Distributed photovoltaic aggregation equivalent model control parameter optimization method and system

Abstract

The invention provides a distributed photovoltaic aggregation equivalent model control parameter optimization method and system, and the method comprises the steps: introducing a dynamic response affinity factor, dividing A / B type photovoltaic clusters according to the dynamic response affinity factor, carrying out the multi-machine equivalence of the divided distributed photovoltaic clusters, and carrying out the multi-machine equivalence of the divided distributed photovoltaic clusters, carrying out secondary equivalence on the multi-machine equivalence result to obtain a secondary equivalence model; accessing the secondary equivalent model into a simulation model of a target power grid to form an equivalent power grid simulation platform; according to the equivalent power grid simulation platform, constructing a multi-dimensional optimization objective function according to voltage stability indexes of key nodes of a power grid; and optimizing the control parameters in the quadratic equivalent model according to the multi-dimensional optimization objective function. According to the method, systematic optimization of distributed photovoltaic cluster control parameters can be realized, so that the voltage stability of a power grid is actively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system modeling and simulation, and particularly relates to a distributed photovoltaic aggregation equivalent model control parameter optimization method and system. BACKGROUND

[0002] The large-scale access of distributed photovoltaics has profoundly changed the power flow distribution and dynamic characteristics of distribution networks and even transmission networks. Its inherent randomness, intermittency, and complex fault response behavior, such as the LVRT support of Class A photovoltaics and the active disconnection of Class B photovoltaics, have brought new challenges to the voltage stability of the power grid, especially the transient voltage stability.

[0003] At present, the power grid dispatching department generally adopts the mode of "observable but uncontrollable" or "extensive control" for distributed photovoltaics, and lacks fine control means. In terms of voltage stability control, it mainly relies on traditional reactive power compensation devices (such as SVC and STATCOM) and the automatic voltage regulator of synchronous generators. However, this control mode fails to fully tap the reactive voltage regulation potential of distributed photovoltaics, especially the large number of Class A photovoltaics. At the same time, the collective disconnection and recovery behavior of Class B photovoltaics during faults can easily cause a secondary voltage drop or oscillation, and existing control strategies are difficult to effectively guide and optimize.

[0004] The root cause lies in the lack of a model that can accurately represent the dynamics of photovoltaic clusters while being simplified enough for optimization calculations. Existing detailed models are too complex and have heavy optimization calculation burdens; while overly simplified models cannot reflect the real relationship between control parameters and system stability. Therefore, there is an urgent need for a technical solution that can bridge "high-precision modeling" and "high-efficiency optimization" to systematically optimize the control parameters of distributed photovoltaic clusters and thus actively improve the voltage stability of the power grid. SUMMARY

[0005] The present application aims to provide a distributed photovoltaic aggregation equivalent model control parameter optimization method and system, which aims to solve the problem that the existing technology cannot effectively utilize the distributed photovoltaic cluster for voltage stability optimization.

[0006] In a first aspect, the present application provides a distributed photovoltaic aggregation equivalent model control parameter optimization method, which comprises:

[0007] A dynamic response affinity factor is introduced, and the A / B photovoltaic cluster is divided according to the dynamic response affinity factor, and the divided distributed photovoltaic cluster is multi-machine equivalent, and the multi-machine equivalent result is twice equivalent, to obtain a twice equivalent model;

[0008] The twice equivalent model is connected to the simulation model of the target power grid to form an equivalent power grid simulation platform;

[0009] Based on the equivalent power grid simulation platform, a multi-dimensional optimization objective function is constructed using the voltage stability index of key power grid nodes;

[0010] The control parameters in the quadratic isometry model are optimized based on the multidimensional optimization objective function.

[0011] In some embodiments, the step of introducing a dynamic response affinity factor includes:

[0012] The dynamic response affinity factor is calculated using the following formula:

[0013]

[0014] Where, β a For the dynamic response affinity factor, ΔQ a (t) represents the reactive power change of the a-th distributed photovoltaic inverter, ΔU a (t) represents the voltage change at the grid connection point of the a-th distributed photovoltaic inverter, and T is the set transient process observation time window.

[0015] In some embodiments, the step of classifying photovoltaic clusters into A / B categories based on the dynamic response affinity factor includes:

[0016] Calculate the grid-connected equivalent impedance of each distributed photovoltaic inverter using the following formula:

[0017]

[0018] Among them, Z eq，a Let Ia be the grid-connected equivalent impedance of the a-th distributed photovoltaic inverter. The first to n distributed photovoltaic inverters are radially connected, and the (n+1)th to (n+m)th distributed photovoltaic inverters are trunk-connected. I0 and Z0 are the total line current and line impedance of the first layer, respectively. a Z a Let I be the line current and line impedance of the line connected to the a-th distributed photovoltaic inverter, respectively; j Z represents the line current of the line connected to the j-th distributed photovoltaic unit; i Let be the line impedance of the line connected to the i-th photovoltaic unit;

[0019] Using the hybrid feature vector F of each distributed photovoltaic inverter a =[Z eq,a ,β a As a clustering index, the Canopy-FCM clustering algorithm was used to perform multi-dimensional feature fusion clustering for Class A distributed photovoltaic inverters with low-voltage ride-through capability and Class B distributed photovoltaic inverters with low-voltage blocking capability, respectively, to obtain c.A one A-type distributed photovoltaic cluster, B one B-type distributed photovoltaic cluster.

[0020] In some embodiments, the step of performing multi-machine equivalence on the divided distributed photovoltaic cluster and performing secondary equivalence on the multi-machine equivalence result to obtain a secondary equivalence model comprises:

[0021] The dynamic voltage difference between each photovoltaic power generation unit in the cluster c1 and the PCC is calculated according to the following formula

[0022] wherein, Z f , P f are the line impedance of the line connected to the fth distributed photovoltaic inverter in the cluster and the active power output by the photovoltaic power generation unit, respectively, U is the PCC point voltage, γ f is a dynamic response coefficient representing the dynamic influence of the power change of the unit on the voltage dynamic characteristic;

[0023] The weighted average voltage difference between each photovoltaic power generation unit in the equivalent cluster c1 and the PCC is calculated and the voltage difference ΔU eq，1 between the equivalent photovoltaic power generation unit after equivalence and the PCC is calculated:

[0024]

[0025] wherein, N1 is the number of photovoltaic power generation units in the cluster c1, Z eq，1 is the equivalent line impedance of the cluster c1;

[0026] By simplifying, the dynamic aggregation impedance network of the cluster c1 is obtained:

[0027]

[0028] The equivalent line impedance of each cluster is calculated by repeating the above steps, and a secondary equivalence model is constructed.

[0029] In some embodiments, the step of constructing a multi-dimensional optimization objective function according to the equivalent power grid simulation platform with the voltage stability index of the key nodes of the power grid comprises:

[0030] The 220kV bus, 110kV bus and 10kV bus of the 220kV substation are taken as key buses, and the average f1(X) of the transient voltage energy deficiency of the key buses is calculated according to the following formula:

[0031]

[0032] wherein, U i(t) represents the time series of the per-unit voltage values ​​of the i-th critical bus, U 0，i T represents the steady-state voltage of the bus before the fault. d The duration of the voltage drop is 0.5 seconds after the fault is cleared, and N is the number of critical buses.

[0033] The mean dynamic recovery response of the critical busbar is calculated using the following formula:

[0034]

[0035] Among them, T recovery，i This is the time required for the voltage of the i-th critical bus to stabilize back to its initial voltage after the fault is cleared.

[0036] The mean convergence of transient oscillations of the critical busbar is calculated using the following formula:

[0037]

[0038] Among them, U i The voltage sequence after the fault ends, U i_0 This is the steady-state voltage value;

[0039] Construct a multi-dimensional optimization objective function based on the following formula:

[0040]

[0041] Among them, U lim As the lower limit of voltage safety, min(U) i ) represents the minimum voltage of the bus during the simulation period, β is the safety priority coefficient, and X is the control parameter vector to be optimized.

[0042] In some embodiments, the step of optimizing the control parameters in the quadratic isometry model according to the multidimensional optimization objective function includes:

[0043] Let the optimization variable vector be:

[0044] X = [x1, x2, ..., x7] = [K] qv K iq U set U block T block K p K q ];

[0045] Among them, K qv K is the reactive power-voltage droop coefficient for Class A equivalent machines. qv ∈[1, 10]; K iq K represents the reactive current coefficient during the LVRT of a Class A equivalent machine. iq∈[0.5, 5.0]; U set For Class A equivalent LVRT start-up voltage threshold, U set ∈[0.85,0.95] in units of pu; U block For Class B equivalent machine-sealed voltage threshold, U block ∈[0.75,0.88] units pu; T block For the duration of the wave sealing of Class B equivalent machine, T block ∈[0.1,2.0] in units of seconds; K p K represents the active power recovery slope of the Class B equivalent machine. p ∈ [0.05, 0.5] in units of pu / s; K q K represents the reactive power recovery slope of the Class B equivalent machine. q ∈[0.05,0.5], unit is pu / s;

[0046] The particle swarm optimization algorithm is used to optimize the variable to be optimized.

[0047] In some embodiments, the Canopy-FCM clustering algorithm is used to perform multi-dimensional feature fusion clustering on Class A distributed photovoltaic inverters with low-voltage ride-through capability and Class B distributed photovoltaic inverters with low-voltage blocking capability, respectively, to obtain c A One Class A distributed photovoltaic cluster, c B The steps for a Class B distributed photovoltaic cluster include:

[0048] The grid-connected equivalent impedance of each distributed photovoltaic inverter is taken as a data point, and the real part of the impedance, resistance R, and the imaginary part, reactance X, are taken as a two-dimensional data point p. i =(R i ,X i ), and select the first and second distance thresholds T1 and T2, where T1>T2;

[0049] Create an empty list to store the generated Canopy, and copy the original dataset as the dataset D to be processed.

[0050] Randomly select a center point, and randomly select a data point P from the dataset D to be processed as the center of the new Canopy; create a Canopy, calculate the Euclidean distance from all other data points in D to point P, include all data points whose distance is less than the first threshold T1 into the Canopy centered at P, and remove all data points whose distance is less than the second threshold T2 from the dataset D to be processed; add the newly generated Canopy to the Canopy list.

[0051] Finally, a set of Canopy clusters is obtained, with each Canopy having a center point, which is used as the initial cluster center for the FCM algorithm;

[0052] Set the final number of clusters c obtained from Canopy coarse clustering as the number of clusters in FCM, use the Canopy center point as the initial cluster center V0, and set the fuzzy coefficient m = 2 and the iteration termination condition ε.

[0053] Based on the current cluster centers, calculate the membership degree of each data point d to each cluster center e;

[0054] Based on the current membership matrix, recalculate the centroid of each cluster, where V is the set of cluster centroid groups.

[0055] Calculate the Euclidean distance difference between the new cluster center V′ and the previous cluster center V. If ||V′-V||<ε, the algorithm converges and stops iterating; otherwise, iterate.

[0056] Output the final cluster center V end Membership matrix U end And based on the membership matrix, each data point is assigned to the Canopy center with the highest membership degree, completing the hard partition and obtaining c. A One Class A distributed photovoltaic cluster, c B A Class B distributed photovoltaic cluster.

[0057] Secondly, the present invention provides a system for optimizing control parameters of a distributed photovoltaic aggregation equivalent model, the system comprising:

[0058] The model building module is used to introduce a dynamic response affinity factor, divide A / B type photovoltaic clusters according to the dynamic response affinity factor, perform multi-machine equivalence on the divided distributed photovoltaic clusters, and perform secondary equivalence on the multi-machine equivalence results to obtain a secondary equivalence model.

[0059] The simulation platform construction module is used to connect the secondary equivalent model to the simulation model of the target power grid to form an equivalent power grid simulation platform;

[0060] The objective function construction module is used to construct a multi-dimensional optimization objective function based on the voltage stability index of key nodes in the power grid, according to the equivalent power grid simulation platform.

[0061] The parameter optimization module is used to optimize the control parameters in the quadratic isometry model according to the multi-dimensional optimization objective function.

[0062] Thirdly, the present invention provides a storage medium that stores one or more programs, which, when executed by a processor, implement the above-described method for optimizing control parameters of a distributed photovoltaic aggregation equivalent model.

[0063] Fourthly, the present invention provides an electronic device, the electronic device comprising a memory and a processor, wherein:

[0064] The memory is used to store computer programs;

[0065] When the processor executes the computer program stored in the memory, it implements the above-mentioned method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model.

[0066] Compared with the prior art, the present invention has the following advantages:

[0067] 1. High control precision: The optimization process is based on a high-precision equivalent model, which ensures that the optimization results can effectively improve the voltage stability of the actual system.

[0068] 2. High optimization efficiency: Optimization calculations are performed using a simplified equivalent model (only two equivalent machines, A and B), which greatly reduces computational complexity and time costs, making online or near-online optimization possible.

[0069] 3. Effective risk avoidance: By optimizing the blocking and recovery parameters of Class B photovoltaic systems, their collective behavior can be smoothed, effectively avoiding secondary impacts on the system.

[0070] 4. A quadratic equivalence strategy is adopted to gradually simplify the model structure, balancing simulation efficiency and accuracy. Attached Figure Description

[0071] Figure 1 This is a flowchart of a method for optimizing control parameters of a distributed photovoltaic aggregation equivalent model according to an embodiment of the present invention;

[0072] Figure 2 A detailed model of a distributed photovoltaic system;

[0073] Figure 3 A schematic diagram of a photovoltaic power generation cluster multi-machine equivalent model;

[0074] Figure 4 Voltage comparison chart of simulation results from detailed model and quadratic equivalent model of distributed photovoltaic system;

[0075] Figure 5 A comparison chart of power simulation results between the detailed model and the quadratic equivalent model of a distributed photovoltaic system;

[0076] Figure 6 Voltage comparison chart of simulation results between equivalent model and equivalent optimized model of distributed photovoltaic system;

[0077] Figure 7 A comparison chart of power simulation results between the equivalent model and the equivalent optimized model of a distributed photovoltaic system;

[0078] Figure 8 This is a schematic diagram of the structure of the distributed photovoltaic aggregation equivalent model control parameter optimization system proposed in an embodiment of the present invention.

[0079] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0080] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects.

[0081] like Figure 1 As shown, an embodiment of the present invention proposes a method for optimizing control parameters of a distributed photovoltaic aggregation equivalent model. This method includes steps S101 to S104, wherein:

[0082] Step S101: Introduce a dynamic response affinity factor, divide A / B type photovoltaic clusters according to the dynamic response affinity factor, perform multi-machine equivalence on the divided distributed photovoltaic clusters, and perform secondary equivalence on the multi-machine equivalence results to obtain a secondary equivalence model.

[0083] It should be noted that in this step, it is necessary to collect data such as line impedance parameters, transformer parameters, rated capacity of distributed photovoltaic inverters, grid connection level and inverter type of the distributed photovoltaic area to be equivalently calculated. Based on the intelligent division of photovoltaic clusters with dynamic response affinity and secondary dynamic equivalent modeling, an equivalent model that accurately reflects the dynamic characteristics of A / B type photovoltaic clusters is constructed.

[0084] Specifically, firstly, based on the power grid topology, the grid-connected equivalent impedance of each distributed photovoltaic inverter is calculated using equation (1):

[0085]

[0086] In equation (1): Z eq，aLet Ia be the grid-connected equivalent impedance of the a-th distributed photovoltaic inverter, where the first to n distributed photovoltaic inverters are radially connected, and the (n+1)th to (n+m)th distributed photovoltaic inverters are trunk-connected; I0 and Z0 are the total line current and line impedance of the first layer, respectively; Ia a Z a Let I be the line current and line impedance of the line connected to the a-th distributed photovoltaic inverter, respectively; j Z represents the line current of the line connected to the j-th distributed photovoltaic unit; i Let be the line impedance of the line connected to the i-th photovoltaic unit.

[0087] Calculate the dynamic response affinity factor β a :

[0088]

[0089] In equation (2): ΔQ a (t) represents the reactive power change of the a-th distributed photovoltaic inverter; ΔU a (t) represents the voltage change at the grid connection point of the a-th distributed photovoltaic inverter; T is the set transient process observation time window.

[0090] Then, the hybrid feature vector F of each distributed photovoltaic inverter is used. a =[Z eq,a ,β a As a clustering index, the Canopy-FCM clustering algorithm was used to perform multi-dimensional feature fusion clustering for Class A distributed photovoltaic inverters with low-voltage ride-through capability and Class B distributed photovoltaic inverters with low-voltage blocking capability, respectively, to obtain c. A One Class A distributed photovoltaic cluster, c B A Class B distributed photovoltaic cluster.

[0091] Furthermore, in some embodiments, the partitioning process is as follows:

[0092] a. Data Processing: The grid-connected equivalent impedance of each distributed photovoltaic inverter is taken as a data point, and the real part of the impedance, resistance R, and the imaginary part, reactance X, are taken as a two-dimensional data point p. i =(R i ,X i And select the first and second distance thresholds T1 and T2, where T1>T2;

[0093] b. Initialization: Create an empty list to store the generated Canopy, and copy the original dataset as the dataset D to be processed;

[0094] c. Execute the following loop until dataset D is empty: ① Randomly select a center point. Randomly select a data point P from the dataset D to be processed as the center of the new Canopy; ② Create a Canopy. Calculate the Euclidean distance from all other data points in D to point P. Include all data points whose distance is less than the first threshold T1 into this Canopy centered at P. Remove all data points whose distance is less than the strict threshold T2 from the dataset D to be processed; ③ Record the Canopy. Add the newly generated Canopy to the Canopy list.

[0095] d. Output the number of clusters c: The final result is a set of Canopies, each with a centroid. These centroids can be used as the initial cluster centers for the FCM algorithm. The number of Canopies is the final number of clusters c.

[0096] e. FCM clustering initialization: Set the final number of clusters c obtained from Canopy coarse clustering as the number of clusters in FCM, use the Canopy center point as the initial cluster center V0, and set the fuzzy coefficient m = 2 and the iteration termination condition ε.

[0097] f. Update the membership matrix U: Based on the current cluster centers, calculate the membership degree of each data point d to each cluster center e according to equation (3). The calculation formula is as follows:

[0098]

[0099] In equation (3): u d，e For data point x d The membership values ​​belonging to the cluster centers of the e-th cluster form the membership matrix U, and the sum of the membership values ​​of each data point in the dataset D is 1; x d =(R d X d );v e =(R e ,X e ); m is the fuzzy coefficient, here we take m = 2; ||x d -v e ||For x d With the center point v of class e e Euclidean distance between; ||x d -v k ||For x d With the center point v of the kth class k Euclidean distance between them;

[0100] g. Update cluster centers V: Based on the current membership matrix, recalculate the center point of each cluster, where V is the set of cluster centers. The new center points are the weighted average of all data points, with weights equal to the m-th power of the membership degree.

[0101]

[0102] h. Determine if convergence: Calculate the Euclidean distance difference between the new cluster center V′ and the previous cluster center V. If ||V′-V|| < ε, the algorithm converges and stops iterating; otherwise, return to step b to continue iterating.

[0103] i. Output: Output the final cluster centers V end Membership matrix U end And based on the membership matrix, each data point is assigned to the Canopy center with the highest membership degree, completing the hard partition and obtaining c. A One Class A distributed photovoltaic cluster, c B A Class B distributed photovoltaic cluster.

[0104] Finally, the intelligent partitioning and secondary dynamic equivalent modeling method of photovoltaic clusters based on dynamic response affinity is used to perform multi-machine equivalent modeling on the distributed photovoltaic cluster, and the results of multi-machine equivalent modeling are then subjected to secondary equivalent modeling to obtain a secondary equivalent model consisting of one type A distributed photovoltaic equivalent model and one type B distributed photovoltaic equivalent model.

[0105] In some embodiments, the construction process of the quadratic equivalent model is as follows:

[0106] a. Use parameter identification methods to identify the simulation data of each cluster and obtain the equivalent control parameters of each cluster;

[0107] b. Considering the dynamic response characteristics, perform dynamic voltage-power sensitivity calculations, and calculate the dynamic voltage difference between each photovoltaic power generation unit in cluster c1 and the PCC.

[0108]

[0109] In equation (5): Z f P f γ represents the line impedance of the line connected to the f-th distributed photovoltaic inverter in this cluster and the active power output of the photovoltaic power generation unit, respectively; U is the voltage at point PCC; f The dynamic response coefficient is used to characterize the dynamic effect of power changes on voltage in this unit.

[0110] c. Calculate the weighted average voltage difference between each photovoltaic power generation unit and the PCC in the equivalent cluster c1. And the voltage difference ΔU between the equivalent photovoltaic power generation unit and the PCC after the conversion. eq，1 :

[0111]

[0112] In the formula: N1 is the number of photovoltaic power generation units in cluster c1; Z eq，1 This is the equivalent line impedance for cluster c1.

[0113] d. Since the voltage difference between the equivalent photovoltaic power generation unit of cluster c1 and the first N1 equivalent photovoltaic units of cluster c1 and the PCC is the same, that is... The simplified result is the dynamic aggregated impedance network of cluster c1:

[0114]

[0115] e. Repeat the above steps to calculate the equivalent line impedance of each cluster and construct a multi-machine equivalent model.

[0116] f. Apply system-level disturbances to the equivalent model of multi-machine type A and the equivalent model of multi-machine type B respectively, record the electrical quantities of the equivalent machine output, and use the multi-convolutional layer hybrid convolutional neural network algorithm to perform secondary parameter identification on the simulation results of the equivalent model of multi-machine type A and the equivalent model of distributed photovoltaic type B respectively, so as to obtain the equivalent control parameters of single machine type A and the equivalent control parameters of single machine type B. The equivalent line impedance parameters of single machine are obtained by equation (8).

[0117] Step S102: Integrate the secondary equivalent model into the simulation model of the target power grid to form an equivalent power grid simulation platform;

[0118] Step S103: Based on the equivalent power grid simulation platform, construct a multi-dimensional optimization objective function using the voltage stability index of key power grid nodes;

[0119] It should be noted that in this step, the 220kV bus, 110kV bus, and 10kV bus of the 220kV substation are first taken as critical buses, and the mean value of transient voltage energy deficit f1(X) of the critical buses is calculated:

[0120]

[0121] In equation (9): U i (t) represents the time series of the per-unit voltage values ​​of the i-th critical bus; U 0，i The steady-state voltage of the bus before the fault (typically 1.0 pu); T d The main duration of the voltage drop is 0.5 seconds after the fault is cleared; N is the number of critical buses (N=3).

[0122] Then, the mean dynamic recovery response of the key busbar is calculated:

[0123]

[0124] In equation (10): T recovery，i This is the time required for the voltage of the i-th critical bus to stabilize back to its initial voltage after the fault is cleared.

[0125] Next, calculate the mean convergence of the transient oscillations of the key bus:

[0126]

[0127] In equation (11): U i The voltage sequence after the fault ends, U i_0 This is the steady-state voltage value.

[0128] Finally, the multi-dimensional optimization objective function F(X) based on the aggregation of dynamic process features is calculated:

[0129]

[0130] In equation (12): U lim As the lower limit of voltage safety, according to "GB / T 29319-2024 Technical Regulations for Photovoltaic Power Generation Systems Access to Distribution Networks", 0.85pu is taken; min(U i ) represents the minimum voltage of the bus during the simulation period; β is the safety priority coefficient, used to balance the relationship between safety constraints and performance optimization, and is set to 0.4; X is the control parameter vector to be optimized.

[0131] Step S104: Optimize the control parameters in the quadratic isometry model according to the multi-dimensional optimization objective function.

[0132] In the optimization process, we first define the optimization variable vector as follows:

[0133] X = [x1, x2, ..., x7] = [K] qv K iq U set U block T block K p K q (13)

[0134] Where: K qv K is the reactive power-voltage droop coefficient for Class A equivalent machines. qv ∈[1, 10]; K iq K represents the reactive current coefficient during the LVRT of a Class A equivalent machine. iq ∈[0.5, 5.0]; U set For Class A equivalent LVRT start-up voltage threshold, U set ∈[0.85, 0.95] in units of pu; U block For Class B equivalent machine-sealed voltage threshold, U block ∈[0.75,0.88] units pu; T block For the duration of the wave sealing of Class B equivalent machine, T block∈[0.1,2.0] in units of seconds; K p K represents the active power recovery slope of the Class B equivalent machine. p ∈[0.05,0.5] in units of pu / s; K q K represents the reactive power recovery slope of the Class B equivalent machine. q ∈[0.05,0.5] The unit is pu / s.

[0135] Then, the variable to be optimized is optimized using the conventional particle swarm optimization algorithm;

[0136] Finally, the optimization results were verified by setting the optimal parameter X. end By incorporating the data into the equivalent model and comparing the voltage response before and after optimization under the same disturbance scenario, the degree of improvement in various voltage stability indicators can be quantitatively calculated.

[0137] For example, 1. Collect the resistance R and reactance X parameters (unit: Ω / km) of all relevant 220kV, 110kV, 35kV, and 10kV lines in the area, as well as the short-circuit impedance, no-load loss, rated capacity, and turns ratio of the distribution transformers. Also collect the grid node corresponding to the installation location of each distributed photovoltaic inverter, its rated capacity (kVA), grid connection type (clearly distinguishing between Class A and Class B distributed photovoltaic inverters), and the inverter's factory control parameters. All line parameters are then reduced to the same voltage level (220kV side), and a detailed simulation model of the area is built in PSASP. The detailed model of the distributed photovoltaic system can be found here. Figure 2 .

[0138] 2. Calculate the grid-connected equivalent impedance of each distributed photovoltaic inverter according to step 1.1. The calculation results are shown in Table 1.

[0139] Table 1 Equivalent Impedance of Distributed Photovoltaic Inverters Connected to the Grid

[0140]

[0141]

[0142] 3. Following step 1.2, the Canopy-FCM clustering algorithm was used to group the distributed photovoltaic inverters of categories A and B respectively. The grouping results are shown in Table 2.

[0143] Table 2. Results of Distributed Photovoltaic Inverter Clustering

[0144] Cluster Photovoltaic power unit number Cluster total capacity / MVA c A1 ]]> 1、3 14 c A2 ]]> 5、7 13.5 c A3 ]]> 10 10 c B1 ]]> 2、4 3.7 c B2 ]]> 6、8、9 2 c B3 ]]> 11、12、14 3.45 c B4 ]]> 13、15 0.77

[0145] 4. Following step 1.3, for each cluster obtained in step 1.2, build a detailed simulation model in PSASP software, apply voltage dip disturbances, and set three operating conditions to momentarily drop the grid connection point voltage to 0.2 pu, 0.5 pu, and 0.8 pu, lasting for 0.15 seconds before recovering. Obtain transient response data such as voltage, current, active power, and reactive power at the cluster outlet. Use a multi-convolutional layer hybrid convolutional neural network algorithm to identify the control parameters of the equivalent model for each cluster. Using minimizing the error between the detailed simulation data and the equivalent simulation data of each cluster as the objective function, finally obtain the PI controller parameters and low-voltage ride-through control parameters of the Class A equivalent machine, and the wave blocking duration T and active current recovery slope K of the Class B equivalent machine. p Reactive current recovery slope K q The equivalent line impedance of each cluster is calculated, and a multi-machine equivalent model is constructed, as shown in the model below. Figure 3 As shown.

[0146] 5. Construct a multi-machine equivalent model. Apply system-level disturbances to the Type A and Type B multi-machine equivalent models constructed in step 1.3, record the electrical quantities at the output of the equivalent machines, and use a multi-convolutional layer hybrid convolutional neural network algorithm to perform secondary parameter identification on the simulation results of the Type A and Type B distributed photovoltaic multi-machine equivalent models, respectively, to obtain a set of Type A single-machine equivalent control parameters and Type B single-machine equivalent control parameters for the final simplified model, and calculate their secondary equivalent impedance.

[0147] 6. Build a simplified model containing only one Class A and one Class B duty machine.

[0148] 7. Define a multi-dimensional optimization objective function F(X) based on dynamic process feature aggregation:

[0149] 7.1 Taking the 220kV busbar, 110kV busbar, and 10kV busbar of a 220kV substation as critical busbars, calculate the average transient voltage energy deficit value f1(X) of the critical busbars:

[0150]

[0151] In the formula: U i (t) represents the time series of the per-unit voltage values ​​of the i-th critical bus; U 0,i The steady-state voltage of the bus before the fault (typically 1.0 pu); T d The main duration of the voltage drop is 0.5 seconds after the fault is cleared; N is the number of critical buses (N=3).

[0152] 7.2 Calculate the mean dynamic recovery response of the critical busbar:

[0153]

[0154] In the formula: T recovery，i This is the time required for the voltage of the i-th critical bus to stabilize back to its initial voltage after the fault is cleared.

[0155] 7.3 Calculate the mean convergence of transient oscillations along the critical busbar:

[0156]

[0157] In the formula: U i The voltage sequence after the fault ends, U i_0 This is the steady-state voltage value.

[0158] 7.4 Calculate the multi-dimensional optimization objective function F(X) based on dynamic process feature aggregation:

[0159]

[0160] In the formula: U lim As the lower limit of voltage safety, according to "GB / T 29319-2024 Technical Regulations for Photovoltaic Power Generation Systems Access to Distribution Networks", 0.85pu is taken; min(U i ) represents the minimum voltage of the bus during the simulation period; β is the safety priority coefficient, used to balance the relationship between safety constraints and performance optimization, and is set to 0.4; X is the control parameter vector to be optimized.

[0161] 8. Let the optimization variable vector be:

[0162] X = [x1, x2, ..., x7] = [K] qv K iq U set U block T block K p K q (13)

[0163] 9. Use the particle swarm optimization algorithm to optimize the variables to be optimized:

[0164] a. Initialize the particle swarm optimization algorithm: Set the population size to 20-50, and the particle position to X. i (0) = X min +rand(0,1)×(X) max -X min The particle velocity is V. i (0) = 0.1·rand(-1, 1)×(X) max -X min );

[0165] b. Iterative updates;

[0166] b1. Speed ​​Update:

[0167] V i (k+1)=ω·V i (k)+c1·r1·(P best，i -X i (k))+c2·r2·(G best -X i (k))

[0168] b2. Location Update:

[0169] X i (k+1)=X i (k)+V i (k+1)

[0170] c. Complete the iteration and obtain the optimization results of the control parameters, as shown in Table 3.

[0171] Table 3 Optimization Results of Distributed Photovoltaic Control Parameters

[0172]

[0173] 10. Optimize the results and verify the optimal parameter X. end By incorporating the data into the equivalent model and comparing the voltage response before and after optimization under the same disturbance scenario, the degree of improvement in various voltage stability indicators is calculated. Detailed comparisons of voltage and power between the model and the quadratic equivalent model can be found in [link to relevant documentation]. Figure 4 , Figure 5 As shown in Table 4, a comparison table of power grid transient stability evaluation indicators before and after optimization is presented. The comparison results of voltage and power before and after optimization are shown in [the table below]. Figure 6 , Figure 7 As shown, the effectiveness of the proposed optimized control strategy is verified.

[0174] Table 4 Comparison of Power Grid Transient Stability Evaluation Indicators Before and After Optimization

[0175]

[0176] In summary, the above-mentioned method for optimizing control parameters in the distributed photovoltaic aggregation equivalent model has the following advantages:

[0177] 1. High control precision: The optimization process is based on a high-precision equivalent model, which ensures that the optimization results can effectively improve the voltage stability of the actual system.

[0178] 2. High optimization efficiency: Optimization calculations are performed using a simplified equivalent model (only two equivalent machines, A and B), which greatly reduces computational complexity and time costs, making online or near-online optimization possible.

[0179] 3. Effective risk avoidance: By optimizing the blocking and recovery parameters of Class B photovoltaic systems, their collective behavior can be smoothed, effectively avoiding secondary impacts on the system.

[0180] 4. A quadratic equivalence strategy is adopted to gradually simplify the model structure, balancing simulation efficiency and accuracy.

[0181] like Figure 8 As shown, this invention proposes a control parameter optimization system for a distributed photovoltaic aggregation equivalent model, the system comprising:

[0182] The model building module 10 is used to introduce a dynamic response affinity factor, divide A / B type photovoltaic clusters according to the dynamic response affinity factor, perform multi-machine equivalence on the divided distributed photovoltaic clusters, and perform secondary equivalence on the multi-machine equivalence results to obtain a secondary equivalence model.

[0183] The simulation platform construction module 20 is used to connect the secondary equivalent model to the simulation model of the target power grid to form an equivalent power grid simulation platform;

[0184] The objective function construction module 30 is used to construct a multi-dimensional optimization objective function based on the voltage stability index of key nodes in the power grid according to the equivalent power grid simulation platform.

[0185] The parameter optimization module 40 is used to optimize the control parameters in the quadratic isometry model according to the multi-dimensional optimization objective function.

[0186] In another aspect, the present invention also proposes a storage medium on which one or more programs are stored, which, when executed by a processor, implement the above-described method for optimizing control parameters of a distributed photovoltaic aggregation equivalent model.

[0187] In another aspect, the present invention also proposes an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to realize the above-mentioned method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model.

[0188] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain stored, communicated, propagated, or transmitted programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0189] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0190] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0191] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A method for optimizing control parameters in a distributed photovoltaic aggregation equivalent model, characterized in that, The method includes: A dynamic response affinity factor is introduced, and photovoltaic clusters of type A / B are divided according to the dynamic response affinity factor. Multi-machine equivalence is performed on the divided distributed photovoltaic clusters, and the multi-machine equivalence results are then subjected to secondary equivalence to obtain a secondary equivalence model. The secondary equivalent model is integrated into the simulation model of the target power grid to form an equivalent power grid simulation platform; Based on the equivalent power grid simulation platform, a multi-dimensional optimization objective function is constructed using the voltage stability index of key power grid nodes; The control parameters in the quadratic isometry model are optimized based on the multidimensional optimization objective function.

2. The method for optimizing control parameters of a distributed photovoltaic aggregation equivalent model according to claim 1, characterized in that, The step of introducing a dynamic response affinity factor includes: The dynamic response affinity factor is calculated using the following formula: Where, β a For the dynamic response affinity factor, ΔQ a (t) represents the reactive power change of the a-th distributed photovoltaic inverter, ΔU a (t) represents the voltage change at the grid connection point of the a-th distributed photovoltaic inverter, and T is the set transient process observation time window.

3. The method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model according to claim 2, characterized in that, The step of classifying photovoltaic clusters into A / B categories based on the dynamic response affinity factor includes: Calculate the grid-connected equivalent impedance of each distributed photovoltaic inverter using the following formula: Among them, Z eq，a Let Ia be the grid-connected equivalent impedance of the a-th distributed photovoltaic inverter. The first to n distributed photovoltaic inverters are radially connected, and the (n+1)th to (n+m)th distributed photovoltaic inverters are trunk-connected. I0 and Z0 are the total line current and line impedance of the first layer, respectively. a Z a Let I be the line current and line impedance of the line connected to the a-th distributed photovoltaic inverter, respectively; j Z represents the line current of the line connected to the j-th distributed photovoltaic unit; i Let be the line impedance of the line connected to the i-th photovoltaic unit; Using the hybrid feature vector F of each distributed photovoltaic inverter a =[Z eq,a ,β a As a clustering index, the Canopy-FCM clustering algorithm was used to perform multi-dimensional feature fusion clustering for Class A distributed photovoltaic inverters with low-voltage ride-through capability and Class B distributed photovoltaic inverters with low-voltage blocking capability, respectively, to obtain c. A One Class A distributed photovoltaic cluster, c B A Class B distributed photovoltaic cluster.

4. The method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model according to claim 3, characterized in that, The steps of performing multi-machine equivalent evaluation on the divided distributed photovoltaic clusters and then performing secondary equivalent evaluation on the multi-machine equivalent evaluation results to obtain a secondary equivalent evaluation model include: The dynamic voltage difference between each photovoltaic power generation unit and the PCC in cluster c1 is calculated using the following formula. Among them, Z f P f These represent the line impedance of the line connected to the f-th distributed photovoltaic inverter in this cluster and the active power output of that photovoltaic power generation unit, respectively; U is the voltage at point PCC; γ f The dynamic response coefficient is used to characterize the dynamic effect of power changes on voltage in this unit. Calculate the weighted average voltage difference between each photovoltaic power generation unit and the PCC in the equivalent cluster c1. And the voltage difference ΔU between the equivalent photovoltaic power generation unit and the PCC after the conversion. eq,1 : Where N1 is the number of photovoltaic power generation units in cluster c1, Z eq,1 The equivalent line impedance for cluster c1; Depend on The simplified result is the dynamic aggregated impedance network of cluster c1: Repeat the above steps to calculate the equivalent line impedance for each cluster and construct a quadratic equivalent model.

5. The method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model according to claim 4, characterized in that, The step of constructing a multi-dimensional optimization objective function based on the voltage stability index of key nodes in the power grid according to the equivalent power grid simulation platform includes: Taking the 220kV busbar, 110kV busbar, and 10kV busbar of a 220kV substation as critical busbars, the mean transient voltage energy deficit value f1(X) of the critical busbars is calculated according to the following formula: Among them, U i (t) represents the time series of the per-unit voltage values ​​of the i-th critical bus, U 0，i T represents the steady-state voltage of the bus before the fault. d The duration of the voltage drop is 0.5 seconds after the fault is cleared, and N is the number of critical buses. The mean dynamic recovery response of the critical busbar is calculated using the following formula: Among them, T recovery，i This is the time required for the voltage of the i-th critical bus to stabilize back to its initial voltage after the fault is cleared. The mean convergence of transient oscillations of the critical busbar is calculated using the following formula: Among them, U i The voltage sequence after the fault ends, U i_0 This is the steady-state voltage value; Construct a multi-dimensional optimization objective function based on the following formula: Among them, U lim As the lower limit of voltage safety, min(U) i ) represents the minimum voltage of the bus during the simulation period, β is the safety priority coefficient, and X is the control parameter vector to be optimized.

6. The method for optimizing control parameters of a distributed photovoltaic aggregation equivalent model according to claim 1, characterized in that, The step of optimizing the control parameters in the quadratic isometry model according to the multi-dimensional optimization objective function includes: Let the optimization variable vector be: X=[x1,x2,...,x7]=[K qv ,K iq ,U set ,U block ,T block ,K p ,K q ]; Among them, K qv K is the reactive power-voltage droop coefficient for Class A equivalent machines. qv ∈[1,10];K iq K represents the reactive current coefficient during the LVRT of a Class A equivalent machine. iq ∈[0.5, 5.0]; U set For Class A equivalent LVRT start-up voltage threshold, U set ∈[0.85, 0.95] in units of pu; U block For Class B equivalent machine shielding voltage threshold, U block ∈[0.75, 0.88] units pu; T block For the duration of the wave sealing of Class B equivalent machine, T block ∈[0.1,2.0] in units of seconds; K p K represents the active power recovery slope of the Class B equivalent machine. p ∈ [0.05, 0.5] in units of pu / s; K q K represents the reactive power recovery slope of the Class B equivalent machine. q ∈[0.05, 0.5], unit is pu / s; The particle swarm optimization algorithm is used to optimize the variable to be optimized.

7. The method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model according to claim 3, characterized in that, The Canopy-FCM clustering algorithm is used to perform multi-dimensional feature fusion clustering on Class A distributed photovoltaic inverters with low-voltage ride-through capability and Class B distributed photovoltaic inverters with low-voltage blocking capability, respectively, to obtain c A One Class A distributed photovoltaic cluster, c B The steps for a Class B distributed photovoltaic cluster include: The grid-connected equivalent impedance of each distributed photovoltaic inverter is taken as a data point, and the real part of the impedance, resistance R, and the imaginary part, reactance X, are taken as a two-dimensional data point p. i =(R i X i ), and select the first and second distance thresholds T1 and T2, where T1>T2; Create an empty list to store the generated Canopy, and copy the original dataset as the dataset D to be processed. Randomly select a center point, and randomly select a data point P from the dataset D to be processed as the center of the new Canopy; create a Canopy, calculate the Euclidean distance from all other data points in D to point P, include all data points whose distance is less than the first threshold T1 into the Canopy centered at P, and remove all data points whose distance is less than the second threshold T2 from the dataset D to be processed; add the newly generated Canopy to the Canopy list. Finally, a set of Canopies is obtained, and each Canopy has a center point, which is used as the initial cluster center of the FCM algorithm; Set the final number of clusters c obtained from Canopy coarse clustering as the number of clusters in FCM, use the Canopy center point as the initial cluster center V0, and set the fuzzy coefficient m = 2 and the iteration termination condition ε. Based on the current cluster centers, calculate the membership degree of each data point d to each cluster center e; Based on the current membership matrix, recalculate the centroid of each cluster, where V is the set of cluster centroid groups. Calculate the Euclidean distance difference between the new cluster center V′ and the previous cluster center V. If ||V′-V||<ε, the algorithm converges and stops iterating; otherwise, iterate. Output the final cluster center V end Membership matrix U end And based on the membership matrix, each data point is assigned to the Canopy center with the highest membership degree, completing the hard partition and obtaining c. A One Class A distributed photovoltaic cluster, c B A Class B distributed photovoltaic cluster.

8. A system for optimizing control parameters of a distributed photovoltaic aggregation equivalent model, characterized in that, The system includes: The model building module is used to introduce a dynamic response affinity factor, divide A / B type photovoltaic clusters according to the dynamic response affinity factor, perform multi-machine equivalence on the divided distributed photovoltaic clusters, and perform secondary equivalence on the multi-machine equivalence results to obtain a secondary equivalence model. The simulation platform construction module is used to connect the secondary equivalent model to the simulation model of the target power grid to form an equivalent power grid simulation platform; The objective function construction module is used to construct a multi-dimensional optimization objective function based on the voltage stability index of key nodes in the power grid, according to the equivalent power grid simulation platform. The parameter optimization module is used to optimize the control parameters in the quadratic isometry model according to the multi-dimensional optimization objective function.

9. A storage medium, characterized in that, The storage medium stores one or more programs, which, when executed by a processor, implement the distributed photovoltaic aggregation equivalent model control parameter optimization method as described in any one of claims 1-7.

10. An electronic device comprising a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the method for optimizing control parameters of the distributed photovoltaic aggregation equivalent model as described in any one of claims 1-7.

Images