A resilience recovery method for energy interconnection distribution system under ice disaster scenario

Abstract

The present invention proposes a method for recovering the resilience of energy interconnected distribution systems in ice disaster scenarios. Based on deep learning theory, historical data features are extracted to generate photovoltaic output scenarios in ice disaster weather, and a time series model of each distributed resource is established. Through multi-stage scheduling of mobile energy storage, fixed energy storage, integrated energy and other distribution network flexibility resources, the optimal energy storage configuration scheme and distributed power output are given under the premise of meeting the emergency microgrid networking constraints and load power consumption characteristics, so as to maximize the restoration of power supply to important loads and improve the resilience of the distribution network. The method of the present invention can not only reduce the large-scale storage of emergency materials, but also help to improve the flexibility of material application and realize the spatiotemporal optimization scheduling of energy.

Description

Technical Field

[0001] The present invention belongs to the field of emergency recovery of power distribution systems, and in particular relates to a method for resilient recovery of energy interconnected power distribution systems in ice disaster scenarios. Background Art

[0002] In order to cope with the dual crises of energy depletion and environmental pollution, a large number of distributed renewable energy and electric energy replacement loads are connected to the distribution network. At the same time, in order to improve energy utilization efficiency, the integrated energy system will be vigorously developed in the distribution network. These flexible resources can not only effectively improve system reliability, but also provide solutions for post-disaster load recovery of the distribution network. How to effectively and fully utilize these flexible resources to improve the resilience of the distribution network when a disaster occurs has received widespread attention.

[0003] At present, scholars at home and abroad have conducted extensive research on the use of distributed resources for post-disaster power grid recovery, mainly including isolated microgrids, emergency generators and mobile energy storage. In terms of isolated microgrids, reference [1] proposed a microgrid formation mechanism that uses distributed power sources and remote control switches to restore important loads after the main grid is paralyzed. Reference [2] uses diesel generators, fixed energy storage and photovoltaic cells to restore important loads when the main grid and some distribution lines fail and the load is cut off. First, the network structure is determined based on the minimum spanning tree algorithm, and then the load to be restored and the power output are solved. Reference [3] proposes a load recovery method based on multiple planned microgrids. When the main grid power supply is interrupted, each microgrid performs rolling optimization scheduling of its fixed energy storage, micro gas turbine and other resources, and then the microgrid with surplus power generation capacity supplies power to the microgrid with load reduction through the tie switch, so as to minimize the load reduction and operating costs. Reference [4] proposed a load recovery strategy that takes into account the distribution of important loads and the location and capacity of fuel-based distributed power sources in the scenario of load outages caused by distribution line failures. First, the location and scale of the pre-disaster fuel power source are determined based on the cost budget and load type. Then, the operating cost in the fault scenario is minimized by scheduling switches and power output, reducing the fixed power reserve. The above studies are all single-stage recovery, without considering the load recovery duration, and require pre-configuration of microgrids or a large number of fixed controllable power sources. However, due to the high cost of microgrid construction and control at this stage, and the large-scale configuration of power sources such as diesel generators is not conducive to carbon emission reduction, the widespread popularization of microgrids is affected.

[0004] With the deepening of research, in addition to considering grid reconstruction and static dispatching resources such as controllable power sources and microgrids at fixed locations, fault optimization recovery considering dynamic resources has recently received attention. In addition, power grid companies have applied mobile resources such as diesel generators to power grid restoration, but the number and resources are limited. Reference [5] pre-places emergency vehicles in temporary storage locations based on load priority and demand scale before a disaster. When the distribution network line fails, emergency diesel generators are dispatched to form a microgrid to restore important loads. Reference [6] In the case of loss of main grid power supply, considering the damage to some lines, the access location of mobile energy storage is optimized with the goal of minimizing the amount of lost load, and a microgrid is formed with local photovoltaics. Reference [7] proposes a restoration method based on multi-stage spatiotemporal scheduling. Based on multiple planned microgrids, the multi-stage dispatching of mobile energy storage minimizes the system load interruption cost and operating cost. The above studies use mobile power sources independently or in conjunction with fixed resources to restore loads. The optimization process of the restoration strategy rarely considers fault scenarios and source-load uncertainty, as well as their impact on the continuous operation of the microgrid. Moreover, with the rapid development of integrated energy, it can not only improve energy utilization efficiency, but also quickly form microgrids with mobile energy storage in disaster situations, and use complementary coupling between resources to reduce fault losses.

[0005] In summary, it can be seen that the rational and effective use of flexible resources such as mobile energy storage and integrated energy systems is of great significance to improving the resilience of distribution systems and the efficiency of emergency resource allocation.

[0006] References:

[0007] [1]Chen C,Wang J,Qiu F,et al.Resilient Distribution System byMicrogrids Formation After Natural Disasters[J].IEEE Transactions on SmartGrid,2016,7(2):958-966.

[0008] [2]Wang Y, Xu Y, He J, et al.Coordinating Multiple Sources for ServiceRestoration to Enhance Resilience of Distribution Systems[J]. IEEETransactions on Smart Grid, 2018.

[0009] [3]Bian Y,Bie Z.Multi-Microgrids for Enhancing Power SystemResilience in Response to the Increasingly Frequent Natural Hazards[J].IFAC-PapersOnLine,2018,51(28):61-66.

[0010] [4] Shi Q, Li F, Kuruganti T, et al. Resilience-Oriented DG Siting and Sizing considering Stochastic Scenario Reduction[J]. IEEE Transactions on Power Systems, 2020, PP(99):1-12.

[0011] [5] Lei S, Wang J, Chen C, et al. Mobile Emergency Generator Pre-Positioning and Real-Time Allocation for Resilient Response to Natural Disasters[J]. IEEE Transactions on Smart Grid, 2018, 9(3): 2030-2041.

[0012] [6] Ren Junzhi, Chen Jian, Jiang Xinyi, et al. Post-disaster recovery strategy of elastic distribution network considering mobile energy storage and network reconstruction [J]. Electric Power Construction, 2020, v.41; No.474(03):88-94.

[0013] [7]Yao S, Wang P, Liu X, et al. Rolling Optimization of Mobile EnergyStorage Fleets for Resilient Service Restoration[J]. IEEE Transactions onSmart Grid, 2019. Summary of the invention

[0014] In response to the above problems, the present invention proposes a method for resilient recovery of energy interconnected distribution systems under ice disaster scenarios. Based on deep learning theory, historical data features are extracted to generate photovoltaic output scenarios under ice disaster weather, and a timing model of each distributed resource is established. Through multi-stage scheduling of flexible distribution network resources such as mobile energy storage and comprehensive energy, the optimal energy storage configuration plan and distributed power output are given under the premise of meeting the networking constraints of emergency microgrids and load power consumption characteristics, thereby maximizing the restoration of power supply to important loads and improving the resilience of the distribution network.

[0015] The present invention proposes a method for resilient recovery of an energy interconnected power distribution system in an ice disaster scenario, comprising:

[0016] Step 1: Extract historical data features based on deep learning theory to generate photovoltaic output scenarios under ice disaster weather;

[0017] Step 2: Establish the spatiotemporal state model of fixed and mobile energy storage and the timing model of the integrated energy system based on the dynamic and static characteristics of the power supply;

[0018] Step 3: Considering the networking constraints of the emergency microgrid and the power consumption characteristics of the load, a multi-stage resilience recovery model of the energy interconnection distribution system is established with the goal of maximizing the continuous power supply time of the emergency load;

[0019] Step 4: Linearize and solve the model using cone optimization and big-M method.

[0020] In the present invention, the conditional generative adversarial network method is used to introduce conditional input items at the input ends of the generator and the discriminator, so that the generator can be effectively applied to scenarios with specific data requirements; the Wasserstein distance is used to characterize the gap between the generated data and the real data, and the gradient penalty item is introduced to replace the gradient truncation strategy of the traditional WGAN.

[0021] Step 1 specifically includes:

[0022] Step 1.1: Preprocess the historical photovoltaic output data and initialize the neural network parameters;

[0023] Step 1.2: The predicted and actual PV output data under freezing weather conditions are used as inputs to the generator neural network and the discriminator neural network, respectively, to train the neural network;

[0024] Step 1.3: Use the photovoltaic day-ahead output forecast data as the input of the generator neural network to generate photovoltaic output scenarios;

[0025] Step 1.4: Use the k-medoids clustering algorithm to reduce the photovoltaic output scenarios, use the silhouette coefficient method to evaluate the clustering effect, and generate photovoltaic output scenarios under ice disaster weather.

[0026] In step 2, a spatiotemporal scheduling model of mobile energy storage, a spatiotemporal scheduling model of fixed energy storage, and an energy time series coupling model of the integrated energy system are established.

[0027] The control variable of the mobile energy storage is the charging power at time t Discharge power And the connection status between mobile energy storage i and node j After receiving the dispatch instruction, if the mobile energy storage needs to be transferred from node j to node k for charging and discharging, the shortest traffic road distance D between the two nodes is first calculated based on the Floyd algorithm. j,k , considering the transportation time of mobile energy storage Installation and configuration time Before the mobile energy storage reaches node k, that is, the scheduling interval △t is less than When its connection status with node k Always 0. In addition, the following constraints must be met: 1) A mobile energy storage can be connected to at most one node at any time; 2) The mobile energy storage can only be charged and discharged when it is in a connected state within a certain scheduling period; 3) The charging and discharging power, charging and discharging state, and charge state constraints of the mobile energy storage.

[0028] During the load recovery phase, fixed-location energy storage charges and discharges energy in the corresponding microgrid as the dispatch instructions change. Its charging and discharging power, charging and discharging state, and capacity need to meet certain constraints.

[0029] Assuming that the natural gas network and equipment in the integrated energy system are not affected by the freezing disaster, after the failure occurs, the energy consumption of natural gas and heat load can be converted into the energy consumption of electric load through the electric, heat and gas coupling equipment. The power supply of important loads is guaranteed by dispatching the output of CHP units and natural gas heat pumps in each period, and the time series model of the integrated energy system is established. The dispatching instructions are the reduction of electric and thermal loads and the output of each unit in each period. The electric and thermal power of CHP units and natural gas heat pumps is used as the control variable, and the natural gas distribution coefficient σ is introduced. When the amount of natural gas input into the integrated energy system per unit time is constant, the output of the units in the integrated energy system can be indirectly controlled by adjusting the natural gas distribution coefficient in each period.

[0030] In step 3, the recovery phase of the system function curve during the disaster process is divided into three periods. In the first period, the remote control equipment is used to quickly identify and isolate the fault; in the second period, the flexible resources are dispatched to achieve reasonable distribution of electricity in time and space to improve system resilience; in the third period, the faulty equipment is repaired to restore the system resilience to normal levels. Therefore, a multi-stage optimization recovery model is established with fixed locations of mobile energy storage. Controllable power output Power supply load power To control the variables, the optimal comprehensive energy, fixed energy storage, mobile energy storage and controllable distributed power output and load shedding strategy are given with the goal of maximizing the continuous power supply time of the emergency load. The constraints of the multi-stage optimization recovery model include load characteristic constraints and emergency microgrid networking constraints. Load characteristic constraints include dual controllable power supply constraints, minimum power supply duration constraints and maximum power supply interruption constraints. Emergency microgrid networking constraints include topology connection constraints, minimum controllable power supply capacity ratio constraints and flow constraints.

[0031] In step 4, the model is linearized using cone optimization and big-M method and solved using the MATLAB toolbox.

[0032] Beneficial effects of the invention: The present invention targets freezing disasters and utilizes improved conditions to generate adversarial networks to identify data deep features. In the present invention, based on the dynamic and static characteristics of the power supply, a fixed and mobile energy storage spatiotemporal state model and an integrated energy system timing model are established, and complementary coupling between resources is used to reduce fault losses; taking into account the controllability of the power supply and the power consumption characteristics of the load, a multi-stage optimization model that considers dynamic and static resources such as photovoltaics, mobile energy storage, and integrated energy is established to solve the optimal load recovery plan. The method of the present invention can not only reduce the large-scale storage of emergency materials, but also help to improve the flexibility of material application and realize the spatiotemporal optimization scheduling of energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Shown is the schematic structure of the energy interconnection distribution system;

[0034] Figure 2 What is shown is the schematic structure of the improved generative adversarial network in the method of the present invention;

[0035] Figure 3 Shown is the spatiotemporal regulation of the mobile energy storage system;

[0036] Figure 4 It is the schematic structure of the integrated energy system;

[0037] Figure 5 It is the function curve of the system during the disaster process;

[0038] Figure 6 It is the model solving flow chart. DETAILED DESCRIPTION

[0039] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0040] From the perspective of improving system resilience, this invention comprehensively considers emergency resources such as photovoltaics, mobile energy storage, fixed energy storage, integrated energy systems and diesel generators, and maximizes the restoration of important loads through multi-stage optimization and scheduling of flexible resources. Figure 1The figure shows the schematic structure of the energy interconnection distribution system. At present, 5G application technology is gradually maturing, and power companies will use advanced communication equipment to support the communication control requirements of distributed resources in the future.

[0041] A method for resilient recovery of an energy interconnected power distribution system in an ice disaster scenario of the present invention comprises:

[0042] Step 1: Extract historical data features based on deep learning theory and generate photovoltaic output scenarios under ice disaster weather.

[0043] Most power outages in power systems are caused by extreme weather, which has a great impact on photovoltaic and comprehensive energy output. Among them, freezing disasters affect a wide area and occur frequently. In freezing disaster scenarios, photovoltaics are affected by factors such as ice and snow coverage and temperature, and their output shows strong temporal and volatility. In order to adapt to the uncertainty of photovoltaic output in this scenario, a photovoltaic output scenario generation method based on data feature mining is adopted to explore the impact characteristics of light intensity and temperature on regional photovoltaic historical output under freezing disasters. In order to study the load recovery strategy of energy interconnected distribution systems under disaster scenarios, the scenario reduction technology is used to extract photovoltaic day-ahead output scenarios, build a refined model and make it have better generalization capabilities.

[0044] (1) Data feature identification

[0045] Traditional scene generation methods such as Markov chain and scene tree generation methods are difficult to describe implicit data features such as weather and space. Generative adversarial networks based on deep learning theory are gradually being applied to renewable energy output scene generation due to their characteristics of data feature mining and unsupervised learning. Through the mutual training of two sets of neural networks, the generator and the discriminator, the implicit features in the massive output historical data are identified, thus generating a new output scene.

[0046] The original generative adversarial network still has some shortcomings in practical applications: mode collapse, the generator only generates data that the discriminator cannot distinguish in subsequent training, and cannot be applied to scenarios with specific data requirements; gradient disappearance, using JS divergence to calculate network loss cannot well measure the distance between generated data and real data, which is not conducive to the back propagation of neural networks.

[0047] The present invention proposes an improved generative adversarial network to improve the above-mentioned shortcomings, and its schematic structure is shown in the attached figure. Figure 2As shown. On the one hand, the conditional generative adversarial network method is used to introduce conditional input items at the input of the generator and the discriminator, so that the generator can be effectively applied to scenarios with specific data requirements; on the other hand, the Wasserstein distance is used to characterize the gap between the generated data and the real data, and the gradient penalty term is introduced to replace the gradient truncation strategy of the traditional WGAN to prevent the gradient vanishing explosion problem and improve the training speed. The loss function of the improved generative adversarial network is shown in the following formula.

[0048]

[0049]

[0050] Among them, G represents the generator network; D represents the discriminator network; x represents the real data; represents the data generated by the generator network; x represents the sampling point between the real data and the generated data; α is the scale parameter; y represents the conditional input item; λ represents the gradient penalty parameter; pr, pg and Represent the distribution functions of real data, generated data and sampling points respectively. The historical output data of regional photovoltaic power is used as real data input to improve the generative adversarial network. After multiple trainings, the generator network will discover the implicit features in a large amount of historical photovoltaic output data. If the day-ahead predicted output data is used as a conditional input item and combined with a set of random noise and input into the generator network, the generated data can be used as a photovoltaic output scenario under the day-ahead predicted conditions.

[0051] (2) Photovoltaic output scenarios are reduced

[0052] Clustering algorithms are often used to improve computational efficiency in the reduction of large-scale scenarios. Among clustering algorithms, the K-means algorithm uses the mean of sample points of the same type as the cluster center, which is very sensitive to abnormal data and initial center selection, and is not conducive to the extraction of extreme scenarios such as disasters. The K-medoids algorithm uses the sample center point as the cluster center, has good robustness to outliers, and effectively overcomes the shortcomings of the K-means algorithm. Therefore, the K-medoids algorithm is used to reduce photovoltaic output scenarios. The process is as follows:

[0053] 1) Randomly select K points from N data objects as the initial center set;

[0054] 2) Calculate the distance between each non-representative object and each center point and assign it to the cluster closest to it;

[0055] 3) For each non-central object, replace one of the central points with the current point and calculate the cost function generated by the replacement. If it is negative, replace the central point, otherwise restore the central point;

[0056] 4) Repeat 2) and 3) processes 2-4 times until the number of iterations is reached or the cost function is within a certain range, and finally K center points are obtained, and all data objects are re-divided;

[0057] Based on the above results, the silhouette coefficient method is used to evaluate the clustering effect of the K-medoids algorithm, so as to optimize the optimal number of cluster centers. For a sample i, the silhouette coefficient is composed of cohesion and separation. Cohesion refers to the average value of the distance from sample i to other sample points of the same class, and separation refers to the minimum value of the average distance from sample i to other clusters. The silhouette coefficient takes into account both intra-cluster and extra-cluster factors. The closer its value is to 1, the better the classification effect of the sample is, and the closer it is to -1, the sample may be misclassified. The quality of the overall clustering effect can be expressed by the average value of the silhouette coefficient of each sample. The larger the value, the better the clustering effect, the closer the distance within the cluster and the farther the distance between clusters. The silhouette coefficient of a single sample point can be expressed by the following formula:

[0058]

[0059] Among them, S i is the silhouette coefficient of sample i; a i b is the cohesion degree; i is the separation degree; C i is the cluster where sample i belongs; n i is the number of samples in the cluster where sample i is located; Dis(j,i) is the distance between sample j and sample i; K is the number of samples.

[0060] Step 2: Establish the spatiotemporal state model of fixed and movable energy storage and the timing model of the integrated energy system based on the dynamic and static characteristics of the power supply.

[0061] In order to describe the spatiotemporal characteristics of energy storage and take into account the dispatch time scale in disaster situations, the dispatch interval is set to 1 hour.

[0062] 2.1 For mobile energy storage

[0063] The temporal and spatial regulation of mobile energy storage is shown in the attached figure. Figure 3 The control variable is the charging power at time t. Discharge power And the connection status between mobile energy storage i and node j

[0064] After receiving the dispatch instruction, if the mobile energy storage needs to be transferred from node j to node k for charging and discharging, the shortest traffic road distance D between the two nodes is first calculated based on the Floyd algorithm. j,k , considering the transportation time of mobile energy storage Installation and configuration time Before the mobile energy storage reaches node k, that is, the scheduling interval △t is less than When its connection status with node k Always 0. In addition, the following constraints must be met: 1) A mobile energy storage can be connected to at most one node at any time; 2) The mobile energy storage can only be charged and discharged when it is in a connected state within a certain scheduling period; 3) The charging and discharging power, charging and discharging state, and charge state constraints of the mobile energy storage.

[0065] At this time, the spatiotemporal scheduling model of mobile energy storage is described as:

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073]

[0074] Among them, ME is the collection of movable energy storage; Transportation speed for mobile energy storage; is a binary variable, indicating the charging state of the mobile energy storage, 1 means charging; is a binary variable, indicating the discharge state of the removable energy storage, 1 indicates discharge; The capacity of the mobile energy storage; Charging efficiency for mobile energy storage; is the discharge efficiency of mobile energy storage; is the upper limit of the mobile energy storage capacity; is the lower limit of the mobile energy storage capacity. Formula (4) determines the shortest transportation time of mobile energy storage i between node j and node k; Formula (5) indicates that when the scheduling interval is less than the sum of the transportation time and the configuration time, the connection state of the mobile energy storage at node k is 0; Formula (6) indicates that the mobile energy storage i can connect to at most one node at the same time; Formulas (7) and (8) give the upper and lower limits of the charging and discharging power of the mobile energy storage; Formula (9) represents the coupling relationship between the charging and discharging state of the mobile energy storage and the spatial state, ensuring that the mobile energy storage can be charged and discharged when it is in the connected state; Formulas (10) and (11) are the charge state constraints of the mobile energy storage.

[0075] The nonlinear constraint shown in equation (9) is transformed into the following equation using the linearization method:

[0076]

[0077] in, is the connection status between mobile energy storage i and node j in the scheduling period t-t+1, 1 indicates connection and 0 indicates disconnection; B is the set of nodes.

[0078] 2.2 Energy storage for fixed locations

[0079] Charging power of fixed location energy storage at time t With discharge power To control the variables, during the load recovery phase, the energy is charged and discharged in the corresponding microgrid as the dispatching instructions change. The charging and discharging power, charging and discharging state, and capacity need to meet the following constraints:

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] Among them, E is the set of fixed energy storage; is a binary variable, indicating the charging state of the fixed energy storage, 1 means charging; is a binary variable, indicating the discharge state of the fixed energy storage, 1 indicates discharge; is the capacity of fixed energy storage; Charging efficiency for stationary energy storage; is the discharge efficiency of fixed energy storage; is the upper limit of fixed energy storage capacity; It is the lower limit of fixed energy storage capacity.

[0086] 2.3 For integrated energy systems

[0087] The small integrated energy system including CHP unit and natural gas heat pump is taken as the object, and the schematic structure is shown in the attached Figure 4 Assuming that the natural gas network and equipment in the integrated energy system are not affected by the freezing disaster, after the failure occurs, the energy consumption of natural gas and heat load can be converted into the energy consumption of electric load through the electric, thermal and gas coupling equipment, and the power supply of important loads can be guaranteed by dispatching the output of CHP units and natural gas heat pumps in different time periods. Therefore, the time series model of the integrated energy system is established.

[0088] In the integrated energy system, the conversion relationship between different forms of energy such as electricity and gas supplying electricity and heat loads through coupling equipment can be expressed in the following matrix form:

[0089]

[0090] Among them, L e Indicates the electrical load output within the comprehensive energy source; L t Indicates the gas load output within the comprehensive energy source; E e Indicates the electrical energy input into the comprehensive energy; E g represents the natural gas energy input in the comprehensive energy; the coupling matrix C describes the energy conversion relationship between input and output, C ee Indicates the energy conversion efficiency of the electricity-to-electricity conversion device, C te Indicates the energy conversion efficiency of gas-to-electricity conversion equipment, C et Indicates the energy conversion efficiency of the electricity-to-gas conversion equipment, C tt Indicates the energy conversion efficiency of gas-to-gas conversion equipment.

[0091] Attachment Figure 4 The integrated energy system shown in the figure has a dispatching instruction for the reduction of electric heat load and the output of each unit in each period. The electric heat power of CHP unit and natural gas heat pump is used as the control variable, and the natural gas distribution coefficient σ, σE is introduced. g represents the amount of natural gas consumed by the CHP unit, (1-σ)E g It represents the amount of natural gas consumed by the natural gas heat pump. Therefore, when the amount of natural gas input into the integrated energy system per unit time is constant, the output of the units in the integrated energy system can be indirectly controlled by adjusting the natural gas distribution coefficient in each period. Therefore, the energy time series coupling model of the integrated energy system can be expressed as:

[0092]

[0093] Among them, η T Indicates transformer efficiency; Indicates the electricity conversion efficiency of the CHP unit; Represents the thermal conversion efficiency of the CHP unit; η HP Indicates the thermal efficiency of the natural gas heat pump.

[0094] Step 3: Considering the networking constraints of the emergency microgrid and the power consumption characteristics of the load, a multi-stage resilience recovery model of the energy interconnection distribution system is established with the goal of maximizing the continuous power supply time of the emergency load.

[0095] After the distribution network loses power supply from the main grid due to a freezing disaster, an emergency microgrid is established through multi-stage dispatching of mobile energy storage, integrated energy systems, fixed energy storage, photovoltaics and other fixed power sources to restore load power supply through the microgrid.

[0096] In the past, during the emergency recovery phase of the power distribution system, repair of faulty equipment and emergency power dispatch were the main measures to improve resilience. However, when the large power grid was paralyzed, it was impossible to restore a large number of power-off loads in a short period of time. At the same time, in disaster scenarios with a longer duration, the continuous characteristics of loads and power sources are very different, resulting in time and space imbalance. This method mainly focuses on the recovery phase of the system's functional curve during the disaster process, as shown in the attached figure. Figure 5 The t3-t5 period is shown. In the t3-t4 period, similar to the traditional recovery strategy, the fault is quickly identified and isolated through remote control equipment; in the t4-t5 period, the flexible resources such as mobile energy storage, integrated energy, and fixed energy storage are dispatched to achieve reasonable distribution of electricity in time and space, which can significantly improve the system resilience compared with the traditional recovery strategy, as shown in the shaded area in the figure. Finally, in the t5-t6 period, the system resilience can be restored to normal levels by repairing the faulty equipment.

[0097] Therefore, considering the multi-stage timing of flexible resources such as mobile energy storage and integrated energy, and taking into account the uncertainty of photovoltaic output scenarios under disaster scenarios, a multi-stage optimization recovery model based on spatiotemporal flexible resources is established.

[0098] 3.1 Multi-stage optimization recovery model objective function

[0099] In disaster situations, the objective function and constraints are usually formulated based on the proposed emergency microgrid networking strategy and the power consumption characteristics of different loads, with the fixed location of the mobile energy storage Controllable power output Power supply load power To control the variables, the optimal controllable distributed power output and load shedding strategy of comprehensive energy, fixed energy storage, mobile energy storage, etc. is given with the goal of maximizing the continuous power supply time of the emergency load. The objective function is:

[0100]

[0101] Where T is the time period during which the fault occurs; B is the set of nodes; w i is the load weight determined according to the load type.

[0102] 3.2 Multi-stage optimization recovery model constraints

[0103] (1) Load characteristic constraints

[0104] The importance and differences of various loads should be considered during load recovery. Electric loads can be divided into primary load, secondary load and tertiary load according to interruption losses. The primary load is the most important. Sudden power outage will cause casualties, huge economic losses and serious disruption of social order. The weight of each load node is divided according to the load level. i For the most important primary load, considering the actual power consumption characteristics and power supply reliability of the load, the following constraints are established for power supply restoration of important users:

[0105] 1) Dual controllable power supply constraints

[0106] In disaster scenarios, for primary loads such as hospitals that affect personal safety, the number of controllable power sources that should be connected in the microgrid k where they are located should not be less than 2 to ensure reliable power supply during the entire fault recovery period. Compared with photovoltaic generator sets, diesel generators, energy storage and cogeneration units have the characteristics of stable and adjustable output and can be used as controllable power sources. Among them, mobile energy storage must be directly connected to this type of load to serve as a controllable power source.

[0107]

[0108] in, is the number of controllable power sources that should be connected in microgrid k.

[0109] 2) Minimum power supply duration constraint

[0110] When a disaster occurs, the duration of the primary load that affects social order, such as emergency lighting in underground stations, fire passages, etc., should be greater than the minimum continuous power supply time specified in the national standard. The power supply status of the load should meet the following conditions:

[0111]

[0112] in, is the minimum continuous power supply time of load i; is the power supply status of load i, 1 means it is in the power supply state.

[0113] 3) Maximum power interruption times constraint

[0114] The number of power outages of the load is an important indicator reflecting the continuity and reliability of power supply. In emergency scenarios, the number of power outages of other primary loads that have an impact on the economy should not be greater than the maximum number of power outages. The power supply status of the load should meet the following conditions:

[0115]

[0116] in is the maximum number of power interruptions for load i.

[0117] (2) Emergency microgrid networking constraints

[0118] In order to restore the power supply of important types of loads as soon as possible, economic costs are generally not a constraint factor for post-disaster recovery, but power supply reliability and safety are still the basic requirements for formulating recovery strategies. Therefore, based on the above microgrid composition principles and load requirements, an emergency microgrid networking constraint model is established:

[0119] 1) Topological connection constraints

[0120] When the distribution network topology is reconstructed, the network needs to satisfy the radial structure. Based on graph theory, the isolated islands are first identified according to the line switch status, that is, the connected subgraph is identified, and then the child and parent node sets of each node in each isolated island are obtained, and the connectivity constraints in the microgrid are established using the virtual power flow method.

[0121]

[0122]

[0123]

[0124]

[0125] Among them, α i,j,t is the switch state of line ij, 1 means closed; L is the set of lines; N B is the total number of nodes; N S is the total number of nodes; δ(i) is the set of child nodes of node i; γ(i) is the set of parent nodes of node i; B S is the set of virtual source nodes in the connected subgraph; F i,j,t is the virtual power flow on line ij; is the virtual power emitted by each source node; M is a very large number.

[0126] 2) Minimum controllable power capacity ratio constraint

[0127] In order to stabilize the voltage and frequency of the microgrid, during the entire period of load recovery, the controllable power capacity in each microgrid should account for a certain proportion of the total power capacity. Therefore, the minimum controllable power capacity proportion principle for forming a feasible microgrid is:

[0128]

[0129] Among them, μ is the proportion of controllable power supply; S i is the rated capacity of the i-th distributed generation in the microgrid; C j is the set of controllable distributed generation sources in the jth microgrid; U iis the set of uncontrollable distributed power sources in the jth microgrid; μ min It is a constant, indicating the minimum proportion of controllable power capacity, and its range is [0-1].

[0130] 3) Power flow constraints

[0131] For radial distribution networks, the DistFlow power flow equation is used. Since the network topology changes with the state of the line switches, the large M method is introduced to relax and improve the voltage equation in the traditional DistFlow power flow model. For the nonlinear constraints between voltage, current and power, the model is linearized by the second-order cone optimization method.

[0132]

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143] in, is the active power of the power supply load at node i; is the active power output of the distributed generation at node i; P ij,t is the active power flowing from node i to node j; Q ij,t is the reactive power flowing from node i to node j; S max is the upper limit of line capacity; is the reactive power of the power supply load at node i; is the reactive power output of the distributed generation at node i; is the square term of the node voltage amplitude at node i; is the square term of the current amplitude on branch ij; R ij is the resistance of branch ij; X ij is the reactance of branch ij.

[0144] Step 4: Linearize the model using cone optimization and big-M method, and solve the model using MATLAB.

[0145] Attached Figure 6 What is shown is the solution flow chart of the entire model. For the multi-stage optimization model established above, the network parameters and node types are first initialized, and the pytorch framework is used to build and train the improved generative adversarial network. The K-medoids clustering algorithm is used to generate the photovoltaic output scenario under the ice disaster. Based on the photovoltaic, mobile energy storage and integrated energy system component models, the objective function and constraint conditions of the multi-stage optimization model are established, and the model is linearized using cone optimization and big M method, and then the YALMIP toolbox in MATLAB is called to solve it.

[0146] The above embodiments are only preferred specific implementations of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for resilient recovery of an energy interconnection distribution system under an ice disaster scenario, comprising: Step 1: Extract historical data features based on deep learning theory to generate photovoltaic output scenarios under ice disaster weather; Step 2: Establish the spatiotemporal state model of fixed and mobile energy storage and the timing model of the integrated energy system based on the dynamic and static characteristics of the power supply; Step 3: Considering the networking constraints of the emergency microgrid and the power consumption characteristics of the load, a multi-stage resilience recovery model of the energy interconnection distribution system is established with the goal of maximizing the continuous power supply time of the emergency load; the recovery stage in the function curve of the system during the disaster process is divided into three periods. In the first period, the remote control equipment is used to quickly identify and isolate the fault; In the second period, flexible resources are dispatched to achieve reasonable distribution of electricity in time and space, thus improving system resilience. In the third period, the system resilience is restored to normal levels by repairing faulty equipment; Connection status between mobile energy storage i and node j at time t Controllable power output Power supply load power To control the variables, with the goal of maximizing the continuous power supply time of the emergency load, the optimal comprehensive energy, fixed energy storage, mobile energy storage controllable distributed power output and load shedding strategy are given, that is, the multi-stage optimization recovery model, and its objective function is: Where T is the time period during which the fault occurs; B is the set of nodes; w i is the load weight determined according to the load type; The multi-stage optimization restoration model constraint conditions include load characteristic constraints and emergency microgrid networking constraints; the emergency microgrid networking constraints include: (1) Topological connection constraints Among them, α i,j,t is the switch state of line ij, 1 means closed; L is the set of lines; N B is the total number of nodes; N S is the total number of nodes; δ(i) is the set of child nodes of node i; γ(i) is the set of parent nodes of node i; B S is the set of virtual source nodes in the connected subgraph; F i,j,t is the virtual power flow on line ij; is the virtual power emitted by each source node; M is a very large number; (2) Minimum controllable power capacity ratio constraint Among them, μ is the proportion of controllable power supply; S i is the rated capacity of the i-th distributed generation in the microgrid; C j is the set of controllable distributed generation in the jth microgrid; U j is the set of uncontrollable distributed power sources in the jth microgrid; μ min is a constant, indicating the minimum proportion of controllable power capacity, and its range is [0-1]; (3) Power flow constraints in, is the active power of the power supply load at node i; is the active power output of the distributed generation at node i; P ij,t is the active power flowing from node i to node j; Q ij,t is the reactive power flowing from node i to node j; S max is the upper limit of line capacity; is the reactive power of the power supply load at node i; is the reactive power output of the distributed generation at node i; is the square term of the node voltage amplitude at node i; is the square term of the current amplitude on branch ij; R ij is the resistance of branch ij; X ij is the reactance of branch ij; Step 4: Linearize and solve the model using cone optimization and big-M method.

2. According to the method for recovering the resilience of an energy interconnection distribution system in an ice disaster scenario according to claim 1, it is characterized in that: In the step 1, a conditional generative adversarial network method is used to introduce conditional input items at the input ends of the generator and the discriminator, so that the generator can be effectively applied to scenarios with specific data requirements; The Wasserstein distance is used to characterize the gap between generated data and real data, and a gradient penalty term is introduced to replace the gradient truncation strategy of traditional WGAN.

3. The method for recovering the resilience of an energy interconnected power distribution system in an ice disaster scenario according to claim 1 or 2, characterized in that: The step 1 specifically includes: Step 1.1: Preprocess the historical photovoltaic output data and initialize the neural network parameters; Step 1.2: The predicted and actual photovoltaic output data under freezing weather conditions are used as inputs to the generator neural network and the discriminator neural network, respectively, to train the neural network; Step 1.3: Use the photovoltaic day-ahead output forecast data as the input of the generator neural network to generate photovoltaic output scenarios; Step 1.4: Use the k-medoids clustering algorithm to reduce the photovoltaic output scenarios, use the silhouette coefficient method to evaluate the clustering effect, and generate photovoltaic output scenarios under ice disaster weather.

4. According to a method for recovering the resilience of an energy interconnection distribution system in an ice disaster scenario according to claim 1, it is characterized in that: The spatiotemporal scheduling model of mobile energy storage is: Among them, ME is the collection of movable energy storage; is the transportation speed of mobile energy storage; D j,k is the shortest distance of the traffic road between two nodes; Transportation time for mobile energy storage; is the connection status between the mobile energy storage i and the node j; Before the mobile energy storage reaches node k, that is, the scheduling interval △ t is less than When , its connection status with node k; To consider the installation and configuration time of mobile energy storage; The control variable is the charging power at time t; is the controlled variable, the discharge power at time t; is a binary variable, indicating the charging state of the removable energy storage, 1 indicates charging; U is a binary variable, indicating the discharging state of the removable energy storage, 1 indicates discharging; The capacity of the mobile energy storage; Charging efficiency for mobile energy storage; is the discharge efficiency of mobile energy storage; is the upper limit of the mobile energy storage capacity; is the lower limit of the capacity of the mobile energy storage; formula (1) determines the shortest time for the mobile energy storage i to travel between nodes j and k; formula (2) indicates that when the scheduling interval is less than the sum of the transportation time and the configuration time, the connection state of the mobile energy storage at node k is 0; formula (3) indicates that the mobile energy storage i can connect to at most one node at the same time; formulas (4) and (5) give the upper and lower limits of the charging and discharging power of the mobile energy storage; formula (6) indicates the coupling relationship between the charging and discharging state of the mobile energy storage and the spatial state, ensuring that the mobile energy storage can be charged and discharged when in the connected state; formulas (7) and (8) are the state of charge constraints of the mobile energy storage; the nonlinear constraint shown in formula (6) is converted into the following formula using the linearization method: in, is the connection status between mobile energy storage i and node j in the scheduling period t-t+1, 1 indicates connection and 0 indicates disconnection; B is the set of nodes.

5. According to the method for recovering the resilience of an energy interconnection distribution system in an ice disaster scenario according to claim 4, it is characterized in that: Charging power of fixed location energy storage at time t With discharge power To control the variables, during the load recovery phase, the energy is charged and discharged in the corresponding microgrid as the dispatching instructions change. The charging and discharging power, charging and discharging state, and capacity need to meet the following constraints: Among them, E is the set of fixed energy storage; and are the charging power and discharging power of the fixed location energy storage at time t respectively; is a binary variable, indicating the charging state of the fixed energy storage, 1 means charging; is a binary variable, indicating the discharge state of the fixed energy storage, 1 indicates discharge; is the capacity of fixed energy storage; Charging efficiency for stationary energy storage; is the discharge efficiency of fixed energy storage; is the upper limit of fixed energy storage capacity; It is the lower limit of fixed energy storage capacity.

6. According to the method of claim 5, the method is characterized in that: The energy time series coupling model of the integrated energy system can be expressed as: Among them, η T Indicates transformer efficiency; Indicates the electricity conversion efficiency of the CHP unit; Represents the thermal conversion efficiency of the CHP unit; η HP Indicates the thermal efficiency of the natural gas heat pump; L e Indicates the electrical load output within the comprehensive energy source; L t Indicates the gas load output within the comprehensive energy source; E e Indicates the electrical energy input into the comprehensive energy; E g represents the natural gas energy input in the comprehensive energy; σ represents the natural gas distribution coefficient, σE g represents the amount of natural gas consumed by the CHP unit, (1-σ)E g Indicates the amount of natural gas consumed by the natural gas heat pump.

7. The method for resilient recovery of energy interconnected power distribution system in ice disaster scenario according to claim 1 is characterized by: The load characteristic constraints include: (1) Dual controllable power supply constraints in, is the number of controllable power sources that should be connected in microgrid k; (2) Minimum power supply duration constraint in, is the minimum continuous power supply time of load i; is the power supply status of load i, 1 means it is in the power supply state; (3) Maximum power interruption times constraint in is the maximum number of power interruptions for load i.

Images