Mobile energy storage pre-layout and dynamic scheduling system and method oriented to toughness improvement of power distribution network

Abstract

The invention discloses a mobile energy storage pre-layout and dynamic scheduling system and method oriented to toughness improvement of a power distribution network, and belongs to the technical field of power systems and intelligent power grids. In the pre-disaster stage, a two-stage robust optimization model fusing photovoltaic output uncertainty is constructed, and the configuration cost of mobile energy storage, the load reduction risk and the traffic accessibility of key nodes are comprehensively considered to generate an optimal deployment scheme; in a post-disaster stage, based on a traffic network state and a multi-source output characteristic, a multi-source collaborative dynamic scheduling model is constructed, a migration path, a charging and discharging strategy and charge state control of mobile energy storage are optimized in real time, and power supply loss of a key load is minimized. The system realizes rapid power supply recovery and safe operation in the island power grid by cooperatively controlling distributed resources such as mobile energy storage, electric vehicles, photovoltaic and diesel generators and the like. According to the invention, the quick response capability and recovery efficiency of the power distribution network under extreme disasters are significantly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems and smart grids, and more specifically, relates to a mobile energy storage pre-layout and dynamic scheduling system and method for improving the resilience of distribution networks. Background Art

[0002] In recent years, extreme disasters have led to frequent large-scale power outages in the power grid. The resilience of the distribution network, which reflects the ability of the distribution system to resist disasters, adapt to them, and restore power supply, has received extensive attention. In addition, to cope with the dual crises of energy depletion and environmental pollution, a large number of distributed generation (DG) and electric energy substitution loads have been connected to the distribution network, providing solutions for load restoration. Therefore, effectively utilizing various distributed resources before and after disasters to reduce power outage losses is of great significance for improving the resilience of the distribution network.

[0003] Traditional distribution network resilience improvement solutions lack the collaborative optimization of mobile energy storage in the pre-disaster prevention stage. They mainly focus on post-disaster recovery and do not fully utilize the pre-layout optimization of mobile energy storage before disasters, resulting in insufficient response speed in the initial stage of disasters. Existing models ignore multi-source collaboration and output uncertainty, rely mostly on single resources (such as diesel generators or fixed energy storage), do not effectively integrate the collaborative operation of multiple types of distributed power sources such as photovoltaic, mobile energy storage, and electric vehicles, and do not consider the impact of photovoltaic output fluctuations on the recovery strategy. They also do not consider the dynamic impact of disasters on the travel time of the transportation network (such as road congestion), leading to a mismatch between the mobile energy storage scheduling plan and the actual traffic conditions and delaying the recovery time.

[0004] Currently, the traditional robust optimization models used in distribution network resilience improvement methods for extreme disaster scenarios may sacrifice economy due to excessive conservatism and do not flexibly adjust the number of islands or topological constraints, limiting the adaptability of the recovery strategy. Moreover, existing strategies mostly adopt static or single scheduling, without realizing the dynamic energy allocation of mobile energy storage in time and space, reducing the recovery efficiency of critical loads. Summary of the Invention

[0005] Aiming at the above deficiencies, the purpose of the present invention is to propose a mobile energy storage pre-layout and dynamic scheduling system and method for improving the resilience of distribution networks, and specifically design a two-stage optimization strategy that integrates pre-disaster mobile energy storage pre-layout and post-disaster multi-source collaborative dynamic scheduling. By constructing an uncertainty-driven resource allocation model and a traffic-power grid coupled dynamic scheduling model, the full-cycle optimization management of mobile energy storage in the time and space dimensions is realized, significantly enhancing the active defense ability and recovery efficiency of the system during disasters.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] According to the first aspect of the present invention, a mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of a distribution network is proposed, including:

[0008] A pre-disaster pre-layout module, configured to:

[0009] Establish a two-stage robust optimization model considering extreme disaster scenarios, where the decision variables in the first stage include the number of energy storage configurations, the nodes of energy storage configurations, and the network topology state, and the optimization objective is to minimize the energy storage configuration cost; the decision variables in the second stage include the load shedding power and the power output of power sources, and the optimization objective is to minimize the critical load shedding cost under the worst photovoltaic power output scenario;

[0010] Solve the two-stage robust optimization model and output the optimal energy storage configuration plan and network topology state;

[0011] A post-disaster dynamic scheduling module, configured to:

[0012] Collect traffic network status, distributed power generation output data, and load demand data in real time;

[0013] Construct a spatio-temporal multi-dimensional dynamic scheduling model, whose decision variables include the migration path sequence of mobile energy storage, the charge and discharge power values, and the node positions where it is connected to the distribution network;

[0014] Dynamically update the scheduling instructions based on the rolling horizon optimization framework;

[0015] A cooperative control module, configured to:

[0016] Establish a multi-source coordinated control strategy, and the priority order is: key load power supply guarantee > maintenance of distribution network topology stability > V2G coordination between mobile energy storage and electric vehicles > output complementarity between diesel generators and photovoltaics;

[0017] Realize power distribution through a distributed control architecture, where mobile energy storage is used as the main regulation unit.

[0018] Furthermore, where ξ is the set of uncertain scenarios, representing the set of actual output powers of all photovoltaic nodes at the current moment; is the set of predicted scenarios, representing the set of predicted output powers of photovoltaic nodes, given by the time series prediction model LSTM; Q is the covariance matrix trained based on historical prediction errors; τ is the scale factor for controlling the boundary of the uncertainty set; is the n-dimensional real number space, is the photovoltaic processing uncertainty set; T is the matrix transpose symbol;

[0019] Select the photovoltaic power output scenario that maximizes the load shedding cost in the uncertainty set as the worst scenario.

[0020] Furthermore, in the post-disaster dynamic scheduling module, the travel time constraint is calculated as:

[0021]

[0022] Among them, τ j,k,t is the travel time from node i to the critical node j at the post-disaster moment t, d jk is the equivalent travel distance, and v j,k,t is the actual vehicle speed under the disaster.

[0023] Furthermore, the collaborative control module executes the following control logic:

[0024] The electric vehicle switches to the discharging mode after the disaster to provide backup power;

[0025] The diesel generator operates in the maximum output mode to supplement the energy storage power gap;

[0026] Dynamically adjust the output power of the photovoltaic system according to the light conditions, and preferentially supply the local load.

[0027] Furthermore, the pre-disaster pre-layout module uses the column and constraint generation algorithm for solution, including:

[0028] The master problem is a mixed-integer second-order cone programming model for determining the mobile energy storage configuration scheme;

[0029] The sub-problem is the search problem of the worst photovoltaic output scenario, which is converted into a linear form through the strong duality principle and iterates alternately with the master problem until convergence.

[0030] Furthermore, the post-disaster dynamic scheduling module also includes the following operation constraints during the scheduling process:

[0031] The node voltage should be maintained within the allowable fluctuation range;

[0032] The branch current shall not exceed the thermal limit value;

[0033] The state of charge of the energy storage device is limited by the set upper and lower limits.

[0034] According to the embodiments of the present invention, the extreme disaster scenarios include power facility damage scenarios caused by at least one natural disaster among typhoons, earthquakes, and ice disasters.

[0035] According to the second aspect of the present invention, a method for pre-layout and dynamic scheduling of mobile energy storage for improving the resilience of the distribution network is proposed, including the following steps:

[0036] Pre-disaster stage: Based on the photovoltaic output prediction and its uncertainty, a two-stage robust optimization model is constructed to generate the optimal deployment plan of mobile energy storage. The decision variables in the first stage include the number of energy storage configurations, the nodes of energy storage configurations, and the network topology state, and the optimization objective is to minimize the energy storage configuration cost. The decision variables in the second stage include the load shedding power and the power output of power sources, and the optimization objective is to minimize the critical load shedding cost under the worst photovoltaic output scenario.

[0037] Post-disaster stage: Based on the traffic network conditions and the power output status of distributed power sources under disaster conditions, a spatio-temporal multi-dimensional dynamic scheduling model is constructed, and the charging and discharging behavior and spatial migration of mobile energy storage are optimized based on the rolling horizon optimization framework. The scheduling process satisfies the operation constraints of node voltage, branch current, and the state of charge of energy storage devices.

[0038] During the scheduling process, in accordance with the priority of critical load power supply guarantee > maintaining the topological stability of the distribution network > V2G coordination of mobile energy storage and electric vehicles > complementary output of diesel generators and photovoltaics, the multi-source collaborative operation of electric vehicles, photovoltaics, and diesel generators is jointly considered, and critical loads are restored layer by layer to ensure system safety constraints.

[0039] Furthermore, the scheduling process in the post-disaster stage is carried out at 1-hour intervals, and the rolling horizon optimization is used to gradually promote the fault recovery.

[0040] Through the above design scheme, the present invention can bring the following beneficial effects: The present invention proposes a mobile energy storage pre-layout and dynamic scheduling system and method for enhancing the resilience of the distribution network. Through the systematic design method proposed by the present invention, the resilience level of the distribution network in extreme disaster situations can be effectively enhanced. In the pre-disaster stage, based on the robust optimization model, the deployment plan of mobile energy storage is accurately formulated, fully considering the volatility of photovoltaic output and the uncertain scenarios induced by disasters, reducing the risk of critical load shedding, and enhancing the initial response ability and resource guarantee ability of the system. In the post-disaster stage, combined with the dynamic change characteristics of the traffic network, a multi-source collaborative scheduling mechanism is constructed to optimize the spatio-temporal migration path and charging and discharging control strategy of mobile energy storage in real time, realizing the efficient spatio-temporal allocation of energy resources and effectively supporting the rapid recovery of critical loads.

[0041] In addition, the system dynamically identifies and adjusts the island operation structure, and combines multi-dimensional operation constraints such as node voltage, current, and the state of charge of energy storage to prevent the spread of secondary risks during the post-disaster operation process, ensuring the operation safety and topological stability of the local power supply system of the distribution network. The overall scheme covers the complete cycle of pre-disaster planning and post-disaster scheduling, and has the triple advantages of fast response ability, resource collaborative efficiency, and operation reliability, providing a set of implementable and popularizable technical paths for the resilience management of the distribution network under the impact of multi-type disasters. Description of the Drawings

[0042] The following drawings are provided to further understand the present invention, and form a part of the present invention application. The schematic embodiments of the present invention and their descriptions are used to understand the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0043] Figure 1 is the overall structural block diagram of the system of the present invention;

[0044] Figure 2 is the improved IEEE 33-node distribution network topology structure diagram adopted in the embodiment of the present invention;

[0045] Figure 3 is the active power prediction curve diagram of various types of loads;

[0046] Figure 4 is the congestion distribution diagram of the traffic network after the disaster occurs;

[0047] Figure 5 is the active power prediction output curve diagram of each photovoltaic unit before the disaster;

[0048] Figure 6 is the schematic diagram of the scheduling behavior of the mobile energy storage device MESS1;

[0049] Figure 7 is the schematic diagram of the scheduling behavior of the mobile energy storage device MESS2;

[0050] Figure 8 is the power supply power and recovery ratio curve diagram of two types of loads during the fault recovery period;

[0051] Figure 9 is the charge and discharge power curve diagram of each electric vehicle charging pile;

[0052] Figure 10 is the active power output curve diagram of each diesel generator during the post-disaster period;

[0053] Figure 11 is the voltage fluctuation curve diagram of each node during the system operation period;

[0054] Figure 12 is the recovery ratio curve diagram of key loads under different recovery strategies. Detailed implementation manners

[0055] The technical solutions of the present invention will be further described below in conjunction with the drawings and embodiments, but are not limited thereto. Any modification or equivalent replacement of the technical solutions of the present invention without departing from the spirit scope of the technical solutions of the present invention shall be covered within the protection scope of the present invention.

[0056] Figure 1Shows the overall structural block diagram of the system of the present invention, demonstrating the functional relationships and information interaction processes among the pre-disaster pre-layout module, post-disaster dynamic scheduling module, and collaborative control module; Figure 2 Shows the improved IEEE 33-node distribution network topology structure diagram adopted in the embodiment of the present invention, displaying the distribution node numbers, power source access positions, and load distributions; Figure 3 Shows the active power prediction curves of various types of loads (critical loads and non-critical loads), which are used to illustrate the time-varying characteristics of the loads. Critical loads refer to the loads that must be prioritized for power restoration during disasters, including but not limited to loads that are crucial for ensuring the basic operation of society, such as hospitals, communication centers, and emergency service facilities. In the present invention, the active power demand curves of critical loads usually have a higher priority and need to be considered first for power restoration during post-disaster scheduling. Non-critical loads refer to ordinary loads that do not belong to critical infrastructure, mainly including the power demands of places such as residential areas and commercial buildings. Although such loads are important, their power restoration priority is relatively low during post-disaster recovery, and usually, power supply can be gradually restored after the critical loads are restored; Figure 4 Shows the congestion distribution map of the transportation network after a disaster, marking the traffic levels and equivalent delays of different paths, which is used for the selection of mobile energy storage scheduling paths; Figure 5 Shows the active power prediction output curves of each photovoltaic unit before a disaster, demonstrating the time-varying characteristics and uncertainty basis of photovoltaic output; Figure 6 Shows the schematic diagram of the scheduling behavior of the mobile energy storage device MESS1, including the access node positions, discharge power, and state of charge changing over time; Figure 7 Shows the schematic diagram of the scheduling behavior of the mobile energy storage device MESS2, including the access node positions, discharge power, and state of charge changing over time; Figure 8 Shows the power supply and restoration ratio curves of two types of loads (critical loads and non-critical loads) during fault recovery, reflecting the load recovery capabilities under different strategies; Figure 9 Shows the charge and discharge power curves of each electric vehicle charging pile, demonstrating the operation strategies of electric vehicles during post-disaster emergency power supply; Figure 10 Shows the active power output curves of each diesel generator during the post-disaster period, reflecting its basic power supply capacity and energy replenishment role; Figure 11 Shows the voltage fluctuation curves of each node during the system operation, verifying the electrical safety and island support capabilities of the system; Figure 12 Is the recovery ratio curve of critical loads under different recovery strategies, which is used to compare the differences in recovery efficiency among Strategy 1 to Strategy 4.

[0057] The following details the specific implementation steps of the mobile energy storage pre-layout and dynamic scheduling method for improving the resilience of the distribution network:

[0058] Step 1: Construct a pre-disaster pre-layout robust optimization model

[0059] Objective function:

[0060] This model comprehensively considers three aspects of objectives:

[0061] 1. The investment cost of mobile energy storage deployment;

[0062] 2. The loss of critical load curtailment under the worst photovoltaic power output scenario, where critical load refers to the load that must be powered on preferentially during a disaster, including but not limited to loads that are crucial for ensuring the basic operation of society, such as hospitals, communication centers, emergency service facilities, etc.;

[0063] 3. The traffic accessibility of critical load nodes after a disaster, where critical load refers to the load nodes that are crucial for the restoration of the entire power grid during the post-disaster restoration stage. These nodes include not only "critical loads" but also distribution hubs and regional power dispatch centers that need to be restored preferentially during the power grid restoration process.

[0064] The three-objective optimization form is as follows:

[0065]

[0066] In the formula: is the set of distribution network nodes, representing all power distribution nodes in the distribution network; x i is a binary decision variable, indicating whether to deploy mobile energy storage at node i. If x i = 1, it means that mobile energy storage is deployed at node i; if x i = 0, it means that it is not deployed; C i is the unit configuration cost of mobile energy storage at the node, representing the cost required to configure a unit of energy storage at node i; is the set of all load nodes, representing the set of all load demand nodes in the distribution network; is the set of critical load nodes, representing the set of critical nodes that need to be powered on preferentially during post-disaster restoration, such as hospitals, communication centers, etc.; is the importance of critical node j under the uncertain scenario set ξ; is the load curtailment power of critical node j under the uncertain scenario set ξ; d ijis the equivalent traffic passing time from node i to critical node j after a disaster (considering factors such as traffic congestion and road damage); u is the photovoltaic processing uncertainty set; f1 represents the total investment cost of deploying mobile energy storage on all nodes in the system, that is, economic efficiency; f2 represents the load shedding risk considering all photovoltaic output uncertainty scenarios during the post-disaster recovery process, that is, risk resilience; f3 represents the weighted sum of the traffic delay from the mobile energy storage node to the critical node during the post-disaster recovery process, that is, traffic responsiveness.

[0067] To simplify the solution, the weighted method is used to unify the three objectives into a single-objective optimization:

[0068] minZ = ω1f1 + ω2f2 + ω3f3 (2)

[0069] In the formula, Z is the simplified objective function, and ω1, ω2, and ω3 are the weight coefficients of economic efficiency, risk resilience, and traffic responsiveness respectively.

[0070] Modeling of photovoltaic output uncertainty:

[0071] Photovoltaic output has significant randomness and volatility. Especially under the influence of disaster weather, its prediction error is significantly amplified. In the present invention, an ellipsoidal uncertainty set constructed based on the prediction results of the LSTM model (long short-term memory network model) is used to characterize its uncertainty, and the specific form is as follows:

[0072]

[0073] In the formula: ξ is the set of uncertain scenarios, representing the set of actual output powers of all photovoltaic nodes at the current moment; is the set of prediction scenarios, representing the set of predicted output powers of photovoltaic nodes, given by the time series prediction model (LSTM); Q is the covariance matrix trained based on historical prediction errors; τ is the scale factor that controls the boundary of the uncertainty set and adjusts the conservatism of the system; is the n-dimensional real number space, is the photovoltaic processing uncertainty set, and T is the matrix transpose symbol.

[0074] Constraint conditions:

[0075] 1. Resource quantity limit: The total number of mobile energy storage that can be deployed before the disaster is limited, and the constraint is as follows:

[0076]

[0077] where M is the maximum number of mobile energy storage that can be deployed before the disaster.

[0078] 2. Single-node unique configuration constraint:

[0079]

[0080] That is, at most one mobile energy storage is arranged at each node.

[0081] 3. Traffic accessibility constraint: To ensure that key nodes can receive support within the shortest time after a disaster, it is necessary to ensure that there is at least one energy storage configuration point within the range of traffic accessibility of key nodes after the disaster:

[0082]

[0083] In the formula, δ ij ∈ {0, 1} indicates whether node i is accessible to key node j after the disaster, which is calculated from the affected graph of the traffic network.

[0084] 4. Load restoration ability constraint: It is required that the system can guarantee the minimum restoration ratio under all uncertain scenarios:

[0085]

[0086] In the formula, η is the set minimum load restoration ratio threshold, and P j is the active power of the load at key node j, representing the power demand of key node j in the distribution network.

[0087] 5. Radial topology constraint of the distribution network:

[0088] The improved single-commodity flow method and virtual source node modeling method are adopted. On the basis of the traditional single-commodity flow method, more optimization elements and constraint conditions are added to adapt to more complex power networks and more complex requirements. Combining with the Big-M relaxation technology, it is ensured that the network structure meets the requirements of radial operation after reconstruction:

[0089]

[0090] In the formula, f ij is the virtual power flow from node i to key node j, s i is the virtual source flag variable used to determine whether the island contains a power source node, and ε is the set of distribution network branches, representing the set of all lines connecting nodes in the distribution network.

[0091] Solution method:

[0092] Due to the second-order uncertainty, non-linear coupling and multi-objective characteristics of the model, the present invention adopts the following two-stage strategy for solution:

[0093] 1. Uncertain scenario generation stage

[0094] Train the LSTM model based on historical operation data and photovoltaic prediction errors, generate multiple typical photovoltaic output scenarios, and construct their covariance matrix for ellipsoidal set description. The covariance matrix is a symmetric matrix, and its elements represent the correlation of prediction errors of different photovoltaic nodes. The ellipsoidal set is shown in Equation (3).

[0095] 2. Multi-objective robust optimization stage

[0096] Use the column and constraint generation algorithm (C&CG) to solve the robust model, where: the master problem is the mobile energy storage configuration problem under a given output scenario, modeled as a mixed integer linear programming (MILP);

[0097] The sub-problem is to determine the worst photovoltaic output situation, which is reconstructed into a single-layer linear model through dual transformation and nested in the master problem for iterative convergence. This belongs to the prior art and will not be elaborated here in detail.

[0098] Step 2: Build a multi-source collaborative post-disaster recovery optimization model

[0099] Objective function:

[0100] The post-disaster scheduling aims to minimize the total system load shedding cost during the fault duration, and dynamically coordinates the charge and discharge states and movement paths of various resources on the premise of maintaining system operation safety and topological stability. The objective function is as follows:

[0101]

[0102] In the formula: is the set of fault times, and the scheduling period is divided by hours; ΔP j,t is the load shedding power at the critical node j at time t; λ j is the critical load cost weight of the critical node j.

[0103] This objective function accumulates the critical load shedding amounts in all time periods in the time domain and comprehensively evaluates the service loss during the system resilience recovery process in a weighted manner.

[0104] Definition of decision variables:

[0105] x i,j,t ∈{0,1}: indicates whether the mobile energy storage i is connected to the critical node j at time t;

[0106] indicates whether the mobile energy storage i is in the charging / discharging state at time t;

[0107] The charging / discharging power of the mobile energy storage i at time t;

[0108] SOC i,t : State of charge of the energy storage;

[0109] z i,j,k,t ∈ {0, 1}: Whether mobile energy storage i is transported from critical node j to node k at time t;

[0110] d jk : The equivalent travel distance from critical node j to node k, in km;

[0111] v j,k,t : The actual travel speed between nodes at time t, affected by the road conditions after the disaster;

[0112] τ j,k,t = d jk / v j,k,t : The estimated travel time in the disaster scenario.

[0113] Constraints:

[0114] Spatio-temporal dynamic scheduling constraints of mobile energy storage

[0115] 1. Initial state constraints of mobile energy storage:

[0116]

[0117] Among them, Whether mobile energy storage i is connected to critical node j at time t0. Is the pre-disaster pre-layout result to ensure that the initial scheduling conditions are consistent with the deployment plan.

[0118] 2. Path and transportation time constraints:

[0119] When the device is transported from critical node j to node k, it cannot participate in energy supply during transportation:

[0120]

[0121] Among them, t s Is the start time of the disaster.

[0122] 3. Node connection uniqueness constraints:

[0123]

[0124] At any time, the mobile energy storage is only connected to one node or in a transportation state.

[0125] 4. Charge and discharge state logic and power constraints:

[0126]

[0127] In the formula, Is the maximum charging power of mobile energy storage i at time t is the maximum discharge power of mobile energy storage i at time t.

[0128] 5. SOC dynamic change constraint:

[0129]

[0130]

[0131] In the formula, η ch is the charging efficiency of the energy storage, representing the proportion of the actually stored energy during the charging process, and η dis is the discharge efficiency, representing the proportion of the actually released energy during the discharge process.

[0132] 6. Load restoration constraint:

[0133] For all node loads, the reduction amount cannot exceed the original load power:

[0134]

[0135] 7. Operation constraints of electric vehicles, photovoltaic and diesel generators:

[0136] Similar to step one, the photovoltaic operates at a fixed power factor, and the electric vehicle supports the V2G function. Its operation mode is uniformly changed to the discharge mode after the disaster. The specific power constraints are as follows:

[0137]

[0138] In the formula, is the discharge power of the electric vehicle at time t at the key node j, is the maximum discharge power of the electric vehicle, and SOC ev,j,t is the state of charge of the electric vehicle at the key node j at time t, and SOC ev,j,t+1 is the state of charge of the electric vehicle at the key node j at time t + 1, and are the minimum / maximum state of charge of the electric vehicle respectively.

[0139] 8. Electrical safety constraints of the distribution network:

[0140] Node voltage limit: The voltage of each node in the distribution network must be maintained within a reasonable range to ensure the normal operation of electrical equipment and avoid overvoltage or undervoltage. The voltage limit is usually set as:

[0141]

[0142] In the formula, V i is the voltage of node i, V min and V maxare the minimum and maximum limits of the voltage, and generally the value range is 0.9 ≤ V i ≤ 1.1 to ensure the safe operation of the equipment.

[0143] Branch current limit: The current of each branch cannot exceed its maximum carrying capacity to prevent equipment overload and system failures. The branch current limit is usually expressed as:

[0144]

[0145] In the formula, I i,j is the current from node i to the critical node j, is the maximum current carrying capacity of branch i,j. ε is the set of distribution network branches.

[0146] Power flow conservation equation (power balance): According to Kirchhoff's current law (KCL) and Kirchhoff's voltage law (KVL), the sum of the active and reactive power inflows and outflows at each node must be balanced. The power flow conservation equation of the power system is used to ensure that the power flow of the system conforms to physical laws:

[0147]

[0148] In the formula, P ij and Q ij respectively represent the active and reactive power flows from node i to the critical node j, P ji and Q ji respectively represent the active and reactive power flows from the critical node j to node i, P i and Q i are the active power and reactive power of node i.

[0149] Branch power limit: The power flow of each branch has a maximum power limit to ensure that it will not be overloaded. The branch power limit is usually expressed as:

[0150]

[0151] In the formula, f i,j is the power flow from node i to the critical node j, is the maximum power flow of branch i,j.

[0152] Finally, the above constraints are linearized by combining the DistFlow model with the Big-M method, which belongs to the prior art and will not be elaborated in detail here.

[0153] Solution strategy

[0154] The model is a large-scale mixed-integer linear programming (MILP) problem. The scheduling variables accumulate and grow over time, with a high degree of coupling. To improve the solution efficiency, the following strategy is adopted:

[0155] 1. Rolling Horizon Optimization

[0156] The entire post-disaster recovery cycle is segmented and scheduled according to an hourly sliding window to reduce the scale of the problem solved in a single instance.

[0157] 2. Multi-source Block Solving Mechanism

[0158] According to the island division structure, different power source islands are modeled and solved in parallel. The "master coordination - sub-solving" architecture is adopted to improve the global efficiency. This part belongs to the prior art and will not be elaborated in detail here.

[0159] 3. Pre-positioning of Priority Load Constraints

[0160] During the post-disaster scheduling process, the power supply to critical load nodes is considered first and restored preferentially to ensure the rapid restoration of basic social functions. Critical load nodes usually include medical facilities, communication centers, and emergency service centers, etc. At the same time, during the scheduling process, by setting higher weights and restoration priorities, it is ensured that these critical load nodes can restore power supply within the shortest time. The specific constraint form is as follows:

[0161]

[0162] In the formula, is the set of critical load nodes, representing the nodes that need to be restored power supply preferentially. x i,j,t represents whether the mobile energy storage i is connected to the critical node j at time t, ensuring that the critical node j can restore power supply as soon as possible after the disaster. The extreme disaster scenarios described in the present invention include power facility damage scenarios caused by at least one of natural disasters such as typhoons, earthquakes, and ice disasters.

[0163] Example Demonstration

[0164] To demonstrate the invention effect, an improved IEEE 33-node distribution network as shown in Figure 2 is constructed for simulation analysis. The power sources, loads, and system operation parameters are shown in Table 1, Table 2, and Table 3 respectively. The loads with a load shedding cost of 10 yuan in Table 2 belong to critical loads, and the remaining loads belong to non-critical loads. The active power prediction curves of the two types of loads are shown in Figure 3 , the traffic network topology is the same as the power grid, and the road distance between adjacent electrical nodes is 2 km. The photovoltaic prediction parameters before the disaster are shown in Table 4. It is assumed that the disaster occurs at 04:00, resulting in faults in the upstream power grid and distribution lines 3-4, 7-8, 12-22, and 28-29, and the expected repair time is 10 h. Before the disaster, the mobile energy storage has enough time to be configured at the pre-layout nodes, and the electric vehicles in each mobile energy storage and charging pile are in a fully charged state. The congestion degree of the traffic network after the disaster is shown in Figure 4, the active power output of the photovoltaic is the predicted output value for the day before, as shown in Figure 5 .

[0165] Table 1 Power Supply Parameters

[0166]

[0167] Table 2 Load Parameters

[0168]

[0169]

[0170] Table 3 System Operating Parameters

[0171]

[0172] Table 4 Predicted Output of Photovoltaic Units Before Disaster

[0173]

[0174] Result Analysis

[0175] Pre-disaster Pre-layout Results

[0176] Under the condition that the uncertainty τ of photovoltaic output is 0.35 and the parameter ε is 4, the pre-disaster pre-layout plan is shown in Table 5. Among them, Plan 1 considers the pre-layout of the number and location of mobile energy storage, and Plan 2 does not pre-layout the location of mobile energy storage.

[0177] Table 5 Pre-disaster Pre-layout Plan

[0178]

[0179] In Plan 2, the mobile energy storage devices are all located at the root node 1. Although it can reduce the pre-configuration cost of the mobile energy storage location, it needs to bear a load shedding cost of 4,452 yuan. Since the duration of the fault in the pre-disaster prevention stage is unknown, this cost belongs to the risk cost faced by the system. Compared with Plan 2, in Plan 1, the mobile energy storage is pre-configured at Node 12 and Node 24 respectively, reducing the pre-layout cost by 40.5%.

[0180] Post-disaster Recovery Results

[0181] After the disaster, multiple distributed resources cooperate to restore the load power consumption. The dynamic scheduling results of 2 mobile energy storages are shown in Table 5. The relationship between their output power, state of charge and the location of the access node is shown in Figure 6 , Figure 7 , where MESS1 and MESS2 represent the 2 pre-configured mobile energy storages. The power of the supply load in each period and the recovery ratios of the two types of loads are as Figure 8 shown.

[0182] Table 6 Dynamic scheduling results of mobile energy storage

[0183]

[0184] Note: Discharge power is represented as a positive number, and charge power is represented as a negative number.

[0185] From Table 2 and Figure 4 it can be seen that at the moment of the disaster, the two mobile energy storages are located at Node 12 and Node 24 respectively, and the distribution network is divided into three islands. Since there is a critical load of 420 kW at Node 24, the two mobile energy storages are connected to the island where the node is located in sequence for discharging. During the disaster recovery period, the recovery ratio of the critical load remains above 85%, the maximum power supply load is 1832 kW, and the charge and discharge powers of each electric vehicle charging pile are as Figure 9 shown, and the output of the diesel generator and the node voltage are shown in Figure 10 and Figure 11 .

[0186] During the period from 04:00 to 06:00, affected by the congestion of the transportation network after the disaster, MESS1 has not yet reached Node 24 and the PV output is small. The minimum power supply load power is 638 kW. Since the electric vehicles in the three charging piles (EVS1 to EVS3) are all in the discharging state, all diesel generators output electric energy at the maximum active power, and the critical load is not severely reduced, and its recovery ratio still reaches 86%. During the period from 07:00 to 12:00, the PV output gradually increases, and the power supply load power shows an increasing trend. During this period, MESS2 is transferred to Node 29 and stores electric energy at a charging power of 140 kW. The electric vehicle in the EVS1 charging pile also switches to the charging state. Except for the stable distributed power nodes such as diesel generators, the node voltages of other nodes fluctuate significantly, but distributed power sources such as mobile energy storage and diesel generators have a certain voltage support force, and the node voltages of each node still remain within a reasonable range. During the period from 12:00 to 14:00, the PV output decreases and the stored energy of the electric vehicles in the charging piles is exhausted, and the overall power supply load power decreases slightly. Since MESS2 is connected to the critical load Node 26, the recovery ratio of the critical load reaches 98% at this time, and the system resilience is improved to a relatively high level.

[0187] Influence of different recovery strategies on the improvement effect of system resilience

[0188] To verify the advantages of the proposed two-stage distribution network resilience improvement strategy, four restoration strategies shown in Table 6 are set according to whether mobile energy storage location pre-layout is adopted before the disaster and the mobile energy storage dispatching method after the disaster, and three distribution network fault scenarios in Table 7 are respectively simulated. Among them, the location of the mobile energy storage in Strategy 1 is fixed at the root node 1. For the convenience of comparison, the number of mobile energy storage in Strategies 1 and 2 during the post-disaster restoration stage is the same as that in Strategies 3 and 4. In Strategy 3, single scheduling means that the mobile energy storage only performs one position scheduling of the pre-layout, and Strategy 4 is the strategy proposed in the present invention. The example parameters are the same as those in the example demonstration part. According to the critical load power restored in each time period, the restoration ratio curves of the critical load in each strategy are obtained, as Figure 12 shown. The load shedding costs of each restoration strategy under different fault scenarios are shown in Table 8.

[0189] Table 7 Comparison of Four Restoration Strategies

[0190]

[0191]

[0192] Table 8 Comparison of Fault Scenarios

[0193]

[0194] Table 9 Load Shedding Costs of Each Strategy in Different Fault Scenarios

[0195]

[0196] It can be seen from Table 8 and Table 9 that in the three fault scenarios, the load shedding costs of the restoration strategies based on mobile energy storage pre-layout and dynamic dispatching are the lowest. In Scenario 1, it is reduced by 34.3%, 6.1%, and 21.7% compared with Strategies 1, 2, and 3 respectively. In terms of the restoration effect of the critical load, since the load spatial distribution and distributed power generation output conditions have been considered in the pre-disaster model, the pre-layout points of the mobile energy storage obtained are near the critical nodes with a higher risk of load loss. Therefore, in Restoration Strategies 3 and 4 that consider the coupling of pre-disaster pre-layout and post-disaster restoration, the mobile energy storage can quickly access the critical nodes, and the critical load shedding volume in the early stage of the fault can be significantly reduced, thus improving the resilience of the distribution network. Because the traffic distance between the root node 1 and the critical load node 24 in the example is relatively close, the mobile energy storage in Strategy 2 can also quickly reach node 24. Therefore, the critical load shedding volume after 2 hours of the fault is significantly reduced compared with Strategy 1. Since Strategies 2 and 4 can dynamically dispatch the mobile energy storage, the optimal allocation of energy in the time and space dimensions can be achieved, and various distributed resources can be fully utilized. Therefore, the restoration ratio of the critical load in the later stage of the fault is relatively high, reaching more than 90% overall.

Claims

1. A mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of distribution networks, characterized in that, Including: Pre-disaster pre-layout module, configured to: Establish a two-stage robust optimization model considering extreme disaster scenarios, where the decision variables in the first stage include the number of energy storage configurations, the nodes for energy storage configuration, and the network topology state, and the optimization objective is to minimize the energy storage configuration cost; the decision variables in the second stage include the load shedding power and the power output of power sources, and the optimization objective is to minimize the critical load shedding cost under the worst photovoltaic output scenario; Solve the two-stage robust optimization model and output the optimal energy storage configuration plan and network topology state; Post-disaster dynamic scheduling module, configured to: Collect traffic network status, distributed power generation output data, and load demand data in real time; Construct a spatio-temporal multi-dimensional dynamic scheduling model, whose decision variables include the migration path sequence of mobile energy storage, the charge and discharge power values, and the node positions for connecting to the distribution network; Dynamically update the scheduling instructions based on the rolling time domain optimization framework; Coordinated control module, configured to: Establish a multi-source coordinated control strategy, and the priority order is: power supply guarantee for critical loads > maintenance of distribution network topology stability > V2G coordination between mobile energy storage and electric vehicles > output complementarity between diesel generators and photovoltaics; Realize power distribution through a distributed control architecture, where mobile energy storage serves as the main regulation unit.

2. The mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of the distribution network according to claim 1, wherein: The worst photovoltaic output scenario is determined in the following way: The uncertainty of photovoltaic output is modeled using an ellipsoidal set, defined as: where $\xi$ is the set of uncertain scenarios, representing the set of actual output powers of all photovoltaic nodes at the current moment; is the set of predicted scenarios, representing the set of predicted output powers of photovoltaic nodes, given by the time series prediction model LSTM; $Q$ is the covariance matrix trained based on historical prediction errors; $\tau$ is the scale factor for controlling the boundary of the uncertainty set; is the n-dimensional real number space, is the photovoltaic processing uncertainty set; is the matrix transpose symbol; Select the photovoltaic output scenario that maximizes the load shedding cost in the uncertainty set as the worst scenario.

3. The mobile energy storage pre-layout and dynamic scheduling system for enhancing the resilience of the distribution network according to claim 1, wherein: In the post-disaster dynamic scheduling module, the travel time constraint is calculated as: Among them, τ j,k,t is the travel time from node i to the critical node j at the post-disaster moment t, d jk is the equivalent travel distance, and v j,k,t is the actual vehicle speed under the disaster.

4. The mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of the distribution network according to claim 1, wherein: The coordinated control module executes the following control logic: Electric vehicles switch to the discharge mode after the disaster to provide standby power; Diesel generators operate in the maximum output mode to supplement the energy storage power gap; Dynamically adjust the output power of the photovoltaic system according to the lighting conditions and give priority to supplying local loads.

5. The mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of the distribution network according to claim 1, characterized in that: The pre-disaster pre-layout module uses the column and constraint generation algorithm for solution, including: The master problem is a mixed-integer second-order cone programming model for determining the mobile energy storage configuration plan; The sub-problem is the worst photovoltaic output scenario search problem, which is transformed into a linear form through the strong duality principle and iteratively alternates with the master problem until convergence.

6. The mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of the distribution network according to claim 1, wherein: In the post-disaster dynamic scheduling process, the following operating constraints are also included in the post-disaster dynamic scheduling module: The node voltage should be maintained within the allowable fluctuation range; The branch current shall not exceed the thermal limit value; The state of charge of the energy storage device is limited to the set upper and lower limits.

7. The mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of the distribution network according to claim 1, characterized in that: The extreme disaster scenarios include power facility damage scenarios caused by at least one of natural disasters such as typhoons, earthquakes, and ice disasters.

8. A method for scheduling based on the mobile energy storage pre-layout and dynamic scheduling system for improving the resilience of the distribution network according to any one of claims 1 to 7, characterized in that, The method includes the following steps: Pre-disaster stage: Based on the photovoltaic output prediction and its uncertainty, construct a two-stage robust optimization model to generate the optimal deployment plan of mobile energy storage; where the decision variables in the first stage include the number of energy storage configurations, the nodes for energy storage configuration, and the network topology state, and the optimization objective is to minimize the energy storage configuration cost; the decision variables in the second stage include the load shedding power and the power output of power sources, and the optimization objective is to minimize the critical load shedding cost under the worst photovoltaic output scenario; Post-disaster stage: Based on the traffic network conditions and the output status of distributed power sources under the disaster situation, a spatio-temporal multi-dimensional dynamic scheduling model is constructed. The charging and discharging behaviors and spatial migration of mobile energy storage are optimized based on the rolling horizon optimization framework, and the operation constraints of node voltage, branch current, and the state of charge of energy storage devices are satisfied during the scheduling process; During the scheduling process, in accordance with the priority order of critical load power supply guarantee > maintaining the topological stability of the distribution network > V2G coordination between mobile energy storage and electric vehicles > complementary output of diesel generators and photovoltaic, the multi-source collaborative operation of electric vehicles, photovoltaic, and diesel generators is jointly considered, and critical loads are restored layer by layer while ensuring system security constraints.

9. The method according to claim 8, wherein: The scheduling process in the post-disaster stage advances the fault recovery step by step using rolling horizon optimization at one-hour intervals.

Images