A method and system for regulating and optimizing a multi-terminal flexible hub switch

Abstract

The application provides a regulation and optimization method for a multi-terminal feeding flexible hub switch, and belongs to the technical field of multi-terminal feeding flexible hub switches. The method comprises the following steps: constructing an adjacency matrix through the switch state and physical connection of the multi-terminal feeding flexible hub switch, outputting space-time fusion features of the adjacency matrix through a hierarchical graph convolution network and a bidirectional long short-term neural network; solving total network loss of each node according to the space-time fusion features of the multi-terminal feeding flexible hub switch; solving node marginal electricity price according to the total network loss of each node; solving node congestion cost according to the node marginal electricity price; constructing a total economic cost regulation and optimization model according to the total network loss and the node congestion cost of all nodes, with the minimum total economic cost as the target; solving the total economic cost regulation and optimization model by using a mixed integer programming, outputting an optimal economic cost regulation strategy, and realizing regulation and optimization of the multi-terminal feeding flexible hub switch. The application improves the reliability, real-time performance and accuracy of the regulation and optimization of the multi-terminal feeding flexible hub switch.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of multi-terminal feeding flexible hub switch, more particularly, to a multi-terminal feeding flexible hub switch regulation optimization method and system. BACKGROUND

[0002] The traditional multi-terminal feeding flexible hub switch based on MMC (Modular Multilevel Converter) directly changes the current path and voltage distribution by changing the switch state inside, but the traditional topology is simplified as a fixed connection, which cannot update the physical connection relationship, and ignoring the influence of real-time switch state on the topology structure will lead to errors in power flow calculation, deviating from the true value when calculating loss, which will lead to incorrect regulation decisions, and further cause economic losses, and the traditional static model cannot capture the topology transient process, and the delay of the central control instruction of the distribution network causes the bridge arm current in the current MMC topology to form resonance or the voltage to exceed the safety threshold, causing switch damage;

[0003] In addition, the traditional node marginal price solving only relies on the system power flow distribution, ignoring the network loss and congestion cost caused by real-time switch action, and the static model also leads to that the node marginal price solving is often based on outdated topology, which cannot guarantee the real-time synchronization of switch action and node marginal price, easily aggravating congestion, and cannot guarantee accurate regulation timing, leading to inaccurate regulation of the multi-terminal feeding flexible hub switch and further leading to the increase of total cost. SUMMARY

[0004] To solve the problems in the prior art, the present application provides a multi-terminal feeding flexible hub switch regulation optimization method and system, which fuses the topology state and electrical quantity by real-time switch state and branch current and voltage, so that the space-time fusion feature reflects the real path, improves the reliability of the multi-terminal feeding flexible hub switch, accurately represents the transient and steady state of the multi-terminal feeding flexible hub switch through the space-time fusion feature input by BiLSTM (Bidirectional Long Short Term Memory), improves the dynamic response speed of the switch action inside the multi-terminal feeding flexible hub switch, synchronizes the topology and switch state in real time through the dynamic adjacency matrix of the multi-terminal feeding flexible hub switch topology, suppresses the congestion cost caused by switch action disturbance by multiplying the number of switch actions, and combines with the real-time synchronized node marginal price to improve the real-time of the switch regulation of the congestion state.

[0005] The present application adopts the following technical solutions.

[0006] The first aspect of the present application provides a multi-terminal feeding flexible hub switch regulation optimization method, comprising:

[0007] An adjacency matrix is constructed by a switch state and a physical connection of the multi-terminal flexible hub switch, and a topological state feature of the multi-terminal flexible hub switch is input into a hierarchical graph convolution network and a bidirectional long short-term neural network as an initial state, and a spatio-temporal fusion feature of the multi-terminal flexible hub switch is output in combination with the adjacency matrix;

[0008] Total network losses of each node are solved according to the spatio-temporal fusion feature of the multi-terminal flexible hub switch;

[0009] A node marginal price is solved according to the total network losses of each node, and a node congestion cost is solved according to the node marginal price;

[0010] A total economic cost regulation and optimization model is constructed according to the total network losses and the node congestion costs of all nodes, with a minimum total economic cost as a target;

[0011] An optimal economic cost regulation and optimization strategy is output by solving the total economic cost regulation and optimization model, and the multi-terminal flexible hub switch is regulated and optimized.

[0012] Preferably, the spatio-temporal fusion feature of the multi-terminal flexible hub switch output in combination with the adjacency matrix comprises:

[0013] A topological state feature of the multi-terminal flexible hub switch is acquired;

[0014] An adjacency matrix is obtained by topological modeling of the multi-terminal flexible hub switch through a switch state and a physical connection, and a topological state feature of the multi-terminal flexible hub switch is input as an initial state, and the topological state feature is aggregated by a hierarchical graph convolution network in combination with the adjacency matrix to obtain a spatial distribution feature of the multi-terminal flexible hub switch;

[0015] A spatio-temporal fusion feature of the multi-terminal flexible hub switch is obtained by spatio-temporal feature fusion of the multi-terminal flexible hub switch through a bidirectional long short-term neural network.

[0016] Preferably, the topological state feature of the multi-terminal flexible hub switch comprises:

[0017] A voltage and a current of the multi-terminal flexible hub switch are acquired and normalized respectively to obtain a voltage feature and a current feature;

[0018] The current feature is vectorized into a one-dimensional current vector;

[0019] A comprehensive state vector is solved according to the one-dimensional current vector, the voltage feature vector and a switch state matrix;

[0020] The comprehensive state vector is input into an activation function to solve the topological state feature.

[0021] Preferably, the topology modeling of the multi-terminal flexible hub switch through the switch state and the physical connection obtains an adjacency matrix, and the topology state characteristics of the multi-terminal flexible hub switch are set as initial states, and the topology state characteristics are aggregated through a hierarchical graph convolution network combined with the adjacency matrix, including:

[0022] The multi-terminal flexible hub switch is topologically modeled according to the switch state matrix and the inherent physical connection matrix, and an adjacency matrix is obtained;

[0023] The topology state characteristics are set as initial state characteristics, and the state characteristics of the Lth layer are solved through an L-layer graph convolution network combined with the adjacency matrix;

[0024] The state characteristics of the Lth layer solved are set as the spatial distribution characteristics of the topology.

[0025] Preferably, the topology modeling of the multi-terminal flexible hub switch includes:

[0026] The inherent physical connection matrix is set according to whether there is physical connection between nodes;

[0027] The switch state matrix is set according to whether the switch is on;

[0028] The switch state matrix and the inherent physical connection matrix are multiplied element by element to obtain a node connectivity result;

[0029] The node connectivity result and the node self-loop matrix are added to obtain an adjacency matrix.

[0030] Preferably, the total network loss of each node is solved according to the spatiotemporal fusion characteristics of the multi-terminal flexible hub switch, including:

[0031] The spatiotemporal fusion characteristics of the multi-terminal flexible hub switch are mapped into loss-sensitive characteristics through a loss-sensitive weight matrix and an activation function;

[0032] The loss-sensitive characteristics and the physical loss are fused to generate a constraint loss matrix of each node;

[0033] The neighbor set loss of all nodes in the constraint loss matrix of each node is aggregated to obtain the total network loss of each node.

[0034] Preferably, the node marginal price is solved according to the total network loss of each node, including:

[0035] The regional average price of the power grid connected to the alternating current output end is multiplied by the system rated capacity to obtain the theoretical income per hour when the system is running at full load, and the historical average congestion cost is divided by the theoretical income per hour of the system when the system is running at full load to obtain the system price sensitivity weight;

[0036] ​Solve marginal network loss cost according to total network loss of each node, solve marginal congestion cost according to system electricity price sensitivity weight, sum marginal network loss cost, marginal congestion cost and marginal power generation cost to generate node marginal electricity price.

[0037] Preferably, the solving node congestion cost according to node marginal electricity price comprises:

[0038] Solve node marginal electricity price deviation by subtracting node marginal electricity price from benchmark electricity price;

[0039] Solve the sensitivity of switch action to topological characteristics by solving the partial derivative of the L2 norm of the space-time fusion characteristics of the multi-terminal feed-in flexible hub switch to the switch state;

[0040] The sensitivity of switch action to topological characteristics is multiplied by the number of switch actions and the node marginal electricity price deviation to obtain the node congestion cost.

[0041] Preferably, the total economic cost regulation optimization model is constructed according to the total network loss and node congestion cost of all nodes with the minimum total economic cost as the target, comprising:

[0042] Multiply the total network loss of all nodes, the node marginal electricity price of all nodes and the set time step to obtain the total network loss cost, sum all node congestion costs to obtain the total congestion cost, multiply the total network loss cost and the total congestion cost by the network loss weight and the congestion weight respectively and sum to construct the objective function of the total economic cost regulation optimization model;

[0043] The power flow balance constraint is constructed by subtracting the active power consumption of each node at time step t from the active power injection of each node at time step t and the active power net outflow of each node, and the total network loss of each node is equal to the total network loss of each node;

[0044] The device safety constraint is constructed by limiting the effective value of alternating voltage of each node at time step t and the current amplitude matrix of each node to the rest of the nodes at time t;

[0045] The topological radiation constraint is constructed by setting the sum of all adjacency matrix elements equal to twice the number of network branches, and the rank of the difference between the degree matrix and the adjacency matrix equal to the number of network branches.

[0046] The second aspect of the present application provides a multi-terminal feed-in flexible hub switch regulation optimization system, which runs the multi-terminal feed-in flexible hub switch regulation optimization method of the first aspect, comprising

[0047] The topology construction module is used for constructing the adjacency matrix by the switch state of the multi-terminal feed-in flexible hub switch and the physical connection, inputting the topological state characteristics of the multi-terminal feed-in flexible hub switch into the hierarchical graph convolution network and the bidirectional long short-term neural network as the initial state, and outputting the space-time fusion characteristics of the multi-terminal feed-in flexible hub switch in combination with the adjacency matrix.

[0048] a network loss solving module configured to solve total network losses of each node according to the space-time fusion feature of the multi-terminal feed-in flexible hub switch;

[0049] a node marginal electricity price solving module configured to solve node marginal electricity prices according to the total network losses of each node;

[0050] a node congestion cost solving module configured to solve node congestion costs according to the node marginal electricity prices;

[0051] a regulation optimization model construction module configured to construct a total economic cost regulation optimization model according to the total network losses and the node congestion costs of all nodes, with the minimum total economic cost as the target;

[0052] a solving module configured to solve the total economic cost regulation optimization model and output an optimal economic cost regulation strategy to implement regulation optimization on the multi-terminal feed-in flexible hub switch.

[0053] Compared with the prior art, the present application has at least the following beneficial effects:

[0054] The present application synchronizes the topology and the switch state in real time through the dynamic adjacency matrix of the multi-terminal feed-in flexible hub switch topology, reflects the topology change after the switch action in real time, calculates the node marginal electricity price in a timely and accurate manner, ensures the real-time synchronization of the switch action regulation instruction and the node marginal electricity price, and improves the real-time performance of the multi-terminal feed-in flexible hub switch regulation.

[0055] The present application fuses the topology state and the electrical quantity through the real-time switch state and the branch current and voltage, makes the space-time fusion feature reflect the real path, solves the loss matrix based on the real topology, makes the total network loss calculation more accurate, combines the corresponding power flow balance constraint, improves the accuracy of the switch action regulation, and further improves the reliability of the multi-terminal feed-in flexible hub switch regulation.

[0056] The present application accurately represents the transient state and the steady state of the multi-terminal feed-in flexible hub switch through the space-time fusion feature input by the BiLSTM, improves the dynamic response speed of the internal switch action of the multi-terminal feed-in flexible hub switch, combines the topology radiation constraint to ensure that no loop current path is generated in the multi-terminal feed-in flexible hub switch when the switch state is dynamically switched, and improves the safety of the multi-terminal feed-in flexible hub switch regulation.

[0057] The present application obtains the switch disturbance by taking the partial derivative of the switch through the space-time fusion feature, suppresses the congestion cost caused by the switch action disturbance by multiplying the switch action times, combines the real-time synchronized node marginal electricity price, improves the real-time performance of the switch regulation on the congestion state, ensures the accurate regulation timing, jointly optimizes the network loss cost and the congestion cost, and improves the real-time performance and the accuracy of the multi-terminal feed-in flexible hub switch regulation. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is a schematic diagram of a multi-terminal feed flexible hub switch control method provided according to an embodiment of the present application;

[0059] Figure 2 is a schematic diagram of a multi-terminal feed flexible hub switch topology provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. The described embodiments are only a part of the embodiments of the present application, not all the embodiments. All other embodiments obtained by those skilled in the art without creative labor based on the spirit of the present application shall fall within the protection scope of the present application.

[0061] As shown in Figure 1 and Figure 2 , embodiment 1 of the present application provides a multi-terminal feed flexible hub switch control optimization method, comprising the following steps:

[0062] Step 1: constructing an adjacency matrix by the switch state and physical connection of the multi-terminal feed flexible hub switch, inputting a topological state feature of the multi-terminal feed flexible hub switch into a hierarchical graph convolution network and a bidirectional long short-term neural network as an initial state level diagram, and outputting a spatiotemporal fusion feature of the multi-terminal feed flexible hub switch in combination with the adjacency matrix.

[0063] In a preferred but non-limiting embodiment of the present application, step 1 comprises:

[0064] Step 1.1, obtaining a topological state feature of the multi-terminal feed flexible hub switch.

[0065] Further preferably, step 1.1 comprises:

[0066] Step 1.1.1, obtaining voltage and current of the multi-terminal feed flexible hub switch and normalizing them respectively to obtain voltage features and current features, wherein the voltage features include battery port voltage and FBSM capacitor voltage, and the current features include bridge arm current.

[0067] Step 1.1.2, vectorizing the current features into a one-dimensional current vector, expressed as follows:

[0068]

[0069] In the formula, vec ( I t ) represents the one-dimensional current vector, I t represents the current features,flatten (.) indicates tensor expansion.

[0070] Step 1.1.3: Solve for the comprehensive state vector based on the one-dimensional current vector obtained in Step 1.1.2, the voltage eigenvector obtained in Step 1.1.1, and the switching state matrix. Z t It can be expressed by the following formula:

[0071]

[0072] In the formula,

[0073] W v Indicates the importance weight of voltage.

[0074] V t Represents the voltage eigenvector.

[0075] W i Indicates the circulation suppression weight.

[0076] W s This indicates the weight of the number of switching actions, used to evaluate the benefit of IGBT switching actions.

[0077] S t This represents the switching state matrix, which includes the on and off states of the IGBTs inside the FBSM, with 1 indicating on and 0 indicating off.

[0078] b represents the bias vector, used to compensate for errors in current and voltage acquisition sensors.

[0079] Step 1.1.4: Solve the topological state features of the comprehensive state vector obtained in Step 1.1.3 using an activation function, as expressed by the following formula:

[0080]

[0081] In the formula, ReLU represents the topological state characteristics, and ReLU represents the activation function.

[0082] Step 1.2: Set the spatial aggregation machine learning algorithm as a hierarchical graph convolutional network, perform topological modeling on the multi-terminal feeder flexible hub switch to obtain the adjacency matrix, and use the hierarchical graph convolutional network to combine the adjacency matrix to spatially aggregate the topological state features obtained in Step 1.1 to obtain the spatial distribution features of the multi-terminal feeder flexible hub switch.

[0083] More preferably, step 1.2 includes:

[0084] Step 1.2.1: Perform topological modeling of the multi-terminal feeder flexible hub switch based on the switch state matrix and the inherent physical connection matrix to obtain the adjacency matrix, including:

[0085] Set an inherent physical connection matrix based on whether there is physical connectivity between nodes;

[0086] Set the switch state matrix according to whether the switch is on or off;

[0087] The node connectivity result is obtained by multiplying the switch state matrix element by element with the inherent physical connection matrix.

[0088] The adjacency matrix is ​​obtained by adding the node connectivity result to the node self-loop matrix, realizing the topology modeling of the multi-terminal feed-in flexible hub switch, as expressed by the following formula:

[0089]

[0090] In the formula, Represents the adjacency matrix; This represents the inherent physical connection matrix, where 1 indicates that there is a physical connection between nodes and 0 indicates that there is no physical connection between nodes. It is used to describe the physical connection of PCB traces. This represents a self-loop matrix of nodes, with 1s on the diagonal and 0s on the rest, used to describe the connection between a node and itself.

[0091] Step 1.2.2: Set the topological state features output in Step 1.1 as the initial state features. Combine the adjacency matrix from Step 1.2.1 with an L-layer graph convolutional network to solve for the state features of the Lth layer, expressed by the following formula:

[0092]

[0093] In the formula, This represents the initial state characteristics, used to describe the state characteristics of the first layer; express Layer state characteristics, express Layer state characteristics, This represents the node degree matrix, used to describe the number of connections for each node. Indicates learnable weights, , This indicates the number of layers in the graph convolutional network.

[0094] Step 1.2.3, the solution obtained in Step 1.2.2 is... The state characteristics of the layer are set as the spatial distribution characteristics of the topology.

[0095] Step 1.3. Time-space feature fusion is performed on the spatial distribution features of the topology obtained in step 1.2 by a bidirectional long-short term neural network to obtain time-space fusion features.

[0096] Further preferably, step 1.3 comprises:

[0097] Step 1.3.1. A forward hidden state at the current time is solved according to the spatial distribution features of the topology, and is expressed in the following formula:

[0098]

[0099] In the formula, represents a forward LSTM function, represents a state feature of the Lth layer, represents a hidden state at the previous time, represents a forward hidden state at the current time.

[0100] Step 1.3.2. A backward hidden state at the current time is solved according to the spatial distribution features of the topology, and is expressed in the following formula:

[0101]

[0102] In the formula, represents a backward LSTM function, represents a hidden state at the next time, represents a backward hidden state at the current time.

[0103] Step 1.3.3. The forward hidden state at the current time in step 1.3.1 and the backward hidden state at the current time in step 1.3.2 are fused and normalized into time-space fusion features, and are expressed in the following formula:

[0104]

[0105] In the formula, LayerNorm represents layer normalization, and || represents splicing, W 0 represents a fusion weight.

[0106] Step 2: Total network loss of node i is solved according to the time-space fusion features of the multi-terminal feed-in flexible hub switch in step 1.

[0107] In the preferred but non-limiting real-time mode of the present application, step 2:

[0108] Step 2.1. The time-space fusion features of the multi-terminal feed-in flexible hub switch in step 1 are mapped into loss-sensitive features by a loss-sensitive weight matrix and an activation function, which are used to quantify the positive node loss overload risk and the negative energy feedback caused by braking to charge the battery, and are expressed in the following formula:

[0109]

[0110] wherein,

[0111] F loss,t denotes the loss-sensitive feature, used to mark the high-loss nodes,

[0112] tanh denotes the hyperbolic tangent activation function, used to constrain the physical range of loss,

[0113] W h denotes the loss-sensitive weight matrix, used to represent the contribution sensitivity of node state to loss,

[0114] b h denotes the measurement error compensation bias.

[0115] Step 2.2, fuse the loss-sensitive feature of step 2.1 with the physical loss to generate the node i constraint loss matrix, denoted as

[0116]

[0117] wherein,

[0118] Φij, t denotes the node i constraint loss matrix, used to describe the constraint loss matrix of branch starting node i to branch ending node j at time t,

[0119] |I ij,t | denotes the current amplitude matrix of node i to node j at time t, denotes the i and j non-connected nodes,

[0120] R denotes the resistance matrix, the resistance matrix R R ij,t denotes the equivalent resistance of branch starting node i to branch ending node j at time t,

[0121] denotes the physical loss term,

[0122] i,j=1,2,3…N , N is the total number of nodes.

[0123] Step 2.3, aggregate the neighbor set loss of all node i in the node i constraint loss matrix of step 2.2 to get the total network loss of node i, denoted as

[0124] ​

[0125] wherein,

[0126] P loss,i,t represents the total network loss of node i,

[0127] Φ ik,t represents the loss from branch starting node i to neighbor node k,

[0128] N (i) represents the neighbor node set of node i.

[0129] Step 3: according to the total network loss of node i solved in step 2, solve the node marginal electricity price.

[0130] In the preferred but non-limiting embodiment of the present application, step 3 comprises:

[0131] Step 3.1, multiply the regional average electricity price of the power grid connected by the alternating current output end by the system rated capacity to obtain the theoretical income per hour when the system is fully loaded, divide the historical average congestion cost by the theoretical income per hour when the system is fully loaded to obtain the system electricity price sensitivity weight W λ , which is expressed as follows:

[0132]

[0133] wherein,

[0134] W λ represents the system electricity price sensitivity weight, which is used to describe the proportion of congestion cost in unit energy income,

[0135] λ avg represents the regional average electricity price of the power grid connected by the alternating current output end,

[0136] P rated represents the system rated capacity, which corresponds to the sum of the three-phase bridge arm rated capacity,

[0137] C cong,avg represents the historical average congestion cost, which is obtained by subtracting the three-phase bridge arm current effective value from the maximum safe current and multiplying the result by the penalty coefficient and the sampling time interval to obtain the instantaneous congestion cost of the kth sampling, and then dividing the sum of all instantaneous congestion costs by the number of sampling points to obtain the historical average congestion cost.

[0138] Step 3.2, according to the total network loss of node i, solve the marginal network loss cost, according to the system electricity price sensitivity weight, solve the marginal congestion cost, sum the marginal network loss cost, the marginal congestion cost and the marginal power generation cost to generate the node marginal electricity priceλ i,t is expressed as follows:

[0139]

[0140] In the formula,

[0141] C G represents the total generation cost of node i, including the capacity degradation cost of battery discharge, operation and maintenance cost and energy loss cost;

[0142] P i represents the output power of node i;

[0143] I d represents the current overrun rate, which is used to describe the proportion of actual three-phase bridge arm current sensor exceeding the safety value;

[0144] represents the network loss sensitivity,

[0145] represents the marginal generation cost;

[0146] W λ ·I d ·λ avg represents the marginal congestion cost;

[0147] represents the marginal network loss cost.

[0148] Step 4, according to the node marginal price obtained in step 3, the node congestion cost is solved.

[0149] In the preferred but non-limiting embodiment of the present application, step 4 comprises:

[0150] The node marginal price deviation is solved by subtracting the reference price from the node marginal price;

[0151] The sensitivity of switch action to topology characteristics is obtained by solving the partial derivative of the L2 norm of the space-time fusion characteristics of the multi-terminal feed-in flexible hub switch to the switch state, which is used to quantify the disturbance intensity of switch action to topology;

[0152] The sensitivity of switch action to topology characteristics is multiplied by the number of switch actions and the node marginal price deviation to obtain the node congestion cost, which is expressed as follows:

[0153]

[0154] In the formula,

[0155] λ i,t represents the node marginal price, λ ref represents the reference price, λ i,t λ ref represents that the node i is in a power congestion state;

[0156] represents the node congestion cost,

[0157] represents the L2 norm,

[0158] represents the number of switch actions, and the more frequent the switch actions, the higher the risk of congestion;

[0159] represents the congestion cost sensitivity, which is used to describe the change in cost per switch action.

[0160] Step 5: A total economic cost regulation and optimization model is constructed by the total network loss of all nodes and the node congestion cost, with the minimum total economic time cost as the target.

[0161] In the preferred but non-limiting embodiments of the present application, step 5 includes:

[0162] Step 5.1, the objective function of the total economic cost regulation and optimization model constructed according to the total network loss of all nodes and the node congestion cost, with the minimum total economic time cost as the target, includes:

[0163] The total network loss cost is obtained by multiplying the total network loss of all nodes, the node marginal price of all nodes and the set time step, the total congestion cost is obtained by summing the congestion cost of all nodes, the total economic cost is obtained by multiplying the total network loss cost and the total congestion cost by the network loss weight and the congestion weight respectively and summing them, and the objective function of the total economic cost regulation and optimization model is expressed as follows:

[0164]

[0165] In the formula,

[0166] J t represents the total economic cost,

[0167] represents the time step;

[0168] and respectively represent the network loss weight and the congestion weight, . ​

[0169] Step 5.2, constructing constraints, including constructing power flow balance constraints, equipment safety constraints, and topological radiality constraints.

[0170] Further preferably, step 5.2 includes:

[0171] Step 5.2.1, constructing power flow balance constraints by subtracting the active power consumption of node i at time step t and the active power net outflow of node i from the active power injection of node i at time step t, which equals the total net loss of node i, is expressed as follows:

[0172]

[0173] In the formula,

[0174] P gen,i,t Pi(t) represents the active power injection of node i at time step t;

[0175] P load,i,t Pi(t) represents the active power consumption of node i at time step t;

[0176] B ik Yik represents the admittance of branch ik of node i and neighbor node k;

[0177] θ it φi(t) represents the voltage phase angle of node i at time step t;

[0178] Pi(t) represents the active power net outflow of node i.

[0179] Step 5.2.2, constructing equipment safety constraints by limiting the AC voltage effective value of node i at time step t to be within the minimum working voltage allowed by the system and the maximum working voltage allowed by the system, and limiting the current amplitude matrix of node i to node j at time t to be within the maximum safety current, is expressed as follows:

[0180]

[0181] In the formula,

[0182] V ij,t Vi(t) represents the AC voltage effective value of node i at time step t;

[0183] V min Vmin represents the minimum working voltage allowed by the system, V max Vmax represents the maximum working voltage allowed by the system;

[0184] I max,ij represents the maximum safe current.

[0185] Step 5.2.3, constructing a topological radiance constraint by taking the difference between the degree matrix and the adjacency matrix, and the rank of the difference is equal to the number of network branches, that is, the sum of all adjacency matrix elements is equal to 2 times the number of network branches N-1 , which is expressed in the following formula:

[0186]

[0187] In the formula,

[0188] represents the adjacency matrix element, 1 represents connection, and 0 represents disconnection;

[0189] N represents the total number of nodes;

[0190] rank(·) represents the rank of the matrix, which is used to describe the number of topological independent paths, and N-1 describes the ideal matrix rank of the radiate network of the present application, when formula (20) is satisfied, the multi-terminal feed flexible hub switch network of the present application is a connected acyclic structure, and there is no risk of circulating current;

[0191] L t represents the Laplace matrix, which is used to describe the network topology at time step t, .

[0192] Step 6, using mixed integer programming to solve the total economic cost regulation and optimization model obtained in step 5, the optimal economic cost regulation strategy, and realizing the regulation and optimization of the multi-terminal feed flexible hub switch.

[0193] In the preferred but non-limiting embodiments of the present application, step 6 includes:

[0194] Step 6.1, linearly transforming the space-time fusion features by the switch tendency weight, compressing the linear transformation result to the interval (0, 1) through the Sigmoid activation function, generating the switch tendency features, taking the integer of the switch tendency features, which is used to represent the switch state and describe the conduction probability of the switch through the switch tendency features, setting the switch times to be less than a set threshold, and generating an initial switch strategy;

[0195] Step 6.2, filtering out the feasibility strategy set of the switch strategy in step 6.1 through the power flow balance constraint, the device safety constraint and the topological radiance constraint.

[0196] Step 6.3, solving the optimal switch strategy according to the feasibility strategy set, which is the optimal economic cost regulation strategy, and realizing the regulation and optimization of the multi-terminal feed flexible hub switch.

[0197] The embodiment 2 of the present application provides a regulation and optimization system of a multi-terminal feed-in flexible hub switch, and the regulation and optimization method of the multi-terminal feed-in flexible hub switch is run, including:

[0198] A topology construction module is configured to construct an adjacency matrix by a switch state and a physical connection of the multi-terminal feed-in flexible hub switch, input a topological state feature of the multi-terminal feed-in flexible hub switch as an initial state into a hierarchical graph convolution network and a bidirectional long short-term neural network, and output a time-space fusion feature of the multi-terminal feed-in flexible hub switch in combination with the adjacency matrix;

[0199] A network loss solving module is configured to solve total network losses of each node according to the time-space fusion feature of the multi-terminal feed-in flexible hub switch;

[0200] A node marginal electricity price solving module is configured to solve node marginal electricity prices according to the total network losses of each node;

[0201] A node congestion cost solving module is configured to solve node congestion costs according to the node marginal electricity prices;

[0202] A regulation and optimization model construction module is configured to construct a total economic cost regulation and optimization model according to the total network losses and the node congestion costs of all nodes with the minimum total economic cost as a target;

[0203] A solving module is configured to solve the total economic cost regulation and optimization model, output an optimal economic cost regulation strategy, and implement regulation and optimization on the multi-terminal feed-in flexible hub switch.

[0204] Compared with the prior art, the present application has at least the following beneficial effects:

[0205] The present application synchronizes a real-time topology and a switch state through a dynamic adjacency matrix of a topology of the multi-terminal feed-in flexible hub switch, reflects a topology change after a switch action in real time, calculates a node marginal electricity price in a timely and accurate manner, ensures real-time synchronization of a switch action regulation instruction and the node marginal electricity price, and improves real-time performance of the multi-terminal feed-in flexible hub switch regulation;

[0206] The present application fuses a topology state and electrical quantities through a real-time switch state and branch current and voltage, makes a time-space fusion feature reflect a real path, solves a constraint loss matrix based on a real topology, makes total network loss calculation more accurate, combines corresponding power flow balance constraints, improves accuracy of the switch action regulation, and further improves reliability of the multi-terminal feed-in flexible hub switch regulation;

[0207] The application accurately characterizes the transient state and steady state of the multi-terminal feeding flexible hub switch through the spatio-temporal fusion feature input by BiLSTM, improves the dynamic response speed of the internal switch action of the multi-terminal feeding flexible hub switch, and guarantees the generation of the multi-terminal feeding flexible hub switch internal loopless path when the switch state is dynamically switched, in combination with the topological radiation constraint, and improves the safety of the multi-terminal feeding flexible hub switch regulation.

[0208] The switch disturbance is obtained by taking the partial derivative of the switch through the spatio-temporal fusion feature, the congestion cost caused by the switch action disturbance is suppressed by multiplication with the number of switch actions, and the real-time of the switch in regulating the congestion state is improved in combination with the real-time synchronized node marginal price, to ensure accurate regulation timing, and the network loss cost and congestion cost are jointly optimized to improve the real-time and accuracy of the multi-terminal feeding flexible hub switch regulation.

[0209] The present disclosure can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium having computer readable program instructions embodied therewith, which cause a processor to carry out various aspects of the present disclosure.

[0210] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit it. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application. Any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered within the protection scope of the claims of the present application.

Claims

1. A method for regulating and optimizing a multi-terminal fed flexible pivot switch, characterized in that, include: An adjacency matrix is ​​constructed by combining the switching state matrix and physical connection matrix of a multi-terminal fed flexible hub switch. The topological state features of the multi-terminal fed flexible hub switch are used as initial state features and input into a hierarchical graph convolutional network and a bidirectional long short-term neural network. The adjacency matrix is ​​then combined to output the spatiotemporal fusion features of the multi-terminal fed flexible hub switch, including: Obtain the topological state characteristics of a multi-terminal feeder flexible hub switch; The adjacency matrix is ​​obtained by performing topological modeling of the multi-terminal feeder flexible hub switch using the switch state matrix and physical connection matrix. The topological state features of the multi-terminal feeder flexible hub switch are used as the initial state features. The topological state features are aggregated by combining the adjacency matrix with a hierarchical graph convolutional network to obtain the spatial distribution features of the topology of the multi-terminal feeder flexible hub switch. By fusing the spatial distribution characteristics of the multi-terminal feed flexible hub switch topology using a bidirectional long short-term neural network, the spatiotemporal fusion characteristics of the multi-terminal feed flexible hub switch are obtained, including: The forward hidden state at the current time step is solved based on the spatial distribution characteristics of the topology; the backward hidden state at the current time step is solved based on the spatial distribution characteristics of the topology; the forward hidden state and the backward hidden state at the current time step are fused and normalized into a spatiotemporal fusion feature, expressed by the following formula: In the formula, LayerNorm represents layer normalization, || represents concatenation, and W0 represents the fusion weight. This indicates the current forward hidden state. This indicates the current backward hidden state. This indicates the spatiotemporal fusion characteristics of a multi-terminal feed-in flexible hub switch; The total network loss of each node is calculated based on the spatiotemporal fusion characteristics of the multi-terminal feeder flexible hub switch. The marginal electricity price of each node is calculated based on the total network loss of each node, and the node congestion cost is calculated based on the marginal electricity price of each node. With the goal of minimizing total economic cost, a total economic cost regulation and optimization model is constructed based on the total network loss and node congestion cost of all nodes. Solve the total economic cost regulation optimization model, output the optimal economic cost regulation strategy, and implement regulation optimization for multi-terminal feed-in flexible hub switch.

2. The method for regulating and optimizing a multi-terminal fed flexible hub switch according to claim 1, characterized in that: The topological state characteristics of the multi-terminal fed flexible hub switch to be acquired include: The voltage and current of the multi-terminal feed flexible hub switch are obtained and normalized respectively to obtain voltage characteristics and current characteristics; Transform the current feature vector into a one-dimensional current vector; The comprehensive state vector is solved based on the one-dimensional current vector, voltage eigenvector, and switching state matrix. The integrated state vector is input into the activation function to solve for the topological state features.

3. The method for regulating and optimizing a multi-terminal fed flexible hub switch according to claim 1, characterized in that: The process of performing topological modeling of the multi-terminal feeder flexible hub switch using the switch state matrix and physical connection matrix to obtain the adjacency matrix, and using the topological state features of the multi-terminal feeder flexible hub switch as the initial state features, aggregating the topological state features through a hierarchical graph convolutional network combined with the adjacency matrix includes: A topology model of the multi-terminal feeder flexible hub switch is performed based on the switch state matrix and the inherent physical connection matrix to obtain the adjacency matrix; The topological state features are set as the initial state features, and the state features of the Lth layer are solved by combining the adjacency matrix with an L-layer graph convolutional network. The state characteristics of the Lth layer are set as the spatial distribution characteristics of the topology.

4. The method for regulating and optimizing a multi-terminal fed flexible hub switch according to claim 1 or 3, characterized in that: The topology modeling of the multi-terminal feeder flexible hub switch includes: Set an inherent physical connection matrix based on whether there is physical connectivity between nodes; Set the switch state matrix according to whether the switch is on or off; The node connectivity result is obtained by multiplying the switch state matrix element by element with the inherent physical connection matrix. The adjacency matrix is ​​obtained by adding the node connectivity result to the node self-loop matrix.

5. The method for regulating and optimizing a multi-terminal fed flexible hub switch according to claim 1, characterized in that: The method for calculating the total network loss of each node based on the spatiotemporal fusion characteristics of multi-terminal fed-in flexible hub switch includes: The spatiotemporal fusion features of the multi-terminal feed-in flexible hub switch are mapped to loss-sensitive features by using a loss-sensitive weight matrix and activation function. By fusing loss-sensitive features with physical losses, a constraint loss matrix for each node is generated. By aggregating the neighbor set losses of all nodes in the constraint loss matrix of each node, the total network loss of each node is obtained.

6. The method for regulating and optimizing a multi-terminal fed flexible hub switch according to claim 1, characterized in that, include: The method for calculating the nodal marginal electricity price based on the total network loss of each node includes: Multiply the regional average electricity price of the power grid connected to the AC output terminal by the rated capacity of the system to obtain the theoretical revenue per hour when the system is running at full load. Divide the historical average congestion cost by the theoretical revenue per hour when the system is running at full load to obtain the system electricity price sensitivity weight. The marginal network loss cost is calculated based on the total network loss of each node, and the marginal congestion cost is calculated based on the system electricity price sensitivity weight. The marginal network loss cost, marginal congestion cost, and marginal generation cost are summed to generate the node marginal electricity price.

7. The method for regulating and optimizing a multi-terminal fed flexible hub switch according to claim 1, characterized in that, include: The calculation of node congestion cost based on the marginal electricity price of the node includes: The deviation of the nodal marginal electricity price is solved by subtracting the nodal marginal electricity price from the benchmark electricity price; By solving the partial derivative of the L2 norm of the spatiotemporal fusion characteristics of the multi-terminal fed-in flexible hub switch with respect to the switch state, the sensitivity of the switch action to the topological features is obtained. The sensitivity of switching actions to topological characteristics is multiplied by the number of switching actions and the deviation of the node's marginal electricity price to obtain the node congestion cost.

8. The method for regulating and optimizing a multi-terminal fed flexible pivot switch according to claim 1, characterized in that, include: The overall economic cost control optimization model, which aims to minimize the overall economic cost and is constructed based on the total network loss and node congestion cost of all nodes, includes: Multiply the total network loss of all nodes, the marginal electricity price of all nodes, and the set time step to obtain the total network loss cost. Sum the congestion costs of all nodes to obtain the total congestion cost. Multiply the total network loss cost and the total congestion cost by the network loss weight and the congestion weight respectively, and sum them to construct the objective function of the total economic cost regulation and optimization model. The total network loss of each node is equal to the active power injection of each node at time step t minus the active power consumption of each node at time step t and the net active power outflow of each node. This constitutes a power flow balance constraint. Equipment safety constraints are constructed by limiting the effective value of AC voltage at each node at time step t and the current amplitude matrix from each node to the other nodes at time t. A topological radial constraint is constructed by summing all elements of the adjacency matrix to twice the number of network branches and setting the rank of the difference between the degree matrix and the adjacency matrix to the number of network branches.

9. A control and optimization system for a multi-terminal fed flexible hub switch, comprising the control and optimization method for a multi-terminal fed flexible hub switch as described in any one of claims 1-8, characterized in that, include: The topology construction module is used to construct an adjacency matrix by using the switch state matrix and physical connection matrix of the multi-terminal feed flexible hub switch. The topological state features of the multi-terminal feed flexible hub switch are used as the initial state features to input the hierarchical graph convolutional network and the bidirectional long short-term neural network, and the spatiotemporal fusion features of the multi-terminal feed flexible hub switch are output by combining the adjacency matrix. The network loss calculation module is used to calculate the total network loss of each node based on the spatiotemporal fusion characteristics of the multi-terminal feed-in flexible hub switch. The nodal marginal electricity price calculation module is used to calculate the nodal marginal electricity price based on the total network loss of each node; The node congestion cost calculation module is used to calculate the node congestion cost based on the node's marginal electricity price. The regulation and optimization model construction module is used to construct a total economic cost regulation and optimization model based on the total network loss and node congestion cost of all nodes, with the goal of minimizing the total economic cost. The solution module is used to solve the total economic cost regulation optimization model, output the optimal economic cost regulation strategy, and implement regulation optimization for multi-terminal feed-in flexible hub switches.

Images