[0086] wherein: represents augmented Lagrange function,

[0087] to and Both Taylor expansion and linearization of nonlinear equations are performed:

[0088] (26)

[0089] In the formula: , Jocabian matrix, respectively, and (simplified as H and C),

[0090] First-order partial derivative of formula (25):

[0091] (27)

[0092] Based on Newton method iteration, the incremental update matrix form is:

[0093] (28)

[0094] In the formula: k represents the number of iterations; matrix ; ; ,

[0095] Based on the exponential weight robust state estimation algorithm, the exponential weight automatically corrects the measurement weight, and suppresses the influence of bad data,

[0096] The detailed calculation steps of the algorithm are as follows:

[0097] ① Initialization, set the initial value of the real and imaginary parts of the balance, load and tie-in endpoint three-phase voltage, take the balance endpoint A phase as the phase reference, and the ground as the potential reference. Set R-1 for the fixed weight matrix, Tmax for the maximum number of iterations, ε for the convergence threshold, and time=1 for the initial iteration;

[0098] ② Calculate the endpoint admittance matrix and Jocabian matrix, calculate the unbalance and , and calculate the diagonal element value of the weight matrix W, form and Jocabian matrix H, C, and calculate ;

[0099] ③ Calculate the state deviation , and update the state value: , time = time + 1;

[0100] ④ Convergence judgment, if , the convergence is calculated and the result is output; if and time > Tmax, the iteration is stopped, and "not convergent!" Otherwise, go to step ⑤;

[0101] 5. Correcting the dynamic scale parameter, if ( ) 2 0.01, ( ) 2 = ; if ( ) 2 <0.01, ( ) 2 = , go to step 2.

[0102] The beneficial effects that can be achieved by the application of the above technical solutions are:

[0103] 1. A three-phase measurement-state model is established for the multi-source SCADA, PMU and AMI data in the active power distribution network, so as to establish a three-phase state estimation model for the active power distribution network.

[0104] 2. The PMU and SCADA data in the waiting area are processed by time delay, so that the PMU, SCADA and AMI data are normalized, and the SE calculation of the section can be performed after the AMI data arrives.

[0105] 3. Based on the constructed IEEE-13 node power distribution system, an active power distribution network containing distributed power sources is constructed, and SCADA, PMU and AMI measurements are configured. The method is simulated in the constructed ADN network, and the superiority of the method in SE calculation is verified in terms of effectiveness and robustness. BRIEF DESCRIPTION OF DRAWINGS

[0106] Figure 1 The branch equivalent structure model in the application;

[0107] Figure 2 The exponential weight robust state estimation algorithm flowchart in the application;

[0108] Figure 3 The IEEE13 node correction system diagram in the application. DETAILED DESCRIPTION

[0109] The active power distribution network robust three-phase state estimation method based on multi-source mixed data has the following method steps:

[0110] a. Establish a three-phase measurement-state model for multi-source SCADA, PMU and AMI data;

[0111] Branch model

[0112] In active power distribution network, the equivalent model of network structure between two endpoints is shown as Figure 1

[0113] In the figure: i, j are the endpoint number; are the current, voltage of endpoint i, j in t phase (set of three phases); are the current of branch i, j in t phase; are the self, mutual impedance of branch i, j; are the admittance of t phase, phase of endpoint i, j.

[0114] The current injection equation of endpoint i is:

[0115] (1)

[0116] In the formula: is the injection current of t phase, is the set of endpoints directly connected with endpoint i, and contains i; are the admittance of endpoint; is the voltage of t phase of endpoint i.

[0117] a1, AMI measurement-state model

[0118] In active power distribution network, the distributed power endpoints and the load endpoints with loads are configured with AMI smart meters. The smart meters are set to collect data every 15 minutes, and collect the three-phase voltage modulus , injection current modulus , injection active power size , injection reactive power size

[0119] According to the injection , the current injection equation is established as follows:

[0120] (2) ​​​​​​​​​​​​​​​​​​​​​​​

[0121] Unfolding the above equation, we get the real part and imaginary part equations respectively:

[0122] (3)

[0123] (4)

[0124] In the formula: is the injection end point power; is the injection end point active power; is the injection end point reactive power; , is the end point of the real and imaginary parts of the phase voltage; , is the end point mutual conductance and mutual susceptance According to the voltage modulus value

[0125] , the voltage modulus equation is established as follows:

[0126] (5) According to the injection current modulus value

[0127] , the current modulus equation is established as follows:

[0128] (6)

[0129] a2, PMU measurement-state model

[0130] At the end point of the active distribution network, the PMU measurement device is configured, and the relevant voltage and current phasors at the end point are obtained through the PMU. The types of measurements mainly include: each phase voltage phasor , ; each branch current phasor associated with the PMU end point , , and ,

[0131] The voltage phasor of the PMU , , and the voltage equation is established as follows:

[0132] (7)

[0133] Unfolding the above equation, we get the real part and imaginary part equations respectively:

[0134] (8) ​

[0135] (9)

[0136] where: , are real, imaginary parts of

[0137] each branch current phasor associated with a PMU endpoint , whose branch current phasor equation is established as follows:

[0138] (10)

[0139] Expanding the above equation, we get the real, imaginary part equations:

[0140] (11)

[0141] (12)

[0142] where: , denote taking the real, imaginary part, respectively;

[0143] a3, SCADA measurement-state model

[0144] At the RTU device in the SCADA system, the injected power , , branch power , and voltage modulus are measured, and the SCADA system measurement is full measurement,

[0145] For the voltage modulus , the voltage modulus equation is established as follows:

[0146] (13)

[0147] For the injected power , , the current injection equation is established as follows:

[0148] (14)

[0149] Expanding the above equation, we get the real, imaginary part equations, respectively:

[0150] (15)

[0151] (16)

[0152] For the branch power , The branch power equation is established as follows:

[0153] (17)

[0154] The above formula is expanded to obtain the real and imaginary equations:

[0155] (18)

[0156] ;(19)

[0157] a4、 Contact endpoint virtual measurement-state model

[0158] In the active power distribution network, there are zero injection endpoints that are not directly connected to the generator or the load, and only play a contact role. The virtual measurement equation is established as follows:

[0159] (20).

[0160] b. Normalize the measurement section by calculating the optimal buffer length of PMU and SCADA data, and combine PMU, SCADA and AMI data for measurement section normalization processing;

[0161] Because the sampling period of AMI measurement and SCADA, PMU measurement has a big difference, so AMI measurement and SCADA, PMU measurement cannot be directly combined together for use due to the section time difference problem, and must be normalized first.

[0162] In the actual operation of the active power distribution network, the PMU sampling period is the shortest about 40ms, and the frequency is the highest; the SCADA sampling period is about 2s; the AMI sampling period is the longest, which needs to be sampled once every 15 minutes, and the sampling frequency is the lowest. Therefore, in order to combine AMI measurement with SCADA and PMU measurement in the same section, the optimal buffer length of SCADA and PMU measurement is calculated before AMI sampling, and the equivalent section data is calculated to combine the three data for state estimation calculation.

[0163] b1. PMU optimal buffer length calculation

[0164] Because PMU is sampled every 40ms, there are a large number of PMU data collected and transmitted to the data center in the gap between two AMI data, so the optimal buffer length of PMU is calculated to combine the equivalent section data with AMI data.

[0165] Since the PMU sampling data series is too long within 15 minutes, the algorithm in this paper sets up PMU data in the 30 seconds before the arrival of AMI as the data to be used in the buffer, and the optimal buffer length is determined based on this data length. By calculating the mean and variance of PMU statistical data within the optimal buffer length range, the normalization of various data sections is performed so that they can be used in combination for state estimation.

[0166] 30 seconds before the AMI occurs, the PMU has already collected 750 measurement data points. The method for calculating the optimal buffer length is to assume that the PMU sampling frequency is... The system updates its status every time an AMI (Automatic Missing Machine) data arrives, 30 seconds before the AMI arrives:

[0167] (1) The number of PMU sampling measurements is ,Will Each sampling measurement is divided into A subset of measurements, typically taken Therefore, the number of each measurement subset is , Each subset is represented as: ,

[0168] (2) The latest PMU measurement subset Perform change point detection:

[0169] ①Calculation The standard deviation of the amplitude is measured, and the calculated standard deviation is compared with the threshold value, with the amplitude threshold value set to 10. -4 PU, phase threshold value set to 10 -3 If the standard deviation is smaller than the threshold value, it indicates There are no changes inside. The entire contents are contained in the buffer; go to (3);

[0170] ② If the standard deviation is larger than the threshold value, If there is a variable point, then take... The latest PMU measurement is used as the equivalent cross-sectional measurement value, and the calculation ends.

[0171] (3) Based on the built-in function ttest2 in Matlab and , , and , and Perform a hypothesis test with a default probability of 95%, and the result is used... Represented as:

[0172] (twenty one)

[0173] wherein: is the PMU optimal buffer length; the overline is the "not" operator,

[0174] Therefore, based on the above steps, the optimal buffer length of each PMU measuring point can be calculated, and the mean and variance of the PMU statistical data in the buffer are calculated, so as to perform section normalization,

[0175] b2, SCADA optimal buffer length calculation

[0176] Similarly, for SCADA measurement, since it is sampled about every 2s, there are more SCADA data collected and transmitted to the data center between two AMI data gaps, so the optimal buffer length of SCADA needs to be calculated to combine the equivalent section data with AMI data for application.

[0177] Within 15 minutes, the SCADA sampling data series has 450 measuring data, and the method for calculating the optimal buffer length is similar to PMU.

[0178] After the measuring section normalization of PMU and SCADA, all the processed data and the incoming AMI data are regarded as data in the same section, so as to be combined together for SE calculation.

[0179] c, based on the exponential weight robust state estimation method, an active distribution network robust state estimation model is established, and an active distribution network robust three-phase state estimation method is proposed,

[0180] Combine AMI, PMU and SCADA three types of data, based on the exponential weight robust estimation method, an active distribution network robust estimation model is established, and the measurement of the active distribution network system is:

[0181] (22)

[0182] wherein: represents the full measurement of the active distribution network; 、 and AMI, PMU and SCADA three types of measurement,

[0183] Based on the above normalized full measurement, an exponential weight robust estimation model of the main distribution network is established, and the Newton method is used to solve, and an exponential weight robust three-phase state estimation method of the active distribution network is proposed,

[0184] c1, exponential weight robust state estimation model of active distribution network

[0185] In the exponential weight robust state estimation model, the objective function is the minimum sum of exponential weight residuals:

[0186] (23)

[0187] wherein: represents the measurement; represents the measurement function; c(x) represents the equality constraint; represents the measurement residual; x represents the state variable; represents the diagonal matrix of the weight function, ,

[0188] The calculation expression is:

[0189] (24)

[0190] wherein: represents the fixed weight; represents the standard deviation, and the initial value calculation formula is , P represents the prior diagonal matrix, and the diagonal element , m represents the number of measurements, and n represents the number of states;

[0191] c2, active distribution network index weight robust state estimation algorithm

[0192] Based on the Lagrange multiplier The equality constraint is converted and processed to obtain an unconstrained objective function, and the optimization solution is calculated as follows:

[0193] (25)

[0194] wherein: represents the augmented Lagrange function,

[0195] The and are both Taylor expanded and linearized to process the nonlinear equation:

[0196] (26)

[0197] wherein: , represents the Jocabian matrix, and is and (simplified as H and C),

[0198] The first-order partial derivative of formula (25) is:

[0199] (27)

[0200] Based on the Newton method iteration solution, the incremental update matrix form is:

[0201] (28)

[0202] where k represents the iteration number; matrix ; ; ,

[0203] Based on the exponential weight robust state estimation algorithm, the exponential weight automatically corrects the measurement weight, and the influence of bad data is suppressed. The calculation process is shown in Fig. 2. Figure 2

[0204] The detailed calculation steps of the algorithm are as follows:

[0205] ① Initialization, set the real and imaginary parts of the initial value of the balance, load and tie point three-phase voltage, take the balance point A phase as the phase reference, and the earth as the potential reference. Set the fixed weight matrix R-1, the maximum iteration number Tmax, the convergence threshold ε, and the initial iteration time = 1.

[0206] ② Calculate the end point admittance matrix and Jocabian matrix, calculate the unbalance and , and calculate the diagonal element value of the weight matrix W, form and Jocabian matrix H, C, and calculate ;

[0207] ③ Calculate the state deviation , and update the state value: , time = time + 1;

[0208] ④ Convergence judgment, if , the convergence is calculated and the result is output; if and time > Tmax, the iteration is stopped, and "not convergent!" Otherwise, go to step ⑤.

[0209] ⑤ Correct the dynamic scale parameter, if( ) 2 0.01,( ) 2 = ; if( ) 2 <0.01,( ) 2 = , go to step ②.

[0210] Example simulation and analysis

[0211] Basic data and simulation conditions

[0212] ​To verify the effectiveness and robustness of the proposed method, simulation experiments are carried out on the IEEE-13 node ADN network. The network connection of the ADN network is shown in Fig. 1. Figure 3

[0213] To adapt to the research of the active distribution network, the standard system is modified as follows:

[0214] ①The branch type of the original system is 501;

[0215] ②The 5th node of the original system is connected to a photovoltaic power supply, and the three-phase output of the power supply is set to 40 kW, and the power factor is 0.98;

[0216] ③SCADA measurement is arranged in the whole system, and node 1 is set as the system balance node at the low-voltage side of the distribution transformer. AMI smart meters are configured at the low-voltage side of the distribution transformer and each load terminal, and PMUs are arranged at nodes 2, 6, 9, and 11.

[0217] To simulate actual measurement, the improved system is calculated by the Newton-Raphson method to obtain the power flow results as the true values of the system physical quantities. On this basis, random distributed errors are superimposed to simulate actual measurement. 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.

[0218] Based on the above improved system and simulated measurement, the following two simulation examples are studied:

[0219] Example 1: The system has no bad data, and each simulated measurement is generated by superimposing random distributed errors on the true value;

[0220] Example 2: The system has bad data, and it is assumed that the bad data is generated by the active power of phase a of node 2 and the voltage modulus of phase C of node 6. The active power error is set to 30% of the original value, and the modulus is set to 55% of the original value, simulating bad data collected due to external reasons.

[0221] In the above examples, the proposed method is compared with the WLS method.

[0222] Simulation results and analysis

[0223] (1) ADN estimation result effectiveness analysis

[0224] To verify the effectiveness of the proposed method, the following two indicators are defined for evaluation:

[0225] (29)

[0226] ​ (30)

[0227] wherein: denotes the system error; denotes the maximum error; denotes the true value; denotes the estimated value; denotes the number of state variables.

[0228] The SE method and the WLS method based on the multi-source mixed data active distribution network method proposed by the present method are simulated in two cases of Example 1 and Example 2, and the simulation results are shown in Table 1.

[0229]

[0230] As shown in Table 1, in Example 1, for the voltage modulus value, the system error and the maximum error of the multi-source mixed data active distribution network SE method of the present method are 1.1322×10 -3 and 5.5124×10 -4 , respectively, and the system error and the maximum error of the WLS are 7.9231×10 -3 and 5.7981×10 -4 , respectively; for the voltage phase angle, the system error and the maximum error of the multi-source mixed data active distribution network SE method of the present method are 0.04913 and 1.6824×10 -3 , respectively, and the system error and the maximum error of the WLS are 0.05537 and 1.9721×10 -3 , respectively. Obviously, in the case of no bad data in Example 1, the system error and the maximum error of the multi-source mixed data active distribution network SE method of the present method and the WLS method are relatively small, but the error of the present method is smaller, and the estimation result is better.

[0231] In Example 2, for the voltage modulus value, the system error and the maximum error of the multi-source mixed data active distribution network SE method of the present method are 3.1765×10 -3 and 8.6133×10 -4 , respectively, and the system error and the maximum error of the WLS are 0.1327 and 5.6521×10 -3 , respectively; for the voltage phase angle, the system error and the maximum error of the multi-source mixed data active distribution network SE method of the present method are 0.0944 and 1.2563×10 -2 , respectively, and the system error and the maximum error of the WLS are 1.7514 and 0.1124. Obviously, in the case of bad data in Example 2, the system error and the maximum error of the multi-source mixed data active distribution network SE method of the present method and the WLS method are relatively larger than those in Example 1, but the error of the present method increases less, and has excellent robustness.

[0232] (2) ADN estimation result accuracy analysis

[0233] The method is based on multi-source mixed data to carry out active distribution network three-phase SE, which can calculate three-phase state. Based on example 1 and example 2, the following two simulation scenarios are set to compare with the true value to verify the accuracy of the calculation results of the method based on multi-source mixed data active distribution network three-phase SE:

[0234] (1) The balanced end point phasor of the active distribution network is set to be symmetric;

[0235] (2) The balanced end point phasor of the active distribution network is set to be asymmetric.

[0236] Table 2 and Table 3 are the comparison of three-phase voltage SE values of active distribution network 1, 2 and 5 between the three end points and the true value under the two simulation scenarios of example 1 and example 2.

[0237] As shown in Table 2, based on example 1, whether the balanced end point is set to be symmetric or not, the results obtained by the multi-source mixed data active distribution network three-phase SE method are consistent with the true value, and the three-phase SE of the active distribution network can be accurately performed. As shown in Table 3, based on example 2, whether the balanced end point is set to be symmetric or not, the results obtained by the multi-source mixed data active distribution network three-phase SE method are also very close to the true value, and the three-phase SE of the active distribution network can be accurately performed. The main reason is that example 2 sets part of the bad data based on example 1, and the multi-source mixed data robust SE method can effectively filter out bad data to obtain the best estimated value closest to the true value, but compared with the case without bad data, there is a slight deviation, but the deviation is still very small, and the three-phase SE of the active distribution network can be accurately performed.

[0238]

[0239]

[0240] (3) Estimation result convergence analysis of active distribution network

[0241] To verify the convergence of the method proposed by the multi-source mixed data active distribution network SE method with the increase of the penetration rate of distributed power, the following three situations are set:

[0242] (1) Access to 1 distributed power source: access photovoltaic power source at the original system 5 node, set the three-phase output of the node power source to be 40kW, and the power factor to be 0.98;

[0243] (2) Access to 2 distributed power sources: access photovoltaic power sources at the original system 5 and 7 nodes, set the three-phase output of the node power source to be 40kW, and the power factor to be 0.98;

[0244] (3) Access to 3 distributed power supply: access to photovoltaic power supply in the original system 5, 7 and 9 nodes, set node power three-phase output is 40kW, power factor 0.98;

[0245] In the above three cases, based on the example 1 and example 2 set by the method respectively, the multi-source hybrid data robust SE method and WLS method proposed in the method are used for SE calculation, and the convergence number and the estimation error under the corresponding situation are shown in table 4 and table 5 respectively.

[0246]

[0247] From table 4, based on example 1, in three kinds of distributed power supply access situation, the iteration number of multi-source hybrid data robust SE method based on the method is 5, 6 and 6 times respectively, and the iteration number of WLS method is 4, 5 and 6 times respectively, the iteration number of WLS method is relatively less than the multi-source hybrid data robust SE method of the method, but almost close, the reason is that example 1 has no bad data, the two methods can realize reliable calculation, and the convergence is good. Based on example 2, in three kinds of distributed power supply access situation, the iteration number of multi-source hybrid data robust SE method based on the method is 6, 7 and 7 times respectively, and the iteration number of WLS method is 8, 9 and 11 times respectively. The iteration number of WLS method is significantly increased than the multi-source hybrid data robust SE method of the method, the reason is that example 1 sets bad data, with the increase of distributed photovoltaic power supply, WLS cannot effectively handle bad data, and the iteration number increases.

[0248]

[0249] From table 5, in example 1, for voltage module value and phase angle, in three situations, the system error of multi-source hybrid data active distribution network SE method of the method is 1.1322×10 -3 , 1.1431×10 -3 , 1.1437×10 -3 and 0.04913, 0.05011, 0.05024, the system error of WLS is 7.9231×10 -3 , 7.9351×10 -3 , 7.9377×10 -3 and 0.05537, 0.05734, 0.05759. In example 2, for voltage module value and phase angle, in three situations, the system error of multi-source hybrid data active distribution network SE method of the method is 3.1765×10 -3 , 3.1861×10 -3 , 3.1951×10 -3The system errors of WLS are 0.1327, 0.1433, 0.1459 and 1.7514, 1.7667, 1.7721, respectively.

[0250] Obviously, in the case of no bad data in Example 1, the system errors of the multi-source hybrid data active distribution network SE method and the WLS method are relatively small, and the convergence is also better. However, in Example 2, after setting the bad data, with the increase of the distributed power supply, WLS cannot effectively handle the bad data, not only the iteration number increases, but also the SE result is poor, while the multi-source hybrid data robust SE method can realize reliable calculation, has good convergence, and can maintain good accuracy.

Claims

1. A robust three-phase state estimation method for active distribution networks based on multi-source mixed data, characterized in that, The method steps are as follows: a. Establishing a three-phase measurement-state model of each of the multi-source SCADA, PMU and AMI data; b. Normalizing the measurement section of the PMU, SCADA and AMI data by calculating the optimal buffer length of the PMU and SCADA data; c. Establishing a robust state estimation model of the active distribution network based on an exponential weight robust state estimation method, and proposing a robust three-phase state estimation method of the active distribution network, The step c is specifically as follows, The AMI, PMU and SCADA data are combined, a robust estimation model of the active distribution network is established based on an exponential weight robust estimation method, and the measurement of the active distribution network system is: (22) wherein: represents full measurement of active distribution network; , and respectively represent AMI, PMU and SCADA three types of measurement, Based on the full measurement after the measurement section normalization processing, an exponential weight robust estimation model of the main distribution network is established, and the Newton method is used for solving, and an exponential weight robust three-phase state estimation method of the active distribution network is proposed, c1. Exponential weight robust state estimation model of the active distribution network In the exponential weight robust state estimation model, the objective function is the minimum sum of the exponential weight residual: (23) where: represents the measurement; represents the measurement function; c(x) represents the equality constraints; represents the measurement residual; x represents the state; represents the diagonal matrix of the weight function, , Computes expression: (24) wherein: represents the fixed weight; represents the standard deviation, the initial value is calculated by , P represents the prior diagonal matrix, the diagonal element m represents the number of measurements, and n represents the number of states. c2. Exponential weight robust state estimation algorithm of the active distribution network Based on the Lagrange multiplier The equality constraint is transformed into an unconstrained objective function, and the optimization solution is calculated as follows: (25) wherein: denotes the augmented Lagrange function, For and both Taylor expanded and linearized about the nonlinear equations: (26) In the formula: , denotes the Jacobian matrix, respectively and , The first-order partial derivative of formula (25) is solved: (27) Based on the Newton method iteration, the incremental update matrix form is: (28) wherein: k represents the number of iterations; matrix ; ; .

2. The method of claim 1, wherein the method is based on multi-source hybrid data of active distribution network robust three-phase state estimation. The step a is specifically as follows, a1. AMI measurement-state model In the active distribution network, the distributed power supply end and the load end connected with the load are configured with AMI smart meters. The smart meters are set to collect data every 15 minutes, and collect the three-phase voltage module value of the end point , the injected current module value , the injected active power size , the injected reactive power size , According to the injection , The current injection equation is established as follows: (2) Expanding the above formula, the real part and imaginary part equations are respectively obtained: (3) (4) wherein: is the injection point power; is the injection point active; is the injection point reactive; , is the end point in real and imaginary parts of the phase voltage; , is the end point mutual conductance, mutual susceptance, According to the voltage module value The voltage module value equation is established as follows: (5) According to the injected current modulus The equation of the current modulus is established as follows: ； (6) a2. PMU measurement-state model On the endpoint of the active distribution network, the PMU measurement device is configured, and the relevant voltage and current phasors on the endpoint are obtained through the PMU. The types of measurement mainly include: each-phase voltage phasor , ; Individual branch current phasors associated with PMU endpoints , , and , Voltage phasor to PMU , The voltage equation is established as follows: (7) Expanding the above formula, the real part and imaginary part equations are obtained: (8) (9) wherein: , are the real, imaginary parts of The branch current phasors associated with the PMU endpoints , The branch current phasor equations are established as follows: (10) Expanding the above formula, the real part and imaginary part equations are obtained: (11) (12) In the formulae: , denote the real and imaginary parts, respectively. a3. SCADA measurement-state model RTU device in a SCADA system, measuring injected power , , branch power , and voltage modulus , SCADA system measures as total measures, For the voltage magnitude The voltage magnitude equation is established as follows: (13) For injection power , Its current injection equation is established as follows: (14) Expanding the above formula, the real part and imaginary part equations are respectively obtained: (15) (16) For branch power , , its branch power equation is established as follows: (17) Expanding the above formula, the real part and imaginary part equations are obtained: (18) ； (19) a4. Virtual measurement-state model of the connection end point In the active distribution network, there are zero injection end points which are not directly connected with the generator or the load, and only serve as a connection function. The virtual measurement equation is established as follows: (20) 。 3. The method of claim 1, wherein: The step b is specifically as follows, b1. PMU optimal buffer length calculation Before 30s of AMI, there are 750 measurements of PMU, the method of calculating the best buffer length is assuming that the sampling frequency of PMU is sample / s, the system updates the state once every time the next AMI data arrives, (1) The number of PMU sampling measurements is The sampling measurements are divided into measurement subsets, and in general , so the number of each measurement subset is , The subsets are expressed as: , (2) on the latest subset of PMU measurements Performing change point detection: ①calculating the standard deviation of the measurements and comparing the calculated standard deviation to a threshold value, the amplitude threshold value being 10 -4 , the phase threshold value being 10 -3 , if the standard deviation is smaller than the threshold value, it is indicated that there is no change point within the buffer, that the entire buffer is contained in the buffer, go to (3); If the standard deviation is greater than the threshold value, the method proceeds to step 2. If the standard deviation is greater than the threshold value, the method proceeds to step 2. The latest PMU in the middle is taken as the measurement value of the equivalent section, and the calculation is completed. (3) Based on the internal function ttest2 in Matlab With , , With , With At the default value of 95% probability, the hypothesis test is performed, and the result is represented as: ​ (21) wherein: is the PMU optimal buffer length; the overline is the "not" operator symbol, Therefore, based on the above steps, the optimal buffer length of each PMU measurement point is calculated, and the mean and variance of the PMU statistical data in the buffer zone are calculated, so that the section normalization is performed, b2. SCADA optimal buffer length calculation Within 15 minutes, there are 450 measurement data in the SCADA sampling data series, and the method for calculating the optimal buffer length is similar to that of the PMU, After the measurement section normalization processing of the PMU and SCADA, all the processed data and the incoming AMI data are regarded as data in the same section, so as to be combined together for SE calculation.

Images