Power-distribution-network self-healing method and system taking photovoltaic output into consideration

Abstract

Provided in the present invention are a power-distribution-network self-healing method and system taking photovoltaic output into consideration. The method comprises: acquiring historical operation data of a photovoltaic power station and irradiance observation data from a meteorological station; on the basis of the historical operation data of the photovoltaic power station and the irradiance observation data from the meteorological station, predicting the generated power of the photovoltaic power station by using a convolutional long-short-term memory recurrent neural network model that takes sparrow search into consideration; on the basis of the generated power of the photovoltaic power station, a segment-switch state of a power distribution network and a network topology of the power distribution network, constructing an objective function and a constraint condition for a power-distribution-network self-healing model, and obtaining the power-distribution-network self-healing model; solving the power-distribution-network self-healing model by using a propagation search algorithm, so as to obtain an optimal recovery strategy; and executing the optimal recovery strategy by means of segmented switches and node loads. The present invention can realize self-healing of a power distribution network while ensuring the minimum power generation cost of a distributed power source, the minimum network loss and the minimum node voltage deviation.

Description

A power distribution network fault self-healing method and system considering photovoltaic output TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid, and in particular to a power distribution network fault self-healing method and system considering photovoltaic output. BACKGROUND

[0002] A large number of renewable distributed power generation systems are connected to the power distribution network, so that the structure of the traditional power distribution network becomes an active power distribution network. The accelerated connection of the distributed power grid to the power distribution network brings major challenges to the operation of the power distribution network. Ensuring that the active power distribution network can quickly recover and reconfigure when a fault occurs and continue to maintain safe, economic and reliable operation has become an important proposition in the field of power distribution network fault recovery technology. The existing power distribution network fault recovery technology does not comprehensively consider the constraint conditions when establishing the fault recovery model, and easily ignores the material constraint conditions, resulting in low efficiency and poor recovery effect of the power distribution network fault self-healing. At the same time, in order to objectively reflect the output range of the photovoltaic power station, it is necessary to predict the distributed energy output, and the existing technology is affected by the hyperparameter setting, which requires a lot of manual parameter adjustment time, thereby resulting in low prediction efficiency. SUMMARY

[0003] The present application provides a power distribution network fault self-healing method and system considering photovoltaic output, which can realize power distribution network fault self-healing under the premise of ensuring minimum distributed power generation cost, minimum network loss and minimum node voltage deviation.

[0004] In order to solve the above technical problems, the present application provides a power distribution network fault self-healing method considering photovoltaic output, comprising:

[0005] Obtaining historical operation data of a photovoltaic power station and light intensity observation data of a weather station;

[0006] Based on the historical operation data of the photovoltaic power station and the light intensity observation data of the weather station, a convolutional long short-term memory recurrent neural network model considering sparrow search is used to predict the power generation of the photovoltaic power station;

[0007] According to the power generation of the photovoltaic power station, the sectional switch state of the power distribution network and the network topology of the power distribution network, a target function and a constraint condition of a power distribution network fault self-healing model are constructed to obtain the power distribution network fault self-healing model; wherein the target function of the power distribution network fault self-healing model includes load recovery amount, distributed power supply operation cost, network loss after self-healing and voltage deviation after self-healing; the constraint conditions of the power distribution network fault self-healing model include power flow constraints, Kirchhoff constraints, radiative constraints, branch current constraints, node voltage constraints, feeder capacity constraints, transformer overload constraints, distributed power supply capacity constraints, fault repair constraints and material constraints;

[0008] The optimal recovery strategy is obtained by using a propagation search algorithm to solve a power distribution network fault self-healing model.

[0009] The optimal recovery strategy is executed through sectional switches and node loads.

[0010] The present application uses a convolutional long short-term memory recurrent neural network model considering sparrow search to predict the output of distributed energy, which can realize the hyperparameter optimization of complex combined models under the premise of ensuring the prediction accuracy, improve the efficiency of distributed energy output prediction; then combine the output of distributed energy and other power distribution network data to construct a power distribution network fault self-healing model, add material constraints to the constraint conditions to make the model constraints more comprehensive; finally, the optimal solution of the power distribution network fault self-healing model is obtained by using a propagation search algorithm, and the on-off state of the normally open switch and the grid-connected and off-grid operation state of the distributed power supply are optimized, so that the active power distribution network is repaired with the minimum load reduction, and the distributed power generation cost, network loss and node voltage deviation are minimized under the premise of ensuring the minimum distributed power generation cost, network loss and node voltage deviation, realizing the power distribution network fault self-healing.

[0011] Further, the convolutional long short-term memory recurrent neural network model considering sparrow search is used to predict the power generation of the photovoltaic power station based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station, including:

[0012] Based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station, the first correlation between the power of the photovoltaic power station and the light condition of the weather station is calculated.

[0013] According to the historical operation data of the photovoltaic power station and the first correlation, the convolutional long short-term memory recurrent neural network model considering sparrow search is used to predict the power generation of the photovoltaic power station.

[0014] Further, the first correlation between the power of the photovoltaic power station and the light condition of the weather station is calculated based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station, including:

[0015] A first multi-dimensional vector of the historical operation data of the photovoltaic power station and a second multi-dimensional vector of the observation data of the light intensity of the weather station are constructed, and the cosine similarity of the first multi-dimensional vector and the second multi-dimensional vector is calculated; wherein the cosine similarity is specifically as follows:

[0016] In the formula, similarity and cos(θ) represent the cosine similarity between the first multi-dimensional vector of the historical operation data of the photovoltaic power station and the second multi-dimensional vector of the observation data of the light intensity of the weather station; A represents the first multi-dimensional vector of the historical operation data of the photovoltaic power station, A i represents the components in the first multi-dimensional vector; B represents the second multi-dimensional vector of the observation data of the light intensity of the weather station, Bi represents a component in the second multi-dimensional vector; n represents the dimension of the multi-dimensional vector;

[0017] Based on the historical operation data of the photovoltaic power station and the observation data of the meteorological station light intensity, a light characteristic variable space of the photovoltaic power station is constructed; wherein the light characteristic variable space of the photovoltaic power station is specifically as follows:

[0018] In the formula, X i,F represents the light characteristic variable space of the photovoltaic power station i; X i,s represents the s-th light characteristic variable, that is, the s-th virtual light; P i,j represents the meteorological station light radiation intensity value of the photovoltaic power station i at the j-th period;

[0019] In combination with the cosine similarity and the light characteristic variable space of the photovoltaic power station, a first correlation between the photovoltaic power station power and the meteorological station light condition is obtained.

[0020] Further, the convolutional long short-term memory recurrent neural network model considering sparrow search is used to predict the photovoltaic power station power according to the historical operation data of the photovoltaic power station and the first correlation, which comprises:

[0021] According to the historical operation data of the photovoltaic power station and the first correlation, an input time matrix is constructed;

[0022] The input time matrix is input into the convolutional long short-term memory recurrent neural network model considering sparrow search; wherein the convolutional long short-term memory recurrent neural network model considering sparrow search comprises a convolutional long short-term memory network layer and a recurrent neural network;

[0023] The input time matrix is convolved layer by layer using the convolutional long short-term memory network layer to obtain convolutional data; wherein the convolutional long short-term memory network layer comprises a plurality of filters; the convolutional operation of the k-th filter on the input time matrix is specifically as follows: h k = RELU(W k *X+b k )

[0024] In the formula, h k represents the feature vector output by the k-th filter convolution kernel; * represents convolution operation, RELU(x) = max(0, x) represents an activation function, X represents the input time matrix, W k represents the weight matrix of the k-th convolution kernel, b k represents the bias vector of the k-th convolution kernel;

[0025] The convolutional data is input into the recurrent network layer for long sequence prediction to obtain the photovoltaic power station power.

[0026] The application uses a convolutional long short-term memory recurrent neural network algorithm considering sparrow search to predict photovoltaic power station power generation, first uses a convolutional long short-term memory network to perform multi-layer convolution on input variables to obtain convolutional data, so as to extract short-term features in the time dimension and local dependence between variables, and then uses a recurrent neural network to perform long sequence prediction on the convolutional data to obtain predicted distributed energy output, that is, predicted photovoltaic power station power generation, realizes hyperparameter optimization of a complex combined model under the premise of ensuring prediction accuracy, and improves the efficiency of distributed energy output prediction.

[0027] Further, the objective function of the power distribution network fault self-healing model comprises:

[0028] The objective function of the power distribution network fault self-healing model comprises load recovery amount, distributed power operation cost, network loss after self-healing and voltage deviation after self-healing, wherein the objective function is specifically as follows:

[0029] In the formula, f represents a multi-objective optimization function, F1 represents load recovery amount, F2 represents distributed power operation cost, F3 represents network loss after self-healing, F4 represents voltage error after self-healing, a, b, c and d respectively represent weight coefficients of load recovery amount, distributed power cost, network loss and voltage deviation;

[0030] The load recovery amount is specifically as follows:

[0031] In the formula, F1 represents load recovery amount, N T represents the total number of time intervals in the grid recovery period, and t represents the time of the grid recovery period; L represents a load set, and L k represents the priority weight of the load k, L represents all nodes on the line, and μ k represents the state quantity of the load k;

[0032] The distributed power operation cost is specifically as follows:

[0033] In the formula, F2 represents distributed power operation cost, N T represents the total number of time intervals in the grid recovery period, and t represents the time of the grid recovery period; DC represents a distributed power load set, and P DG,k represents the active output power of the kth distributed power in the power distribution network, represents the rated output power of the kth distributed power in the power distribution network;

[0034] The network loss after self-healing is specifically as follows:

[0035] In the formula, F3 represents the network loss; N T represents the total number of time intervals within the grid recovery period, t represents the time of the grid recovery period, N i represents the number of branches; x i represents the state of the i-th switching node, R i represents the resistance of branch i, I i represents the current passing through branch i, P TX,norm represents the rated power of the distribution transformer;

[0036] The voltage offset after self-healing is specifically as follows:

[0037] In the formula, F4 represents the voltage error after self-healing; N T represents the total number of time intervals within the grid recovery period, t represents the time of the grid recovery period, N z represents the number of buses; V m represents the actual voltage value of point z, V0 represents the rated voltage value of point z.

[0038] Further, the constraint conditions of the power distribution network fault self-healing model include:

[0039] The constraint conditions of the power distribution network fault self-healing model include power flow constraints, Kirchhoff constraints, radial constraints, branch current constraints, node voltage constraints, feeder capacity constraints, transformer overload constraints, distributed power capacity constraints, fault repair constraints and material constraints;

[0040] The power flow constraints are specifically as follows:

[0041] In the formula, IE t and OE t respectively represent the edge set flowing into node t and the edge set flowing out of node t, and respectively represent the total inflow power of node t and the total outflow power of node t, P t represents the power demand of node t; represents the imaginary part of the current flowing into the distributed power source k; and respectively represent the real part and the imaginary part of the voltage of the distributed power source k; P DG,k and Q DG respectively represent the active power and the reactive power of the distributed power source k;

[0042] The Kirchhoff constraints are specifically as follows:

[0043] wherein, and denote the imaginary and real parts of the load current flowing out of node k, and denote the imaginary and real parts of the distributed source current flowing into node k, and denote the imaginary and real parts of the current flowing into branch mk, and denote the imaginary and real parts of the current flowing out of branch km, k denote the set of nodes k;

[0044] The radiative constraints are specified as follows:

[0045] wherein γ is a binary variable, B denote the set of system branches, N denote the set of system nodes, SW denote the set of open system branches, sub denote the set of substation nodes of the system, B / Γ SW denote the set of branches after removing open branches, N ∣Γ sub denote the set of system nodes after removing substation nodes, s(k) denotes the set of all nodes connected to node k; y km denotes the state of the kmth branch, y km = 1 when the kmth branch is closed, y km = 0 when the kmth branch is open; denotes the state at node k on the kmth branch, denotes the state at node m on the kmth branch;

[0046] The fault repair constraints are specified as follows:

[0047] wherein F denotes the set of vectors composed of the start and end nodes of all fault lines, N T denotes the total number of time intervals within the grid recovery period; denotes the switching state of the branch with node i as the start point and node j as the end point within the fault recovery time period t; denotes the initial state of the fault; denotes the switching state of the branch with node i as the start point and node j as the end point within the fault recovery time period t m ; denotes the switching state of the branch with node i as the start point and node j as the end point within the fault recovery time period t nswitching state of the branch with node i as the start point and node j as the end point; denotes the fault recovery time point t+T f switching state of the branch with node i as the start point and node j as the end point; h denotes the maximum number of lines that can be simultaneously repaired in each time slot during the fault recovery process; T f denotes the time spent on repairing the fault line;

[0048] The material constraint is specifically as follows:

[0049] In the formula, sf i,t denotes the operation state of the fault line i at time t, sf i,t-1 denotes the operation state of the fault line i at time t-1; ma i , mb i and mc i respectively denote the number of materials a, material b and material c required in a single maintenance period of the fault line i, and respectively denote the upper limit of the callable number of material a, material b and material c in a single maintenance period, material a is a metal material, material b is an insulating material, and material c is a consumable substance; IL denotes the number of fault lines.

[0050] The application takes the load recovery maximization, the distributed power generation cost minimization, the line network loss minimization and the voltage deviation minimization as objective functions, adds a material constraint condition in the constraint condition, combines the dynamic network reconstruction and the active island operation of the distributed power supply to construct a power distribution network fault self-healing model, and solves the problem of incomplete consideration of constraints when the model is constructed.

[0051] Further, the use of the propagation search algorithm to solve the power distribution network fault self-healing model and obtain the optimal recovery strategy comprises the following steps.

[0052] Obtain the network parameters and fault data of the fault lines in the power distribution network, and set the segment switch state on the line on which the fault node is located to zero;

[0053] Initialize the first search agent and the population according to the objective function of the power distribution network fault self-healing model, and construct a search propagation algorithm model based on the first search agent;

[0054] Based on the network parameters and fault data of the fault lines in the power distribution network, the search propagation algorithm model is used to optimally solve the objective function of the power distribution network fault self-healing model, and the optimal recovery strategy is obtained.

[0055] Further, the initialization of the first search agent and the population according to the objective function of the power distribution network fault self-healing model and the construction of the search propagation algorithm model based on the first search agent comprise the following steps.

[0056] Initialize a first search agent and a population according to an objective function of a power distribution network fault self-healing model; wherein the first search agent is specifically as follows: X v = LB + r x (UB - LB) X i = LB + R x (UB - LB)

[0057] In the formula, X v and X i represent a search agent, X v = [X v1 , X v2 , X v3 , … X vd ] T , X i = [X i1 , X i2 , X i3 , … X id ] T ; LB and UB represent a vector upper bound and a vector lower bound respectively, UB = [ub1, ub2, ub3, … ub d ] T , LB = [lb1, lb2, lb3, … lb d ] T ; r represents a random number; R = [r1, r2, r3, … r d ] T ;

[0058] Construct a search propagation algorithm model based on the first search agent; wherein the search propagation algorithm model is specifically as follows:

[0059] In the formula, the search agent X v represents a voltage, and the search agent X i represents a current; and represent a best search agent, V R represents a voltage, I R represents a current, γ represents a propagation constant, and l represents a transmission line length; t represents a current iteration, T represents a total iteration number, r1, r2 and r3 represent random numbers uniformly distributed between 0 and 1, and r n represents a random number with a mean value of 0.

[0060] Further, the search propagation algorithm model is used to optimally solve the objective function of the power distribution network fault self-healing model based on network parameters and fault data of a fault line in the power distribution network, to obtain an optimal recovery strategy, including:

[0061] Calculate the fitness values of all individuals in the population, obtain a first optimal fitness value and a first suboptimal fitness value;

[0062] Iteratively update the population using the search propagation algorithm model until a preset stopping condition is reached, stop the iterative update and output the current search propagation algorithm model output result, and obtain an optimal recovery strategy; Wherein, during each iteration update, according to the self update of all individuals in the population, the first search agent, the first optimal fitness value and the first suboptimal fitness value are updated using the search propagation algorithm model, and the second search agent, the second optimal fitness value and the second suboptimal fitness value are obtained; According to the second optimal fitness value and the second suboptimal fitness value, the individuals in the population are crossed.

[0063] The present application uses the search propagation algorithm to solve the fault self-healing model optimally, so as to seek the optimal scheme of power distribution network self-healing, and according to the model solving result, the on-off state of the normally open switch and the grid-connected and off-grid operation state of the distributed power supply are optimized, so that the active power distribution network is repaired with the minimum load reduction, and on the premise of ensuring the minimum generation cost of the distributed power supply, the minimum network loss and the minimum node voltage deviation, the power distribution network fault self-healing is realized.

[0064] On the basis of the above method embodiment, the present application correspondingly provides a system embodiment, and provides a power distribution network fault self-healing system considering photovoltaic output, comprising: a historical data module, a power generation power module, a model construction module, a model solving module and a fault self-healing module;

[0065] The historical data module is used to obtain historical operation data of a photovoltaic power station and illumination intensity observation data of a meteorological station;

[0066] The power generation power module is used to predict the power generation power of the photovoltaic power station based on the historical operation data of the photovoltaic power station and the illumination intensity observation data of the meteorological station, and use a convolution long short-term memory recurrent neural network model considering sparrow search;

[0067] The model construction module is used to construct the objective function and the constraint condition of the power distribution network fault self-healing model according to the power generation power of the photovoltaic power station, the state of the power distribution network sectional switch and the network topology of the power distribution network, and obtain the power distribution network fault self-healing model; Wherein, the objective function of the power distribution network fault self-healing model includes load recovery amount, distributed power supply operation cost, network loss after self-healing and voltage deviation after self-healing; The constraint condition of the power distribution network fault self-healing model includes power flow constraint, Kirchhoff constraint, radiative constraint, branch current constraint, node voltage constraint, feeder capacity constraint, transformer overload constraint, distributed power supply capacity constraint, fault repair constraint and material constraint;

[0068] The model solving module is configured to solve the power distribution network fault self-healing model by using a propagation search algorithm to obtain an optimal recovery strategy.

[0069] The fault self-healing module is configured to execute the optimal recovery strategy by using sectional switches and node loads. BRIEF DESCRIPTION OF DRAWINGS

[0070] FIG. 1 is a flowchart of one embodiment of the power distribution network fault self-healing method considering photovoltaic output provided by the present application;

[0071] FIG. 2 is a structure diagram of one embodiment of the convolutional long short-term memory recurrent neural network model considering sparrow search provided by the present application;

[0072] FIG. 3 is a flowchart of one embodiment of the search propagation algorithm used to solve the objective function provided by the present application;

[0073] FIG. 4 is a flowchart of one embodiment of the power distribution network fault self-healing provided by the present application;

[0074] FIG. 5 is a structure diagram of one embodiment of the power distribution network fault self-healing system considering photovoltaic output provided by the present application. DETAILED DESCRIPTION

[0075] The technical solutions in the present application will be described clearly and completely below in conjunction with the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of the present application.

[0076] In the description of the present application, it should be understood that the terms "first", "second" and "third" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" and "third" can be explicitly or implicitly included one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "several" is two or more.

[0077] The present application provides a power distribution network fault self-healing method and system considering photovoltaic output, which can realize power distribution network fault self-healing under the premise of ensuring minimum distributed power generation cost, minimum network loss and minimum node voltage deviation.

[0078] Embodiment 1

[0079] Based on the above needs, an embodiment of the present application provides a power distribution network fault self-healing method considering photovoltaic output, and the method flow is shown in FIG. 1. The method includes steps S1 to S5, and each step is specifically as follows:

[0080] S1, obtain historical operation data of the photovoltaic power station and observation data of the light intensity of the weather station.

[0081] In order to objectively reflect the output range of the photovoltaic power station, the embodiment needs to obtain historical operation data of the photovoltaic power station and observation data of the light intensity of the weather station from the operation data at the past time, which is beneficial to the subsequent prediction of the distributed resource output.

[0082] S2, based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station, use the convolution long short-term memory recurrent neural network model considering sparrow search to predict the power generation of the photovoltaic power station. This step includes steps S2.1 to S2.2, each of which is as follows:

[0083] S2.1, based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station, calculate the first correlation of the power of the photovoltaic power station and the light condition of the weather station. This step includes steps S2.1.1 to S2.1.3, each of which is as follows:

[0084] S2.1.1, construct a first multi-dimensional vector of the historical operation data of the photovoltaic power station and a second multi-dimensional vector of the observation data of the light intensity of the weather station, and calculate the cosine similarity of the first multi-dimensional vector and the second multi-dimensional vector; wherein the cosine similarity is as follows:

[0085] In the formula, similarity and cos(θ) represent the cosine similarity between the first multi-dimensional vector of the historical operation data of the photovoltaic power station and the second multi-dimensional vector of the observation data of the light intensity of the weather station; A represents the first multi-dimensional vector of the historical operation data of the photovoltaic power station, A i represents the components in the first multi-dimensional vector; B represents the second multi-dimensional vector of the observation data of the light intensity of the weather station, B i represents the components in the second multi-dimensional vector; n represents the dimension of the multi-dimensional vector.

[0086] In this embodiment, based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station at the past time, a multi-dimensional vector is constructed, the first multi-dimensional vector of the historical operation data of the photovoltaic power station and the second multi-dimensional vector of the observation data of the light intensity of the weather station are obtained, and the cosine similarity of the first multi-dimensional vector and the second multi-dimensional vector is calculated.

[0087] In this embodiment, the historical operation data of the photovoltaic power station is represented as a vector a=(x1, y1), the observation data of the light intensity of the weather station is represented as b=(x2, y2), and then according to the two-dimensional vectors a and b, multi-dimensional vectors A=(x1, y1, z1, …) and B=(x2, y2, z2, …) are introduced.

[0088] S2.1.2, based on the historical operation data of the photovoltaic power station and the observation data of the light intensity of the weather station, a light characteristic variable space of the photovoltaic power station is constructed; wherein the light characteristic variable space of the photovoltaic power station is specifically as follows:

[0089] In the formula, X i,F represents the light characteristic variable space of the photovoltaic power station i; X i,s represents the s-th light characteristic variable, that is, the s-th virtual light; P i,j represents the light intensity radiation value of the weather station of the photovoltaic power station i at the j-th time period.

[0090] In order to consider the correlation between the photovoltaic power station power and the weather station light condition, the selected light intensity data is reconstructed in space in this embodiment, that is, the virtual light near the target photovoltaic power station site is constructed, and the light characteristic variable space of the photovoltaic power station is obtained. After the space reconstruction is completed, the operation state of the photovoltaic power station can be predicted according to the historical operation data of the photovoltaic power station and the reconstructed state variable space.

[0091] S2.1.3, the first correlation between the photovoltaic power station power and the weather station light condition is obtained in combination with the cosine similarity and the light characteristic variable space of the photovoltaic power station.

[0092] S2.2, the sparrow search convolutional long short-term memory recurrent neural network model is used to predict the photovoltaic power station power according to the historical operation data of the photovoltaic power station and the first correlation. This step includes steps S2.2.1 to S2.2.4, and each step is specifically as follows:

[0093] S2.2.1, an input time matrix is constructed according to the historical operation data of the photovoltaic power station and the first correlation.

[0094] In this embodiment, the input time matrix is constructed according to the historical operation data of the photovoltaic power station and the first correlation between the photovoltaic power station power and the weather station light condition. Wherein the input time matrix is represented as X=[X i ,X i,f ,X i,tl ,X i,th ], X i and X i,f represent the historical input variables of the photovoltaic power station, that is, the historical operation data of the photovoltaic power station; X i,tl and X i,th represent the light intensity radiation values of the main weather stations near the location of the photovoltaic power station.

[0095] S2.2.2, input the input time matrix into a convolutional long short-term memory recurrent neural network model considering sparrow search; wherein the convolutional long short-term memory recurrent neural network model considering sparrow search comprises a convolutional long short-term memory network layer and a recurrent neural network.

[0096] The convolutional long short-term memory recurrent neural network model considering sparrow search in the embodiment is shown in FIG. 2. In the embodiment, the input time matrix X = [X i ,X i,f ,X i,tl ,X i,th ] is input into the convolutional long short-term memory recurrent neural network model considering sparrow search (SSA-CNN-LSTM model), and the target is to rollingly predict the power generation of a photovoltaic power station in a future period of time.

[0097] S2.2.3, performing layer-by-layer convolution on the input time matrix using the convolutional long short-term memory network layer to obtain convolutional data.

[0098] In the SSA-CNN-LSTM model, the convolutional long short-term memory network layer is used to extract spatial feature information of the reconstruction space. In the embodiment, the input time matrix X = [X i ,X i,f ,X i,tl ,X i,th ] is input into the convolutional long short-term memory network layer, and the input time matrix is subjected to layer-by-layer convolution by the convolutional long short-term memory network layer to extract short-term features in the time dimension and local dependency relationships between variables, thereby obtaining convolutional data.

[0099] The convolutional long short-term memory network layer comprises a plurality of filters. The convolutional operation of the kth filter on the input time matrix is specifically as follows: h k = RELU(W k *X + b k )

[0100] In the formula, h k represents a feature vector output by the convolutional kernel of the kth filter; * represents convolution operation, RELU(x) = max(0, x) represents an activation function, X represents the input time matrix, W k represents a weight matrix of the kth convolutional kernel, and b k represents a bias vector of the kth convolutional kernel.

[0101] S2.2.4, inputting the convolutional data into a recurrent network layer for long sequence prediction to obtain power generation of the photovoltaic power station.

[0102] In the embodiment, the convolution data is input into the recursive network layer as a whole for long sequence prediction to obtain the power generation of the photovoltaic power station. The recursive neural network layer is specifically as follows:

[0103] In the formula, i t represents the input gate, f t represents the forget gate, o t represents the output gate, c t represents the cell state of the network, represents the candidate state memory unit, h t represents the output of the current state; σ represents the sigmoid activation function, tanh represents the hyperbolic tangent activation function, W xi , W xf , W xo and W xc respectively represent the weights of x t in the activation functions of the input gate, the forget gate, the output gate and the cell state, W hi , W hf , W ho and W hc respectively represent the weights of h t-1 in the activation functions of the input gate, the forget gate, the output gate and the cell state, b i , b f , b o and b c respectively represent the bias vectors of the activation functions of the input gate, the forget gate, the output gate and the cell state, and ⊙ represents the element-wise multiplication of vectors according to the corresponding positions.

[0104] S3, according to the power generation of the photovoltaic power station, the state of the distribution network sectional switch and the network topology of the distribution network, a target function and a constraint condition of a distribution network fault self-healing model are constructed to obtain the distribution network fault self-healing model; wherein the target function of the distribution network fault self-healing model includes the load recovery amount, the distributed power operation cost, the network loss after self-healing and the voltage deviation after self-healing; the constraint conditions of the distribution network fault self-healing model include the power flow constraint, the Kirchhoff constraint, the radiation constraint, the branch current constraint, the node voltage constraint, the feeder capacity constraint, the transformer overload constraint, the distributed power capacity constraint, the fault repair constraint and the material constraint. The step includes step S3.1 and step S3.2, and each step is specifically as follows:

[0105] S3.1, a target function of a distribution network fault self-healing model is constructed.

[0106] When the distribution network detects a multi-point short-line fault, the embodiment constructs a target function of a distribution network fault self-healing model according to the power generation of the photovoltaic power station, the state of the distribution network sectional switch and the network topology of the distribution network.

[0107] The objective function of the power distribution network fault self-healing model includes load recovery amount, distributed power operation cost, network loss after self-healing, and voltage deviation after self-healing; wherein the objective function is specifically as follows:

[0108] In the formula, f represents a multi-objective optimization function; F1 represents a load recovery amount, F2 represents a distributed power operation cost, F3 represents a network loss after self-healing, F4 represents a voltage error after self-healing, and a, b, c, and d respectively represent weight coefficients of the load recovery amount, the distributed power cost, the network loss, and the voltage deviation.

[0109] The load recovery amount is specifically as follows:

[0110] In the formula, F1 represents a load recovery amount; N T represents a total number of time intervals in a grid recovery period, and t represents a time of the grid recovery period; Γ L represents a load set, and L k represents a priority weight of the load k, L represents all nodes on a line, and μ k represents a state amount of the load k.

[0111] In the above load recovery amount objective, the load importance levels represented as 1, 2, and 3 correspond to the priority weights L k When the objective of the power distribution network fault self-healing is only to maximize the load recovery, the weight coefficient of the load recovery amount can be set to 1.

[0112] The distributed power operation cost is specifically as follows:

[0113] In the formula, F2 represents a distributed power operation cost; N T represents a total number of time intervals in a grid recovery period, and t represents a time of the grid recovery period, Γ DC represents a distributed power load set, and P DG,k represents an active output power of the kth distributed power in the power distribution network, represents a rated output power of the kth distributed power in the power distribution network.

[0114] The network loss after self-healing is specifically as follows:

[0115] In the formula, F3 represents a network loss; N T represents a total number of time intervals in a grid recovery period, and t represents a time of the grid recovery period, N i represents a number of branches; and x iRi represents the state of the ith switching node, R i Ri represents the resistance of branch i, I i Ii represents the current through branch i, P TX,norm Pi represents the rated power of the distribution transformer.

[0116] The voltage offset after self-healing is specifically as follows:

[0117] In the formula, F4 represents the voltage error after self-healing; N T N represents the total number of time intervals in the grid recovery period, and t represents the time of the grid recovery period. z V represents the number of buses; V m Vz represents the actual voltage value of point z, and V0 represents the rated voltage value of point z.

[0118] S3.2, constraint condition for constructing a distribution network fault self-healing model.

[0119] When the distribution network detects a multi-point short-line fault, the embodiment constructs a constraint condition for the distribution network fault self-healing model according to the power generation of the photovoltaic power station, the state of the distribution network sectional switch, and the network topology of the distribution network.

[0120] The constraint condition for the distribution network fault self-healing model includes power flow constraints, Kirchhoff constraints, radial constraints, branch current constraints, node voltage constraints, feeder capacity constraints, transformer overload constraints, distributed power capacity constraints, fault repair constraints, and material constraints.

[0121] The power flow constraints are specifically as follows:

[0122] In the formula, IE t and OE t respectively represent the edge set flowing into node t and the edge set flowing out of node t, and respectively represent the total inflow power of node t and the total outflow power of node t, P t represents the power demand of node t; represents the imaginary part of the current flowing into the distributed power source k; and respectively represent the real part and the imaginary part of the voltage of the distributed power source k; P DG,k and Q DG respectively represent the active power and the reactive power of the distributed power source k.

[0123] The Kirchhoff constraints are specifically as follows:

[0124] In the formula, and and and and and and and and k denotes the set of nodes k.

[0125] The radiance constraint is specifically as follows:

[0126] where γ is a binary variable, and B denotes the set of system branches, and N denotes the set of system nodes, and SW denotes the set of disconnected branches of the system, and sub denotes the set of substation nodes of the system, and B / Γ SW denotes the set of branches after removing disconnected branches, and N |Γ sub denotes the set of system nodes after removing substation nodes, and s(k) denotes the set of all nodes connected to node k; and y km denotes the state of the kmth branch, and y km = 1 when the kmth branch is closed, and y km = 0 when the kmth branch is disconnected. denotes the state of node k on the kmth branch, denotes the state of node m on the kmth branch.

[0127] In the above radiance constraint, the radiance constraint is ensured by introducing a binary variable to correspond the topology of the distribution network and the spanning tree connected to the main substation, and each node has only one parent node except the root node (substation node).

[0128] The fault repair constraint is specifically as follows:

[0129] where F denotes a vector set composed of starting nodes and ending nodes of all fault lines, and N T denotes the total number of time intervals in the grid recovery period; denotes the switching state of the branch with node i as the starting point and node j as the ending point in the fault recovery time period t; denotes the initial state of the fault; denotes the fault recovery time period t mswitching state of the branch from node i to node j; denotes the fault recovery time period t n switching state of the branch from node i to node j; denotes the fault recovery time period t+T f switching state of the branch from node i to node j; h denotes the maximum number of lines that can be repaired simultaneously in each time slot during the fault recovery process; T f denotes the time spent on repairing the fault line.

[0130] In the above fault repair constraint condition, is a 0-1 variable. Before the fault is removed from the distribution network, the fault line opening state variable (verification expression) is forced to be set to 0. Expression describes the initial state of the fault, and all fault lines are in the open state. Expression ensures that the maximum number of h fault lines can be serviced in each time slot during the fault recovery process.

[0131] The material constraint is as follows:

[0132] In the formula, sf i,t denotes the operating state of the fault line i at time t, sf i,t-1 denotes the operating state of the fault line i at time t-1; ma i , mb i , and mc i respectively denote the amount of material a, material b, and material c required in a single repair cycle of the fault line i, and respectively denote the upper limit of the callable amount of material a, material b, and material c in a single repair cycle. Material a is a metal material, material b is an insulating material, and material c is a consumable substance; IL denotes the number of fault lines.

[0133] In the process of removing faults from the distribution network, three types of materials are usually used, including metal materials, insulating materials, and low-value consumable substances. In the above material constraint condition, material a represents a metal material such as copper, aluminum, etc.; material b represents an insulating material such as rubber, plastic, etc.; and material c represents a low-value, consumable substance such as a welding rod, welding wire, etc. For the operating state of the fault line i at time t, sf i,t = 0 indicates that the fault line is in an open state, and sf i,t = 1 indicates that the fault line is in a closed state.

[0134] S4, solving the power distribution network fault self-healing model using a propagation search algorithm to obtain an optimal restoration strategy. This step includes steps S4.1 to S4.3, each of which is as follows:

[0135] S4.1, obtaining network parameters and fault data of the fault line in the power distribution network, and setting the sectionalizing switch state on the line on which the fault node is located to zero.

[0136] This embodiment obtains network parameters and fault data of the fault line in the power distribution network, and then sets the sectionalizing switch state on the line on which the fault node is located to zero according to the network parameters and fault data of the fault line.

[0137] S4.2, initializing a first search agent and a population according to the objective function of the power distribution network fault self-healing model, and constructing a search propagation algorithm model based on the first search agent. This step includes steps S4.2.1 to S4.2.2, each of which is as follows:

[0138] S4.2.1, initializing a first search agent and a population according to the objective function of the power distribution network fault self-healing model; wherein the first search agent is as follows: X v = LB + r x (UB - LB) X i = LB + R x (UB - LB)

[0139] In the formula, X v and X i represent a search agent, X v = [X v1 , X v2 , X v3 ,... X vd ] T , X i = [X i1 , X i2 , X i3 ,... X id ] T ; LB and UB represent the upper and lower bounds of the vector, respectively, UB = [ub1, ub2, ub3,...ub d ] T , LB = [lb1, lb2, lb3,...lb d ] T ; r represents a random number; R = [r1, r2, r3,...r d ] T .

[0140] The embodiment uses a propagation search algorithm (PSA algorithm) to solve the power distribution network fault self-healing model. First, the basic calculation parameters of the PSA algorithm are initialized, and then the search agent is initialized using the PSA algorithm: first, the dimension (1*D) of the cost function is used to randomly initialize the vector of the search agent between the upper bound (LB) and the lower bound (UB). n search agents can be initialized, and the dimension of all search agents is (n*D).

[0141] S4.2.2, constructing a search propagation algorithm model based on the first search agent; wherein the search propagation algorithm model is as follows:

[0142] In the formula, the search agent X v represents the voltage, the search agent X i represents the current. and represents the best search agent, V R represents the voltage, I R represents the current, γ represents the propagation constant, l represents the length of the transmission line; t represents the current iteration, T represents the total number of iterations, r1, r2 and r3 represent random numbers uniformly distributed between 0 and 1, and r n represents a random number with a mean value of 0.

[0143] S4.3, based on the network parameters and fault data of the fault line in the power distribution network, using the search propagation algorithm model to optimally solve the objective function of the power distribution network fault self-healing model, and obtaining the optimal recovery strategy. This step includes steps S4.3.1 to S4.3.2, and each step is as follows:

[0144] S4.3.1, calculating the fitness value of all individuals in the population to obtain the first optimal fitness value and the first optimal fitness value.

[0145] The embodiment uses the search propagation algorithm model to calculate the fitness value of all individuals in the population, obtain the first optimal fitness value and the first optimal fitness value, and record the first optimal fitness value as the first optimal fitness value is recorded as

[0146] S4.3.2, iteratively update the population using the search propagation algorithm model until a preset stopping condition is reached, stop the iterative update and output a current search propagation algorithm model output result, and obtain an optimal recovery strategy; wherein, in each iteration update, according to self-updating of all individuals in the population, the first search agent, the first optimal fitness value and the first suboptimal fitness value are updated using the search propagation algorithm model to obtain the second search agent, the second optimal fitness value and the second suboptimal fitness value; and according to the second optimal fitness value and the second suboptimal fitness value, the individuals in the population are subjected to a crossover operation.

[0147] In each iteration update process of the search propagation algorithm for solving the target function, the embodiment first updates the first search agent, the first optimal fitness value and the first suboptimal fitness value according to the search propagation algorithm model of step S4.2.2 to obtain the second search agent, the second optimal fitness value and the second suboptimal fitness value; and then, according to the updated second search agent, the second optimal fitness value and the second suboptimal fitness value, excellent individuals are selected from the population for a crossover operation, wherein the excellent individuals are the first N individuals in the population arranged in descending order of fitness value, and N is a preset positive integer.

[0148] The specific process of using the search propagation algorithm to optimally solve the target function is shown in FIG. 3. After the optimal solution of the target function using the search propagation algorithm is completed, the optimal recovery strategy under the conditions of load balancing and optimal operation benefit is obtained.

[0149] S5, execute the optimal recovery strategy through sectional switches and node loads.

[0150] The embodiment distributes the optimal recovery strategy obtained according to step S4 to sectional switches and loads in each fault recovery stage, such as on-off partial line sectional switches or off-supply partial non-essential loads.

[0151] The specific process of implementing power distribution network fault self-healing of the embodiment is shown in FIG. 4. The embodiment uses the PSA algorithm to solve the power distribution network fault self-healing model to obtain an optimal recovery strategy with minimum distributed power generation cost, minimum network loss and minimum node voltage deviation, and implements power distribution network fault self-healing by executing the optimal recovery strategy.

[0152] The above embodiments of the present application have the following beneficial effects:

[0153] The application provides a power distribution network fault self-healing method considering photovoltaic output, uses a convolution long short-term memory recurrent neural network model considering sparrow search to predict the output of distributed energy, can realize hyperparameter optimization of a complex combination model under the premise of ensuring prediction accuracy, and improves the efficiency of distributed energy output prediction; then, a power distribution network fault self-healing model is constructed in combination with the output of distributed energy and other power distribution network data, a material constraint condition is added to the constraint condition to make the model constraint more comprehensive; finally, the power distribution network fault self-healing model is optimally solved through a propagation search algorithm, the on-off state of a normally open switch and the grid-connected and off-grid operation state of a distributed power supply are optimized, the active power distribution network is repaired with minimum load reduction, and the power distribution network fault self-healing is realized under the premise of ensuring that the distributed power generation cost, network loss and node voltage deviation are minimum.

[0154] Embodiment 2

[0155] Based on the above-mentioned content of the embodiment, an embodiment of the application provides a power distribution network fault self-healing system considering photovoltaic output, which comprises a historical data module 101, a power generation power module 102, a model construction module 103, a model solving module 104 and a fault self-healing module 105. The system structure is shown in Figure 5.

[0156] The historical data module 101 is used to obtain historical operation data of a photovoltaic power station and illumination intensity observation data of a meteorological station.

[0157] The power generation power module 102 is used to predict the power generation power of the photovoltaic power station based on the historical operation data of the photovoltaic power station and the illumination intensity observation data of the meteorological station using a convolution long short-term memory recurrent neural network model considering sparrow search.

[0158] The model construction module 103 is used to construct a target function and a constraint condition of a power distribution network fault self-healing model according to the power generation power of the photovoltaic power station, the state of a power distribution network sectional switch and the network topology of the power distribution network, and obtain the power distribution network fault self-healing model; wherein the target function of the power distribution network fault self-healing model comprises a load recovery amount, a distributed power supply operation cost, a network loss after self-healing and a voltage deviation after self-healing; and the constraint condition of the power distribution network fault self-healing model comprises a power flow constraint, a Kirchhoff constraint, a radiative constraint, a branch current constraint, a node voltage constraint, a feeder capacity constraint, a transformer overload constraint, a distributed power supply capacity constraint, a fault repair constraint and a material constraint.

[0159] The model solving module 104 is used to solve the power distribution network fault self-healing model using a propagation search algorithm, and obtain an optimal recovery strategy.

[0160] The fault self-healing module 105 is used to execute the optimal recovery strategy through sectional switches and node loads.

[0161] The working principle and step flow of the embodiment can be seen in the related description of embodiment 1, but are not limited thereto.

[0162] The above embodiment of the present application has the following beneficial effects:

[0163] The power distribution network fault self-healing system considering photovoltaic output provided by the application comprises a historical data module, a power generation power module, a model construction module, a model solving module and a fault self-healing module, and can realize power distribution network fault self-healing under the premise of guaranteeing minimum distributed power generation cost, minimum network loss and minimum node voltage deviation.

[0164] The above is the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled persons in the art, without departing from the principle of the present application, a number of improvements and refinements can be made, and these improvements and refinements are also considered within the protection scope of the present application.

Claims

1. A power distribution grid fault self-healing method considering photovoltaic output, characterized in that, The method comprises the following steps: acquiring historical operation data of a photovoltaic power station and light intensity observation data of a weather station; predicting power generation of the photovoltaic power station based on the historical operation data of the photovoltaic power station and the light intensity observation data of the weather station by using a convolutional long short-term memory recurrent neural network model considering sparrow search; constructing an objective function and constraint conditions of a power distribution network fault self-healing model according to the power generation of the photovoltaic power station, states of sectional switches of the power distribution network and a network topology of the power distribution network, and obtaining the power distribution network fault self-healing model; wherein the objective function of the power distribution network fault self-healing model comprises load recovery amount, distributed power operation cost, network loss after self-healing and voltage deviation after self-healing; the constraint conditions of the power distribution network fault self-healing model comprise power flow constraint, Kirchhoff constraint, radiative constraint, branch current constraint, node voltage constraint, feeder capacity constraint, transformer overload constraint, distributed power capacity constraint, fault repair constraint and material constraint; solving the power distribution network fault self-healing model by using a propagation search algorithm to obtain an optimal recovery strategy; executing the optimal recovery strategy through sectional switches and node loads.

2. A power distribution network fault self-healing method considering photovoltaic output power as claimed in claim 1, characterized by, The method of predicting power generation of the photovoltaic power station based on the historical operation data of the photovoltaic power station and the light intensity observation data of the weather station by using the convolutional long short-term memory recurrent neural network model considering sparrow search comprises the following steps: calculating a first correlation between power of the photovoltaic power station and light conditions of the weather station based on the historical operation data of the photovoltaic power station and the light intensity observation data of the weather station; predicting power generation of the photovoltaic power station by the convolutional long short-term memory recurrent neural network model considering sparrow search according to the historical operation data of the photovoltaic power station and the first correlation.

3. A power distribution network fault self-healing method that considers photovoltaic output as claimed in claim 2, characterized by, The method of calculating the first correlation between power of the photovoltaic power station and light conditions of the weather station based on the historical operation data of the photovoltaic power station and the light intensity observation data of the weather station comprises the following steps: A first multi-dimensional vector of historical operation data of a photovoltaic power station and a second multi-dimensional vector of light intensity observation data of a meteorological station are constructed, and a cosine similarity of the first multi-dimensional vector and the second multi-dimensional vector is calculated; wherein the cosine similarity is specifically as follows: In the formula, similarity and cos(0) represent the cosine similarity between the first multi-dimensional vector of the photovoltaic power station historical operation data and the second multi-dimensional vector of the meteorological station light intensity observation data; A represents the first multi-dimensional vector of the photovoltaic power station historical operation data, A i represents the components in the first multi-dimensional vector; B represents the second multi-dimensional vector of the meteorological station light intensity observation data, B i represents the components in the second multi-dimensional vector; n represents the dimension of the multi-dimensional vector; Based on historical operation data of a photovoltaic power station and light intensity observation data of a weather station, a light characteristic variable space of the photovoltaic power station is constructed; wherein the light characteristic variable space of the photovoltaic power station is specifically as follows: wherein X i,F represents the light characteristic variable space of the photovoltaic power station i; X i,s represents the s-th light characteristic variable, i.e. the s-th virtual light; P i,j represents the meteorological station light radiation intensity value of the photovoltaic power station i at the j-th time period; in combination with the cosine similarity and the light characteristic variable space of the photovoltaic power station, a first correlation between the photovoltaic power station power and the meteorological station light condition is obtained.

4. A power distribution network fault self-healing method considering photovoltaic output power as claimed in claim 2, characterized by, The method of predicting power generation of the photovoltaic power station by the convolutional long short-term memory recurrent neural network model considering sparrow search according to the historical operation data of the photovoltaic power station and the first correlation comprises the following steps: constructing an input time matrix according to the historical operation data of the photovoltaic power station and the first correlation; inputting the input time matrix into the convolutional long short-term memory recurrent neural network model considering sparrow search; wherein the convolutional long short-term memory recurrent neural network model considering sparrow search comprises a convolutional long short-term memory network layer and a recurrent neural network; performing layer-by-layer convolution on the input time matrix by using the convolutional long short-term memory network layer to obtain convolution data; wherein the convolutional long short-term memory network layer comprises a plurality of filters; a convolution operation of a kth filter on the input time matrix is specifically as follows: h k = RELU(W k *X + b k ) In the formula, h k represents the feature vector output by the kth filter convolution kernel; * represents a convolution operation, RELU(x) = max(0, x) represents an activation function, X represents an input time matrix, W k represents the weight matrix of the kth convolution kernel, b k represents the bias vector of the kth convolution kernel; inputting the convolution data into the recurrent neural network layer to perform long sequence prediction and obtain power generation of the photovoltaic power station.

5. A power distribution network fault self-healing method considering photovoltaic output power as claimed in claim 1, characterized by, The objective function of the power distribution network fault self-healing model comprises: The objective function of the power distribution network fault self-healing model comprises a load recovery amount, a distributed power supply operation cost, a network loss after self-healing, and a voltage deviation after self-healing; wherein the objective function is specifically as follows: In the formula, f represents a multi-objective optimization function; F1 represents a load recovery amount, F2 represents a distributed power operation cost, F3 represents a network loss after self-healing, F4 represents a voltage error after self-healing, a, b, c and d respectively represent weight coefficients of the load recovery amount, the distributed power cost, the network loss and the voltage deviation; The load recovery amount, in particular, is as follows: wherein F1 represents the amount of load recovery; N T denotes the total number of time intervals within the grid recovery period, t denotes the time of the grid recovery period; Γ L denotes the set of loads, L k denotes the priority weight of load k, L denotes all nodes on the line, μ k denotes the state quantity of load k; The distributed power supply operation cost is specifically as follows: In the formula, F2 represents the distributed power operation cost; N T represents the total number of time intervals within the grid recovery period, t represents the time of the grid recovery period, Γ DC represents the distributed power load set, P DG,k represents the active output power of the kth distributed power in the power distribution network, P k represents a rated output power of the kth distributed power in the distribution network; The network loss after the self-healing is specifically as follows: where F3 represents network loss; N T represents the total number of time intervals within the grid recovery period, t represents the time of the grid recovery period, N i represents the number of branches; x i represents the state of the i-th switching node, R i represents the resistance of branch i, I i represents the current through branch i, P TX,norm represents the rated power of the distribution transformer; The voltage offset after the self-healing is specified as follows: In the formula, F4 represents the voltage error after self-recovery; N T represents the total number of time intervals within the grid recovery period, t represents the time of the grid recovery period, N z represents the total number of buses; V m represents the actual voltage value of the z point, V0 represents the rated voltage value of the z point.

6. A power distribution network fault self-healing method that considers photovoltaic output as claimed in claim 1, characterized by, The constraint condition of the distribution network fault self-healing model comprises: The constraint condition of the distribution network fault self-healing model comprises power flow constraints, Kirchhoff constraints, radial constraints, branch current constraints, node voltage constraints, feeder capacity constraints, transformer overload constraints, distributed power capacity constraints, fault repair constraints and material constraints; The tidal current constraint is specifically as follows: In the formula, IE t and OE t respectively represent the edge set flowing into node t and the edge set flowing out of node t, and respectively denote the total inflow power of node t and the total outflow power of node t, P t denotes the power demand of node t; Im(k) represents the imaginary part of the current flowing into the distributed power source k; and Re(k) and Im(k) represent the real and imaginary parts of the voltage of distributed generator k, respectively; P DG,k and Q DG P(k) and Q(k) represent the active and reactive power of distributed generator k, respectively; The Kirchhoff constraint, in particular, is as follows: In the formulae, and respectively denote the imaginary and real parts of the outgoing node k load current, and respectively denote the imaginary and real parts of the distributed generator current flowing into node k, and respectively denote the imaginary and real parts of the branch mkinflow current, and respectively denote the imaginary and real parts of the branch km outgoing current, Γ k denotes the set of nodes k; The radiation constraint is specified as follows: where γ is a binary variable, Γ B denotes the set of system branches, Γ N denotes the set of system nodes, Γ SW denotes the set of system disconnected branches, Γ sub denotes the set of substation nodes of the system, Γ B / Γ SW denotes the set of branches with disconnected branches removed, Γ N ∣Γ sub denotes the set of system nodes with substation nodes removed, s(k) denotes the set of all nodes connected to node k; y km denotes the state of the kth branch, y km = 1 when the kth branch is closed, y km = 0 when the kth branch is disconnected; represents the state at node k on the km branch, S km represents a state of node m on km branch; The failure recovery constraint is specified as follows: In the formula, F represents a vector set composed of all fault line start nodes and end nodes, N T represents the total number of time intervals within the grid recovery period; represents the switching state of the branch from node i to node j in the fault recovery period t; an initial state indicating occurrence of a fault; represents a fault recovery time period t m the switching state of the branch with node i as the start point and node j as the end point; represents a fault recovery time period t n the switching state of the branch with node i as the start point and node j as the end point; denotes the time to repair a failed link f the switching state of the branch from node i to node j; h denotes the maximum number of links that can be repaired simultaneously in each time slot during the failure recovery process; T f denotes the time to repair a failed link The material constraints, in particular, are as follows: wherein sf i,t denotes the operating state of the faulty line i at time t, sf i,t-1 denotes the operating state of the faulty line i at time t-1; ma i , mb i and mc i denote the amount of material a, material b and material c, respectively, required for a single repair cycle of the faulty line i, and IL represents a number of fault lines.

7. A power distribution network fault self-healing method that considers photovoltaic output as claimed in claim 1, characterized by, The method for solving the distribution network fault self-healing model by using the propagation search algorithm to obtain an optimal recovery strategy comprises: Obtaining network parameters and fault data of fault lines in the distribution network, and setting a sectionalizing switch state on a line on which a fault node is located to zero; Initializing a first search agent and a population according to an objective function of the distribution network fault self-healing model, and constructing a search propagation algorithm model based on the first search agent; Optimally solving the objective function of the distribution network fault self-healing model by using the search propagation algorithm model based on the network parameters and the fault data of the fault lines in the distribution network, and obtaining an optimal recovery strategy.

8. A power distribution network fault self-healing method that considers photovoltaic output as claimed in claim 7, characterized by, The method for initializing the first search agent and the population according to the objective function of the distribution network fault self-healing model, and constructing the search propagation algorithm model based on the first search agent comprises: Initializing the first search agent and the population according to the objective function of the distribution network fault self-healing model; wherein the first search agent is as follows: X v = LB + r x (UB - LB) X i = LB + R x (UB - LB) where X v and X i represent search agents, X v = [X v1 , X v2 , X v3 ,... X vd ] T , X i = [X i1 , X i2 , X i3 ,... X id ] T ; LB and UB represent vector upper bound and vector lower bound, respectively, UB = [ub1, ub2, ub3,... ub d ] T , LB = [lb1, lb2, lb3,... lb d ] T ; r represents random number; R = [r1, r2, r3,... r d ] T ; constructing a search propagation algorithm model based on the first search agent; wherein the search propagation algorithm model is specifically as follows: wherein the search agent X v represents a voltage, the search agent X i represents a current; and denotes the best search agent, V R denotes the voltage, I R denotes the current, γ denotes the propagation constant, l denotes the transmission line length; t denotes the current iteration, T denotes the total number of iterations, r1, r2, and r3 denote random numbers uniformly distributed between 0 and 1, r n denotes a random number with mean 0.

9. A power distribution network fault self-healing method that considers photovoltaic output as claimed in claim 7, characterized by, The method for optimally solving the objective function of the distribution network fault self-healing model by using the search propagation algorithm model based on the network parameters and the fault data of the fault lines in the distribution network, and obtaining an optimal recovery strategy comprises: Calculating fitness values of all individuals in the population, and obtaining a first optimal fitness value and a first suboptimal fitness value; Iteratively updating the population by using the search propagation algorithm model until a preset stop condition is reached, stopping the iterative updating and outputting a current search propagation algorithm model output result, and obtaining an optimal recovery strategy; wherein, at each time of iterative updating, the first search agent, the first optimal fitness value and the first suboptimal fitness value are updated by using the search propagation algorithm model according to self-updating of all individuals in the population, to obtain a second search agent, a second optimal fitness value and a second suboptimal fitness value; and individuals in the population are subjected to a crossover operation according to the second optimal fitness value and the second suboptimal fitness value.

10. A power distribution grid fault self-healing system that considers photovoltaic output, characterized by, The method comprises: The historical data module is configured to obtain historical operation data of a photovoltaic power station and light intensity observation data of a weather station. The historical data module is configured to obtain historical operation data of a photovoltaic power station and light intensity observation data of a weather station. The power generation module, based on historical operation data of a photovoltaic power station and light intensity observation data of a weather station, uses a convolutional long short-term memory recurrent neural network model considering sparrow search to predict power generation of the photovoltaic power station; The model construction module, according to the power generation of the photovoltaic power station, a sectional switch state of a distribution network and a network topology of the distribution network, constructs an objective function and a constraint condition of a distribution network fault self-healing model to obtain the distribution network fault self-healing model; wherein the objective function of the distribution network fault self-healing model includes a load recovery amount, a distributed power operation cost, a network loss after self-healing and a voltage deviation after self-healing; and the constraint condition of the distribution network fault self-healing model includes a power flow constraint, a Kirchhoff constraint, a radiative constraint, a branch current constraint, a node voltage constraint, a feeder capacity constraint, a transformer overload constraint, a distributed power capacity constraint, a fault repair constraint and a material constraint; The model solution module, using a propagation search algorithm to solve the distribution network fault self-healing model, obtains an optimal recovery strategy; The fault self-healing module, executing the optimal recovery strategy through sectional switches and node loads.

Images