A multi-period power distribution network fault recovery method considering electric vehicles

Abstract

The application discloses a kind of multi-period distribution network failure recovery strategies considering electric vehicle (EV), and the recovery thought that the strategy considers individualized randomness influence EV as mobile energy storage unit, including the following steps: establishing the failure recovery model and objective function of distribution network, establishing the constraint model of distribution network failure recovery;Simulate the driving path of actual operation EV and battery state of charge (SOC) state quantity, establish the operation constraint model considering EV space-time distribution;Establish the optimization objective function of charging pile planning, and plan charging pile considering load level priority;Based on the failure recovery model, EV operation constraint model and corresponding charging pile planning, the failure recovery of distribution network is carried out;The application is based on wind and light and the like fluctuating distributed power, considers the planning of EV dispersity recovery capability and corresponding charging pile, carries out more efficient and comprehensive power supply recovery to load node, improves the toughness of distribution network, and can be popularized and used in actual recovery.

Description

Technical Field

[0001] This invention relates to the field of fault recovery, and more particularly to a multi-time period distribution network fault recovery method considering electric vehicles, belonging to the field of distribution network fault recovery. Background Technology

[0002] In recent years, the frequent occurrence of natural disasters such as typhoons, snowstorms, sustained high temperatures, and lightning has had a significant impact on the normal operation of power systems. Compared with transmission networks, distribution networks operate in a more complex environment and face various safety threats under extreme events, making it imperative to enhance the resilience of distribution networks. With changes in distribution network structure, traditional fault recovery methods are no longer applicable. Islanding schemes using distributed power sources such as wind and solar power are affected by environmental factors, impacting load recovery levels. Therefore, flexible and stable mobile energy storage devices are needed to assist in fault recovery of the distribution network.

[0003] Therefore, scholars both domestically and internationally have conducted a series of studies on distribution network fault recovery strategies that consider mobile energy storage. According to statistics from 2022, the number of new energy vehicles in China has reached 11.49 million. As a dispatchable distributed power source, EVs have a very promising application prospect. The emergency load support strategy for integrated charging, swapping, discharging, and energy storage power stations can serve as an emergency power source to support important loads. It is also possible to consider deploying electric buses and electric taxis, constrained by the transportation system, as mobile energy storage to participate in fault recovery. However, electric buses and electric taxis have spatiotemporal regularities, and their recovery capability for sudden events caused by extreme conditions is relatively weaker than that of electric private cars.

[0004] To address this issue, some scholars have proposed treating charging EV parking lots as distributed energy storage units for power distribution network fault recovery. However, this approach only considers EVs as a whole for planning, failing to adequately account for the impact of individual randomness and the recovery capability of EVs with corresponding charging pile planning for power distribution network faults.

[0005] If the recovery capabilities of EVs can be fully utilized for the distribution network, it will be of great significance to improving the fault recovery capabilities of the distribution network. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a multi-time period distribution network fault recovery method that fully considers the individualized recovery capability of EVs, so as to effectively improve the fault recovery capability of distribution networks.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention includes the following steps:

[0008] 1) Establish a fault recovery model and objective function for the distribution network, and establish a constraint model for fault recovery in the distribution network;

[0009] 2) Simulate the actual driving path of the EV and the state of charge (SOC) of the battery, and establish an operating constraint model that considers the spatiotemporal prediction of the EV.

[0010] 3) Based on the EV spatiotemporal prediction model, establish an optimization objective function for charging pile planning, plan charging piles considering load level priority, and perform fault recovery for the distribution network.

[0011] Firstly, establish a fault recovery model and objective function for the distribution network, and establish a constraint model for distribution network fault recovery, specifically as follows:

[0012] The objective function for distribution network fault recovery consists of load recovery amount and network loss, with load recovery amount being the primary recovery objective. The distribution network fault recovery model includes: a load model for determining load controllability; a radial topology model for post-fault network reconfiguration; branch power flow constraints for constraining branch power flow; and limit safety constraints for imposing safety limits on variables.

[0013] Firstly, the actual driving path of the EV and the state of charge (SOC) of the battery are simulated to establish an operational constraint model that considers the spatiotemporal prediction of the EV, specifically:

[0014] By combining EV operation data models and traffic road models, the node location and real-time operation data of a unit EV at each time period can be obtained, resulting in an EV spatiotemporal prediction model. EV operation constraints include battery charging and discharging constraints and energy transmission constraints. The difference between EV and traditional mobile energy vehicles lies in the fact that EVs have their own spatiotemporal state and a random SOC state. The unique constraints are as follows:

[0015]

[0016]

[0017] In the formula: These are the initial instantaneous state of charge and the minimum state of charge of the j-th EV, respectively; Let be the initial instantaneous SOC value of the j-th EV; Let SOC be the minimum limit for the j-th EV; The value is a binary integer variable ranging from 0 to 1. When it is 1, it means that the initial instantaneous SOC of the j-th EV has reached the callable limit and can participate in the fault recovery process. When it is 0, the opposite is true.

[0018] Prioritizing the use of an EV spatiotemporal prediction model, an optimization objective function for charging pile planning is established. This involves planning charging piles based on load level priority and fault recovery of the distribution network.

[0019] The planning essentially involves allocating the location of each charging station; a fitting matrix B = [B1, B2, ..., B...] is set. M [ ] represents the set of EV start and end point locations. Considering load level priority, nodes containing important loads are weighted. The algorithm's solution set is used to optimize the fit T of the fitted matrix B, with the objective function being:

[0020]

[0021] In the formula: M represents the number of nodes in the transportation network; D represents the dimension and length of the algorithm's solution set; F represents the optimal solution set for each iteration of the algorithm.

[0022] Particle swarm optimization (PSO) boasts advantages such as simple structure and fast convergence speed, but its local search capability is weak, making it prone to getting trapped in local optima and thus failing to achieve global optima. To address this, this invention designs a hybrid algorithm that balances sparse and concentrated search, effectively mitigating the tendency of single algorithms to get trapped in local optima or diverge. A two-stage algorithm model is employed for charging station planning: the first stage uses PSO, suitable for global search, to obtain initial values ​​for the next stage; the second stage uses Tabu Search, which has stronger local search capabilities, to improve upon this.

[0023] Considering the fault recovery model, EV operation constraint model, and corresponding charging pile planning, fault recovery is carried out on the power distribution network. Attached Figure Description

[0024] Figure 1 This is a flowchart of the present invention;

[0025] Figure 2 This is a distribution diagram of the EV start and end positions;

[0026] Figure 3 It is the EV spatiotemporal prediction model framework;

[0027] Figure 4 This is a flowchart of the tabu search algorithm;

[0028] Figure 5 This is a diagram of the IEEE 33-node distribution network structure.

[0029] Figure 6 This is a diagram of the 29-node transportation network structure;

[0030] Figure 7 This is a comparison chart of weighted load power recovery ratios; Detailed Implementation

[0031] To highlight the advantages of this invention and clarify the calculation steps, the specific implementation scheme of this invention is described in detail below with reference to the accompanying drawings.

[0032] Figure 1 The flowchart illustrating the specific implementation of the recovery strategy of this invention is as follows:

[0033] 1) Establish a fault recovery model and objective function for the distribution network, and establish a constraint model for fault recovery of the distribution network.

[0034] A distribution network fault recovery objective is established, constraining the radial topology, branch power flow, and variable limits of the distribution network. The objective function for distribution network fault recovery consists of load restoration and network losses, with load restoration being the primary recovery objective. The fault recovery time is divided into T time periods, and the objective function is a weighted sum of load restoration and network losses for each time period, followed by normalization. The specific function is as follows:

[0035]

[0036] In the formula: N load The set of node loads; z i To indicate the importance of load node i, the corresponding weights for first-, second-, and third-level loads are 10, 5, and 1, respectively; c i,t The value is a binary integer variable ranging from 0 to 1, where 0 indicates that the load at node i in time period t has not been restored, and 1 indicates the opposite. si represents the load recovery power at node i during time period t; s0 is the network loss coefficient, used to ensure that the primary and secondary targets are not affected, and its value is taken as 0.5; I ij,t R is the branch current on line (i,j) during time period t; ij The resistance on line (i,j); ε is the set of available lines during recovery; S base This is the reference power.

[0037] The radial topology uses a single-commodity flow constraint, which has the advantages of simple form and few introduced variables and constraints, and can ensure the radial operation of the distribution network. The model is as follows:

[0038] ∑l ij,t =|N node |-1 (2)

[0039]

[0040]

[0041]

[0042] In the formula: l ij,t N is a binary integer variable consisting of 0 and 1, where 0 indicates that line (i,j) has not been restored during time period t, and 1 indicates the opposite; node H represents the number of nodes in the distribution network. ij,tD is a continuous variable representing the virtual power flow magnitude of line (i,j); i The virtual demand for non-root nodes is represented by 1; r is the root node number; M is a very large positive real number, set as the number of distribution network nodes.

[0043] The distribution network is a three-phase symmetrical radial network. A second-order cone relaxation method is used for linearization, transforming the fault recovery model into a second-order cone programming model. The single-phase model is described as follows:

[0044]

[0045]

[0046]

[0047]

[0048] In the formula: P ij,t Q ij,t R represents the active and reactive power flowing through line (i,j) during time period t; ij X ij Let J be the resistance and reactance of line (i,j) during time period t; ij,t It is the square of the current amplitude flowing through line (i,j) during time period t; These represent the active and reactive power of the gas turbine connected to node i during time period t; These represent the active and reactive power of the distributed power source connected to node i during time period t. Let t be the active power of the distributed power source connected to node i during time period t; The value is a binary integer variable ranging from 0 to 1. When it is 1, it means that the initial instantaneous SOC of the j-th EV in time period t has reached the callable limit and can participate in the fault recovery process. When it is 0, the opposite is true. These represent the active and reactive power restored at node i during time period t; v i,t a is the square of the voltage amplitude at node i during time period t; ij.t It is an auxiliary variable.

[0049] The ultimate safety constraints for active and reactive power, as well as voltage and current in the distribution network are:

[0050]

[0051]

[0052]

[0053]

[0054] In the formula: Let be the upper and lower limits of the active and reactive power of the generator at node i, respectively; G is the set of nodes where the generator is located. , represent the upper and lower limits of the active and reactive power of the distributed generation at node i, respectively. The distributed generation in this paper includes wind turbines and photovoltaics; DG is the set of nodes containing the distributed generation; V i,max V i,min These are the upper and lower limits of the voltage amplitude at node i, respectively; J ij,max This represents the upper limit that the square of the current amplitude can be taken.

[0055] 2) Simulate the actual driving path of the EV and the state of charge (SOC) of the battery, and establish an operating constraint model that considers the spatiotemporal prediction of the EV.

[0056] The initial SOC, start time, and return time were simulated using the Monte Carlo method: the initial SOC was set to the state before charging started. Considering the lifespan of the EV battery, the allowable range of SOC was set to (0.2, 0.9), and its dynamic change trend conformed to the characteristics of a normal distribution. The initial SOC distribution was taken as N(0.45, 0.15). The dynamic change trends of the start time and return time both met the characteristics of a log-normal distribution, and their probability density functions are shown in Equations (1) and (2).

[0057]

[0058]

[0059] In the formula: f(T) g ) for EV in T g The probability density function of travel at any given time; μ g Let μ be the mathematical expectation of the start time of EV travel. g =8.92; σ g Let σ be the standard deviation of the start time of EV travel. g =3.24; f(T) b ) for EV in T b The probability density function of travel at any given time; μ b Let μ be the mathematical expectation of the start time of EV travel. b =17.47; σ b Let σ be the standard deviation of the start time of EV travel. b =3.41.

[0060] The origin-destination (OD) matrix describes the travel distribution characteristics of EVs. The initial location of the EV is obtained using the Monte Carlo method, and then the EV's travel destination can be generated through random sampling by combining the OD matrix. The distribution of origin and destination locations is as follows: Figure 2As shown. If the generated initial location and destination node are the same, set this EV to have no travel plans for today.

[0061] The traffic road model references the traffic road topology of a specific region and uses graph theory to represent the adjacency relationships between nodes in the traffic network. Using G = (N... P ,ε L ) represents a transportation network, where: N P ε is the edge set, that is, the set of road segments in the transportation network; L Matrix D is the set of vertices, that is, the set of endpoints of road segments or the set of intersections of multiple road segments in a transportation network. R This is an adjacency matrix of road weights, used to describe the length of each road segment and the connection relationships between nodes, where the element d ij The assignment function is

[0062]

[0063] In the formula: is the length of the road segment between node i and node j; inf indicates that there is no road segment connecting the two nodes.

[0064] In a transportation network, the driving speed of an EV is mainly affected by road capacity and traffic flow. To simulate the actual operation of an EV, the following model is introduced.

[0065]

[0066] In the formula: v ij (t) represents the speed of EV between adjacent nodes (i,j) in the transportation network at time t; v ij,o q represents the zero-velocity between adjacent nodes (i,j) in the transportation network; ij (t) represents the road flow rate of segment (i,j) of the traffic network at time t; C ij The traffic capacity of segment (i,j) in the transportation network is affected by the road grade; q ij (t) / C ij Let be the road saturation of segment (i,j) of the traffic network at time t; a, b, and n are coefficients for different road grades. Roads are divided into two types: main roads and secondary roads. For main roads, a, b, and n take values ​​of 1.726, 3.15, and 3, respectively; for secondary roads, a, b, and n take values ​​of 2.076, 2.870, and 3, respectively.

[0067] To simplify the complexity of path selection during actual runtime, the EV driving path selection uses the Floyd algorithm, which utilizes the adjacency matrix D in the road topology. R Find the shortest path, meaning all EVs travel along the shortest path. Assume that the set of shortest paths between an EV's initial position i and its destination j is R = {i,…,e,f,…,j}, containing H direct road segments. The travel time model is as follows:

[0068]

[0069]

[0070] Where: ΔT ij Let ΔT be the total travel time for the shortest path between (i,j); h The time taken for the h-th direct connection segment; d h The mileage of the h-th directly connected road segment can be obtained from equation (3); V h (t) represents the driving speed of the h-th direct-connection segment calculated by the speed-flow model.

[0071] By combining EV operation data models and traffic road models, the node location and real-time operation data of a unit EV at each time period can be obtained. The EV spatiotemporal prediction model framework is as follows: Figure 3 As shown in the diagram, the framework simulates the initial SOC, start time, and return time using the Monte Carlo method. The initial position and destination of the EV are obtained through the OD probability matrix, and a probabilistic distribution method is used to establish an EV operation data model with uncertainty. The traffic road model references the traffic road topology of a specific area, uses the Floyd algorithm to obtain the shortest path for the EV, and obtains the corresponding running time and speed through a speed-flow model, thereby enabling vehicle path selection.

[0072] EV operating constraints include battery charging and discharging constraints and energy transfer constraints. The difference between EVs and traditional mobile energy vehicles lies in the fact that EVs possess their own spatiotemporal states and stochastic SOC states. The constraint model is as follows:

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] In the formula: These are the charging and discharging state variables of the j-th EV during time period t; E evThis represents the total number of EVs. These represent the charging and discharging power of the j-th EV during time period t; These are the maximum charging and discharging power of the j-th EV during time period t; These represent the state of charge of the j-th EV at time intervals t and t+1, respectively. These are the initial instantaneous state of charge and the minimum state of charge of the j-th EV, respectively; Let be the charging and discharging efficiencies of the j-th EV, respectively, both taken as 0.9; Let be the initial instantaneous SOC value of the j-th EV; Let SOC be the minimum limit for the j-th EV; This is the flag indicating that the j-th EV is parked and connected at node s during time period t. The value is determined by the spatiotemporal state of the EV itself; Let be the active power of the j-th EV in time period t; M is a very large positive real number.

[0082] 3) Based on the EV spatiotemporal prediction model, establish an optimization objective function for charging pile planning, plan charging piles considering load level priority, and perform fault recovery for the distribution network.

[0083] The planning essentially involves allocating the location of each charging station. The fitting matrix is ​​set as B = [B1, B2, ..., B...]. M [ ] represents the set of EV start and end point locations, with weights applied to nodes containing significant loads. The algorithm's solution set is used to optimize the fit T of the fitted matrix B, and the objective function can be expressed as:

[0084]

[0085] In the formula: M represents the number of nodes in the transportation network; D represents the dimension and length of the algorithm's solution set; F represents the optimal solution set for each iteration of the algorithm.

[0086] Particle swarm optimization (PSO) boasts advantages such as simple structure and fast convergence speed, but its local search capability is weak, making it prone to getting trapped in local optima and thus failing to achieve global optima. To address this, this invention designs a hybrid algorithm that balances sparse and concentrated search, effectively mitigating the tendency of single algorithms to get trapped in local optima or diverge. A two-stage algorithm model is employed for charging station planning: the first stage uses PSO, suitable for global search, to obtain initial values ​​for the next stage; the second stage uses tabu search, which has stronger local search capabilities, to improve upon this.

[0087] The first stage: Particle Swarm Optimization (PSO) is a heuristic algorithm that simulates biological activity. It treats each individual within the swarm as a particle with no volume or mass in a D-dimensional search space. These particles move within the search space at a certain speed and direction, ultimately yielding the swarm's optimal solution. Considering a single charging station as a particle, each particle possesses a position vector x = [x1, x2, ..., x...]. N ] and velocity vector v = [v1, v2, ..., v N All of them have memory function. The optimal position of the i-th particle is The global optimal position of the population is G = [G 1 G 2 ,…,G D The particle swarm iteration process and objective function are:

[0088]

[0089]

[0090] In the formula: v i The particle velocity is in the range [v] min ,v max ]; w is the inertia weighting factor; c1 and c2 are acceleration constants; r1 and r2 are random numbers between [0,1].

[0091] The second stage: Tabu Search is a global stepwise optimization algorithm that simulates human intelligence and is based on local neighborhood search. It avoids roundabout and repetitive searches by setting up a tabu table storage structure and tabu criteria, thus escaping local optima. The disdain criterion effectively exempts some good solutions from the tabu list, thereby achieving global optimization within the search range. However, its drawback is that the optimization performance of Tabu Search is greatly affected by the initial solution, which can be improved by the first stage. Its algorithm flowchart is as follows: Figure 4 As shown.

[0092] Using an improved IEEE 33-node distribution network coupled with a 29-node transportation network as the test system, a cplex solver was employed to verify the effectiveness of the proposed EV and its charging station planning in improving the distribution network's fault recovery capability, i.e., its resilience. The distribution network node structure diagram, transportation network structure diagram, and comparison diagram of the distribution network recovery effect are shown below. Figure 5 , Figure 6 and Figure 7 As shown.

Claims

1. A method for multi-time period distribution network fault recovery considering electric vehicles, wherein the recovery method enhances the resilience of the distribution network by combining EVs as mobile energy storage with charging pile planning that considers load level priority, characterized in that... Includes the following steps: 1) Establish a fault recovery model and objective function for the distribution network, and establish a constraint model for fault recovery in the distribution network; The objective function for distribution network fault recovery consists of load recovery amount and network loss, with load recovery amount being the primary recovery objective. The distribution network fault recovery model includes: a load model for determining load controllability; a radial topology model for post-fault network reconfiguration; branch power flow constraints for constraining branch power flow; and limit safety constraints for imposing safety limits on variables. 2) Simulate the actual driving path of the EV and the state of charge (SOC) of the battery, and establish an operating constraint model that considers the spatiotemporal prediction of the EV. 3) Based on the EV spatiotemporal prediction model, an optimization objective function for charging pile planning is established. Charging piles are planned considering load level priority, and fault recovery is performed on the distribution network. The planning essentially involves allocating the location of each charging pile. A fitting matrix is ​​set. Given the set of EV start and end point locations, the nodes containing important loads are weighted according to load level priority; an algorithm is used to solve the set and fit the matrix. goodness of fit As the optimization objective, the objective function is: In the formula: , , This refers to the number of nodes in the transportation network. , , Let be the dimension and length of the solution set of the algorithm; This represents the optimal solution set for each iteration of the algorithm. Particle swarm optimization (PSO) has the advantages of simple structure and fast convergence speed, but its local search ability is weak and it is prone to getting trapped in local optima, thus failing to achieve global optimum. A hybrid algorithm design can achieve a balance between sparse and concentrated search, effectively improving the situation where a single algorithm is prone to getting trapped in local optima or diverging. A two-stage algorithm model is used for charging pile planning: the first stage uses PSO, which is suitable for global search, to obtain the initial value for the next stage; the second stage uses Tabu Search, which has stronger local search ability, to improve upon this. Considering the fault recovery model, EV operation constraint model, and corresponding charging pile planning, fault recovery is carried out on the power distribution network.

2. The method for multi-time period distribution network fault recovery considering electric vehicles according to claim 1, characterized in that, Step 2 includes: By combining EV operation data models and traffic road models, the node location and real-time operation data of a unit EV at each time period can be obtained, resulting in an EV spatiotemporal prediction model. EV operation constraints include battery charging and discharging constraints and energy transmission constraints. The difference between EV and traditional mobile energy vehicles lies in the fact that EVs have their own spatiotemporal state and a random SOC state. The unique constraints are as follows: In the formula: , The first The initial instantaneous state of charge and minimum state of charge of each EV; For the first The initial instantaneous SOC value of each EV; For the first Minimum SOC value for each EV; A binary integer variable with values ​​ranging from 0 to 1; a value of 1 indicates the first... When the initial instantaneous SOC of an EV reaches the callable limit, it can participate in the fault recovery process; when it is 0, the opposite is true.

Images