Distribution network reactive power coordinated regulation and control method based on data driving

Through a data-driven method, a multi-layer perceptron model is constructed and the trend constraints are linearized, which solves the problem of low resolution efficiency of distribution network reactive power optimization model, realizes efficient and accurate reactive power coordinated regulation, and improves the economicality of power grid operation and reliability.

CN118889587APending Publication Date: 2024-11-01SUQIAN POWER SUPPLY COMPANY OF JIANGSU PROVINCE POWER +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410914353.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-09
Publication Date
2024-11-01

AI Technical Summary

Technical Problem

The existing distribution network reactive power optimization model is not very effective in the case of a large number of system nodes and complex structures, making it difficult to effectively adjust the distribution network reactive power equipment to ensure the balance and voltage stability of the power system.

Method used

Using a data-driven method, by establishing a physical model of the distribution network, generating expert samples and building a multi-layer perceptron model, linearizing the current constraints, and obtaining a data-driven linear model to realize the problem of reactive power distribution network reactive power coordination. Efficient and accurate solution.

Benefits of technology

The solution efficiency and accuracy of the distribution network reactive power optimization model is improved, and the coordinated optimization of the minimum active network loss and the planning and operation of reactive equipment is achieved, which improves the economic and reliability of system operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118889587A_ABST
    Figure CN118889587A_ABST
Patent Text Reader

Abstract

The invention discloses a distribution network reactive power coordination regulation and control method based on data driving. The method comprises the following steps: acquiring power grid, photovoltaic power generation, load and various controllable equipment technology and economic characteristic data; establishing a power distribution network system planning operation optimization model P1 considering the reactive compensation equipment by taking the minimum power distribution network line loss as an optimization target; processing nonlinear constraints in the model P1 by using a relaxation technology to obtain a relaxation model P2; solving the operation scenes of the power distribution network system under different photovoltaic and load conditions by using the model P2, and generating a large number of expert samples; constructing a multi-layer perceptron model according to a large number of obtained expert samples; taking power flow data output by the multi-layer perceptron model as a linearization base point, and relaxing the model P2 to obtain a linear model P3; and the model P3 is used to solve the reactive power coordinated regulation problem of the power distribution network, and a scheduling scheme of each adjustable device of the power distribution network is obtained.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power system operation and dispatching automation, and in particular to a data-driven coordinated reactive power control method for a distribution network. Background Art

[0002] Against the background of vigorous development of new energy, the penetration rate of new energy in distribution networks will continue to increase, while the proportion of traditional fossil energy such as thermal power in the power grid will inevitably continue to decrease. The access of a large number of distributed photovoltaic and other new energy sources to the power grid will bring new challenges to the operation of the power grid. The demand for regulation of reactive equipment in distribution networks will increase significantly, and it is particularly important to ensure the reactive balance and voltage stability of the power system.

[0003] At present, reactive power optimization of distribution networks mostly focuses on the connection between the constructed model and the actual physical structure of the distribution network. For example, the second-order cone programming model is currently widely used in the physical modeling of distribution networks. Although most of the constraints of the model have been processed and converted into linear constraints, it still contains nonlinear constraints. In scenarios with a large number of system nodes and complex structures, there is still a problem of low solution efficiency. In order to solve these problems, a data-driven coordinated reactive power control method for distribution networks is proposed. Summary of the invention

[0004] In order to overcome the shortcomings of relying entirely on physical structure models, the purpose of the present invention is to provide a data-driven method for reactive power coordinated control of distribution networks. A physical model of the distribution network is established, and then a large number of expert samples are generated by solving in different scenarios to construct a multi-layer perceptron model. The output of the multi-layer perceptron model is used as the basis for power flow linearization to obtain a linear model based on data-driven, thereby achieving efficient and accurate solution to the problem of reactive power coordinated control of distribution networks.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A data-driven coordinated control method for reactive power distribution network, characterized in that it comprises the following steps:

[0007] Step 1: Obtain technical and economic characteristic data of power grid, photovoltaic power generation, load and various controllable equipment;

[0008] Step 2: Based on the data obtained in step 1, with the minimum distribution network line loss as the optimization goal, a distribution network system planning and operation optimization model P1 considering reactive power compensation equipment is established;

[0009] Step 3: According to the optimization model P1 established in step 2, the nonlinear constraints in the model are processed using relaxation technology to obtain a relaxation model P2;

[0010] Step 4: According to the relaxed model P2 in step 3, solve different distribution network system operation scenarios to generate a large number of expert samples;

[0011] Step 5: Based on the large number of expert samples obtained in step 4, a multi-layer perceptron model is constructed and the model is trained;

[0012] Step 6: According to the multi-layer perceptron model obtained in step 5, linearize P2 to obtain a linear model P3;

[0013] Step 7: Solve the problem of coordinated reactive power regulation of the distribution network according to the linear model P3 obtained in step 6.

[0014] The data in step 1 include: the safe operating range of voltage and line current, the predicted interval of photovoltaic active output, the active and reactive load requirements of each node, the impedance of each line, the ratio step size of the on-load tap-changing transformer, the upper limit of the gear position, the upper limit of the number of actions, the switching capacity of a single group of group switching capacitors, the upper limit of the number of switching groups, the upper limit of the number of actions, the upper and lower limits of the active power of energy storage charging and discharging, the upper and lower limits of the reactive power of charging and discharging, the upper and lower limits of the stored power, the charging and discharging efficiency, the upper and lower limits of the reactive output of the static VAR compensator, the photovoltaic apparent power and the maximum allowable power factor angle.

[0015] The modeling process of the distribution network system planning and operation optimization model P1 considering the reactive compensation equipment described in step 2 is as follows:

[0016] 1) Establish the objective function as

[0017]

[0018] Where T is the total number of time periods, is the set of all branches of the distribution network, I ij,t is the current flowing through branch (i, j) during period t, r ij is the resistance of branch (i,j), and Δt is the duration of each time period.

[0019] 2) Introducing branch flow model equation

[0020]

[0021] in, is the total power flowing through branch (i, j) during period t, is the set of branch end nodes with j as the head node, is the total power injected into node j during period t, which is equal to the power generated by all generators on node j minus the load power of node j, ij is the impedance on branch (i,j), is the current flowing through branch (i, j) during period t, is the voltage on node i during period t.

[0022] 3) Introducing on-load tap-changing transformer operation constraints

[0023]

[0024] Among them, U 0,base It is the voltage on the high-voltage side of the on-load tap-changing transformer, and its value is usually a constant. is the transformation ratio of the voltage on both sides of the on-load tap-changing transformer during period t, n oltc,0 is the minimum value of the transformation ratio, and Δn is the step size of the transformation ratio. is the gear position of the on-load tap-changing transformer during period t, It is the minimum value of the on-load tap-changing transformer gear. It is the maximum value of the on-load tap-changing transformer gear. It is a 0-1 variable, which is used to indicate whether the gear position of the on-load tap-changing transformer changes at time t. It is the maximum number of times the on-load tap-changing transformer can operate in one day.

[0025] 4) Introducing group switching capacitor operation constraints

[0026]

[0027]

[0028] Where Q0 is the reactive power output by a single group of capacitors, It is the reactive power output by the group-switched capacitor at node i during period t. is the number of group-switched capacitors at node i during period t, is the minimum number of switching groups of grouped switching capacitors at node i, It is the maximum number of switching groups of group-switched capacitors at node i. is a 0-1 variable, which is used to indicate whether the number of group switching capacitors at node i at time t has changed. It is the maximum number of times the group switching capacitor at node i can operate in one day.

[0029] 5) Introducing energy storage operation constraints

[0030]

[0031] in, and are 0-1 variables, representing the charging and discharging status of the energy storage at node i during period t. They represent the active power of energy storage charging at node i in period t, its minimum and maximum values ​​respectively. They respectively represent the active power discharged by the energy storage at the node i during period t, and its minimum and maximum values. They represent the reactive power of energy storage charging at node i during period t and its maximum value respectively. They represent the reactive power discharged by the energy storage at node i during period t and its maximum value respectively. They represent the amount of energy stored at node i during period t, its minimum value and maximum value respectively. ess,c , α ess,d are the efficiency of energy storage charging and discharging, respectively, and they are constants between 0 and 1.

[0032] 6) Introducing static VAR compensator constraints

[0033]

[0034] in, They are the reactive power output by the static VAR compensator at node i in period t, the minimum value and the maximum value of the reactive power respectively.

[0035] 7) Introducing PV inverter operation constraints

[0036]

[0037] in, is the reactive power output of the photovoltaic power plant at node i during period t. It is the maximum power factor angle allowed by the photovoltaic inverter. is the active power output of the photovoltaic power plant at node i during period t. is the rated apparent capacity of the PV at node i.

[0038] 8) Introducing safe operation constraints

[0039]

[0040] Among them, U i,min , U i,max is the minimum and maximum voltage amplitude at node i. ij,min ,I ij,max are the minimum and maximum current amplitudes flowing through the (i, j) branch.

[0041] The relaxation process described in step 3 includes:

[0042] 1) Processing of branch flow model

[0043] Multiply both ends of equation (2-3) by their respective conjugates, retain the amplitude of the variable, relax the phase angle of the variable, and relax equation (2-3) to

[0044]

[0045] Substitute equation (2-3) into equation (2-2) and multiply both ends of the equation by their respective conjugates to relax equation (2-2) to

[0046]

[0047] Decouple equation (2-1) and decompose it into active power balance and reactive power balance

[0048]

[0049] 2) Introduce the auxiliary variable shown in equation (10) and replace all the square terms of voltage amplitude and current amplitude in P1 with separate variables

[0050]

[0051] 3) Handle the operating constraints of on-load tap-changing transformers

[0052] Introduce two auxiliary variables Rewrite equation (3-4)-(3-5) into equation (11-1)-(11-4)

[0053]

[0054] in, It is a 0-1 variable used to represent whether the gear of the on-load tap-changing transformer increases or decreases.

[0055] Introducing auxiliary variables Rewrite formula (3-2) into formula (11-5)

[0056]

[0057] in, is a 0-1 variable used to Perform binary exponential expansion.

[0058] Introducing auxiliary variables Multiply both sides of equation (11-5) by u 0,base , we get formula (11-6)

[0059]

[0060] Using the Big M method After processing, we get formula (11-7)-(11-8)

[0061]

[0062] Wherein, M is a sufficiently large positive number.

[0063] Introducing auxiliary variables Multiply both ends of equation (11-6) by We get formula (11-9)

[0064]

[0065] Using the Big M method After processing, we get formula (11-10)-(11-11)

[0066]

[0067] 4) Handle the operation constraints of group switching capacitors

[0068] Introduce two auxiliary variables Rewrite equation (4-3)-(4-4) into equation (12-1)-(12-4)

[0069]

[0070] in, It is a 0-1 variable used to represent whether the number of switching groups of grouped switching capacitors increases or decreases.

[0071] 5) Handling of photovoltaic inverter operation constraints

[0072] Replace the actual value of the photovoltaic active output with the upper and lower limits of its active output, and rewrite equations (7-1)-(7-2) into equations (13-1)-(13-2)

[0073]

[0074] in, is the upper limit of the photovoltaic active output at node i, It is the lower limit of the photovoltaic active output at node i.

[0075] The specific process of generating expert samples described in step 4 is as follows:

[0076] The distribution network system operation scenarios with different photovoltaic and load conditions are solved to generate a large number of expert samples (S, A), where S represents the system status data, including the load data of each node and the photovoltaic power generation data of the installed photovoltaic nodes; A represents the system flow data, including the voltage of each node, the active power, reactive power and current flowing through each branch.

[0077] The specific process of building a multi-layer perceptron model described in step 5 is as follows:

[0078] Based on the large number of expert samples obtained, a multi-layer perceptron model is constructed, and the expert sample S is used as input and A is used as the label for training. By adjusting the model's hyperparameters such as the learning rate, a model is obtained that minimizes the difference between the predicted value and the label value.

[0079] The specific process of linearizing P2 described in step 6 is as follows:

[0080] 1) Use the output of the multi-layer perceptron model as the basis for linearization

[0081]

[0082] in, are the reference values ​​of active power, reactive power, current amplitude square and voltage amplitude square of node i flowing through branch (i, j) in period t, which are obtained by the multi-layer perceptron model. ΔP ij,t , ΔQ ij,t , Δl ij,t , Δu i,t They are respectively the active power, reactive power, square of current amplitude flowing through branch (i, j) in period t and square of voltage amplitude at node i adjusted based on the reference value.

[0083] 2) Linearize the power flow constraint (9-1) according to formula (14)

[0084]

[0085]

[0086] Solving the linear model described in step 7 specifically includes: using model solving tools such as gurobi to obtain the dispatching plan of each reactive device and the minimum network loss of the system.

[0087] The present invention provides an electronic device, comprising one or more processors for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors execute the aforementioned data-driven coordinated control method for reactive power of a distribution network.

[0088] The present invention provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the aforementioned data-driven coordinated control method for reactive power distribution networks when running.

[0089] Compared with the prior art, the present invention has at least the following beneficial effects:

[0090] 1. The present invention provides a distribution network system planning and operation optimization model considering reactive power compensation equipment. It can achieve the coordinated optimization of the distribution network active power loss minimization and the planning and operation of adjustable reactive equipment. It can improve the node voltage level, reduce the system operation line loss, and improve the economy and reliability of the distribution network system operation, which has certain practical value for relevant departments such as power grid planning and construction.

[0091] 2. The present invention provides a data-driven method for solving the reactive power optimization model of the distribution network. The method does not completely rely on the physical structure of the distribution network. It obtains the relationship between photovoltaic output, load demand and distribution network flow through a large amount of data. It can improve the efficiency of solving the reactive power optimization model of the distribution network while ensuring a high solution accuracy.

[0092] The application prospects of this method in the field of power system operation planning include: power flow linearization processing and distribution network reactive power coordinated optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is a step flow chart of a data-driven distribution network reactive power coordinated control method provided by the present invention.

[0094] Figure 2 It is a schematic diagram of the improved 33-node distribution network system topology structure in an embodiment of the present invention.

[0095] Figure 3 It is a comparison diagram of system network losses obtained by the second-order cone programming algorithm and the data-driven algorithm in an embodiment of the present invention.

[0096] Figure 4 It is a comparison diagram of system voltages obtained by the second-order cone programming algorithm and the data-driven algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0097] In order to make the above-mentioned objects, features and advantages of the present invention more clearly understood, the following is a brief description of the present invention in conjunction with the attached drawings. Figure 1-4 The present invention is further described in detail with reference to the accompanying drawings and specific embodiments.

[0098] The present invention proposes a data-driven coordinated control method for reactive power distribution network, the flow chart of which is as follows: Figure 1 As shown, the following steps are included:

[0099] Step 1: Obtain technical and economic characteristic data of power grid, photovoltaic power generation, load and various controllable equipment, including distribution network topology, voltage and line current safe operation range, photovoltaic active output prediction interval, active and reactive load demand of each node, impedance of each line, on-load tap-changing transformer ratio step, gear upper limit, action number upper limit, group switching capacity of single group of capacitor, upper limit of switching group number, action number upper limit, upper and lower limits of energy storage charging and discharging active power, upper and lower limits of charging and discharging reactive power, upper and lower limits of stored power, charging and discharging efficiency, upper and lower limits of reactive output of static VAR compensator, photovoltaic apparent power and maximum allowable power factor angle;

[0100] Step 2: Based on the data obtained in step 1, with the minimum distribution network line loss as the optimization goal, a distribution network system planning and operation optimization model P1 considering reactive power compensation equipment is established;

[0101] The optimization goal of model P1 is to minimize the line loss of the distribution network. During the transmission process, the distribution network will generate line loss on each branch. Line loss will waste resources and reduce the operating efficiency of the distribution network. By reducing line loss, energy utilization can be improved, making the distribution network safer and more economical. Therefore, the objective function of the model is described as

[0102]

[0103] Where T is the total number of time periods, is the set of all branches of the distribution network, I ij,t is the current flowing through branch (i, j) during period t, r ij is the resistance of branch (i,j), and Δt is the duration of each time period.

[0104] The branch power flow model is a complete description of the power flow distribution during the operation of the distribution network. It consists of three constraints, namely node power balance, branch Ohm's law and node head-end power.

[0105]

[0106] in, is the total power flowing through branch (i, j) during period t, is the set of branch end nodes with j as the head node, is the total power injected into node j during period t, which is equal to the power generated by all generators on node j minus the load power of node j, ij is the impedance on branch (i,j), is the current flowing through branch (i, j) during period t, is the voltage on node i during period t.

[0107] The on-load tap-changing transformer needs to meet the transformation ratio constraints at both ends of the transformer, the upper and lower limit constraints of the gear position, and the daily dispatch frequency limit constraints.

[0108]

[0109] Among them, U 0,base It is the voltage on the high-voltage side of the on-load tap-changing transformer, and its value is usually a constant. is the transformation ratio of the voltage on both sides of the on-load tap-changing transformer during period t, n oltc,0 is the minimum value of the transformation ratio, and Δn is the step size of the transformation ratio. is the gear position of the on-load tap-changing transformer during period t, It is the minimum value of the on-load tap-changing transformer gear. It is the maximum value of the on-load tap-changing transformer gear. It is a 0-1 variable, which is used to indicate whether the gear position of the on-load tap-changing transformer changes at time t. It is the maximum number of times the on-load tap-changing transformer can operate in one day.

[0110] The group switching capacitors need to meet the upper and lower limit constraints on the number of switching groups and the daily scheduling limit constraints.

[0111]

[0112]

[0113] Where Q0 is the reactive power output by a single group of capacitors, It is the reactive power output by the group-switched capacitor at node i during period t. is the number of group-switched capacitors at node i during period t, is the minimum number of switching groups of grouped switching capacitors at node i, It is the maximum number of switching groups of group-switched capacitors at node i. is a 0-1 variable, which is used to indicate whether the number of group switching capacitors at node i at time t has changed. It is the maximum number of times the group switching capacitor at node i can operate in one day.

[0114] Energy storage operation needs to meet charging and discharging limit constraints, upper and lower limit constraints of charging and discharging active and reactive power, and storage capacity constraints.

[0115]

[0116] in, and are 0-1 variables, representing the charging and discharging status of the energy storage at node i during period t. They represent the active power of energy storage charging at node i in period t, its minimum and maximum values ​​respectively. They respectively represent the active power discharged by the energy storage at the node i during period t, and its minimum and maximum values. They represent the reactive power of energy storage charging at node i during period t and its maximum value respectively. They represent the reactive power discharged by the energy storage at node i during period t and its maximum value respectively. They represent the amount of energy stored at node i during period t, its minimum value and maximum value respectively. ess,c , α ess,d are the efficiency of energy storage charging and discharging, respectively, and they are constants between 0 and 1.

[0117] The static VAR compensator needs to meet the upper and lower limits of reactive power output.

[0118]

[0119] in, They are the reactive power output by the static VAR compensator at node i in period t, the minimum value and the maximum value of the reactive power respectively.

[0120] The reactive power output of the photovoltaic inverter is restricted by the active power and power factor angle on the one hand, and by the rated apparent capacity on the other.

[0121]

[0122] in, is the reactive power output of the photovoltaic power plant at node i during period t. It is the maximum power factor angle allowed by the photovoltaic inverter. is the active power output of the photovoltaic power plant at node i during period t. is the rated apparent capacity of the PV at node i.

[0123] The safety operation constraints mainly refer to the upper and lower limits of the voltage amplitude of each node and the current amplitude of each line.

[0124]

[0125] Among them, U i,min , U i,max is the minimum and maximum voltage amplitude at node i. ij,min ,I ij,max are the minimum and maximum current amplitudes flowing through the (i, j) branch.

[0126] Therefore, the distribution network system planning and operation optimization model considering reactive compensation equipment can be expressed as

[0127]

[0128] st(2)-(8)

[0129] Step 3: According to the optimization model P1 established in step 2, the nonlinear constraints in the model are processed using relaxation technology to obtain a relaxation model P2;

[0130] The branch flow model is difficult to solve due to the existence of complex number operations. At the same time, the distribution network operates in a radial manner in most cases. Some scholars have proved that for radial networks, phase angle relaxation is always tight, that is, the solution of the relaxed model is consistent with the solution of the original model, and the solution to the original problem can be restored from the solution of the relaxed model. Therefore, the branch flow model is subjected to phase angle relaxation.

[0131] Multiply both ends of equation (2-3) by their respective conjugates, retain the amplitude of the variable, relax the phase angle of the variable, and relax equation (2-3) to

[0132]

[0133] Substitute equation (2-3) into equation (2-2) and multiply both ends of the equation by their respective conjugates to relax equation (2-2) to

[0134]

[0135] Decouple equation (2-1) and decompose it into active power balance and reactive power balance

[0136]

[0137] Since constraints with square terms are slow to solve, replace all square terms of voltage and current amplitudes in P1 with separate variables.

[0138]

[0139] The main nonlinearity in the operation constraints of the on-load tap-changing transformer is reflected in the absolute value of the upper and lower limit constraints of the gear, and in the fact that the square of the current amplitude is used as the decision variable, which will lead to the transformation ratio in the constraints. and gear The square term of .

[0140] For absolute value constraints, two auxiliary variables are introduced to remove the absolute value.

[0141] Introduce two auxiliary variables Rewrite equation (3-4)-(3-5) into equation (11-1)-(11-4)

[0142]

[0143] in, It is a 0-1 variable used to represent whether the gear of the on-load tap-changing transformer increases or decreases.

[0144] For discrete variables and gear The square of Perform binary expansion and then use the big-M method to transform the nonlinear constraints into mixed integer linear constraints.

[0145] Introducing auxiliary variables Rewrite formula (3-2) into formula (11-5)

[0146]

[0147] in, is a 0-1 variable used to Perform binary exponential expansion.

[0148] Introducing auxiliary variables Multiply both sides of equation (11-5) by u 0,base , we get formula (11-6)

[0149]

[0150] Using the Big M method After processing, we get formula (11-7)-(11-8)

[0151]

[0152] Wherein, M is a sufficiently large positive number.

[0153] Introducing auxiliary variables Multiply both ends of equation (11-6) by We get formula (11-9)

[0154]

[0155] In formula (11-9), That is, the voltage on the low voltage side of the transformer. After processing, we get formula (11-10)-(11-11)

[0156]

[0157] There are absolute values ​​in the operating constraints of group switching capacitors, and the processing method is similar to that of on-load tap-changing transformers.

[0158] Introduce two auxiliary variables Rewrite equation (4-3)-(4-4) into equation (12-1)-(12-4)

[0159]

[0160] in, It is a 0-1 variable used to represent whether the number of switching groups of grouped switching capacitors increases or decreases.

[0161] The reactive power output of photovoltaic is related to its active power output. However, in fact, the active power output of photovoltaic is an uncertain variable. Use its upper and lower limits Instead, on the one hand, the original nonlinear constraint is converted into a linear constraint, which is convenient for solving; on the other hand, (13-1)-(13-2) actually sets the photovoltaic reactive output power to the lower limit of its maximum value (the upper limit of the minimum value). The photovoltaic inverter can always output reactive power corresponding to the active power. This treatment is equivalent to reserving a certain amount of reserve for the photovoltaic reactive power output scheduled in the day before, so that the photovoltaic reactive power output can be adjusted again during the intraday scheduling stage.

[0162] Replace the actual value of the photovoltaic active output with the upper and lower limits of its active output, and rewrite equations (7-1)-(7-2) into equations (13-1)-(13-2)

[0163]

[0164] in, is the upper limit of the photovoltaic active output at node i, It is the lower limit of the photovoltaic active output at node i.

[0165] Step 4: According to the relaxed model P2 in step 3, solve different distribution network system operation scenarios to generate a large number of expert samples;

[0166] The distribution network system operation scenarios with different photovoltaic and load conditions are solved to generate a large number of expert samples (S, A), where S represents the system status data, including the load data of each node and the photovoltaic power generation data of the installed photovoltaic nodes; A represents the system flow data, including the voltage of each node, the active power, reactive power and current flowing through each branch.

[0167] Step 5: Based on the large number of expert samples obtained in step 4, a multi-layer perceptron model is constructed and the model is trained;

[0168] Based on the large number of expert samples obtained, a multi-layer perceptron model was constructed, and 60% of the data was divided into a training set, 20% of the data was divided into a validation set, and 20% of the data was divided into a test set. Before the data was input into the model, in order to prevent the different magnitudes of the data from affecting the model training and testing, the linear function normalization method was used to normalize the original data so that the result was mapped into a decimal between [0, 1]. The hidden layer of the model was set to 2 layers, with 256 neurons in the first layer and 128 neurons in the second layer. Sigmoid was selected as the activation function for each layer, the learning rate was set to 0.01, the batch size of the model was set to 32, and the mean square error (MSE) was selected as the loss function. The learning rate was adjusted according to the MSE on the validation set. The learning rate was adjusted after every 50 rounds of training to obtain a model that minimized the difference between the predicted value and the label value.

[0169] Step 6: According to the multi-layer perceptron model obtained in step 5, linearize P2 to obtain the linear model P3; use the output of the multi-layer perceptron model as the base point of linearization

[0170]

[0171] in, are the reference values ​​of active power, reactive power, current amplitude square and voltage amplitude square of node i flowing through branch (i, j) in period t, which are obtained by the multi-layer perceptron model. ΔP ij,t , ΔQ ij,t , Δl ij,t , Δu i,t They are respectively the active power, reactive power, square of current amplitude flowing through branch (i, j) in period t and square of voltage amplitude at node i adjusted based on the reference value.

[0172] According to formula (14), the power flow constraint formula (9-1) is linearized

[0173]

[0174] Therefore, all constraints in model P3 are linear constraints and can be solved using commercial solution packages such as gurobi.

[0175] Step 7: Solve the problem of coordinated reactive power regulation of the distribution network according to the linear model P3 obtained in step 6.

[0176] Using the optimization software package gurobi, the model P3 is solved and the following solution results can be obtained: the voltage of each node in the distribution network, the network loss of the distribution network, the scheduling plan of each adjustable equipment such as the on-load tap-changing transformer, etc.

[0177] The present invention also provides an electronic device, comprising one or more processors for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors execute the aforementioned data-driven distribution network reactive power coordinated control method.

[0178] The present invention also provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the aforementioned data-driven coordinated control method for reactive power distribution networks when running.

[0179] Example

[0180] In order to enable those skilled in the art to better understand the present invention, this embodiment uses an improved 33-node distribution network system to perform reactive power optimization of the distribution network. The system has 33 nodes and 32 branches. The substation reference voltage is 12.66 kV, the voltage safety range is set at [0.95 pu, 1.05 pu], and the installation area of ​​each reactive adjustable device is as follows: Figure 2 As shown, the load data and photovoltaic power generation data used in this embodiment are derived from the photovoltaic and load data of the AT node from January 1, 2019 to December 31, 2019 in the open source data set openpowerdatasystemtime_series_60min. The photovoltaic data is scaled with 600kW as the maximum value, and the load data is scaled with the active and reactive loads of each node of the test system as the maximum value.

[0181] The second-order cone programming algorithm and the data-driven algorithm proposed in this invention are used to solve the optimal power flow of the system for 24 hours respectively. Different photovoltaic and load data within a year are selected for solution. The distribution network losses obtained by the two algorithms are as follows: Figure 3 As shown, it can be seen that the total system network loss obtained by the data-driven algorithm and the second-order cone programming algorithm is almost the same. In all test cases, the maximum relative difference in network loss obtained by the two algorithms is 11.2%, and the average relative difference is 1.98%. It can be seen that this method has good accuracy in solving the reactive power optimization problem of distribution network.

[0182] The system voltage calculated by the data-driven method and the second-order cone programming algorithm at 12 noon on a certain day was selected for comparison. The results are as follows: Figure 4 As shown in the figure, the voltage deviation of each node in the system obtained by the data-driven algorithm is generally lower than that of each node in the system obtained by the second-order cone programming algorithm, which shows that the solution obtained by the data-driven algorithm can reduce the fluctuation of the system voltage to a certain extent and improve the voltage level of the distribution network system.

Claims

1. A data-driven coordinated control method for reactive power in distribution networks, characterized in that: The following steps are involved: Step 1: Obtain data on the technical and economic characteristics of the power grid, photovoltaic power generation, loads, and various controllable equipment; Step 2: Based on the data obtained in step 1, with the minimum distribution network line loss as the optimization goal, a distribution network system planning and operation optimization model P1 considering reactive power compensation equipment is established; Step 3: According to the optimization model P1 established in step 2, the nonlinear constraints in the model are processed using relaxation technology to obtain a relaxation model P2; Step 4: According to the relaxed model P2 in step 3, solve the distribution network system operation scenarios with different photovoltaic and load conditions, and generate a large number of expert samples S and expert samples A, where S represents the system status data, including the load data of each node and the photovoltaic power generation data of the photovoltaic nodes installed, and A represents the system flow data, including the voltage of each node, the active power, reactive power and current flowing through each branch; Step 5: Based on the large number of expert samples obtained in step 4, a multi-layer perceptron model is constructed, and the expert samples S are used as input and A is used as labels for training. By adjusting the model's learning rate and other hyperparameters, a model is obtained that minimizes the difference between the predicted value and the label value. Step 6: According to the multi-layer perceptron model obtained in step 5, the power flow data output by the multi-layer perceptron model is used as the base point of linearization, and the model P2 is relaxed to obtain the linear model P3; Step 7: According to the linear model P3 obtained in step 6, the problem of coordinated reactive power control of the distribution network is solved to obtain a dispatching plan for each adjustable device in the distribution network.

2. According to a data-driven coordinated control method for reactive power distribution network according to claim 1, it is characterized in that: The data described in step 1 include: the safe operating range of voltage and line current, the forecast interval of photovoltaic active output, the active and reactive load requirements of each node, the impedance of each line, the ratio step of the on-load tap-changing transformer, the gear upper limit, the upper limit of the number of actions, the switching capacity of a single group of group switching capacitors, the upper limit of the number of switching groups, the upper limit of the number of actions, the upper and lower limits of the active power of energy storage charging and discharging, the upper and lower limits of the reactive power of charging and discharging, the upper and lower limits of the stored power, the charging and discharging efficiency, the upper and lower limits of the reactive output of the static VAR compensator, the photovoltaic apparent power and the maximum allowable power factor angle.

3. According to a data-driven coordinated control method for reactive power distribution network according to claim 1, it is characterized in that: In step 2, the process of establishing the distribution network system planning and operation optimization model P1 considering reactive power compensation equipment is as follows: 2.1) Establish the objective function; 2.2) Introduce branch flow model equation; 2.3) Introduce operating constraints of on-load tap-changing transformers; 2.4) Introducing group switching capacitor operation constraints; 2.5) Introducing energy storage operation constraints; 2.6) Introducing static VAR compensator constraints; 2.7) Introducing PV inverter operation constraints; 2.8) Introduce safe operation constraints.

4. A data-driven coordinated control method for reactive power distribution network according to claim 3, characterized in that: Step 2.1) Establish the objective function as Where T is the total number of time periods, is the set of all branches of the distribution network, I ij,t is the current flowing through branch (i, j) during period t, r ij is the resistance of branch (i,j), and Δt is the duration of each time period.

5. The data-driven coordinated control method for reactive power distribution network according to claim 3 is characterized in that: Step 2.2) Introduce the branch flow model equation in, is the total power flowing through branch (i, j) during period t, is the set of branch end nodes with j as the head node, is the total power injected into node j during period t, which is equal to the power generated by all generators on node j minus the load power of node j, z ij is the impedance on branch (i,j), is the current flowing through branch (i, j) during period t, is the voltage on node i during period t.

6. A data-driven coordinated control method for reactive power distribution network according to claim 3, characterized in that: Step 2.3) Introduce the operating constraints of the on-load tap-changing transformer Among them, U 0,base It is the voltage on the high-voltage side of the on-load tap-changing transformer, and its value is usually a constant; is the transformation ratio of the voltage on both sides of the on-load tap-changing transformer during period t, n oltc,0 is the minimum value of the transformation ratio, Δn is the step size of the transformation ratio; is the gear position of the on-load tap-changing transformer during period t, It is the minimum value of the on-load tap-changing transformer gear. It is the maximum value of the gear position of the on-load tap-changing transformer; It is a 0-1 variable, which is used to indicate whether the gear position of the on-load tap-changing transformer changes at time t. It is the maximum number of times the on-load tap-changing transformer can operate in one day.

7. The data-driven coordinated control method for reactive power distribution network according to claim 3 is characterized in that: Step 2.4) Introduce group switching capacitor operation constraints Where Q0 is the reactive power output by a single group of capacitors, is the reactive power output by the group-switched capacitor at node i during period t; is the number of group-switched capacitors at node i during period t, is the minimum number of switching groups of grouped switching capacitors at node i, is the maximum number of switched groups of grouped switched capacitors at node i; is a 0-1 variable, which is used to indicate whether the number of group switching capacitors at node i at time t has changed. It is the maximum number of times the group switching capacitor at node i can operate in one day.

8. The data-driven coordinated control method for reactive power distribution network according to claim 3 is characterized in that: Step 2.5) Introduce energy storage operation constraints in, and are 0-1 variables, representing the charging and discharging states of the energy storage at node i during period t; They represent the active power of energy storage charging at node i in period t, its minimum and maximum values ​​respectively; They represent the active power of energy storage discharge at node i in period t, and its minimum and maximum values ​​respectively; They represent the reactive power of energy storage charging at node i during period t and its maximum value respectively; They represent the reactive power discharged by the energy storage at the node i during period t and its maximum value respectively; They represent the amount of energy stored at node i during period t, its minimum and maximum values ​​respectively; α ess,c , α ess,d are the efficiency of energy storage charging and discharging, respectively, and they are constants between 0 and 1.

9. The data-driven coordinated control method for reactive power distribution network according to claim 3 is characterized in that: Step 2.6) Introduce SVAC constraints in, They are the reactive power output by the static VAR compensator at node i in period t, the minimum value and the maximum value of the reactive power respectively.

10. The data-driven coordinated control method for reactive power distribution network according to claim 3, characterized in that: Step 2.7) Introduce PV inverter operation constraints in, is the reactive power output of the photovoltaic power plant at the i node during period t; is the maximum power factor angle allowed by the photovoltaic inverter; is the active power output of the photovoltaic power plant at the i node in period t; is the rated apparent capacity of the PV at node i.

11. The data-driven coordinated control method for reactive power distribution network according to claim 3, characterized in that: Step 2.8) Introduce safe operation constraints Among them, U i,min , U i,max is the minimum and maximum voltage amplitude at node i; I ij,min ,I ij,max are the minimum and maximum current amplitudes flowing through the (i, j) branch.

12. The data-driven coordinated control method for reactive power distribution network according to claim 1, characterized in that: Step 3: Based on the optimization model P1 established in step 2, the relaxation technology is used to process the nonlinear constraints in the model to obtain the relaxation model P2 as follows: 3.1) Process the branch power flow model; 3.2) Introduce the auxiliary variable shown in equation (10) and replace all the square terms of voltage amplitude and current amplitude in P1 with separate variables; 3.3) Processing the operating constraints of the on-load tap-changing transformer; 3.4) Processing the operation constraints of group switching capacitors; 3.5) Processing of photovoltaic inverter operation constraints; Step 3.1) Process the branch flow model Multiply both ends of equation (2-3) by their respective conjugates, retain the amplitude of the variable, relax the phase angle of the variable, and relax equation (2-3) to Substitute equation (2-3) into equation (2-2) and multiply both ends of the equation by their respective conjugates to relax equation (2-2) to Decouple equation (2-1) and decompose it into active power balance and reactive power balance Step 3.2) Introduce the auxiliary variable shown in equation (10) and replace all the square terms of the voltage amplitude and current amplitude in P1 with a separate variable Step 3.3) Processing the operating constraints of the on-load tap-changing transformer Introduce two auxiliary variables Rewrite equation (3-4)-(3-5) into equation (11-1)-(11-4) in, It is a 0-1 variable, used to indicate whether the gear of the on-load tap-changing transformer increases or decreases; Introducing auxiliary variables Rewrite formula (3-2) into formula (11-5) in, is a 0-1 variable used to Perform binary exponential expansion; Introducing auxiliary variables Multiply both sides of equation (11-5) by u 0,base , we get formula (11-6) Using the Big M method After processing, we get formula (11-7)-(11-8) Among them, M is a sufficiently large positive number; Introducing auxiliary variables Multiply both ends of equation (11-6) by We get formula (11-9) Using the Big M method After processing, we get formula (11-10)-(11-11) Step 3.4) Processing the operating constraints of group switching capacitors Introduce two auxiliary variables Rewrite equation (4-3)-(4-4) into equation (12-1)-(12-4) in, It is a 0-1 variable, used to represent the increase or decrease of the number of group switching capacitors; Step 3.5) Processing of PV inverter operation constraints Replace the actual value of the photovoltaic active output with the upper and lower limits of its active output, and rewrite equations (7-1)-(7-2) into equations (13-1)-(13-2) in, is the upper limit of the photovoltaic active output at node i, It is the lower limit of the photovoltaic active output at node i.

13. The data-driven coordinated control method for reactive power distribution network according to claim 1, characterized in that: Step 6: According to the multi-layer perceptron model obtained in step 5, the power flow data output by the multi-layer perceptron model is used as the basis point of linearization, and the linear model P3 obtained by relaxing the model P2 includes: 6.1) Use the output of the multi-layer perceptron model as the basis for linearization; 6.2) Linearize the flow constraint equation (9-1) according to equation (14).

14. A data-driven coordinated control method for reactive power distribution network according to claim 13, characterized in that: Step 6.1) Use the output of the multilayer perceptron model as the basis for linearization in, are the reference values ​​of active power, reactive power, current amplitude square and voltage amplitude square of node i flowing through branch (i, j) in period t, which are obtained by the multi-layer perceptron model. ΔP ij,t , ΔQ ij,t , Δl ij,t , Δu i,t They are respectively the active power, reactive power, square of current amplitude flowing through branch (i, j) in period t and square of voltage amplitude at node i adjusted based on the reference value.

15. The data-driven coordinated control method for reactive power distribution network according to claim 13, characterized in that: 6.2) Linearize the power flow constraint (9-1) according to equation (14)