Distribution network intra-day distributed rolling scheduling method based on DGC division

Abstract

The invention relates to a DGC division-based intra-day distributed rolling scheduling method for a power distribution network. The method comprises the following steps of S1, decoupling each DGC of the power distribution network; s2, updating each DGC virtual boundary tie line variable and a Lagrange multiplier rule, and realizing an ADMM algorithm based on the DGC; s3, establishing each DGC objective function by taking the day-ahead start-stop condition of the MGT and the day-ahead switching condition of the CB as fixed values; s4, performing calculation under the constraint of each DGC scheduling optimization model condition; s5, example analysis is carried out; the method has the advantages that the alternating direction multiplier method is adopted, the intra-day distributed rolling scheduling model is established, and the prediction error influence is eliminated.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network dispatching, and in particular relates to a distributed rolling dispatching method for a distribution network based on DGC division. Background Art

[0002] At present, there are many research works on cluster division, but most of them are aimed at voltage control and use a single electrical coupling degree to evaluate the DGC division results. The ADN dispatching flexibility supply is closely related to the operating status of each dispatching flexibility resource. Therefore, the distribution network DGC division research considering the dispatching flexibility demand is closely related to the day-ahead dispatching. Ignoring the day-ahead dispatching will not accurately characterize the ADN dispatching flexibility supply capacity, making it difficult to obtain a reliable DGC. In addition, due to the large impedance of the distribution network, the strong uncertainty of DPV output will also lead to the instability of node voltage in the distribution network, which will endanger the safe operation of the distribution network. Therefore, it is very necessary to provide a distribution network intraday distributed rolling dispatching method based on DGC division that adopts the alternating direction multiplier method, establishes an intraday distributed rolling dispatching model, and eliminates the influence of prediction errors. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a distribution network intraday distributed rolling scheduling method based on DGC division, which adopts an alternating direction multiplier method, establishes an intraday distributed rolling scheduling model, and eliminates the influence of prediction errors.

[0004] The object of the present invention is achieved by: a distributed rolling dispatching method for a distribution network based on DGC division, the method comprising the following steps:

[0005] Step S1: Decouple each DGC of the distribution network;

[0006] Step S2: updating the virtual boundary tie line variables and Lagrange multiplier rules of each DGC to implement the DGC-based ADMM algorithm;

[0007] Step S3: Taking the start and stop status of MGT and the switching status of CB as constants, establish the objective functions of each DGC;

[0008] Step S4: performing calculations under the constraints of the DGC scheduling optimization model conditions;

[0009] Step S5: case analysis.

[0010] The ADMM algorithm in step S2 adopts synchronous ADMM, i.e., SADMM, to change the rules of alternating iteration of each variable. When optimizing a certain DGC, the average value of the virtual boundary contact line variable of the previous iteration of the DGC and the adjacent DGC is used as the reference value in this iteration process for calculation: In the formula, are the reference values ​​of the virtual tie line variables of cluster ci and cluster cj in the m+1th iteration respectively; are the virtual tie line variable values ​​calculated for cluster ci and cluster cj at the mth iteration respectively.

[0011] The update rules of the DGC virtual boundary tie line variables and Lagrange multipliers in step S2 are:

[0012]

[0013] The convergence conditions of the SADMM algorithm include the original residual and the dual residual, as shown in the following formula: In the formula, are the original residuals of cluster ci and cluster cj at the m+1th iteration respectively; are the dual residuals of cluster ci and cluster cj at the mth iteration respectively; ε is the convergence condition, and its value is usually taken as N is the total number of distribution network nodes.

[0014] The establishment of each DGC objective function in step S3 is specifically as follows: considering the situation that the active and reactive power of the distribution network are not decoupled, active-reactive coordinated optimization is performed, and the synchronous ADMM suitable for DGC is used for solving the problem. The distributed power generation cluster division result with multiple control objectives based on the day-ahead scheduling result is considered. The economic goal of minimum daily operating cost and the goal of minimum voltage deviation are considered, and the consistency cost is introduced. After the distribution network DGC is divided, the objective functions of each DGC are basically the same. The start and stop situation of MGT and the switching situation of CB are taken as fixed values, and the objective functions of each DGC are established as follows:

[0015] In the formula, the superscript c represents the cth DGC; is the cost of purchasing electricity from the external network at time t; c buy (t), P buy (t) is the cost coefficient and power of electricity purchased from the external network at time t; is the MGT operating cost at time t; c gi,op 、c gi,of is the output cost parameter and start-stop cost parameter of the gi-th MGT; gi (t) is the output status of the gi-th MGT at time t; is the ESS operating cost at time t; c si is the charge and discharge loss cost parameter of the si-th ESS; P si,char (t), P si,disc (t) is the charge and discharge power of the si-th ESS at time t; is the network loss cost at time t; ij represents the branch with i as the head end and j as the end end; is the voltage deviation penalty cost of cluster c at time t; c V is the voltage deviation penalty cost coefficient; V i (t) is the voltage value of node i at time t; V i,ref (t) is the reference voltage value of node i at time t, i.e., the day-ahead scheduling value; is the consistency penalty cost of cluster c; ck is the set of adjacent DGCs of cluster ci; x ij′ (t) is the virtual tie line variable between cluster ci and its adjacent cluster cj at time t, is the reference value of the virtual tie line variable in the mth iteration at time t, and its initial value is the day-ahead scheduling value of the actual line ij; is the Lagrange multiplier between cluster ci and its adjacent cluster cj during the mth iteration; ρ is the iteration step size.

[0016] The constraints of each DGC dispatch optimization model in step S4 are basically consistent with the day-ahead global model, and the constraints include power flow constraints, power balance constraints, purchased power constraints, MGT operation constraints, ESS operation constraints, and SVC operation constraints;

[0017] The power flow constraints are specifically: (I ij (t)) 2 =((P ij (t)) 2 +(Q ij (t)) 2 ) / V i (t),ij∈Nl(15), ((P ij (t)) 2 +(Q ij (t)) 2 )≤((I ij (t)) 2 (V i (t)) 2 ),ij∈Nl(18), where V i,min 、V i,max Respectively represent the minimum and maximum voltage amplitude of node i; V i (t) represents the voltage amplitude of node i at time t; r ij 、x ij Respectively represent the resistance and reactance of branch ij; P ij (t), Qij (t) represent the active and reactive power flowing through branch ij at time t respectively; I ij,max Indicates the maximum current amplitude allowed to flow through branch ij; P j (t), Q j (t) represents the active and reactive power injected by node j; Nl represents the node set; m(j) and n(j) represent the end / head node sets of the branch with node j as the head / end respectively.

[0018] The power balance constraints are specifically: Q Load,i (t) = Q buy,i (t)+Q CB,i (t)+Q SVC,i (t)+(I ij (t) 2 )x ij (20), where P Load,i (t), Q Load,i (t) respectively represent the active and reactive load power of node i at time t; P buy,i (t), Q buy,i (t) represents the active and reactive power purchased by node i from the upper power grid at time t. If the node is not connected to the upper power grid, its value is 0; P pv,i (t), P g,i (t) represents the output value of the DPV device and MGT device connected to node i at time t. If the node is not connected to the above devices, its value is 0; They represent the discharge and charging power of the ESS connected to node i at time t, respectively. If the node is not connected to the ESS, its value is 0; They represent the discharge and charging power of the EVCS connected to node i at time t. If the node is not connected to EVCS, its value is 0; Q CB,i (t), Q SVC,i (t) represents the reactive power output of the CB and SVC equipment connected to node i at time t. If the node is not connected to the above equipment, its value is 0.

[0019] The power purchase power constraint is specifically: In the formula, Indicates the maximum value of active power purchase; Respectively represent the minimum and maximum values ​​of reactive power purchase.

[0020] The MGT operation constraints are as follows: The MGT operation constraints in the intraday distributed scheduling include the upper and lower limits of the output and the maximum climbing / slope power constraints, as shown in the following formula. The start and stop states are determined by the day-ahead scheduling arrangement: Where Ng is the MGT set in ADN; P gi(t) is the output value of the gi-th MGT at time t; are the maximum climbing and sliding powers of the gi-th MGT, respectively; are the maximum and minimum outputs of the gi-th MGT respectively; T is the scheduling period; u gi (t) is the start / stop status of the gi-th MGT at time t.

[0021] The ESS operation constraints are specifically as follows: ESS constraints include charge and discharge power constraints and capacity constraints, namely: u si,char (t)+u si,disc (t)≤1,si∈Ns(27), Where Ns is the set of ESS in ADN; P si,char (t), P si,disc (t) are the charge and discharge power of the si-th ESS at time t; are the maximum charge and discharge power of the si-th ESS; u si,char (t),u si,disc (t) is the charge and discharge state 0-1 variable of the si-th ESS at time t, u si,disc (t) = 1, it is in the discharge state, u si,char (t) = 1, it is in charging state; are the maximum and minimum SOC of the si-th ESS respectively; E si , η si are the maximum capacity and charge / discharge efficiency of the si-th ESS respectively; E si (t) is the capacity and charge-discharge efficiency of the si-th ESS at time t.

[0022] The SVC operation constraints are specifically: SVC constraints include SVC capacity constraints, namely: Where, Nsvc is the SVC set in ADN; They respectively represent the upper and lower limits of the SVC operating capacity of the svci station.

[0023] Beneficial effects of the present invention: The present invention is a distributed rolling scheduling method for distribution networks within a day based on DGC division. During use, compared with the traditional centralized optimization method, the SADMM algorithm adopted by the present invention can improve the calculation efficiency while ensuring the validity of the calculation results, and realize the efficient optimization scheduling of the DG high-penetration distribution network; the present invention performs SADMM distributed optimization solution according to different DGC division schemes, and the results obtained show that a reasonable DGC division scheme can effectively improve the calculation efficiency of the SADMM algorithm and the active and reactive autonomy capabilities of each DGC; the present invention has the advantages of adopting the alternating direction multiplier method, establishing a distributed rolling scheduling model within a day, and eliminating the influence of prediction errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of the present invention.

[0025] Figure 2 It is a schematic diagram of the DGC decoupling process of the present invention.

[0026] Figure 3 This is a flow chart of the SADMM calculation based on DGC partitioning of the present invention.

[0027] Figure 4 It is the DPV day-ahead and day-intraday prediction curve diagram of the present invention.

[0028] Figure 5 It is the load forecast curve diagram before and during the day of the present invention.

[0029] Figure 6 It is a schematic diagram of the intraday distributed scheduling results of the present invention.

[0030] Figure 7 This is the SADMM optimization convergence curve diagram of the DGC partitioning result based on Scheme 1 of the present invention.

[0031] Figure 8 This is the SADMM optimization convergence curve diagram of the DGC partitioning result based on Scheme 2 of the present invention.

[0032] Fig. 9 This is a schematic diagram of the power changes of the DGC interconnection line in the distributed optimization method of Scheme 2 of the present invention.

[0033] Fig.10 This is a schematic diagram of the power changes of the DGC interconnection line in the distributed optimization method of Scheme 1 of the present invention.

[0034] Fig.11 This is a schematic diagram of node voltage distribution of the distributed optimization method of Scheme 2 of the present invention.

[0035] Fig.12 This is a schematic diagram of node voltage distribution of the distributed optimization method of Scheme 1 of the present invention.

[0036] Fig.13 This is a schematic diagram of the node voltage and day-ahead result deviation rate of the distributed optimization method of Scheme 2 of the present invention.

[0037] Fig.14 This is a schematic diagram of the node voltage and day-ahead result deviation rate of the distributed optimization method of Scheme 1 of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings.

[0039] Example 1

[0040] like Figure 1-14 As shown, a distributed rolling dispatching method for a distribution network within a day based on DGC division is provided, wherein the method comprises the following steps:

[0041] Step S1: Decouple each DGC of the distribution network;

[0042] Step S2: updating the virtual boundary tie line variables and Lagrange multiplier rules of each DGC to implement the DGC-based ADMM algorithm;

[0043] Step S3: Taking the start and stop status of MGT and the switching status of CB as constants, establish the objective functions of each DGC;

[0044] Step S4: performing calculations under the constraints of the DGC scheduling optimization model conditions;

[0045] Step S5: case analysis.

[0046] In this embodiment, in order to implement the ADMM algorithm based on DGC, the DGCs of the distribution network are firstly decoupled. The decoupling process is as follows: Figure 2 As shown in the figure, x ij′ 、x i′j Respectively represent the reactance value of branch ij′, i′j; P ij′ , Q ij′ Respectively represent the active and reactive power flowing through the branch ij′; I ij′ represents the current flowing through the branch ij′; P i′j , Q i′j Respectively represent the active and reactive power flowing through branch i′j; I i′j Represents the current flowing through branch i′j.

[0047] The ADMM algorithm in step S2 adopts synchronous ADMM, i.e., SADMM, to change the rules of alternating iteration of each variable. When optimizing a certain DGC, the average value of the virtual boundary contact line variable of the previous iteration of the DGC and the adjacent DGC is used as the reference value in this iteration process for calculation: In the formula, are the reference values ​​of the virtual tie line variables of cluster ci and cluster cj in the m+1th iteration respectively; are the virtual tie line variable values ​​calculated for cluster ci and cluster cj at the mth iteration respectively.

[0048] The update rules of the DGC virtual boundary tie line variables and Lagrange multipliers in step S2 are:

[0049]

[0050] The convergence conditions of the SADMM algorithm include the original residual and the dual residual, as shown in the following formula: In the formula, are the original residuals of cluster ci and cluster cj at the m+1th iteration respectively; are the dual residuals of cluster ci and cluster cj at the mth iteration respectively; ε is the convergence condition, and its value is usually taken as N is the total number of distribution network nodes.

[0051] The establishment of each DGC objective function in step S3 is specifically as follows: considering the situation that the active and reactive power of the distribution network are not decoupled, active-reactive coordinated optimization is performed, and the synchronous ADMM suitable for DGC is used for solving the problem. The distributed power generation cluster division result with multiple control objectives based on the day-ahead scheduling result is considered. The economic goal of minimum daily operating cost and the goal of minimum voltage deviation are considered, and the consistency cost is introduced. After the distribution network DGC is divided, the objective functions of each DGC are basically the same. The start and stop situation of MGT and the switching situation of CB are taken as fixed values, and the objective functions of each DGC are established as follows: In the formula, the superscript c represents the cth DGC; is the cost of purchasing electricity from the external network at time t; c buy (t), P buy (t) is the cost coefficient and power of electricity purchased from the external network at time t; is the MGT operating cost at time t; c gi,op 、c gi,of is the output cost parameter and start-stop cost parameter of the gi-th MGT; gi (t) is the output status of the gi-th MGT at time t; is the ESS operating cost at time t; c si is the charge and discharge loss cost parameter of the si-th ESS; P si,char (t), P si,disc (t) is the charge and discharge power of the si-th ESS at time t; is the network loss cost at time t; ij represents the branch with i as the head end and j as the end end; is the voltage deviation penalty cost of cluster c at time t; c V is the voltage deviation penalty cost coefficient; V i (t) is the voltage value of node i at time t; V i,ref (t) is the reference voltage value of node i at time t, i.e., the day-ahead scheduling value; is the consistency penalty cost of cluster c; ck is the set of adjacent DGCs of cluster ci; x ij′ (t) is the virtual tie line variable between cluster ci and its adjacent cluster cj at time t, is the reference value of the virtual tie line variable in the mth iteration at time t, and its initial value is the day-ahead scheduling value of the actual line ij; is the Lagrange multiplier between cluster ci and its adjacent cluster cj during the mth iteration; ρ is the iteration step size.

[0052] The constraints of each DGC dispatch optimization model in step S4 are basically consistent with the day-ahead global model, and the constraints include power flow constraints, power balance constraints, purchased power constraints, MGT operation constraints, ESS operation constraints, and SVC operation constraints;

[0053] The power flow constraints are specifically: (I ij (t)) 2 =((P ij (t)) 2 +(Q ij (t)) 2 ) / V i (t),ij∈Nl(15), ((P ij (t)) 2 +(Q ij (t)) 2 )≤((I ij (t)) 2 (V i (t)) 2 ),ij∈Nl(18), where V i,min 、V i,max Respectively represent the minimum and maximum voltage amplitude of node i; V i (t) represents the voltage amplitude of node i at time t; r ij 、x ij Respectively represent the resistance and reactance of branch ij; P ij (t), Q ij (t) represent the active and reactive power flowing through branch ij at time t respectively; I ij,max Indicates the maximum current amplitude allowed to flow through branch ij; P j (t), Q j (t) represents the active and reactive power injected by node j; Nl represents the node set; m(j) and n(j) represent the end / head node sets of the branch with node j as the head / end respectively.

[0054] The power balance constraints are specifically: Q Load,i (t) = Q buy,i (t)+Q CB,i (t)+QSVC,i (t)+(I ij (t) 2 )x ij (20), where P Load,i (t), Q Load,i (t) respectively represent the active and reactive load power of node i at time t; P buy,i (t), Q buy,i (t) represents the active and reactive power purchased by node i from the upper power grid at time t. If the node is not connected to the upper power grid, its value is 0; P pv,i (t), P g,i (t) represents the output value of the DPV device and MGT device connected to node i at time t. If the node is not connected to the above devices, its value is 0; They represent the discharge and charging power of the ESS connected to node i at time t, respectively. If the node is not connected to the ESS, its value is 0; They represent the discharge and charging power of the EVCS connected to node i at time t. If the node is not connected to EVCS, its value is 0; Q CB,i (t), Q SVC,i (t) represents the reactive power output of the CB and SVC equipment connected to node i at time t. If the node is not connected to the above equipment, its value is 0.

[0055] The power purchase power constraint is specifically: In the formula, Indicates the maximum value of active power purchase; Respectively represent the minimum and maximum values ​​of reactive power purchase.

[0056] The MGT operation constraints are as follows: The MGT operation constraints in the intraday distributed scheduling include the upper and lower limits of the output and the maximum climbing / slope power constraints, as shown in the following formula. The start and stop states are determined by the day-ahead scheduling arrangement: Where Ng is the MGT set in ADN; P gi (t) is the output value of the gi-th MGT at time t; are the maximum climbing and sliding powers of the gi-th MGT, respectively; are the maximum and minimum outputs of the gi-th MGT respectively; T is the scheduling period; u gi (t) is the start / stop status of the gi-th MGT at time t.

[0057] The ESS operation constraints are specifically as follows: ESS constraints include charge and discharge power constraints and capacity constraints, namely: u si,char (t)+u si,disc (t)≤1,si∈Ns(27), Where Ns is the set of ESS in ADN; Psi,char (t), P si,disc (t) are the charge and discharge power of the si-th ESS at time t; are the maximum charge and discharge power of the si-th ESS; u si,char (t),u si,disc (t) is the charge and discharge state 0-1 variable of the si-th ESS at time t, u si,disc (t) = 1, it is in the discharge state, u si,char (t) = 1, it is in charging state; are the maximum and minimum SOC of the si-th ESS respectively; E si , η si are the maximum capacity and charge / discharge efficiency of the si-th ESS respectively; E si (t) is the capacity and charge-discharge efficiency of the si-th ESS at time t.

[0058] The SVC operation constraints are specifically: SVC constraints include SVC capacity constraints, namely: Where, Nsvc is the SVC set in ADN; They respectively represent the upper and lower limits of the SVC operating capacity of the svci station.

[0059] In this embodiment, the SADMM calculation process based on DGC partitioning includes the following steps:

[0060] Step 1: Initialize the Lagrange multiplier and the reference value of each DGC virtual tie line variable. The initial value is the day-ahead scheduling result.

[0061] Step 2: Solve the optimization model of each DGC to obtain the variable value of the virtual tie line of each DGC of the mth generation;

[0062] Step 3: Calculate the original residual and the dual residual according to equations (6)-(7) to determine whether the residual meets the convergence condition. If so, the iteration ends and the scheduling optimization result is output. If not, proceed to the next step of calculation.

[0063] Step 4: Calculate the reference values ​​of the virtual tie lines and the Lagrange multipliers of the m+1th generation DGCs according to equations (1), (4)-(5), and return to step 2;

[0064] In summary, the calculation process of the distributed rolling dispatching method for distribution network based on DGC division can be obtained as follows: Figure 3 shown.

[0065] The present invention is a distributed rolling dispatching method for distribution networks based on DGC division. In use, based on day-ahead dispatching and cluster division, the present invention aims at the problem that traditional centralized control is difficult to meet the performance requirements of intraday dispatching time of DG high-penetration distribution networks. According to ultra-short-term forecast data of DPV and load rolling report, the alternating direction multiplier method (ADMM) is adopted to establish a distributed rolling dispatching model based on DGC division of distribution networks, so as to eliminate the influence of DPV and load forecast errors to the greatest extent. Compared with the traditional centralized optimization method, the SADMM algorithm adopted by the present invention can improve the calculation efficiency and realize the efficient optimization dispatching of DG high-penetration distribution networks while ensuring the validity of the calculation results. The present invention performs SADMM distributed optimization solution according to different DGC division schemes, and the obtained results show that a reasonable DGC division scheme can effectively improve the calculation efficiency of the SADMM algorithm and the active and reactive autonomy of each DGC. The present invention has the advantages of adopting the alternating direction multiplier method, establishing an intraday distributed rolling dispatching model, and eliminating the influence of prediction errors.

[0066] Example 2

[0067] like Figure 1-14 As shown, a distributed rolling dispatching method for a distribution network within a day based on DGC division is provided, wherein the method comprises the following steps:

[0068] Step S1: Decouple each DGC of the distribution network;

[0069] Step S2: updating the virtual boundary tie line variables and Lagrange multiplier rules of each DGC to implement the DGC-based ADMM algorithm;

[0070] Step S3: Taking the start and stop status of MGT and the switching status of CB as constants, establish the objective functions of each DGC;

[0071] Step S4: performing calculations under the constraints of the DGC scheduling optimization model conditions;

[0072] Step S5: case analysis.

[0073] In this embodiment, the example analysis: taking the IEEE33-node system as an example, but not considering EVCS, the intraday distributed optimization scheduling is studied;

[0074] The computing power analysis of the present invention adopts the improved IEEE33 node system, ignoring EVCS, DPV and load intraday ultra-short-term forecast results and load day-ahead and intraday forecast curves. Figure 4-5As shown in the figure, the DGC partitioning result adopts the cluster modularity index and the scheduling flexibility supply and demand robust balance index to perform cluster partitioning, that is, ω1=1, ω2=1, ω3=0, ω4=0, and performs intraday distributed rolling optimization scheduling analysis: the intraday scheduling cycle is 1 hour for rolling scheduling optimization, and the scheduling time scale is 15 minutes. In order to realize the efficient solution of the intraday distributed rolling scheduling model, the solution calculation is performed on the Cplex of the Matlab platform;

[0075] In order to verify the effectiveness of the SADMM distributed algorithm and its advantages in solution performance, the traditional centralized optimization scheduling method and the scheme 1 distributed optimization scheduling method based on the obtained distribution network DGC division results (hereinafter referred to as the scheme 1 distributed optimization method) are used to solve the intraday distributed rolling optimization scheduling example.

[0076] The scheduling results obtained by the distributed optimization method of Scheme 1 are as follows Figure 6 As shown, the total system scheduling cost is 30303.8881 yuan, and the total system scheduling cost obtained by the centralized optimization method is 30241.6342 yuan. The error rate of the total system scheduling cost obtained by the distributed optimization method and the centralized optimization method in scheme 1 is only 0.2%;

[0077] Figure 7 The convergence curves of the centralized optimization method and the distributed optimization method of Scheme 1 are calculated. It can be seen from the figure that after 9 iterations, the distributed optimization method of Scheme 1 gradually converges to the results of the centralized optimization method. In terms of solution efficiency, the centralized solution algorithm takes 188.3215 seconds, while the distributed optimization method of Scheme 1 takes a total of 676.1367 seconds after 9 iterations. Considering that the optimization calculation on a single computer is a serial calculation, in the actual parallel computing environment, the average solution time of each DGC of the distributed optimization method of Scheme 1 is 676.1367 / 4=169.0342 seconds, which is 10.2% less than that of the centralized optimization method. It can be seen that the distributed optimization method of Scheme 1 has a significant advantage in solution efficiency over the centralized optimization method.

[0078] In order to analyze the impact of cluster partitioning results on the SADMM algorithm, the present invention further performs distributed optimization scheduling method (hereinafter referred to as Scheme 2 distributed optimization method) calculation based on the distribution network DGC partitioning results obtained in Scheme 2 (only considering the modularity index for cluster partitioning, that is, ω1=1, ω2=0, ω3=0, ω4=0);

[0079] The total system scheduling cost calculated by the distributed optimization method of Scheme 2 is 30339.414 yuan. Compared with the centralized optimization method, the error rate is 0.3%. In terms of calculation efficiency, Figure 8The convergence curve of the distributed optimization method of Scheme 1 is calculated. It can be seen from the figure that Scheme 2, which only considers the modularity index for DGC division, converges after 15 iterations. The distributed optimization method of Scheme 2 takes a total of 1080.81 seconds. In the parallel computing environment, the average solution time of each DGC is 1080.81 / 6=180.135 seconds. Compared with the centralized optimization method, the solution time is reduced by 4.3%;

[0080] Comparing the distributed optimization method of Scheme 2 with that of Scheme 1, the calculation error of Scheme 2 increased slightly, and the solution time increased by 6.6%. This is because Scheme 2 only considers the modularity index when dividing DGCs, and divides the distribution network into 6 DGCs. The SADMM algorithm needs to interact with more virtual boundary variables when solving, and the convergence speed is slower.

[0081] The changes of DGC interconnection line power in each distributed optimization method within a day compared with the day-ahead optimization results are investigated. Figure 9-10 shown; according to Fig. 9 and Fig.10 During the peak period of DPV output, the power of the interconnection lines between DGCs in the distributed optimization methods of each scheme has changed significantly compared with the day-ahead optimization results. Each scheme needs to coordinate each DGC to meet the power change. However, since the distributed optimization method of Scheme 2 adopts the DGC division result that only considers the modularity, the frequency of power change of the interconnection lines between DGCs is significantly faster than that of the distributed optimization method of Scheme 1. The autonomy of each DGC is weak and it is more dependent on the coordination between DGCs.

[0082] Figure 11-12 Figure 13-14 are the node voltage distribution diagrams of the distributed optimization methods of each scheme, and the deviation rate between the voltage amplitude of each node in the system and the voltage amplitude of each node scheduled before the day is obtained by the distributed optimization method of each scheme. It can be seen from the figure that the overall distribution trend of the voltage of each node obtained by the distributed optimization method of Scheme 1 is more stable, and the deviation rate of the voltage of each node and the day-ahead result is significantly smaller than that of the distributed optimization method of Scheme 2. This is because the DGC division scheme 1 takes into account the ability of each cluster to deal with the voltage changes caused by DPV and load forecasting errors. The DGC obtained by the division has a stronger voltage autonomy ability, which further verifies the effectiveness of the DGC division method considering multiple control objectives.

[0083] The present invention is a distributed rolling dispatching method for distribution networks based on DGC division. In use, based on day-ahead dispatching and cluster division, the present invention aims at the problem that traditional centralized control is difficult to meet the performance requirements of intraday dispatching time of DG high-penetration distribution networks. According to ultra-short-term forecast data of DPV and load rolling report, the alternating direction multiplier method (ADMM) is adopted to establish a distributed rolling dispatching model based on DGC division of distribution networks, so as to eliminate the influence of DPV and load forecast errors to the greatest extent. Compared with the traditional centralized optimization method, the SADMM algorithm adopted by the present invention can improve the calculation efficiency and realize the efficient optimization dispatching of DG high-penetration distribution networks while ensuring the validity of the calculation results. The present invention performs SADMM distributed optimization solution according to different DGC division schemes, and the obtained results show that a reasonable DGC division scheme can effectively improve the calculation efficiency of the SADMM algorithm and the active and reactive autonomy of each DGC. The present invention has the advantages of adopting the alternating direction multiplier method, establishing an intraday distributed rolling dispatching model, and eliminating the influence of prediction errors.

Claims

1. A distributed rolling dispatching method for distribution network based on DGC division, characterized by: The method comprises the following steps: Step S1: Decouple each DGC of the distribution network; Step S2: updating the virtual boundary tie line variables and Lagrange multiplier rules of each DGC to implement the DGC-based ADMM algorithm; Step S3: Taking the start and stop status of MGT and the switching status of CB as constants, establish the objective functions of each DGC; Step S4: performing calculations under the constraints of the DGC scheduling optimization model conditions; Step S5: case analysis.

2. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 1 is characterized by: The ADMM algorithm in step S2 adopts synchronous ADMM, i.e., SADMM, to change the rules of alternating iteration of each variable. When optimizing a certain DGC, the average value of the virtual boundary contact line variable of the previous iteration of the DGC and the adjacent DGC is used as the reference value in this iteration process for calculation: In the formula, are the reference values ​​of the virtual tie line variables of cluster ci and cluster cj in the m+1th iteration respectively; are the virtual tie line variable values ​​calculated for cluster ci and cluster cj at the mth iteration respectively.

3. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 2 is characterized by: The update rules of the DGC virtual boundary tie line variables and Lagrange multipliers in step S2 are: The convergence conditions of the SADMM algorithm include the original residual and the dual residual, as shown in the following formula: In the formula, are the original residuals of cluster ci and cluster cj at the m+1th iteration respectively; are the dual residuals of cluster ci and cluster cj at the mth iteration respectively; ε is the convergence condition, and its value is usually taken as N is the total number of distribution network nodes.

4. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 1 is characterized by: The establishment of each DGC objective function in step S3 is specifically as follows: considering the situation that the active and reactive power of the distribution network are not decoupled, active-reactive coordinated optimization is performed, and the synchronous ADMM suitable for DGC is used for solving the problem. The distributed power generation cluster division result with multiple control objectives based on the day-ahead scheduling result is considered. The economic goal of minimum daily operating cost and the goal of minimum voltage deviation are considered, and the consistency cost is introduced. After the distribution network DGC is divided, the objective functions of each DGC are basically the same. The start and stop situation of MGT and the switching situation of CB are taken as fixed values, and the objective functions of each DGC are established as follows: In the formula, the superscript c represents the cth DGC; is the cost of purchasing electricity from the external network at time t; c buy (t), P buy (t) is the cost coefficient and power of electricity purchased from the external network at time t; is the MGT operating cost at time t; c gi,op 、c gi,of is the output cost parameter and start-stop cost parameter of the gi-th MGT; gi (t) is the output status of the gi-th MGT at time t; is the ESS operating cost at time t; c si is the charge and discharge loss cost parameter of the si-th ESS; P si,char (t), P si,disc (t) is the charge and discharge power of the si-th ESS at time t; is the network loss cost at time t; ij represents the branch with i as the head end and j as the end end; is the voltage deviation penalty cost of cluster c at time t; c V is the voltage deviation penalty cost coefficient; V i (t) is the voltage value of node i at time t; V i,ref (t) is the reference voltage value of node i at time t, i.e., the day-ahead scheduling value; is the consistency penalty cost of cluster c; ck is the set of adjacent DGCs of cluster ci; x ij′ (t) is the virtual tie line variable between cluster ci and its adjacent cluster cj at time t, is the reference value of the virtual tie line variable in the mth iteration at time t, and its initial value is the day-ahead scheduling value of the actual line ij; is the Lagrange multiplier between cluster ci and its adjacent cluster cj during the mth iteration; ρ is the iteration step size.

5. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 1 is characterized by: The constraints of each DGC dispatch optimization model in step S4 are basically consistent with the day-ahead global model, and the constraints include power flow constraints, power balance constraints, purchased power constraints, MGT operation constraints, ESS operation constraints, and SVC operation constraints; The power flow constraints are specifically: (I ij (t)) 2 =((P ij (t)) 2 +(Q ij (t)) 2 ) / V i (t),ij∈Nl(15), ((P ij (t)) 2 +(Q ij (t)) 2 )≤((I ij (t)) 2 (V i (t)) 2 ),ij∈Nl(18), where V i,min 、V i,max Respectively represent the minimum and maximum voltage amplitude of node i; V i (t) represents the voltage amplitude of node i at time t; r ij 、x ij Respectively represent the resistance and reactance of branch ij; P ij (t), Q ij (t) represent the active and reactive power flowing through branch ij at time t respectively; I ij,max Indicates the maximum current amplitude allowed to flow through branch ij; P j (t), Q j (t) represents the active and reactive power injected by node j; Nl represents the node set; m(j) and n(j) represent the end / starting node sets of the branch with node j as the starting / ending node, respectively.

6. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 5 is characterized by: The power balance constraints are specifically: Q Load,i (t) = Q buy,i (t)+Q CB,i (t)+Q SVC,i (t)+(I ij (t) 2 )x ij (20), where P Load,i (t), Q Load,i (t) respectively represent the active and reactive load power of node i at time t; P buy,i (t), Q buy,i (t) represents the active and reactive power purchased by node i from the upper power grid at time t. If the node is not connected to the upper power grid, its value is 0; P pv,i (t), P g,i (t) represents the output value of the DPV device and MGT device connected to node i at time t. If the node is not connected to the above devices, its value is 0; They represent the discharge and charging power of the ESS connected to node i at time t, respectively. If the node is not connected to the ESS, its value is 0; They represent the discharge and charging power of the EVCS connected to node i at time t. If the node is not connected to EVCS, its value is 0; Q CB,i (t), Q SVC,i (t) represents the reactive power output of the CB and SVC equipment connected to node i at time t. If the node is not connected to the above equipment, its value is 0.

7. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 5 is characterized by: The power purchase power constraint is specifically: In the formula, Indicates the maximum value of active power purchase; Respectively represent the minimum and maximum values ​​of reactive power purchase.

8. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 5 is characterized by: The MGT operation constraints are as follows: The MGT operation constraints in the intraday distributed scheduling include the upper and lower limits of the output and the maximum climbing / slope power constraints, as shown in the following formula. The start and stop states are determined by the day-ahead scheduling arrangement: Where Ng is the MGT set in ADN; P gi (t) is the output value of the gi-th MGT at time t; are the maximum climbing and sliding powers of the gi-th MGT, respectively; are the maximum and minimum outputs of the gi-th MGT respectively; T is the scheduling period; u gi (t) is the start / stop status of the gi-th MGT at time t.

9. The distributed rolling dispatching method for distribution network based on DGC division in a day as claimed in claim 5, characterized in that: The ESS operation constraints are specifically as follows: ESS constraints include charge and discharge power constraints and capacity constraints, namely: u si,char (t)+u si,disc (t)≤1,si∈Ns(27), Where Ns is the set of ESS in ADN; P si,char (t), P si,disc (t) are the charge and discharge power of the si-th ESS at time t; are the maximum charge and discharge power of the si-th ESS; u si,char (t),u si,disc (t) is the charge and discharge state 0-1 variable of the si-th ESS at time t, u si,disc (t) = 1, it is in the discharge state, u si,char (t) = 1, it is in charging state; are the maximum and minimum SOC of the si-th ESS respectively; E si , η si are the maximum capacity and charge / discharge efficiency of the si-th ESS respectively; E si (t) is the capacity and charge-discharge efficiency of the si-th ESS at time t.

10. The distributed rolling dispatching method for distribution network based on DGC division in a day according to claim 5, characterized in that: The SVC operation constraints are specifically: SVC constraints include SVC capacity constraints, namely: Where, Nsvc is the SVC set in ADN; They respectively represent the upper and lower limits of the SVC operating capacity of the svci station.

Images