Comprehensive energy distribution network frame reconstruction multi-objective optimization method based on BAS algorithm

Abstract

The invention discloses an integrated energy distribution network frame reconstruction multi-objective optimization method based on a BAS algorithm, and relates to the technical field of distribution network optimization, and the method comprises the steps: introducing a fault prediction model based on a convolutional neural network, training fault data under different topological structures, carrying out the hyper-parameter adjustment and optimization through the combination of cross validation and grid search, and carrying out the reconstruction of a network frame of an integrated energy distribution network. The fault type can be accurately predicted; according to the method, active power loss minimization, voltage deviation index minimization and load balancing are taken as optimization objectives, and a comprehensive objective function is constructed based on a plurality of optimization objectives; by introducing a switch state constraint equation, the feasibility of an optimization scheme in actual operation is ensured, in addition, a BAS algorithm is adopted for solving, the optimization scheme is converted into a specific control instruction, the specific control instruction is issued to a power distribution network control system, switch operation and reactive compensation operation are executed, dynamic reconstruction optimization of the power distribution network is achieved, and the power distribution network optimization efficiency is improved. And the optimization efficiency and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network optimization, and in particular to a multi-objective optimization method for grid reconstruction of an integrated energy distribution network based on a BAS algorithm. Background Art

[0002] The distribution network refers to the medium and low voltage distribution system, which consists of substations, distribution lines, switchgear, transformers, etc. The main task of the distribution network is to receive electricity from the transmission network, reduce the voltage, and distribute it to end users, such as residents, commercial and industrial users. The distribution network reconstruction, also known as network reconstruction or topology optimization, is to reconfigure the connection mode of the power grid by changing the switch state to achieve the purpose of optimizing the power grid performance. This optimization can be carried out without affecting the reliability of power supply and can bring many benefits.

[0003] Existing network optimization and reconstruction strategies are mainly used in the distribution network planning stage, which is difficult to effectively stimulate the potential performance of the network in real-time operation; and traditional distribution network reconstruction algorithms generally have the problems of low optimization accuracy and long calculation time. Most methods only focus on a single optimization goal, such as minimizing active power loss or improving voltage quality, and lack effective integration of multi-objective collaborative optimization, which directly affects the improvement of reconstruction efficiency. When dealing with complex and changeable actual networks, it is often difficult to find the global optimal solution within a reasonable time. In addition, the optimization goal is single and the multi-objective optimization problem is not fully considered, which affects the efficiency and reliability of distribution network reconstruction to a certain extent.

[0004] In order to solve this problem, we propose a multi-objective optimization method for comprehensive energy distribution network reconstruction based on the BAS algorithm. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a multi-objective optimization method for comprehensive energy distribution network reconstruction based on the BAS algorithm, which can effectively solve the problems in the above-mentioned background technology.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm includes the following steps:

[0008] S1. Obtain node information and branch information of the distribution network grid, deploy distribution automation terminals, and obtain distribution network operation data, including real-time distribution network operation data and distribution network historical operation data sets;

[0009] S2. Obtain distribution network operation fault data set, and build a distribution network fault prediction model based on the operation fault data of distribution networks with different topological structures;

[0010] S3. Based on the historical operation data set of the distribution network, the historical load change curve of each node of the distribution network is obtained, and the corresponding historical meteorological data, historical social data and electricity type data are obtained based on the historical load change curve. The historical meteorological data, historical social data and electricity type data are analyzed for correlation factors with the historical load change curve, and a dynamic load prediction model is constructed. Based on the dynamic load prediction model, the predicted total load of the branch connected to the same power equipment in the current time period is obtained;

[0011] S4. Based on the real-time operation data of the distribution network, the current topological structure information of the distribution network grid is obtained, and the distribution network operation status indicators are calculated based on the real-time operation data of the distribution network. The distribution network operation status indicators include the active power loss of each node, the load balance index and the voltage deviation index;

[0012] S5. Based on the current topology information, the active power loss limit, load balancing index limit and voltage deviation index limit are set. Based on the over-limit conditions of the active power loss, load balancing index and voltage deviation index, it is determined whether the distribution network needs to be reconstructed and optimized. If it is determined that reconstruction and optimization are not required, exit; if it is determined that reconstruction and optimization are required, enter S6;

[0013] S6. Determine whether there is a fault in the current distribution network based on the distribution network fault prediction model. If there is a fault, output the fault information, process the fault branch of the distribution network based on the fault information, and then proceed to S7; if there is no fault, directly proceed to S7;

[0014] S7. Obtain the current topological structure information of the distribution network, take minimization of active power loss, minimization of voltage deviation index and load balancing as optimization goals, and respectively construct an active power loss degree objective function, a voltage stability degree objective function and a load balancing degree objective function;

[0015] S8. Based on the active power loss degree objective function, the voltage stability degree objective function and the load balance degree objective function, a comprehensive objective function is constructed, and based on the predicted total load of the same group of branches, a switch state constraint equation is established. Under the condition that the switch state constraint equation is established, based on the BAS algorithm, the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are solved;

[0016] S9. Determine the optimization plan for each node of the distribution network according to the switch planning results and reactive power compensation results, and reconstruct and optimize the distribution network based on the optimization plan.

[0017] Preferably, the distribution network fault prediction model is constructed based on the operation fault data of distribution networks with different topological structures in S2, which specifically includes:

[0018] Based on the operation fault data set of the distribution network, the topological structure information of the distribution network when the operation fault occurs is obtained, wherein the operation fault data set includes the fault type, fault location, fault duration, fault occurrence time and operation parameters, and the topological structure information includes the switch state, line connection relationship and node type;

[0019] The above data are preprocessed and the operating fault data set is divided into a training set and a test set. The convolutional neural network model is trained using the training set, and the model parameters are adjusted through the back propagation algorithm. The trained distribution network fault prediction model is used to predict the fault types of the distribution network under different topological structures.

[0020] During the training process of the distribution network fault prediction model, cross validation and grid search are used to perform hyperparameter tuning to optimize the performance of the distribution network fault prediction model;

[0021] Preferably, S3 specifically includes:

[0022] Extract the load value of each node in each time period from the historical operation data set of the distribution network;

[0023] Arrange the load values ​​of each node in chronological order to form time series data;

[0024] With time as the horizontal axis and load value as the vertical axis, use visualization tools to draw the historical load change curve of each node;

[0025] Obtain the load value of each time period in the historical load change curve, remove the node data with a load value of 0, and obtain the historical meteorological data, historical social data and electricity consumption type data of each node corresponding to each time period. The historical meteorological data includes temperature, humidity, light intensity and wind speed. The historical social data includes holiday information and large-scale event information. The electricity consumption type data includes residential electricity consumption, commercial electricity consumption, industrial electricity consumption and agricultural electricity consumption;

[0026] Preprocess the above data, integrate historical meteorological data, historical social data and data on electricity consumption types of each node, generate relevant impact data, conduct correlation analysis on the load values ​​of each node obtained in the same time period and the relevant impact data, and use the Pearson correlation coefficient to calculate the correlation between the relevant impact data and the load values ​​of each node;

[0027] Based on the correlation between the relevant impact data and the load value of each node, the correlation index of the relevant impact data is obtained, and a dynamic load prediction model is constructed based on the correlation index;

[0028] Specifically, a neural network model is constructed through a multi-layer perceptron, and the correlation index and related impact data are used as input and the load value as output to train the dynamic load forecasting model.

[0029] Obtain real-time meteorological data, real-time social data and power consumption data of each node in the distribution network in the current time period, input the above data into the dynamic load forecasting model, and obtain the predicted load of the branch in each operating state in the current time period;

[0030] The branches connected to each power device in the distribution network are obtained, the branches connected to the same power device are recorded as a group, and the predicted loads of the branches in the same group are added together to be recorded as the predicted total load.

[0031] Preferably, the S4 specifically includes:

[0032] Based on the real-time operation data of the distribution network, the node-branch association matrix is ​​obtained, and the topological structure information of the distribution network grid is generated using a graph theory method. The real-time operation data includes the state of each switch, branch capacity, branch apparent power, branch active power, branch reactive power, node current, node voltage, transformer state, line connection relationship and node type;

[0033] Calculate the distribution network operation status indicators based on the real-time operation data of the distribution network. The distribution network operation status indicators include the active power loss, load balance index and voltage deviation index of each node;

[0034] The active power loss calculation formula is:

[0035]

[0036] Where P loss,m is the active power loss of node m, N is the number of branches connected to node m in the distribution network, R n is the resistance of branch n, P m and Q n are the active power and reactive power flowing through branch n, V m is the voltage amplitude of node m;

[0037] The voltage deviation index calculation formula is:

[0038]

[0039] Where ΔV is the voltage deviation index, V m,rated is the reference voltage of node m, M is the total number of nodes, V m is the voltage amplitude at node m.

[0040] The load balancing index calculation formula is:

[0041]

[0042] Where, L b is the load balancing index, B is the total number of branches, S nis the apparent power of branch n, S max,n is the capacity of branch n, S l is the apparent power of branch l, S max,l is the capacity of branch l.

[0043] Preferably, the S5 specifically includes:

[0044] Set the active power loss limit P based on the current topology loss，max , Load balancing index limit L b，max And voltage deviation index limit ΔV max ;

[0045] Determine the current active power loss P of each node loss Is it more than P loss，max ;

[0046] Determine the current load balancing index L b Is it more than L b，max ;

[0047] Determine whether the current voltage deviation index ΔV exceeds ΔV max ;

[0048] If all indicators are within the limit, there is no need to restructure and optimize, and the program can be exited;

[0049] If any indicator exceeds the limit, reconstruction optimization is required.

[0050] Preferably, judging whether there is a fault in the current distribution network based on the distribution network fault prediction model in S6 specifically includes:

[0051] Obtain the current topological structure information of the distribution network, input the real-time operation data and topological structure information of the distribution network into the distribution network fault prediction model, output the suspected fault type and fault probability of the distribution network under the current topological structure, and calculate the fault impact index of the suspected fault type;

[0052] If the fault impact index exceeds the set impact threshold, it is determined that there is a fault in the current distribution network, and abnormal real-time operation data is obtained based on the suspected fault type. The fault location algorithm is used to locate the fault area based on the abnormal real-time operation data. The fault area refers to the branch / node affected by the fault. The fault branch is disconnected by the isolation switch, and the load is transferred from the fault branch to other normally operating branches through the tie switch to ensure that the fault area is isolated from other normally operating areas;

[0053] If the fault impact index does not exceed the set impact threshold, it is determined that there is no fault in the current distribution network.

[0054] Preferably, the S7 specifically includes:

[0055] The current topological structure information of the distribution network is obtained, and the minimization of active power loss, minimization of voltage deviation index and load balancing are taken as optimization goals. The active power loss degree objective function f1, load balancing degree objective function f2 and voltage stability degree objective function f3 are constructed respectively.

[0056] Preferably, the S8 specifically includes:

[0057] Taking the active power loss degree objective function, the voltage stability degree objective function and the load balancing degree objective function as optimization targets, a comprehensive objective function is constructed;

[0058] The switch state and reactive power compensation amount are used as decision variables;

[0059] Based on the predicted total load of the same group of branches, a switch state constraint equation is established, and each branch in the same group satisfies the switch state constraint equation;

[0060] Under the condition that the switch state constraint equation is established, the solution is performed based on the BAS algorithm, and the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are obtained according to the solution results;

[0061] The comprehensive objective function is:

[0062]

[0063] In the formula, are the weight parameters of each objective function.

[0064] Preferably, the solution is performed based on the BAS algorithm, and the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are obtained according to the solution results, specifically including:

[0065] The switch state and reactive compensation amount of the distribution network are combined into a high K-dimensional vector K, where the switch state is represented by a binary variable and the reactive compensation amount is represented by a continuous variable;

[0066] Based on the beetle whisker search algorithm, the initial switch state and reactive compensation amount are combined as the position of the beetle as x, and the coordinates of the left and right beetles are represented as x and l and x r , the distance between the two whiskers of the longicorn is d0;

[0067] Calculate the comprehensive objective function value corresponding to the position of the longicorn;

[0068] Based on the principle of longhorn beetle search behavior, the antennae movement of longhorn beetles is simulated, the maximum number of iterations of the outer loop and the inner loop is set, a direction vector is randomly generated, and the direction vector is normalized to obtain a random vector

[0069] Determine the coordinates of the left and right whiskers of the longhorn beetle and decide whether to accept the new position according to the Metropolis criterion;

[0070] According to the change of the coordinates of the left and right whiskers of the longhorn beetle, the movement direction and distance of the longhorn beetle in the next iteration are determined, and the inner loop step length is updated according to the demand to determine whether the cooling times have been reached. If not, the inner loop search is continued; if so, the outer loop step length is updated according to the demand;

[0071] After reaching the maximum number of iterations, the optimal solution is output;

[0072] Based on the switch planning results and reactive power compensation results, the optimization scheme for each node and branch of the distribution network is determined, and the optimization scheme is converted into specific control instructions, which are sent to the distribution network control system to execute switch operations and reactive power compensation operations to achieve distribution network reconstruction optimization.

[0073] Preferably, the switch state constraint equation is:

[0074]

[0075] In the formula, G is the set of branches in the same group, W pred To predict the total load, S n is the switch state of branch n, S n ∈{0,1}, 0 means open, 1 means closed, W n is the load of branch n, γ1 and γ2 are adjustment coefficients.

[0076] Compared with the prior art, the present invention provides a multi-objective optimization method for comprehensive energy distribution network reconstruction based on the BAS algorithm, which has the following beneficial effects:

[0077] The present invention introduces a fault prediction model based on a convolutional neural network, trains fault data under different topological structures, and combines cross-validation and grid search to perform hyperparameter tuning, thereby realizing the ability to accurately predict fault types in real-time operation data; the present invention takes minimization of active power loss, minimization of voltage deviation index and load balancing as goals, and constructs a comprehensive objective function based on multiple goals, adopts the BAS algorithm for solving, converts the optimization scheme into specific control instructions, and sends it to the distribution network control system to execute switching operations and reactive power compensation operations, thereby realizing dynamic reconstruction optimization of the distribution network and improving the operation efficiency and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a flow chart of a multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm of the present invention;

[0079] Figure 2 It is a schematic diagram of the process of step S3 of the present invention. DETAILED DESCRIPTION

[0080] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the present invention will be further described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] In order to address the shortcomings of the existing technology, such as Figure 1 , 2 As shown, the present invention provides a multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm, comprising the following steps:

[0082] S1. Obtain node information and branch information of the distribution network grid, deploy distribution automation terminals, and obtain distribution network operation data, including real-time distribution network operation data and distribution network historical operation data sets;

[0083] S2. Obtain distribution network operation fault data set, and build a distribution network fault prediction model based on the operation fault data of distribution networks with different topological structures;

[0084] S3. Based on the historical operation data set of the distribution network, the historical load change curve of each node of the distribution network is obtained, and the corresponding historical meteorological data, historical social data and electricity type data are obtained based on the historical load change curve. The historical meteorological data, historical social data and electricity type data are analyzed for correlation factors with the historical load change curve, and a dynamic load prediction model is constructed. Based on the dynamic load prediction model, the predicted total load of the branch connected to the same power equipment in the current time period is obtained;

[0085] S4. Based on the real-time operation data of the distribution network, the current topological structure information of the distribution network grid is obtained, and the distribution network operation status indicators are calculated based on the real-time operation data of the distribution network. The distribution network operation status indicators include the active power loss of each node, the load balance index and the voltage deviation index;

[0086] S5. Based on the current topology information, the active power loss limit, load balancing index limit and voltage deviation index limit are set. Based on the over-limit conditions of the active power loss, load balancing index and voltage deviation index, it is determined whether the distribution network needs to be reconstructed and optimized. If it is determined that reconstruction and optimization are not required, exit; if it is determined that reconstruction and optimization are required, enter S6;

[0087] S6. Determine whether there is a fault in the current distribution network based on the distribution network fault prediction model. If there is a fault, output the fault information, process the fault branch of the distribution network based on the fault information, and then proceed to S7; if there is no fault, directly proceed to S7;

[0088] S7. Obtain the current topological structure information of the distribution network, take minimization of active power loss, minimization of voltage deviation index and load balancing as optimization goals, and respectively construct an active power loss degree objective function, a voltage stability degree objective function and a load balancing degree objective function;

[0089] S8. Based on the active power loss degree objective function, the voltage stability degree objective function and the load balance degree objective function, a comprehensive objective function is constructed, and based on the predicted total load of the same group of branches, a switch state constraint equation is established. Under the condition that the switch state constraint equation is established, based on the BAS algorithm, the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are solved;

[0090] S9. Determine the optimization plan for each node of the distribution network according to the switch planning results and reactive power compensation results, and reconstruct and optimize the distribution network based on the optimization plan.

[0091] Specifically, the distribution network fault prediction model is constructed based on the operation fault data of distribution networks with different topological structures in S2, which specifically includes:

[0092] Based on the operation fault data set of the distribution network, the topological structure information of the distribution network when the operation fault occurs is obtained, wherein the operation fault data set includes the fault type, fault location, fault duration, fault occurrence time and operation parameters, and the topological structure information includes the switch state, line connection relationship and node type;

[0093] The fault types include but are not limited to short circuit fault, open circuit fault, ground fault, equipment fault, overvoltage fault and harmonic fault;

[0094] The above data are preprocessed and the operating fault data set is divided into a training set and a test set. The convolutional neural network model is trained using the training set, and the model parameters are adjusted through the back propagation algorithm. The trained distribution network fault prediction model is used to predict the fault types of the distribution network under different topological structures.

[0095] During the training process of the distribution network fault prediction model, cross validation and grid search are used to perform hyperparameter tuning to optimize the performance of the distribution network fault prediction model;

[0096] It should be noted that the fault prediction model constructed based on the distribution network operation fault data of different topological structures can effectively utilize convolutional neural networks to extract features in topological structures and operating parameters, and combine cross-validation and grid search for hyperparameter tuning, thereby achieving accurate prediction of distribution network fault types based on real-time operation data of the distribution network.

[0097] Specifically, the S3 specifically includes:

[0098] Extract the load value of each node in each time period from the historical operation data set of the distribution network;

[0099] Arrange the load values ​​of each node in chronological order to form time series data;

[0100] With time as the horizontal axis and load value as the vertical axis, use visualization tools to draw the historical load change curve of each node;

[0101] Obtain the load value of each time period in the historical load change curve, remove the node data with a load value of 0, and obtain the historical meteorological data, historical social data and electricity consumption type data of each node corresponding to each time period. The historical meteorological data includes temperature, humidity, light intensity and wind speed. The historical social data includes holiday information and large-scale event information. The electricity consumption type data includes residential electricity consumption, commercial electricity consumption, industrial electricity consumption and agricultural electricity consumption;

[0102] Preprocess the above data, integrate historical meteorological data, historical social data and data on electricity consumption types of each node, generate relevant impact data, conduct correlation analysis on the load values ​​of each node obtained in the same time period and the relevant impact data, and use the Pearson correlation coefficient to calculate the correlation between the relevant impact data and the load values ​​of each node;

[0103] The removal of node data with a load value of 0 can be understood as removing the influence of branches without power transmission on relevant influencing data, so as to provide accurate correlation analysis;

[0104] Based on the correlation between the relevant impact data and the load value of each node, the correlation index of the relevant impact data is obtained, and a dynamic load prediction model is constructed based on the correlation index;

[0105] Obtain real-time meteorological data, real-time social data and power consumption data of each node in the distribution network in the current time period, input the above data into the dynamic load forecasting model, and obtain the predicted load of the branch in each operating state in the current time period;

[0106] The branches connected to each power device in the distribution network are obtained, the branches connected to the same power device are recorded as a group, and the predicted loads of the branches in the same group are added together to be recorded as the predicted total load.

[0107] Specifically, the S4 specifically includes:

[0108] Based on the real-time operation data of the distribution network, a node-branch association matrix is ​​obtained, and the topological structure information of the distribution network grid is generated using a graph theory method. The real-time operation data includes but is not limited to each switch state, branch capacity, branch apparent power, branch active power, branch reactive power, node current, node voltage, transformer state, line connection relationship and node type;

[0109] Calculate the distribution network operation status indicators based on the real-time operation data of the distribution network. The distribution network operation status indicators include the active power loss, load balance index and voltage deviation index of each node;

[0110] The active power loss calculation formula is:

[0111]

[0112] Where P loss,m is the active power loss of node m, N is the number of branches connected to node m in the distribution network, R n is the resistance of branch n, P m and Q n are the active power and reactive power flowing through branch n, V m is the voltage amplitude of node m;

[0113] The voltage deviation index calculation formula is:

[0114]

[0115] Where ΔV is the voltage deviation index, V m,rated is the reference voltage of node m, M is the total number of nodes, V m is the voltage amplitude at node m.

[0116] The load balancing index calculation formula is:

[0117]

[0118] Where, L b is the load balancing index, B is the total number of branches, S n is the apparent power of branch n, S max,n is the capacity of branch n, S l is the apparent power of branch l, S max,l is the capacity of branch l. It should be noted that, Used to indicate the average load rate of all branches.

[0119] Specifically, the S5 specifically includes:

[0120] Set the active power loss limit P based on the current topology loss，max , Load balancing index limit L b，max And voltage deviation index limit ΔV max ;

[0121] It should be noted that the active power loss limit P loss，max , Load balancing index limit L b，max And voltage deviation index limit ΔV max Set according to the fault type and operating parameters in the operating fault data set, the operating parameters include but are not limited to each switch state, branch capacity, branch apparent power, branch active power, branch reactive power, node current, node voltage, transformer state and line connection relationship;

[0122] Substitute the operating parameters in the operating fault data set into the function D(P m ,V m ), calculate the values ​​of α0, α1, α2 and α3;

[0123] The function D(P m ,V m ) is expressed as:

[0124]

[0125] Substitute α0, α1, α2 and α3 into the function D(P m ,V m ), calculate the function D(P m ,V m ) and output P from the function loss，m , ΔV and L b The corresponding values ​​are respectively used as the power loss limit P loss，max , Load balancing index limit L b，max And voltage deviation index limit ΔV max ;

[0126] Determine the current active power loss P of each node loss Is it more than P loss，max ;

[0127] Determine the current load balancing index L b Is it more than L b，max ;

[0128] Determine whether the current voltage deviation index ΔV exceeds ΔV max ;

[0129] If all indicators are within the limit, there is no need to restructure and optimize, and the program can be exited;

[0130] If any indicator exceeds the limit, reconstruction optimization is required.

[0131] Specifically, judging whether there is a fault in the current distribution network based on the distribution network fault prediction model in S6 specifically includes:

[0132] Obtain the current topological structure information of the distribution network, input the real-time operation data and topological structure information of the distribution network into the distribution network fault prediction model, output the suspected fault type and fault probability of the distribution network under the current topological structure, and calculate the fault impact index of the suspected fault type;

[0133] If the fault impact index exceeds the set impact threshold, it is determined that there is a fault in the current distribution network, and abnormal real-time operation data is obtained based on the suspected fault type. The fault location algorithm is used to locate the fault area based on the abnormal real-time operation data. The fault area refers to the branch / node affected by the fault. The fault branch is disconnected by the isolation switch, and the load is transferred from the fault branch to other normally operating branches through the tie switch to ensure that the fault area is isolated from other normally operating areas;

[0134] If the fault impact index does not exceed the set impact threshold, it is determined that there is no fault in the current distribution network.

[0135] Specifically, the S7 specifically includes:

[0136] The current topological structure information of the distribution network is obtained, and the active power loss minimization, voltage deviation index minimization and load balancing are taken as optimization goals, and the active power loss degree objective function f1, load balancing degree objective function f2 and voltage stability degree objective function f3 are constructed respectively;

[0137] The active power loss degree objective function f1, the load balancing degree objective function f2 and the voltage stability degree objective function f3 are:

[0138]

[0139] Specifically, the S8 specifically includes:

[0140] Taking the active power loss degree objective function, the voltage stability degree objective function and the load balancing degree objective function as optimization targets, a comprehensive objective function is constructed;

[0141] The switch state and reactive power compensation amount are used as decision variables;

[0142] Based on the predicted total load of the same group of branches, a switch state constraint equation is established, and each branch in the same group satisfies the switch state constraint equation;

[0143] Under the condition that the switch state constraint equation is established, the solution is performed based on the BAS algorithm, and the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are obtained according to the solution results;

[0144] The comprehensive objective function is:

[0145]

[0146] In the formula, are the weight parameters of each objective function, which are used to balance the importance of different objectives.

[0147] Specifically, the solution is performed based on the BAS algorithm, and the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are obtained according to the solution results, which specifically include:

[0148] The switch state and reactive compensation amount of the distribution network are combined into a high K-dimensional vector K. The switch state is represented by a binary variable (0 for open, 1 for closed), and the reactive compensation amount is represented by a continuous variable (indicating the size of the compensated reactive power).

[0149] Set the position of the beetle to x, that is, the initial switch state and reactive compensation amount combination, and the coordinates of the left and right whiskers of the beetle are represented by x l and x r , the distance between the two whiskers of the longicorn is d0;

[0150] Calculate the comprehensive objective function value corresponding to the position of the longicorn;

[0151] Based on the principle of longhorn beetle search behavior, the antennae movement of longhorn beetles is simulated, the maximum number of iterations of the outer loop and the inner loop is set, a direction vector is randomly generated, and the direction vector is normalized to obtain a random vector

[0152] Determine the coordinates of the left and right whiskers of the longhorn beetle and decide whether to accept the new position according to the Metropolis criterion;

[0153] According to the change of the coordinates of the left and right whiskers of the longhorn beetle, the movement direction and distance of the longhorn beetle in the next iteration are determined, and the inner loop step length is updated according to the demand to determine whether the cooling times have been reached. If not, the inner loop search is continued; if so, the outer loop step length is updated according to the demand;

[0154] After reaching the maximum number of iterations, the optimal solution is output;

[0155] Based on the switch planning results and reactive power compensation results, the optimization scheme for each node and branch of the distribution network is determined, and the optimization scheme is converted into specific control instructions, which are sent to the distribution network control system to execute switch operations and reactive power compensation operations to achieve distribution network reconstruction optimization;

[0156] The random vector The calculation formula for normalization is:

[0157]

[0158] Where rand(·) represents a randomly generated direction vector.

[0159] Specifically, the switch state constraint equation is:

[0160]

[0161] In the formula, G is the set of branches in the same group, W pred To predict the total load, S n is the switch state of branch n, S n ∈{0,1}, 0 means open, 1 means closed, W n is the load of branch n, γ1 and γ2 are adjustment coefficients, which are usually 0.8 and 1.2, and can also be adjusted based on actual conditions.

[0162] In summary, the advantages of the present invention are: taking "minimization of active power loss", "minimization of voltage deviation index" and "load balancing" as comprehensive optimization goals, constructing a multi-objective optimization model, compared with the traditional single-objective optimization method, it can comprehensively improve the optimization performance of distribution network grid reconstruction; by introducing a fault prediction model based on a convolutional neural network, training fault data under different topological structures, and combining cross-validation and grid search for hyperparameter tuning, the function of accurately predicting the fault type in real-time operation data is realized; the three key goals of minimizing active power loss, minimizing voltage deviation index and load balancing are comprehensively considered, and a comprehensive objective function is constructed based on these goals; by introducing a switch state constraint equation, the feasibility of the optimization scheme in actual operation is ensured. In addition, the BAS algorithm is used for solving, and the algorithm has a strong global search capability, improves the global optimality of the optimization result, converts the optimization scheme into a specific control instruction, and sends it to the distribution network control system to execute switch operations and reactive power compensation operations, thereby realizing dynamic reconstruction optimization of the distribution network and improving the operation efficiency and reliability of the system.

[0163] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm, characterized in that: The steps include: S1. Obtain node information and branch information of the distribution network grid, deploy distribution automation terminals, and obtain distribution network operation data, including real-time distribution network operation data and distribution network historical operation data sets; S2. Obtain distribution network operation fault data set, and build a distribution network fault prediction model based on the operation fault data of distribution networks with different topological structures; S3. Based on the historical operation data set of the distribution network, the historical load change curve of each node of the distribution network is obtained, and the corresponding historical meteorological data, historical social data and electricity type data are obtained based on the historical load change curve. The historical meteorological data, historical social data and electricity type data are analyzed for correlation factors with the historical load change curve, and a dynamic load prediction model is constructed. Based on the dynamic load prediction model, the predicted total load of the branch connected to the same power equipment in the current time period is obtained; S4. Based on the real-time operation data of the distribution network, the current topological structure information of the distribution network grid is obtained, and the distribution network operation status indicators are calculated based on the real-time operation data of the distribution network. The distribution network operation status indicators include the active power loss of each node, the load balance index and the voltage deviation index; S5. Set the active power loss limit, load balancing index limit and voltage deviation index limit based on the current topology information, and determine whether the distribution network needs to be reconstructed and optimized based on the over-limit conditions of active power loss, load balancing index and voltage deviation index. If it is determined that reconstruction and optimization are not required, exit; If it is determined that reconstruction optimization is needed, proceed to S6; S6. Determine whether there is a fault in the current distribution network based on the distribution network fault prediction model. If there is a fault, output the fault information, process the fault branch of the distribution network based on the fault information, and then proceed to S7; If there is no fault, go directly to S7; S7. Obtain the current topological structure information of the distribution network, take minimization of active power loss, minimization of voltage deviation index and load balancing as optimization goals, and respectively construct an active power loss degree objective function, a voltage stability degree objective function and a load balancing degree objective function; S8. Based on the active power loss degree objective function, the voltage stability degree objective function and the load balance degree objective function, a comprehensive objective function is constructed, and based on the predicted total load of the same group of branches, a switch state constraint equation is established. Under the condition that the switch state constraint equation is established, based on the BAS algorithm, the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are solved; S9. Determine the optimization plan for each node of the distribution network according to the switch planning results and reactive power compensation results, and reconstruct and optimize the distribution network based on the optimization plan.

2. According to claim 1, the multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm is characterized in that: In S2, a distribution network fault prediction model is constructed based on the operation fault data of distribution networks with different topological structures, which specifically includes: Based on the operation fault data set of the distribution network, the topological structure information of the distribution network when the operation fault occurs is obtained, wherein the operation fault data set includes the fault type, fault location, fault duration, fault occurrence time and operation parameters, and the topological structure information includes the switch state, line connection relationship and node type; The above data are preprocessed and the operating fault data set is divided into a training set and a test set. The convolutional neural network model is trained using the training set, and the model parameters are adjusted through the back propagation algorithm. The trained distribution network fault prediction model is used to predict the fault types of the distribution network under different topological structures. During the training process of the distribution network fault prediction model, cross validation and grid search are used for hyperparameter tuning to optimize the performance of the distribution network fault prediction model.

3. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 2 is characterized in that: The S3 specifically includes: Extract the load value of each node in each time period from the historical operation data set of the distribution network; Arrange the load values ​​of each node in chronological order to form time series data; With time as the horizontal axis and load value as the vertical axis, use visualization tools to draw the historical load change curve of each node; Obtain the load value of each time period in the historical load change curve, remove the node data with a load value of 0, and obtain the historical meteorological data, historical social data and electricity consumption type data of each node corresponding to each time period. The historical meteorological data includes temperature, humidity, light intensity and wind speed. The historical social data includes holiday information and large-scale event information. The electricity consumption type data includes residential electricity consumption, commercial electricity consumption, industrial electricity consumption and agricultural electricity consumption; Preprocess the above data, integrate historical meteorological data, historical social data and data on electricity consumption types of each node, generate relevant impact data, conduct correlation analysis on the load values ​​of each node obtained in the same time period and the relevant impact data, and use the Pearson correlation coefficient to calculate the correlation between the relevant impact data and the load values ​​of each node; Based on the correlation between the relevant impact data and the load value of each node, the correlation index of the relevant impact data is obtained, and a dynamic load prediction model is constructed based on the correlation index; Obtain real-time meteorological data, real-time social data and power consumption data of each node in the distribution network in the current time period, input the above data into the dynamic load forecasting model, and obtain the predicted load of the branch in each operating state in the current time period; The branches connected to each power device in the distribution network are obtained, the branches connected to the same power device are recorded as a group, and the predicted loads of the branches in the same group are added together to be recorded as the predicted total load.

4. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 3 is characterized in that: The S4 specifically includes: Based on the real-time operation data of the distribution network, the node-branch association matrix is ​​obtained, and the topological structure information of the distribution network grid is generated using a graph theory method. The real-time operation data includes the state of each switch, branch capacity, branch apparent power, branch active power, branch reactive power, node current, node voltage, transformer state, line connection relationship and node type; Calculate the distribution network operation status indicators based on the real-time operation data of the distribution network. The distribution network operation status indicators include the active power loss, load balance index and voltage deviation index of each node; The active power loss calculation formula is: Where P loss,m is the active power loss of node m, N is the number of branches connected to node m in the distribution network, R n is the resistance of branch n, P m and Q n are the active power and reactive power flowing through branch n, V m is the voltage amplitude of node m; The voltage deviation index calculation formula is: Where ΔV is the voltage deviation index, V m,rated is the reference voltage of node m, M is the total number of nodes, V m is the voltage amplitude at node m. The load balancing index calculation formula is: Where, L b is the load balancing index, B is the total number of branches, S n is the apparent power of branch n, S max,n is the capacity of branch n, S l is the apparent power of branch l, S max,l is the capacity of branch l.

5. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 4 is characterized in that: The S5 specifically includes: Set the active power loss limit P based on the current topology loss，max , Load balancing index limit L b，max And voltage deviation index limit ΔV max ; Determine the current active power loss P of each node loss Is it more than P loss，max ; Determine the current load balancing index L b Is it more than L b，max ; Determine whether the current voltage deviation index ΔV exceeds ΔV max ; If all indicators are within the limit, there is no need to restructure and optimize, and the program can be exited; If any indicator exceeds the limit, reconstruction optimization is required.

6. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 5 is characterized in that: The step S6 of determining whether there is a fault in the current distribution network based on the distribution network fault prediction model specifically includes: Obtain the current topological structure information of the distribution network, input the real-time operation data and topological structure information of the distribution network into the distribution network fault prediction model, output the suspected fault type and fault probability of the distribution network under the current topological structure, and calculate the fault impact index of the suspected fault type; If the fault impact index exceeds the set impact threshold, it is determined that there is a fault in the current distribution network, and abnormal real-time operation data is obtained based on the suspected fault type. The fault location algorithm is used to locate the fault area based on the abnormal real-time operation data. The fault area refers to the branch / node affected by the fault. The fault branch is disconnected by the isolation switch, and the load is transferred from the fault branch to other normally operating branches through the tie switch to ensure that the fault area is isolated from other normally operating areas; If the fault impact index does not exceed the set impact threshold, it is determined that there is no fault in the current distribution network.

7. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 6 is characterized in that: The S7 specifically includes: The current topological structure information of the distribution network is obtained, and the minimization of active power loss, minimization of voltage deviation index and load balancing are taken as optimization goals. The active power loss degree objective function f1, load balancing degree objective function f2 and voltage stability degree objective function f3 are constructed respectively.

8. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 6 is characterized in that: The S8 specifically includes: Taking the active power loss degree objective function, the voltage stability degree objective function and the load balancing degree objective function as optimization targets, a comprehensive objective function is constructed; The switch state and reactive power compensation amount are used as decision variables; Based on the predicted total load of the same group of branches, a switch state constraint equation is established, and each branch in the same group satisfies the switch state constraint equation; Under the condition that the switch state constraint equation is established, the solution is performed based on the BAS algorithm, and the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are obtained according to the solution results; The comprehensive objective function is: In the formula, are the weight parameters of each objective function.

9. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 6 is characterized in that: The solution is performed based on the BAS algorithm, and the switch planning results and reactive power compensation results of each branch of the distribution network in the current time period are obtained according to the solution results, specifically including: The switch state and reactive compensation amount of the distribution network are combined into a high K-dimensional vector K, where the switch state is represented by a binary variable and the reactive compensation amount is represented by a continuous variable; Set the position of the beetle to x, that is, the initial switch state and reactive compensation amount combination, and the coordinates of the left and right whiskers of the beetle are represented by x l and x r , the distance between the two whiskers of the longicorn is d0; Calculate the comprehensive objective function value corresponding to the position of the longicorn; Based on the principle of longhorn beetle search behavior, the antennae movement of longhorn beetles is simulated, the maximum number of iterations of the outer loop and the inner loop is set, a direction vector is randomly generated, and the direction vector is normalized to obtain a random vector Determine the coordinates of the left and right whiskers of the longhorn beetle, and decide whether to accept the new position according to the Metropolis criterion; According to the change of the coordinates of the left and right whiskers of the longhorn beetle, the movement direction and distance of the longhorn beetle in the next iteration are determined, and the inner loop step length is updated according to the demand to determine whether the cooling times have been reached. If not, the inner loop search is continued; if so, the outer loop step length is updated according to the demand; After reaching the maximum number of iterations, the optimal solution is output; Based on the switch planning results and reactive power compensation results, the optimization scheme for each node and branch of the distribution network is determined, and the optimization scheme is converted into specific control instructions, which are sent to the distribution network control system to execute switch operations and reactive power compensation operations to achieve distribution network reconstruction optimization.

10. The multi-objective optimization method for comprehensive energy distribution network reconstruction based on BAS algorithm according to claim 8 is characterized in that: The switch state constraint equation is: In the formula, G is the set of branches in the same group, W pred To predict the total load, S n is the switch state of branch n, S n ∈{0,1}, 0 means open, 1 means closed, W n is the load of branch n, γ1 and γ2 are adjustment coefficients.

Images