Distribution network source site collaborative planning method and system considering V2G mode

Abstract

The present invention relates to the technical field of coordinated planning of distribution network source stations, and specifically discloses a method and system for coordinated planning of distribution network source stations taking into account the V2G mode. The method comprises: obtaining the existing topological structure of the distribution network, traffic network structure data, spatiotemporal distribution data of the number of electric vehicles, photovoltaic resource distribution data, and a set of historical daily load power curves; predicting the node load power set, the maximum dispatchable power set of the V2G charging and discharging station, and the maximum output power set of the PVG site under various typical daily scenarios in the future; establishing an upper-level planning model with the goal of minimizing the total economic cost of the system, and establishing a lower-level planning model with the goal of minimizing the annual comprehensive operating cost; and using a nested iterative algorithm to solve the two-level planning model and output a coordinated planning scheme for the distribution network source stations. The present invention improves the economy and reliability of the coordinated planning of the distribution network, PVG sites, and V2G charging and discharging stations, significantly improves resource utilization efficiency, and reduces the total economic cost.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network source site collaborative planning, and in particular to a distribution network source site collaborative planning method and system considering a V2G mode. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] V2G (Vehicle-to-Grid) is a technology that allows electric vehicles (EVs) to not only charge from the grid but also discharge back to the grid.

[0004] With the development of new power systems and the advancement of energy transformation, distribution networks are gradually evolving toward a high proportion of renewable energy access, a high degree of power electronics, and high flexibility and interactivity. The explosive growth of electric vehicles and the large-scale grid-connection of photovoltaic (PVG) power plants have brought new challenges to the planning and operation of distribution networks. On the one hand, the maturity of V2G (Vehicle-to-Grid) technology has transformed electric vehicles from a single load to a dual role of "adjustable load + mobile energy storage," changing the load structure and characteristics of the distribution network. On the other hand, the volatility and intermittency of PVG power output, combined with the spatiotemporal randomness of electric vehicle charging demand, exacerbate peak-to-valley and voltage fluctuations in the distribution network, complicating grid power flows.

[0005] Traditional planning methods typically plan PVG sites, distribution networks, and V2G charging stations independently. This involves optimizing PVG site selection and sizing, distribution network expansion, and V2G charging station layout separately, each optimizing its own objective function. While this approach simplifies computational complexity, it also dissociates the temporal and spatial complementarity between distributed PVG sites and V2G charging stations. It also ignores the potential for matching charging and discharging loads with PV output in the temporal dimension, as well as the potential for dynamic grid reconfiguration in the spatial dimension. Given the significant impact of EV charging and discharging behavior on distribution network voltage, power flow, and system stability, this can lead to wasted grid resources or improper system scheduling.

[0006] In recent years, research has attempted to collaboratively plan V2G charging and discharging stations with distributed generation (DGs) to improve grid planning and operational efficiency. However, most existing joint planning methods are based on a fixed distribution network. Few studies simultaneously address distribution network expansion planning, nor do they incorporate PVG station output, V2G charging and discharging behavior, and distribution network topology optimization into a unified model.

[0007] In addition, when jointly planning PVG stations, distribution networks, and V2G charging and discharging stations, it is necessary to establish a unified optimization model and consider a large number of decision variables, such as the site selection and charging and discharging power capacity of charging stations, the installation location and rated power of photovoltaic power sources, and the distribution network line upgrade plan and network reconstruction strategy. At the same time, it is also necessary to deal with the nonlinear and dynamic changes of the constraints. This leads to a combinatorial explosion in the problem dimension, and traditional optimization algorithms face the dual challenges of computational efficiency and solution accuracy. Summary of the Invention

[0008] To address the above issues, the present invention proposes a method and system for collaborative planning of distribution network sources and sites considering the V2G mode, establishes a collaborative planning model for the distribution network, photovoltaic power sources (PVG), and V2G charging and discharging station expansion, and improves the economy and reliability of the distribution network planning scheme.

[0009] In some embodiments, the following technical solutions are adopted:

[0010] A distribution network source-site collaborative planning method considering a V2G mode includes:

[0011] Obtain the existing topology of the distribution network, traffic network structure data, spatiotemporal distribution data of electric vehicles, photovoltaic resource distribution data, and historical daily load power curve sets;

[0012] Confirm the preliminary spatial location constraints of each facility in the source site collaborative planning; cluster the historical daily load power curves into typical daily load power curves, and then predict the node load power set, the maximum dispatchable power set of V2G charging and discharging stations, and the maximum output power set of PVG stations under various typical daily scenarios in the future;

[0013] The upper-level planning model is established with the goal of minimizing the total economic cost of the system, and the lower-level planning model is established with the goal of minimizing the annual comprehensive operating cost;

[0014] A nested iterative algorithm is used to solve the two-level planning model. The upper-level planning model determines the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. A preliminary planning scheme that meets load growth needs and minimizes investment costs is output, and the distribution network topology and equipment information are passed to the lower-level model. The lower-level planning model solves the annual optimal operating results, and simultaneously generates the over-limit branches, node marking sets, and penalty costs, which are fed back to the upper-level planning model.

[0015] Iterate the calculation until the fluctuation of total investment-operation cost is stable and the safety indicators meet the standards, and output the distribution network source site collaborative planning plan.

[0016] In other embodiments, the following technical solutions are adopted:

[0017] A distribution network source-site collaborative planning system considering the V2G mode includes:

[0018] The data acquisition module is used to obtain the existing topology of the distribution network, traffic network structure data, spatiotemporal distribution data of electric vehicles, photovoltaic resource distribution data, and historical daily load power curve sets;

[0019] The scenario generation module is used to confirm the preliminary spatial location constraints of each facility in the source site collaborative planning; cluster the historical daily load power curves into typical daily load power curves, and then predict the node load power set, the maximum dispatchable power set of V2G charging and discharging stations, and the maximum output power set of PVG stations under various typical daily scenarios in the future;

[0020] The model building module is used to establish the upper-level planning model with the goal of minimizing the total economic cost of the system, and to establish the lower-level planning model with the goal of minimizing the annual comprehensive operating cost;

[0021] The model optimization solution module uses a nested iterative algorithm to solve the two-level planning model. The upper-level planning model determines the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. It then outputs a preliminary planning solution that meets the load growth demand and minimizes the investment cost. The distribution network topology and equipment information are then passed to the lower-level model. The lower-level planning model solves the annual optimal operation results and simultaneously generates the over-limit branches, node marking sets, and penalty costs, which are fed back to the upper-level planning model.

[0022] Iterate the calculation until the fluctuation of total investment-operation cost is stable and the safety indicators meet the standards, and output the distribution network source site collaborative planning plan.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] (1) The present invention collects data from multiple systems (intelligent vehicle networking, geographic information system, energy management platform) and combines it with an improved K-means clustering algorithm and Monte Carlo algorithm to accurately simulate and predict future load changes and the dispatching capabilities of V2G charging and discharging stations and photovoltaic stations, providing accurate basic data for distribution network optimization planning and ensuring the rationality of the planning scheme;

[0025] Based on the V2G mode, a two-layer collaborative planning model is used to improve the economy and reliability of the collaborative planning of distribution networks, PVG stations and V2G charging and discharging stations, significantly improve resource utilization efficiency and reduce total economic costs.

[0026] (2) This invention constructs a two-layer collaborative planning model, which effectively optimizes resource allocation by separating the planning layer from the operation layer. The upper planning layer ensures the minimization of total economic costs, including annual construction investment costs and annual comprehensive operation costs; the lower operation layer calculates the optimal state of the distribution network in actual operation through multi-scenario optimization, reduces node branch over-limit and resource waste, and achieves a balance between the economy and reliability of the distribution network.

[0027] Adopting the strategy of "decision transmission-status feedback-scheme correction", the planning scheme is adjusted in real time through the feedback mechanism, and the limit-exceeding problems are discovered and resolved in a timely manner, making the planning and operation process more flexible, and making rapid corrections based on actual conditions to ensure the robustness and feasibility of the planning scheme.

[0028] (3) The present invention adopts an adaptive hybrid particle swarm algorithm to solve the upper-level planning model, flexibly adjusts the hybrid particle search strategy during the upper-level planning process, and dynamically adjusts the search range according to the severity of the off-limit branches and node parameters transmitted from the lower layer, thereby ensuring the efficiency of the algorithm and improving the convergence speed, and better solving complex decision-making problems.

[0029] A parallel particle swarm optimization algorithm is used to solve the underlying planning model. The combination of scenario-based parallel optimization and the particle swarm algorithm effectively improves computational efficiency. The master node in the parallel computing framework distributes the parallel solution tasks for multiple typical daily scenarios to different sub-nodes, significantly accelerating computational efficiency. Each scenario is independently optimized, ultimately achieving the optimal global running cost solution.

[0030] (4) The present invention is based on a V2G implementation method in a microgrid mode. The lower operating layer achieves efficient operation of the V2G charging and discharging station in the microgrid mode. Combining centralized and decentralized dispatching and control methods, the lower operating layer includes a centralized dispatching layer and a decentralized execution layer. The centralized dispatching layer is the intelligent vehicle network system, responsible for global economic optimization, while the decentralized execution layer is the V2G charging and discharging station, responsible for flexible adjustment according to the charging and discharging needs of electric vehicles, thereby improving the operating efficiency and safety of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a diagram of the architecture of a two-layer collaborative planning model in an embodiment of the present invention;

[0032] Figure 2 Schematic diagram of the process of the distribution network source site collaborative planning method considering the V2G mode in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those skilled in the art to which the present application belongs.

[0034] Example 1

[0035] In one or more embodiments, a distribution network source site collaborative planning method considering the V2G mode is disclosed, combined with Figure 2 , specifically including the following process:

[0036] S101: Obtain the existing topological structure of the distribution network, traffic network structure data, spatiotemporal distribution data of the number of electric vehicles, photovoltaic resource distribution data, and historical daily load power curve sets.

[0037] Specifically, this embodiment collects data from the smart Internet of Vehicles system, geographic information system and energy management platform.

[0038] The existing topology of the distribution network includes the geographic coordinates and capacity of substation nodes, detailed equipment parameters such as line impedance parameters and main transformer capacity. A preliminary determination of the set of distribution network lines and substation nodes that can be newly built or upgraded is also made.

[0039] The traffic network structure data includes: coupling the actual geographical location in the traffic network with the distribution network topology, and preliminarily determining the set of candidate nodes where electric vehicle V2G charging and discharging stations can be installed.

[0040] The spatiotemporal distribution data of electric vehicles includes charging period preferences, driving trajectories, and battery capacity characteristics. The Monte Carlo algorithm is used to generate the maximum dispatchable power set for V2G charging and discharging under various typical daily scenarios.

[0041] Photovoltaic resource distribution data includes: PVG maximum output power set under typical daily scenarios within the distribution network area generated by Monte Carlo algorithm simulation.

[0042] The historical daily load power curve set is the distribution network node collected i Historical daily load power curve set .

[0043] S102: Confirm the preliminary spatial location constraints of each facility in the source site collaborative planning; cluster the historical daily load power curves into typical daily load power curves, and then predict the node load power set, the maximum dispatchable power set of the V2G charging and discharging station, and the maximum output power set of the PVG site under various typical daily scenarios in the future.

[0044] In this embodiment, the mapping relationship between the distribution network topology and the transportation network is analyzed. In combination with the urban planning scheme, the preliminary geographical locations of each facility in the source site collaborative planning and the mapped distribution network node location constraints are determined to obtain a series of candidate sets, including: I Line,NU A collection of candidate branches for new or upgraded lines. ij It is the system area node; Oh TR 、 Oh EVS 、 Oh PVG They are the candidate sets of new substations, electric vehicle V2G charging and discharging stations, and PVG station nodes.

[0045] In this embodiment, an improved K-means clustering algorithm is used to cluster historical daily load curves into a set of typical daily load power curves. A recursive method is then used to derive the node load power sets for various future typical daily scenarios. A Monte Carlo algorithm and an improved K-means clustering algorithm are then used to generate the maximum dispatchable power sets for V2G charging and discharging stations and the maximum output power sets for PVG stations for various future typical daily scenarios.

[0046] In the lower operation layer of the model, different typical daily scenarios correspond to different power flow calculation results, which can more accurately reflect the actual operation of the power grid and increase the reliability of the model.

[0047] In this embodiment, the distribution network node is obtained i Daily load curve collection The interpolation method is used to correct the missing load data of each node. The typical daily load curve of each node is clustered separately by the improved K-Means clustering algorithm. Finally, the recursive algorithm is used to obtain the daily load curve and its probability under the future typical day scenario, which are as follows:

[0048] S1021: In the initial stage, a curve is randomly selected from the original sample as the initial cluster center curve. Then, two threshold parameters are set. T 1 and T 2, T 1< T 2. Distance less than T The curve with a value of 1 is the curve within the cluster, and the distance is T 1 to T The curves within the range of 2 are suspicious curves, and the distance is T The curves outside the second circle are cluster-outside curves. New cluster center curves are selected from the data marked as cluster-outside curves. Repeat the process until all cluster-outside curves are correctly classified. After removing smaller clusters, the number of clusters becomes the initial K value, and K initial cluster center curves are selected.

[0049] In this embodiment, by controlling the threshold parameter T 1 and T The size of 2 can control the granularity of the clustering clusters, so that the number of clusters is automatically determined by the data distribution without manual preset; it alleviates the sensitivity of the K-Means clustering algorithm to the initial value and can avoid the clustering results being affected by the initial K value and falling into the local optimum problem.

[0050] S1023: Once the number of clusters K and the initial cluster center curves are known, the K-means clustering method is used: the shortest Euclidean distance between each object and each initial cluster center curve is calculated, and the center curves of all objects in each cluster are updated to obtain K corresponding cluster center curves. This iteration is repeated until the cluster center curves no longer change or the maximum number of iterations is reached.

[0051] The distribution network nodes are clustered by the improved K-means clustering method. i of s Typical day scenario load curve and its probability of occurrence .

[0052] S1024: Typical daily load recursion model. Assume that the overall load growth rate in the future planning year is relative to the historical year. and scenario occurrence probability correction factor It is known that the formula of the simple daily load recursion model is:

[0053] ;

[0054] ;

[0055] Where, 、 Distribution network nodes i Typical day scene s download curves and their probabilities; and Respectively for scenes s Next node i The predicted values ​​of the future typical daily load curve set and its probability set.

[0056] Then normalize the probability to ensure that the probability of occurrence satisfies the following formula:

[0057] .

[0058] In this embodiment, the Monte Carlo algorithm is used to generate the maximum dispatchable power set of V2G charging and discharging under various typical daily scenarios in the future, combined with the future growth of the number of electric vehicles at each node in the distribution network area. Specifically, the distribution network nodei The number of electric vehicles in the planning year is expressed as The duration of each vehicle's charging and the time it starts are usually fixed in a certain period of time and start at a random time of the day. Assuming that the daily charging start time follows the time range ( 、 ) is uniformly distributed, 、 are the earliest and latest time for electric vehicle charging to start. Assume that the charging and discharging power of each vehicle is 、 , are fixed values. Assume that the charging and discharging time are fixed values, expressed as 、 Monte Carlo simulation is used to obtain the charging and discharging power of each node, and the corresponding charging and discharging power is superimposed on the nodes of the planned charging and discharging stations according to the charging service range.

[0059] The Monte Carlo algorithm generates the maximum dispatchable power set for V2G charging and discharging under various typical daily scenarios in the future as follows:

[0060] (1) Obtaining distribution network nodes i Number of electric vehicles in the planning year .

[0061] (2) For nodes i Each electric vehicle is subject to 、 ) Generate the charging start time based on the uniform distribution rule and obtain the charging power and charging time , generating the charging power of a single electric vehicle. i For all electric vehicles within the service range, the charging power of each vehicle is added up to obtain the total charging power that can be dispatched at the node in a day. By traversing all nodes in the distribution network, the total charging power that can be dispatched at each node in a day can be obtained.

[0062] (3) For nodes i After each electric vehicle is charged, it starts to generate its discharge start time according to the uniform distribution rule at the time other than charging, and obtains the discharge power and discharge duration , generating the discharge power of a single electric vehicle. For the same node i For all electric vehicles within the service range, the discharge power of each vehicle is added up to obtain the total discharge power that can be dispatched at the node in a day. By traversing all nodes in the distribution network, the total discharge power that can be dispatched at each node in a day can be obtained.

[0063] (4) Repetition MNext steps (2) and (3), we get M charging and discharging scenarios.

[0064] (5) Using the improved K-means clustering algorithm mentioned above, K The value is set to S ,Will M The charging and discharging power of each scene is clustered as follows: S Typical daily nodes i The maximum dispatchable power set of electric vehicle charging and discharging hourly is expressed as 、 .

[0065] In this embodiment, in order to further explore the PVG output characteristics of the region, the Monte Carlo algorithm is used to generate the maximum output power set of PVG stations under various typical daily scenarios in the future, specifically:

[0066] Assuming that factors such as temperature, module efficiency, and PV module installation angle are taken into account, only the impact of solar irradiance on PVG output is considered. t Solar irradiance The specific steps for simulating solar irradiance and photovoltaic power generation output for one day (24 hours) are as follows:

[0067] (1) The solar irradiance under different scenarios conforms to the normal distribution. The mean and variance of the normal distribution parameters are estimated by historical data or meteorological models. A typical day is divided into 24 hours ( t =1, 2,…,24), fitting the historical data to get the time t The expected irradiance and the standard deviation of irradiance parameter.

[0068] (2) Generate irradiance using Monte Carlo simulation, using the Monte Carlo method at each hour t Generate a random irradiance value .

[0069] ;

[0070] Where, is a random variable generated from a standard normal distribution.

[0071] (3) Repeat step (2) M times, generate M different solar irradiance scenarios.

[0072] (4) Calculate the PVG station output power for each simulated periodt PVG station output power is obtained by irradiance calculation , the formula is:

[0073] ;

[0074] Where, is the efficiency of the photovoltaic module; is the rated power of the PVG station; is the standard irradiance; is the randomly generated irradiance at hour t.

[0075] (5) Using the improved K-means clustering algorithm mentioned above, K The value is set to S ,Will M The PVG station output power of each scenario is clustered as follows: S Typical daily nodes i The maximum dispatchable power set of PVG hourly is expressed as .

[0076] In this embodiment, the maximum dispatchable power of V2G charging and discharging and the maximum dispatchable output power of PVG stations under various typical daily scenarios in the future obtained by the Monte Carlo algorithm are only a description of the generation method; the specific V2G charging and discharging station curves and PVG station output curves need to be regenerated under different scenarios based on the site selection location and capacity specified by the upper-level planning results, and then the optimal operation strategy of the operation layer is solved.

[0077] S103: Establish an upper-level planning model with the goal of minimizing the total economic cost of the system.

[0078] Combine Figure 1 In the two-layer collaborative planning model shown, the objective function of the upper-layer planning model is to minimize the total economic cost, and the formula is:

[0079] min F = C inv + C ope ;

[0080] Where, annual construction investment cost Including the annualized sum of investment costs for new line construction / upgrades, new substation construction, PVG stations, and V2G charging and discharging stations; annual operating costs C ope The calculations will be fed back to the upper planning layer after being calculated in the lower operational layer.

[0081] Among them, the annual construction investment cost C invThe annual value is calculated by using the annual cost conversion coefficient of the entire life cycle investment, including the new line , line upgrade , new construction of the area , new V2G charging and discharging stations and PVG station newly built The annualized investment cost is calculated as follows:

[0082] ;

[0083] Where, A collection of candidate branches for new or upgraded lines. ij It is the system area node; Oh TR 、 Oh EVS 、 Oh PVG They are the candidate sets of nodes for new substations, electric vehicle V2G charging and discharging stations, and PVG station nodes; K Line,N 、 K Line,U 、 K TR 、 K EVS 、 K PVG They are a collection of alternative types for new line construction and upgrade, new substation construction, new V2G power station construction, and new PVG station construction. k It is an optional type; R Line 、 R TR 、 R EVS 、 R PVG They are the annual cost conversion coefficients for the full life cycle investment of lines, substations, V2G power stations and PVG stations. r is the discount rate.

[0084] 、 、 、 、 are 0-1 decision variables for new construction of lines, upgraded lines, new construction of substations, new construction of V2G power stations, and new construction of PVG stations, where 0 means no new construction and 1 means new construction; 、 are the lengths of new and upgraded lines, 、 、 Node i New substations, new V2G power stations and new PVG stationsk Type equipment rated capacity, both are continuous decision variables greater than 0; 、 、 They are the new substations, new V2G power stations and new PVG stations. k Type of equipment i The number of nodes is an integer variable greater than 0; these are the decision variables of the upper planning layer.

[0085] 、 、 、 、 They are k Unit investment and operation and maintenance cost coefficients for new construction and upgrades of lines, new substations, new V2G power stations and new PVG stations of different types.

[0086] The constraints of the upper planning layer include constraints on the construction and upgrading of line types (constraints on the construction and upgrading of line types that take into account the growth of conventional loads and the feedback of new current and voltage over-limits); constraints on the construction type and capacity of substations (constraints on the construction and upgrading of substation types that take into account the growth of conventional loads and the feedback of load over-limits); constraints on the construction of V2G charging and discharging stations (constraints on the site selection, type, quantity and service capacity of new V2G charging and discharging stations); and constraints on the construction of PVG stations (constraints on the site selection and type, installed capacity and quantity of PVG stations).

[0087] The above constraints can improve the reliability and flexibility of the power grid, reduce the risk of load overload, and optimize resource allocation; ensure the rational allocation of substation capacity to prevent equipment damage or unstable power supply caused by excessive load; optimize the construction and operation of V2G charging and discharging stations, improve the load regulation capability of the power grid and the efficiency of battery utilization, and reduce overinvestment and resource waste; improve the energy utilization efficiency of photovoltaic power stations, while reasonably controlling the installed capacity and number to avoid excessive or insufficient investment.

[0088] The new construction and upgrade constraints for conventional load growth require that each branch can only choose to build or upgrade one type of line at most to avoid duplicate investment. The specific formula is:

[0089] ;

[0090] When the current or voltage exceeds the limit feedback in the branch or terminal node, it is necessary to modify the line type construction and upgrade constraints, and aggregate the voltage and current exceeding the limit feedback branches in the lower layer. Related branches are updated synchronously to The middle and upper planning layers focus on the planning or upgrade planning of the line type of the branch.

[0091] Considering the type and capacity constraints of newly built substations for conventional load growth, the substation capacity must cover the base load and future new load growth in the substation (including base load growth and V2G charging and discharging requirements), while also meeting the main transformer capacity constraints. The specific formula is:

[0092] ;

[0093] ;

[0094] Where, Oh Total are the sets of all nodes in the distribution network, and the number is N Total ; The basic capacity of the area. If it is a new area, the basic capacity is 0; is the safety margin coefficient of the substation area, which is usually set to [0.6, 0.8] according to the importance of the node; 、 They are the current basic load of the substation and the future increase in load of the substation; E SS is the capacity of the main transformer of the distribution network.

[0095] When a node has a load over-limit feedback, it is necessary to modify the new type and capacity constraints of the substation, and collect the node load over-limit feedback from the lower layer. Oh Node,over Related nodes are updated synchronously to Oh TR In the set, the focus is on the planning of the node area type and capacity. The specific formula is:

[0096] ;

[0097] Where, Node for lower-level feedback i Power exceeds the limit. Power shortage.

[0098] New V2G charging and discharging stations need to consider the site selection constraints, type constraints, quantity constraints, and service capacity constraints of new V2G charging and discharging stations. Specifically:

[0099] To avoid overlapping service areas, improve planning economy and enhance user experience, charging stations must maintain a minimum geographical distance from each other. d min Get a charging station i and charging stations j The actual geographical coordinates between and calculate the geographical distance d ij , the V2G charging and discharging station site selection constraint formula is obtained as:

[0100] ;

[0101] Due to limited urban land resources, it is necessary to avoid duplicate construction and optimize space utilization. Only one type of V2G charging and discharging station is allowed to be built at the same node, and the number shall not exceed one. The formula for the type and number of new V2G charging and discharging stations is as follows:

[0102] ;

[0103] The capacity of the charging station must match the total charging demand of electric vehicles within the service area. i Local demand weight coefficient It is determined by the population density, traffic flow, and commercial activities at the node and is a constant. i The attraction weight for all electric vehicle charging demands within the distribution network supply area It is the node i Local demand weight and nodes i and j The actual geographical distance between them is jointly determined.

[0104] ;

[0105] ;

[0106] Where, Nodes within the distribution network power supply area i EV charging demand at charging stations; For nodes i The attraction weight of the charging station to all electric vehicle charging demands in the power supply area; The total attraction weight of all charging stations at all nodes to all electric vehicles charging demands in the region; For the l The attraction weight of each node to all electric vehicle charging demands in the distribution network power supply area; The intelligent vehicle networking system first collects basic data, and then uses clustering algorithm and extrapolation algorithm to obtain the total charging demand of electric vehicles in the power supply area of ​​the distribution network in the future.

[0107] Constraints for new PVG stations include site selection and type constraints, and installed capacity constraints, specifically:

[0108] Only one type of PVG station is allowed to be installed at a node, simplifying operations and management and reducing equipment compatibility issues. PV capacity must match the capacity of the substation to prevent over-output and curtailment.

[0109] The constraint formula for PVG station site selection and installation type is:

[0110] ;

[0111] The constraint formula for PVG station installed capacity is:

[0112] .

[0113] S104: Establish a lower-level planning model with the goal of minimizing the annual comprehensive operating cost.

[0114] In this embodiment, the annual comprehensive operating cost objective function of the lower-level planning model is:

[0115] ;

[0116] Where, Oh S A collection of typical daily scenes that may occur in a year; is the natural daily number of the year, =365; A typical day scene s The probability of occurrence ; 、 、 、 、 They represent the power purchase cost of the distribution network, network loss cost, penalty cost for curtailed solar power, V2G battery degradation cost, and penalty cost for safety over-limit. is the adaptive coefficient of the penalty factor.

[0117] The annual comprehensive operating cost of the lower operating layer includes the distribution network power purchase cost, network loss cost, curtailment penalty cost, V2G battery degradation cost, and safety limit penalty cost, specifically:

[0118] Figure 1 In the two-layer collaborative planning model shown, the grid power purchase cost formula of the lower operation layer model is:

[0119] ;

[0120] Where, T is the number of hours in a typical day, T =24; for t The main grid time-of-use electricity price at the moment; A typical day scene s middle t The active power transmitted by the main transformer of the distribution network at the moment.

[0121] Quantify the economic impact of line losses and reduce line impedance and voltage drop. The formula is:

[0122] ;

[0123] Where, I Line is the set of all lines in the distribution network, ij Indicates that from the first node i To the end node j branch road; Unit network loss cost; For the scene s Next time t branch road ij Current; For branch ij resistance.

[0124] Abandoned solar power refers to the amount of power generated by the system that is forced to be abandoned due to technical or economic reasons. The formula is:

[0125] ;

[0126] Where, is the unit light abandonment penalty coefficient; For the scene s Next, node i At the moment t Theoretical maximum photovoltaic active output power; Indicates the scene s Next, node i At the moment t The actual scheduling allows the output of photovoltaic active power.

[0127] Quantify the impact of charging and discharging on battery life and the cost of battery degradation in V2G mode The formula is:

[0128] ;

[0129] Where, is the unit charge and discharge degradation cost and coefficient of the battery; 、 Respectively for scenes s Next node i V2G charging and discharging station time t Discharge and charging power.

[0130] When the distribution network operates under different typical daily scenarios, if voltage, current or load exceeds the limit, the safety limit penalty should be incorporated into the operating cost. The safety limit penalty cost formula is:

[0131] ;

[0132] Where, cVol 、 c Cur 、 c Load are the unit penalty coefficients for voltage, current and load over-limit respectively; D.U. i,t,s 、 D I ij,t,s 、 ΔS i,t,s They are respectively voltage, current and load exceeding the limit.

[0133] In this embodiment, in a variety of typical day scenes s In the above example, due to the different light radiation and the response degree of electric vehicles, the upper limit of the decision variable is different. s Next time t node i The PVG output power allowed at and its power factor angle , and scenes s Next node i V2G charging and discharging station time t Discharge and charging power 、 will also be different (these parameters are the decision variables of the lower operating layer); resulting in inconsistent minimum operating costs for each scenario. By counting the probability of each scenario occurring and taking a weighted sum, the minimum annual comprehensive operating cost can be obtained.

[0134] In this embodiment, the constraints of the lower operation layer include distribution network flow and safety constraints (DistFlow branch flow constraints, current and voltage over-limit and equipment load overload safety constraints); V2G charging and discharging station operation constraints (maximum dispatchable real-time charging and discharging power constraints); PVG station operation constraints (PVG station output constraints, reactive power output constraints).

[0135] These constraints effectively mitigate potential safety hazards in power system operation, such as line overloads and voltage overruns, improving the reliability and stability of the distribution network and reducing the risk of power outages or equipment damage. They maximize the charging and discharging efficiency of V2G sites while enhancing the grid's regulatory capabilities and balancing supply and demand. They effectively improve the stability of photovoltaic power generation and its support for the grid, avoiding grid voltage instability caused by reactive power imbalances and improving the overall operational efficiency of the power system.

[0136] The distribution network flow constraint is the DistFlow branch flow constraint, specifically:

[0137] ;

[0138] Where, Oh H(j) Indicates j The set of branch headend nodes that are end nodes; Oh B(j) Indicates j The set of branch end nodes of the head end node; P ij,t,s 、 Q ij,t,s 、 P jl,t,s 、 Q jl,t,s Respectively for scenes s Down t Time branch ij and branch roads jl Active and reactive power; P i,t,s 、 Q i,t,s Respectively for scenes s Down t Node of moment i The injected active and reactive power; U i,t,s 、 U j,t,s Respectively for scenes s Down t Time Node i and j voltage; 、 They are photovoltaic power source, basic and incremental load reactive power; For branch ij reactance. 、 Respectively for scenes s Down t Time Node i Base and growth load active power.

[0139] The lower-level operational safety constraints of the distribution network include voltage and current exceeding limits or equipment overload. These safety constraints have been modified to soft constraints, which allow violations but incur penalties. This requires continuous iteration and replanning of upper-level equipment types and capacities. The soft safety constraints for current and voltage exceeding limits or equipment overload are as follows:

[0140] ;

[0141] If the above safety constraints occur, the relevant nodes or branches are immediately marked and summarized into the voltage and current limit-exceeding branch set. I Line,over and node load limit set Oh Node,over , the specific formula is:

[0142] ;

[0143] Where, 、 They are marked voltage and current exceeding limit branches and load exceeding limit nodes respectively; 、 、 They are respectively voltage, current and load exceeding the limit.

[0144] The operating constraints of V2G charging and discharging stations include the maximum dispatchable real-time charging and discharging power constraints, specifically:

[0145] The Monte Carlo algorithm and the improved K-means clustering algorithm are used to generate the maximum schedulable real-time charging and discharging power of V2G charging and discharging stations under various typical daily scenarios in the future. The charging and discharging power actually called by the smart car networking system should be less than or equal to the maximum schedulable real-time charging and discharging power. The specific formula is:

[0146] ;

[0147] Where, 、 Respectively for scenes s Down t Time Node i The maximum dispatchable charging and discharging power of the V2G charging and discharging station. 、 Respectively for scenes s Down t Time Node i The actual charging and discharging power used by the V2G charging and discharging station.

[0148] PVG station operation constraints include PVG station output constraints and reactive power output constraints, specifically:

[0149] The Monte Carlo algorithm and the improved K-means clustering algorithm are used to generate the maximum dispatchable photovoltaic output power of the PVG station under various typical daily scenarios in the future. Actual dispatched photovoltaic output power Not exceeding the maximum dispatchable photovoltaic output power , the PVG station output constraint formula is:

[0150] ;

[0151] Where, β The abandoned light rate must be lower than the 5% threshold.

[0152] When the system needs to abandon light due to safety or economic reasons, is limited to less than , allowing the use of the remaining capacity of the inverter to provide reactive power support. The reactive power output constraint formula of the PVG station is:

[0153] ;

[0154] ;

[0155] Where, is the power factor angle under the limit state of the PVG station, take ; For the scene s Next node i The power factor angle of the PVG station during actual operation.

[0156] S105: A nested iterative algorithm is used to solve the two-level planning model. The upper-level planning model determines the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. A preliminary planning scheme that meets the load growth demand and has the lowest investment cost is output, and the distribution network topology and equipment information are passed to the lower-level model. The lower-level planning model solves the annual optimal operation results, and simultaneously generates the over-limit branches, node marking sets, and penalty costs, which are fed back to the upper-level planning model.

[0157] Iterate the calculation until the fluctuation of total investment-operation cost is stable and the safety indicators meet the standards, and output the distribution network source site collaborative planning plan.

[0158] In this embodiment, an adaptive hybrid particle swarm algorithm is used to solve the upper-level planning model to decide the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. A preliminary planning scheme that meets the load growth demand and has the lowest investment cost is output, and the distribution network topology and equipment information are passed to the lower-level model.

[0159] Adaptive hybrid particle swarm optimization algorithm Figure 2 As shown, the specific steps and processes are:

[0160] S1. Algorithm initialization. This includes setting algorithm parameters, initializing the mixed particle swarm, and preliminarily evaluating fitness.

[0161] Set the algorithm parameters. Set the mixed particle swarm size (population size) as ; The maximum number of iterations is expressed as ; The current iteration count is represented by g , g =1,2,…, ; The particle index is expressed as p , p =1,2,…, ; rrand 、 r 1. r 2 Random numbers drawn from a uniform distribution in the interval [0,1], used for processes such as the mixed particle swarm velocity update; inertia weight oh (g) , in g Used in generations to balance global search and local convergence; local search scaling factor g d , when the node / branch corresponding to a certain dimension exceeds the limit seriously, the factor will be reduced to reduce the moving step of the particle in that direction; the target convergence threshold e , if the rate of change of the global optimum within several generations is less than e , the iteration can be ended early; For the g Daizhongdi p The position vector of the hybrid particle contains the values ​​of all decision variables of the upper model; For the g Daizhongdi p The velocity vector of the hybrid particle is used to update the position during iteration; Represents the upper particles p Record of the best position vector in history; Indicates the optimal particle position in the history of the entire population in the upper layer; Represents the optimal value of the upper layer global fitness.

[0162] Initialize the hybrid particle swarm. Set the 0-1 location variables, integer variables, and continuous variables included in the plan. 0-1 location variables: new line flag , Line upgrade sign , installation node sign of the substation , V2G charging and discharging station site selection signs PVG station site selection sign Integer variable: number of installation areas 、Number of V2G charging and discharging stations 、Number of PVG stations Continuous variable: Length of new and upgraded lines 、 , installation capacity of the substation , V2G charging and discharging station capacity PVG station capacity The encoding design will be carried out for various decision variable particles of the upper planning layer. In order for each hybrid particle to fully express an upper planning scheme, it is necessary to write all decision variables in the same hybrid particle vector, including 0-1 variables, integer variables and continuous variables. Initialize the velocity vector , which can usually be 0; mixed particle index p=1,2,…, , randomly generate initial particles in the feasible region , expressed as:

[0163] ;

[0164] Where, d is the hybrid particle vector dimension variable, hybrid particle The total dimension of the vector is D up , d =1,2,…, ; For the p The first mixed particle d The initial vector of dimensions; 、 、 、 They represent the initialization mixed particle position sets of newly built and upgraded lines, newly built substations, newly built V2G charging and discharging stations, and newly built PVG stations respectively; the superscript (0) of other parameters represents the initialization mixed particle positions of the corresponding parameters; || represents vector splicing; initialization mixed particles Temporarily the optimal position of the particle .

[0165] Calculate the initial fitness. Based on the initial mixed particle solution Calculated annualized investment cost C inv ( ); Set initial operating costs C ope ( ) is a large constant; the initial fitness of each mixed particle is obtained ; Count the mixed particles corresponding to the global optimal fitness and set it as the optimal mixed particle position , the initial optimal fitness is .

[0166] S2. Iterative solution algorithm. The iterative solution includes three steps: dynamically adjusting the inertia weight, updating the velocity and position of each particle, and calling the underlying model and calculating the fitness.

[0167] Dynamically adjust the inertia weight. Use linear decreasing strategy to dynamically adjust the inertia weight. , larger in the early stage oh Enhanced global search, smaller in the later stage oh Improve local convergence accuracy, the formula is:

[0168] ;

[0169] Where, 、 are the maximum and minimum values ​​within the inertia weight setting range respectively; the maximum number of iterations is expressed as ; is the number of iterations.

[0170] Update the velocity and position of each mixed particle. p =1,2,…, , respectively update the continuous, discrete and integer variables and their adaptive local search scaling factors. For continuous variables, each dimension in the hybrid particle d Speed ​​update and its location movement updates , the formula is:

[0171] ;

[0172] ;

[0173] Where, is the adaptive local search scaling factor, which indicates the adaptive adjustment of the limit violation degree (the more serious the limit violation, the larger the value, the more >1, prompting particles to make larger corrections in the out-of-limit area); if the out-of-limit problem is minor or has been alleviated, it can be appropriately reduced ,Pick <1, the search range is more refined; if there is no out-of-limit node or branch return in the running layer, then =1, perform normal update. c 1. c 2 is the acceleration factor, generally ranging from 1 to 2; r 1. r 2 is a random number drawn from a uniform distribution in the interval [0,1]; 、 For each dimension of the mixed particle d speed and position; For the p Particle No. d The optimal parameters of each dimension; is the particle population d The optimal parameters for each dimension.

[0174] Integer and 0-1 variables. For integer variables, the nearest integer strategy can be used to correct them to integers. For 0-1 variables, the speed can be mapped to the flip probability. , the formula is:

[0175] ;

[0176] ;

[0177] Where, The changing rules of are consistent with those of continuous variables.

[0178] Call the lower model and calculate the fitness. Pass it to the lower operation layer to perform multi-scenario flow and operation scheduling calculations, and return the optimal operation results under the planning scheme C ope ( ), fitness function The updated calculation formula is:

[0179] = C inv ( ) + C ope ( );

[0180] S3, update the individual best and global best. If it is better than the historical optimal of the hybrid particle itself, the optimal individual hybrid particle is refreshed, which is expressed as:

[0181] = , = ;

[0182] If the fitness of the mixed particle is better than the current global optimum, the optimal mixed particle position in the history of the entire population will be refreshed. and the global optimal fitness value .

[0183] S4. Termination judgment

[0184] If the maximum number of iterations has been reached Or the global optimum improves within a number of iterations to less than the target convergence threshold e up Then stop the iteration.

[0185] V2G control in microgrid mode adopts a "centralized-decentralized combination" scheduling method, and realizes orderly scheduling and control of the entire distribution network through the centralized scheduling layer.

[0186] In this embodiment, V2G control in a microgrid model involves controlling the orderly charging and discharging of electric vehicles as "dispatchable energy storage units" within the microgrid, enabling them to form a local power system alongside other resources such as photovoltaics, energy storage, and the power grid. In microgrid mode, electric vehicles can support various functions within the microgrid, including voltage control, supply and demand balancing, and peak load shifting, through their charging and discharging behaviors. The core concept is to view V2G not simply as "two-way interaction between vehicles and the larger power grid," but to further integrate it into the overall scheduling and operation of the microgrid, achieving a combination of micro-local autonomy and global coordination.

[0187] In this embodiment, the orderly scheduling and control of the entire distribution network is achieved through the centralized scheduling layer and the decentralized execution layer. The centralized scheduling layer and the decentralized execution layer are the main objects in the lower-level operation model, namely the smart Internet of Vehicles system and each V2G charging and discharging station; specifically:

[0188] Centralized Dispatch Layer: This layer makes unified dispatch decisions based on the overall economic and safety objectives of the microgrid. By collecting the maximum dispatchable data from all V2G charging and discharging stations, the centralized dispatch layer performs unified optimized dispatch, determines the charging and discharging requirements of electric vehicles at distribution network nodes in different scenarios, and sends charging and discharging instructions to the decentralized execution layer.

[0189] Decentralized Execution Layer: The microgrid model means that each charging and discharging station or node can make a certain degree of local decision-making autonomy. If the current node voltage is low and the load is high, V2G can prioritize discharge to support the local voltage. Conversely, if the load is low or the electricity price is low, vehicles can be charged or surplus photovoltaic power can be used for G2V. Monte Carlo simulation methods are used to generate the maximum dispatchable charge and discharge power of each V2G charging and discharging station under different scenarios. The decentralized execution layer summarizes the data and uploads it to the centralized scheduling layer.

[0190] Taking into account a variety of typical daily scenarios, the lower operation layer uses a parallel particle swarm algorithm to solve the annual optimal operation results to calculate the operation costs such as network loss, curtailment, and battery degradation, and simultaneously detect safety hazards such as voltage exceeding the limit and line overload, and generate the exceeding limit branch, node marking set and penalty cost to feed back to the upper layer.

[0191] In this embodiment, the parallel particle swarm optimization algorithm specifically includes two parts: scene parallel optimization and particle swarm optimization. Figure 2 shown.

[0192] The Spark parallel computing framework is used for scene parallel optimization. The computing master node will s Typical daily scenarios are assigned to multiple computing nodes, and the computing slave nodes independently run the particle swarm algorithm sub-solver. The main tasks of the computing master node include assigning different typical daily scenarios to slave nodes, monitoring the computing progress of the slave nodes, and calculating the minimum annual operating cost under the planning scheme based on the probability of the scenario occurring. Cope The calculation slave node is responsible for executing specific calculation tasks and running the particle swarm algorithm to optimize the calculation for each typical daily scenario. Each calculation slave node independently processes the scenario assigned to it and calculates the optimal operating cost under the scenario. C ope , s In addition, for each scenario, each computing slave node checks the grid status under the optimal operating state. If there are any out-of-limit problems (such as current overload, node voltage too low, substation load rate too high, etc.), these out-of-limit branches and nodes need to be recorded.

[0193] Particle Swarm Optimization: In the optimization of the underlying operation, it is generated based on the improved K-means clustering method and Monte Carlo simulation s Based on the typical daily scenarios of the distribution network and their probabilities, as well as the distribution network structure and equipment configuration parameters determined by the upper planning layer, a parallelized particle swarm solver is constructed. The continuous decision variables are encoded into multidimensional particle position vectors in time series, and the optimal particle position and its fitness are iteratively searched to solve the minimum operating cost under each scenario.

[0194] The specific process of the parallel particle swarm algorithm is as follows:

[0195] S1. Input data processing. The input data is the upper-level planning parameters and the typical daily scenario set of the corresponding nodes. The upper-level planning parameters include line topology, substation capacity, V2G charging and discharging station and PVG station site selection and equipment capacity upper limit planning parameters. The typical daily scenario set of the corresponding node includes the typical daily load curve set corresponding to the planning parameters, the maximum dispatchable power set of electric vehicle charging and discharging, and the maximum dispatchable power set of photovoltaic output. S Typical day scene s =1,2,…, S and its probability π s , in the calculation slave node, for each scenario s Start a particle swarm algorithm solver to search for the optimal operating cost in the scenario. After each parallel particle swarm completes the optimization, it uploads the results (including the optimal operating cost and over-limit conditions) to the master node for aggregation.

[0196] S2, particle encoding and initialization. s The particle swarm algorithm under the following conditions is used as an example to solve the operation layer model. The decision variables of the lower operation layer are the actual charging and discharging power, the actual output power of the PVG station and its power factor angle. s Under this condition, the decision variables are still 2-dimensional variables, which are not suitable for particle swarm solution. Therefore, the particle swarm initial particle position encoding is performed one by one on the stations obtained by the upper planning. , expressed as:

[0197] ;

[0198] In the formula, the superscript (0) of the variables represents the initialization position of the hybrid particle of the corresponding parameters; d is the dimension variable of the particle vector, and the total dimension of the particle vector is D down ,Right now d =1,2,…, ; For the p Particle No. d The initial vector of dimensions; N EVS Indicates the number of V2G charging and discharging stations obtained from the upper-level planning; N PVG It represents the number of PVG stations obtained from the upper-level planning; 、 Respectively N EVS EVS stations in the scene s Down t Charging and discharging power at each moment; 、 Respectively N PVG PVG stations in the scene s Down t Active power and power factor angle at the moment.

[0199] Assume the number of particles is ; The maximum number of iterations is ; The particle index is p =1,2,…, Individual Optimum and global optimal ; Every dimension in the particle d of g +1 speed update and location updates ;No. g The particle velocity in a generation is defined as For ease of understanding, the inertia weight , acceleration constant c 1. c The basic parameters of the second-level particle swarm remain consistent with the upper-level solution algorithm.

[0200] S3, fitness function calculation. For any particle position In the scene s DistFlow branch flow calculation is required to obtain the minimum operating cost (fitness) , the formula is:

[0201] ;

[0202] Where, 、 、 、 、 In the scene s The particle position under The distribution network power purchase cost, network loss cost, curtailment penalty cost, V2G battery degradation cost, and safety limit penalty cost at that time; For the scene s Adaptive penalty coefficient under .

[0203] S4, speed and position update. Scene s Down g The position and velocity of particles in the +1 generation are updated according to the formula:

[0204] ;

[0205] ;

[0206] S5, adaptive penalty coefficient Run the scenario through the power flow calculation s Check whether there are any branches or nodes exceeding the limit. If so, record the location of the branches and nodes exceeding the limit and the voltage, current and load exceeding the limit. If there is a serious limit exceeding, increase , which makes the fitness of the defaulting particles significantly worse, pushing the population away from infeasible solutions; if the over-limit situation improves, then reduce , making the algorithm search widely.

[0207] S6, update individual optimal and global optimal. s The position of each particle Calculate the new fitness .like Better than the best particle in history , then Updated to .like Better than the global optimal fitness , then Updated to If the maximum number of iterations has been reached Or the global optimum improves within a number of iterations to less than the target convergence threshold e down Then stop the iteration.

[0208] S7, calculate the master node layer for all scenarios πs The probability-weighted calculation formula for the global annual operating cost is:

[0209] ;

[0210] In addition, the computing master node feeds back to the upper planning layer the locations of branches and nodes that exceed the limits, as well as the voltage, current, and load limits, under the optimal operating mode in all scenarios.

[0211] This example uses a nested iterative algorithm to solve a two-layer collaborative planning model. Through a "decision transfer - cost feedback - solution revision" strategy, iteratively optimizes the collaborative planning solution for the source site. Iterative calculations are performed until the fluctuations in total investment and operating costs stabilize and safety indicators meet standards. Ultimately, a collaborative planning solution with optimal economic efficiency is output.

[0212] In this embodiment, the "decision transfer - state feedback - plan modification" strategy optimizes the source website collaborative planning solution as follows:

[0213] Decision transfer: Based on the load in the area covered by the future distribution network and the electric vehicle charging load demand, and assuming that the lower-level operating cost is a large constant, the upper planning layer preliminarily plans the decision variable parameters and transfers them to the lower operating layer.

[0214] State feedback: Based on multiple typical daily scenarios and their occurrence probabilities, the lower operation layer performs distribution network optimization operation solutions, calculates the lower annual operation cost considering safety over-limit penalties, and records the over-limit branch set I Line,over and the set of out-of-limit nodes Oh Node,over , and feedback to the upper planning layer.

[0215] Plan revision: For out-of-limit branches and node clusters that appear in the lower-level operational scenarios, the upper planning layer adjusts the position and velocity update mechanism of the corresponding hybrid particles, recalculates the planning cost, and transmits the new distribution network planning parameters to the lower-level operational layer. This cycle repeats until both the economic and safety conditions meet the convergence threshold of the upper planning layer, ultimately outputting a robust and economically efficient collaborative planning solution.

[0216] Example 2

[0217] In one or more embodiments, a distribution network source-site collaborative planning system considering a V2G mode is disclosed, including:

[0218] The data acquisition module is used to obtain the existing topology of the distribution network, traffic network structure data, spatiotemporal distribution data of electric vehicles, photovoltaic resource distribution data, and historical daily load power curve sets;

[0219] The scenario generation module is used to confirm the preliminary spatial location constraints of each facility in the source site collaborative planning; cluster the historical daily load power curves into typical daily load power curves, and then predict the node load power set, the maximum dispatchable power set of V2G charging and discharging stations, and the maximum output power set of PVG stations under various typical daily scenarios in the future;

[0220] The model building module is used to establish the upper-level planning model with the goal of minimizing the total economic cost of the system, and to establish the lower-level planning model with the goal of minimizing the annual comprehensive operating cost;

[0221] The model optimization solution module uses a nested iterative algorithm to solve the two-level planning model. The upper-level planning model determines the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. It then outputs a preliminary planning solution that meets the load growth demand and minimizes the investment cost. The distribution network topology and equipment information are then passed to the lower-level model. The lower-level planning model solves the annual optimal operation results, generates the over-limit branches, node marking sets, and penalty costs, and feeds them back to the upper-level planning model.

[0222] Iterate the calculation until the fluctuation of total investment-operation cost is stable and the safety indicators meet the standards, and output the distribution network source site collaborative planning plan.

[0223] Visualization module: Generates a coordinated distribution network planning diagram, annotates the site selection and sizing plans for V2G charging and discharging stations and photovoltaic power sources, and identifies line upgrade / new construction paths; and dynamically displays node voltage distribution, V2G charging and discharging power curves, and safety over-limit alarm information under multiple typical daily scenarios.

[0224] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A distribution network source site collaborative planning method considering the V2G mode, characterized in that: include: Obtain the existing topology of the distribution network, traffic network structure data, spatiotemporal distribution data of electric vehicles, photovoltaic resource distribution data, and historical daily load power curve sets; Confirm the preliminary spatial location constraints of each facility in the source site collaborative planning; cluster the historical daily load power curves into typical daily load power curves, and then predict the node load power set, the maximum dispatchable power set of V2G charging and discharging stations, and the maximum output power set of PVG stations under various typical daily scenarios in the future; The upper-level planning model is established with the goal of minimizing the total economic cost of the system, and the lower-level planning model is established with the goal of minimizing the annual comprehensive operating cost; A nested iterative algorithm is used to solve the two-level planning model. The upper-level planning model determines the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. A preliminary planning scheme that meets load growth needs and minimizes investment costs is output, and the distribution network topology and equipment information are passed to the lower-level model. The lower-level planning model solves the annual optimal operating results, and simultaneously generates the over-limit branches, node marking sets, and penalty costs, which are fed back to the upper-level planning model. Iterate the calculation until the fluctuation of total investment and operating cost stabilizes and safety indicators meet the standards, and then output the distribution network source site collaborative planning plan; The upper-level planning model adopts an adaptive hybrid particle swarm algorithm, and the specific steps and processes are as follows: S1. Algorithm initialization; including setting algorithm parameters, initializing mixed particle swarm, and preliminarily evaluating fitness; Setting the size of the hybrid particle swarm ; Maximum number of iterations ; Current iteration count g, g =1,2,…, ; Particle index p , p =1,2,…, ; r rand 、 r 1. r 2A random number drawn from a uniform distribution in the interval [0,1]; inertia weight ω (g) ; Local search scaling factor ζ d ; Target convergence threshold ε ; For the g Daizhongdi p The position vector of the hybrid particle; For the g Daizhongdi p The velocity vector of the mixed particles; Represents the upper particles p Record of the best position vector in history; Indicates the optimal particle position in the history of the entire population in the upper layer; Indicates the optimal value of the upper layer global fitness; Set the 0-1 location variables included in the plan: New line flag , Line upgrade sign , installation node sign of the substation , V2G charging and discharging station site selection signs PVG station site selection sign ; Integer variable: number of installation areas 、Number of V2G charging and discharging stations 、Number of PVG stations ; Continuous variable: Length of new and upgraded lines 、 , installation capacity of the substation , V2G charging and discharging station capacity PVG station capacity ; Encode the decision variable particles of the upper planning layer; Initialize the velocity vector , hybrid particle index p , randomly generate initial particles in the feasible region for: ； Where, d is the hybrid particle vector dimension variable, hybrid particle The total dimension of the vector is D up , d =1,2,…, ; For the p The first mixed particle d The initial vector of dimensions; 、 、 、 are the initialization mixed particle position sets for new and upgraded lines, new substations, new V2G charging and discharging stations, and new PVG stations respectively; the superscript (0) is the initialization mixed particle position of the corresponding parameter; || is vector splicing; initialization mixed particle Temporarily the optimal position of the particle ; Based on the initial mixed particle scheme Calculated annual investment cost C inv ( ); Get the initial fitness of each mixed particle ; Count the mixed particles corresponding to the global optimal fitness and set it as the optimal mixed particle position , the initial optimal fitness is ; S2. Iterative algorithm solution; including dynamic adjustment of inertia weight, particle-by-particle velocity and position update, calling the underlying model and calculating fitness; Dynamically adjust inertia weight using linear decreasing strategy ; Particles p , updating continuous, discrete and integer variables and their adaptive local search scaling factors; For continuous variables, each dimension in the mixed particle d Speed ​​update and location movement updates for: ； ； Where, Scaling factor for adaptive local search; c 1. c 2 is the acceleration factor; 、 For each dimension of the mixed particle d speed and position; For the p Particle No. d The optimal parameters of each dimension; is the particle population d The optimal parameters for each dimension; For integer variables, the nearest integer strategy is used to correct them to integers; for 0-1 variables, the speed is mapped to the flip probability , the formula is: ； ； The updated Pass it to the lower operation layer to perform multi-scenario flow and operation scheduling calculations, and return the optimal operation results under the planning scheme C ope ( ), fitness function The updated calculation formula is: = C inv ( ) + C ope ( )； S3. Update individual best and global best; S4. Terminate the determination.

2. A distribution network source-site collaborative planning method considering the V2G mode according to claim 1, characterized in that: The improved K-means clustering algorithm is used to cluster the historical daily load power curves into typical daily load power curves, and the recursive method is used to obtain the node load power sets under various typical daily scenarios in the future. The Monte Carlo algorithm is used to generate the maximum dispatchable power set of V2G charging and discharging stations and the maximum output power set of PVG stations under various typical daily scenarios in the future.

3. A distribution network source-site collaborative planning method considering the V2G mode as claimed in claim 2, characterized in that: The improved K-means clustering algorithm is used to cluster the historical daily load power curves into typical daily load power curves, specifically: Randomly select a curve as the initial cluster center curve and set the threshold parameter T 1 and T 2. Set the distance to be less than T The curve with a value of 1 is the curve within the cluster, and the distance is T 1 to T The curves within the range of 2 are suspicious curves, and the distance is T The curves outside the 2 circles are cluster-outside curves; select new cluster center curves from the data marked as cluster-outside curves, and repeat the iteration until all cluster-outside curves are correctly classified; get the number of clusters as the initial K value; and randomly select K initial cluster center curves; By iteratively updating the center curves of all objects in each cluster, K corresponding cluster center curves are obtained; thus, the distribution network nodes are obtained. i Typical day scene s Load curve under and its probability of occurrence .

4. A distribution network source-site collaborative planning method considering the V2G mode as claimed in claim 2, characterized in that: The node load power set under various typical daily scenarios in the future is obtained by using the recursive method, specifically: Assuming the overall load growth rate in the future planning year relative to the historical year and scenario occurrence probability correction factor It is known that the formula of the simple daily load recursion model is: ； ； in, 、 Distribution network nodes i Typical day scene s download curves and their probabilities; and Respectively for scenes s Next node i The predicted values ​​of the future typical daily load curve set and its probability set.

5. The method for collaborative planning of distribution network source and site considering V2G mode according to claim 1, characterized in that: The upper-level planning model is established with the goal of minimizing the total economic cost of the system, specifically: min F = C inv + C ope ; in, F is the total economic cost of the system, C inv The annual construction investment cost includes the annualized sum of the investment costs for new lines / upgrades, new substations, PVG stations, and V2G charging and discharging stations. C ope The annual operating cost.

6. A distribution network source-site collaborative planning method considering the V2G mode as claimed in claim 5, characterized in that: The decision variables of the upper-level planning model include: 、 、 、 、 are 0-1 decision variables for new construction of lines, upgraded lines, new construction of substations, new construction of V2G power stations, and new construction of PVG stations, where 0 means no new construction and 1 means new construction; 、 are the lengths of new and upgraded lines, 、 、 Node i New substations, new V2G power stations and new PVG stations k Type equipment rated capacity, both are continuous decision variables greater than 0; 、 、 They are the new substations, new V2G power stations and new PVG stations. k Type of equipment i The number of nodes, which is an integer variable greater than 0.

7. The method for collaborative planning of distribution network source and site considering V2G mode according to claim 1, characterized in that: The lower-level planning model is established with the goal of minimizing the annual comprehensive operating cost, specifically: ； in, is the annual comprehensive operating cost, A collection of typical daily scenes that may occur in a year; is the natural daily number of the year; A typical day scene s The probability of occurrence ; 、 、 、 、 They represent the power purchase cost of the distribution network, network loss cost, penalty cost for curtailed solar power, V2G battery degradation cost, and safety limit penalty cost respectively; is the adaptive coefficient of the penalty factor.

8. A method for collaborative planning of distribution network source and site considering V2G mode according to claim 7, characterized in that: The decision variables of the lower operation layer include: multiple typical daily scenarios s Next node i At the moment t The actual charging power used by the V2G charging and discharging station and discharge power , PVG actual output power and its power factor angle .

9. The method for collaborative planning of distribution network source and site considering V2G mode according to claim 1, characterized in that: Adaptive hybrid particle swarm optimization is used to solve the upper-level planning model, and parallel particle swarm optimization is used to solve the lower-level planning model; Based on the load volume in the area covered by the future distribution network and the demand for electric vehicle charging load, and assuming that the lower-level operating cost is a large constant, the upper-level planning model preliminarily plans the decision variable parameters and passes them to the lower-level planning model. Based on multiple typical daily scenarios and their occurrence probabilities, the lower-level planning model solves the optimal operation of the distribution network, calculates the lower-level annual operating cost taking into account safety over-limit penalties, records the set of over-limit branches and over-limit nodes, and feeds them back to the upper-level planning model. The upper-level planning model adjusts the update mechanism of the position and speed of the corresponding hybrid particles, recalculates the planning cost, and passes the new distribution network planning parameters to the lower-level planning model, repeating the cycle. The iteration continues until both the economic and safety conditions meet the convergence threshold of the upper planning layer, and the optimal collaborative planning solution is finally output.

10. A distribution network source site collaborative planning system considering the V2G mode, characterized in that: include: The data acquisition module is used to obtain the existing topology of the distribution network, traffic network structure data, spatiotemporal distribution data of electric vehicles, photovoltaic resource distribution data, and historical daily load power curve sets; The scenario generation module is used to confirm the preliminary spatial location constraints of each facility in the source site collaborative planning; cluster the historical daily load power curves into typical daily load power curves, and then predict the node load power set, the maximum dispatchable power set of V2G charging and discharging stations, and the maximum output power set of PVG stations under various typical daily scenarios in the future; The model building module is used to establish the upper-level planning model with the goal of minimizing the total economic cost of the system, and to establish the lower-level planning model with the goal of minimizing the annual comprehensive operating cost; The model optimization solution module uses a nested iterative algorithm to solve the two-level planning model. The upper-level planning model determines the type of new or upgraded lines, the capacity and location of substations, and the site selection and sizing of V2G charging and discharging stations and PVG stations. It then outputs a preliminary planning solution that meets the load growth demand and minimizes the investment cost. The distribution network topology and equipment information are then passed to the lower-level model. The lower-level planning model solves the annual optimal operation results and simultaneously generates the over-limit branches, node marking sets, and penalty costs, which are fed back to the upper-level planning model. Iterate the calculation until the fluctuation of total investment and operating cost stabilizes and safety indicators meet the standards, and then output the distribution network source site collaborative planning plan; The upper-level planning model adopts an adaptive hybrid particle swarm algorithm, and the specific steps and processes are as follows: S1. Algorithm initialization; including setting algorithm parameters, initializing mixed particle swarm, and preliminarily evaluating fitness; Setting the size of the hybrid particle swarm ; Maximum number of iterations ; Current iteration count g, g =1,2,…, ; Particle index p , p =1,2,…, ; r rand 、 r 1. r 2A random number drawn from a uniform distribution in the interval [0,1]; inertia weight ω (g) ; Local search scaling factor ζ d ; Target convergence threshold ε ; For the g Daizhongdi p The position vector of the hybrid particle; For the g Daizhongdi p The velocity vector of the mixed particles; Represents the upper particles p Record of the best position vector in history; Indicates the optimal particle position in the history of the entire population in the upper layer; Indicates the optimal value of the upper layer global fitness; Set the 0-1 location variables included in the plan: New line flag , Line upgrade sign , installation node sign of the substation , V2G charging and discharging station site selection signs PVG station site selection sign ; Integer variable: number of installation areas 、Number of V2G charging and discharging stations 、Number of PVG stations ; Continuous variable: Length of new and upgraded lines 、 , installation capacity of the substation , V2G charging and discharging station capacity PVG station capacity ; Encode the decision variable particles of the upper planning layer; Initialize the velocity vector , hybrid particle index p , randomly generate initial particles in the feasible region for: ； Where, d is the hybrid particle vector dimension variable, hybrid particle The total dimension of the vector is D up , d =1,2,…, ; For the p The first mixed particle d The initial vector of dimensions; 、 、 、 are the initialization mixed particle position sets for new and upgraded lines, new substations, new V2G charging and discharging stations, and new PVG stations respectively; the superscript (0) is the initialization mixed particle position of the corresponding parameter; || is vector splicing; initialization mixed particle Temporarily the optimal position of the particle ; Based on the initial mixed particle scheme Calculated annual investment cost C inv ( );Get the initial fitness of each mixed particle ; Count the mixed particles corresponding to the global optimal fitness and set it as the optimal mixed particle position , the initial optimal fitness is ; S2. Iterative algorithm solution; including dynamic adjustment of inertia weight, particle-by-particle velocity and position update, calling the underlying model and calculating fitness; Dynamically adjust inertia weight using linear decreasing strategy ; Particles p , updating continuous, discrete and integer variables and their adaptive local search scaling factors; For continuous variables, each dimension in the mixed particle d Speed ​​update and location movement updates for: ； ； Where, Scaling factor for adaptive local search; c 1. c 2 is the acceleration factor; 、 For each dimension of the mixed particle d speed and position; For the p Particle No. d The optimal parameters of each dimension; is the particle population d The optimal parameters for each dimension; For integer variables, the nearest integer strategy is used to correct them to integers; for 0-1 variables, the speed is mapped to the flip probability , the formula is: ； ； The updated Pass it to the lower operation layer to perform multi-scenario flow and operation scheduling calculations, and return the optimal operation results under the planning scheme C ope ( ), fitness function The updated calculation formula is: = C inv ( ) + C ope ( )； S3. Update individual best and global best; S4. Terminate the determination.

Images