Problem-driven distributed robust capacity configuration method for optical storage and charging system

Abstract

The application discloses a problem-driven distribution robust capacity configuration method of a light storage and charging system, which comprises the following steps: a problem-driven loss distance caused by a capacity configuration scheme between scenes is constructed; a comprehensive norm probability ambiguity set is introduced to construct a problem-driven distribution robust scene reduction model; the scene reduction model is solved to obtain a representative scene set, a nominal probability and a corresponding probability offset boundary; a light storage and charging system distribution robust capacity optimization configuration model is established; the light storage and charging system distribution robust capacity optimization configuration model is solved to obtain an optimization configuration scheme containing photovoltaic installed capacity, energy storage rated power, energy storage rated capacity, electric vehicle charging station capacity and charging pile quantity. The application has the beneficial effects that the opportunity loss is used to construct a problem-driven distance, so that the scene reduction result directly serves the capacity configuration decision, instead of only keeping the curve shape similar; the application also keeps the consistency of the probability uncertainty representation before and after the scene reduction.

Description

Technical Field

[0001] This invention relates to the field of transportation energy system technology, and in particular to a problem-driven method for configuring the capacity of distributed rods in a photovoltaic-energy storage-charging system (PV-energy storage-charging station system). Background Technology

[0002] With the increasing number of electric vehicles on the road each year, the demand for centralized fast charging is growing rapidly. At the same time, my country has huge potential for photovoltaic resources along its highways. Building photovoltaic-storage-charging systems can effectively improve the local absorption capacity of distributed photovoltaic power, reduce the peak-hour power supply pressure on the distribution network, and enhance the charging guarantee capacity for long-distance travel.

[0003] Photovoltaic-storage-charging systems are affected by factors such as weather conditions, holiday traffic flow, seasonal passenger flow, and changes in base load. Among these, photovoltaic output, electric vehicle charging load, and infrastructure load all exhibit significant uncertainties. If these uncertainties are not fully considered during the capacity configuration phase, it can easily lead to problems such as overly conservative configuration of photovoltaic, energy storage, or charging facilities, resulting in excessively high investment costs, or insufficient capacity and a decline in charging service quality.

[0004] Current optimization methods for uncertainty mainly include stochastic optimization, robust optimization, and partial Brutal optimization. Stochastic optimization relies on accurate probability distribution assumptions and has poor robustness when historical samples are insufficient or the distribution shifts. Robust optimization typically makes decisions based on the worst-case scenario, which may lead to high system investment costs. Partial Brutal optimization can balance economy and robustness when the probability distribution is not completely certain. However, directly introducing a large number of original scenarios into the optimization model will significantly increase the size of mixed-integer linear programming and the complexity of iterative solutions.

[0005] In the uncertainty planning problem of green energy systems, scenario reduction is usually performed based on typical days, clustered scenarios, or random scenarios, such as K-means clustering, K-medoids clustering, and scenario reduction methods based on statistical or geometric distance. These methods mainly focus on the similarity of source-load curve shapes or statistical distributions, but fail to directly characterize the impact of different scenarios on capacity allocation decisions, operating costs, electricity purchase demand, curtailment levels, and the degree to which charging loads are met.

[0006] For photovoltaic-storage-charging systems, some source-load curves may have similar shapes, but their corresponding energy storage capacity requirements, grid interaction power boundaries, and charging security risks may differ significantly. Conversely, scenarios with different curve shapes may have similar decision-making effects in terms of capacity configuration. Therefore, there is an urgent need for an integrated method for scenario reduction and sub-Bruker optimal configuration, which selects representative scenarios that have a key impact on capacity configuration results while considering scenario probability shifts, thereby enabling system capacity configuration. Summary of the Invention

[0007] The purpose of this invention is to provide a problem-driven, distributed capacity configuration method for a photovoltaic storage and charging system, which overcomes the shortcomings of traditional scenario reduction methods that only focus on the statistical similarity of source-load curves and ignore the impact of capacity configuration decisions, and solves the problem of difficulty in balancing economy and robustness in uncertain optimization configuration.

[0008] The technical solution for achieving the objective of this invention is as follows:

[0009] A problem-driven method for configuring the capacity of distributed power supply rods in an optical storage and charging system includes the following steps:

[0010] S1 obtains the original source-load scenario set of the optical storage-charging system, uses the single-scenario capacity optimization configuration model to obtain the capacity configuration scheme for each scenario in the original source-load scenario set, and uses the inter-scenario back-substitution method to calculate the problem-driven loss distance between scenarios.

[0011] S2 constructs a scenario reduction model based on the comprehensive norm probability ambiguity set and the problem-driven loss distance;

[0012] S3 solves the scene reduction model to obtain the representative scene set and nominal probability, and maps the original source load scene probability boundary to the representative scene probability space to obtain the representative scene probability offset boundary;

[0013] S4 constructs a multi-bar capacity optimization configuration model for the optical storage and charging system;

[0014] S5 inputs the representative scenario set, nominal probability, and representative scenario probability offset boundary into the sub-Blu-ray rod capacity optimization configuration model, solves the sub-Blu-ray rod capacity optimization configuration model, and obtains the capacity configuration scheme of the optical storage and charging system.

[0015] Preferably, the original source load scene set is as follows:

[0016]

[0017] In the formula, N represents the number of original source load scenes, and the th... A scenario Includes unit photovoltaic output for T time periods within a day EV charging load and base load ; The nominal probability set of the original source-load scenario; ;

[0018] The single-scenario capacity optimization configuration model is as follows: In the formula, Indicates in the scene The following capacity configuration scheme, Represents the set of capacity configuration constraints. Indicating in capacity configuration scheme and scene The total system cost is as follows , Indicates photovoltaic installed capacity, Indicates the rated power of energy storage, Indicates the rated capacity of energy storage, Indicates the configuration capacity of EV charging stations, Indicates the number of charging stations;

[0019] The method employing inter-scene back-substitution to calculate the problem-driven loss distance between scenes is as follows:

[0020] Scene Capacity configuration scheme Return to scene ,get In the scene Total system cost ; to the scene Capacity configuration scheme Return to scene ,get In the scene Total system cost Scene With Scene The problem of distance driving loss In the formula, , Scenes Scene Capacity configuration scheme , The total system cost.

[0021] Preferably, the comprehensive norm probability ambiguity set is

[0022]

[0023] In the formula, For probability ambiguity set, Let be the scene probability distribution vector. For the scene The probability, The nominal probability distribution vector; , They are respectively the 1-norm radius and - Norm radius, used to constrain the overall probability shift and the maximum shift of a single scene probability; N is the number of original source load scenes;

[0024] in,

[0025]

[0026] In the formula, , They are respectively 1-norm and - The confidence level parameter corresponding to the norm probability offset bound; The scaling factor for the radius of the ambiguity set; This is the number of effective samples used to estimate the probability distribution of the scene;

[0027] The problem-driven scenario reduction model for the sub-bar is as follows:

[0028]

[0029] In the formula, The binary variable chosen to represent the scenario. Assign variables to the scene;

[0030] ;

[0031] For the scene The representative error is, , For the scene Assigned to scene The problem of time-driven loss distance; As a weighting factor, Representing candidate scenarios The self-allocation state, , These represent the minimum and maximum number of scenarios.

[0032] Preferably, the objective function of the distributed bar capacity optimization configuration model of the photovoltaic storage and charging system is: In the formula, This represents the average daily total cost of the photovoltaic energy storage and charging system. This refers to the daily value of investment costs, including the daily value of investment costs for energy storage systems, photovoltaic systems, and EV charging stations. , , ; The operating cost is the daily value under the worst-case probability distribution. and Let $\mathbf{k}$ be the running cost of the k-th representative scenario and its corresponding scenario probability after offset. This indicates the number of scenes. Represents the set of probability constraints for the representative scenario; the operating cost of the k-th representative scenario. This includes the operation and maintenance costs of each unit. Operation and maintenance penalty costs Distribution network interaction costs and the cost of power fluctuations in the tie line The constraints of the distributed rod capacity optimization configuration model of the photovoltaic-storage-charging system include: configuration constraints of the energy storage system, photovoltaic system and EV charging station; operation constraints of the energy storage system, photovoltaic system and EV charging station; power purchase and sales constraints; tie line power fluctuation constraints; and system operation power balance constraints.

[0033] Furthermore, the aforementioned operation and maintenance penalty cost Including the total cost of system light curtailment penalty during the operating cycle and system base load reduction penalty costs It also includes the penalty costs for ensuring the quality of EV charging station services: EV charging queuing penalty costs. EV unservice demand penalty cost and the cost of penalties for exceeding service limits ;

[0034] The penalty cost for ensuring the service quality of EV charging stations is determined according to the following method:

[0035] S1 establishes an EV charging model, which characterizes the EV charging load queuing backlog caused by insufficient charging station capacity and limits the maximum queuing charging time:

[0036] Let the energy required for EV charging during time period t be... , For immediate service energy, To meet the delay requirements;

[0037] , This represents the service volume in time period t for the delayed demand from the previous time period. For charging demands that remain unmet even after waiting for the maximum queuing time, the actual charging power of the EV charging station is... , , For EV charging station capacity;

[0038] S2 determines the penalty cost for poor EV charging station service quality.

[0039] ;

[0040] In the formula, For Japan's total EV demand; For the final unserved demand; It is a non-negative auxiliary variable; To achieve the target service rate for Japanese EV charging stations, , , These are the additional penalty cost coefficients for EV charging queues, EV unserved demand, and service rates below the target value.

[0041] Furthermore, the energy storage unit of the energy storage system is a lithium battery; the operating cost of the k-th representative scenario... It also includes the cost of energy storage cycle aging. The energy storage cycle aging cost It is determined by the following method:

[0042] Energy storage in cycle depth The cycle life is as follows:

[0043]

[0044] In the formula, The number of charge-discharge cycles (CDRs) required for a lithium battery to reach 100% depth of discharge and reach the end of its lifespan. These are the fitting parameters;

[0045] The equivalent number of cycles at 100% depth of discharge is:

[0046]

[0047] When energy storage changes from a discharging state to a charging state, the cyclic variable for:

[0048]

[0049] set up If the accumulated battery-side discharge energy is the current continuous discharge phase, then during the discharge phase:

[0050]

[0051] When a cyclic event occurs, the accumulated discharge energy from the previous moment is extracted:

[0052]

[0053] Therefore, the loop depth for:

[0054]

[0055] Determine the energy storage cycle aging cost during the operating cycle. for:

[0056] .

[0057] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0058] 1. By using opportunity loss to construct problem-driven distance, the scenario reduction results can directly serve capacity configuration decisions, rather than simply maintaining a similar curve shape.

[0059] 2. The scene reduction stage and the capacity configuration stage adopt the same probability uncertainty description, and the nominal probability and probability offset boundary are mapped to the representative scene space according to the allocation relationship between the original scene and the representative scene, so as to maintain the consistency of the probability uncertainty representation before and after scene reduction.

[0060] 3. By reducing the scenario scale and related optimization variable scale in the subsequent capacity configuration stage through problem-driven scenario reduction, the solvability of subsequent capacity configuration is improved while maintaining the representativeness of the scenario decision. Attached Figure Description

[0061] Figure 1 This is a flowchart of the present invention.

[0062] Figure 2 This is a schematic diagram of the structure of a photovoltaic energy storage and charging system. Detailed Implementation

[0063] This invention provides a problem-driven method for configuring the capacity of a distributed power supply in an optical storage and charging system, such as... Figure 1 As shown, it includes:

[0064] Step S1: Obtain the original source-load scenario data, solve the single-scenario capacity optimization configuration model, and substitute the obtained single-scenario capacity configuration scheme into other scenarios for execution. Based on the above execution results, construct the problem-driven loss distance caused by the substitution of capacity configuration schemes between scenarios, thereby constructing a symmetric problem-driven loss distance matrix.

[0065] Step S2: Considering the changes in the original scene probability, introduce a comprehensive norm probability ambiguity set and construct a problem-driven sub-bar scene reduction model.

[0066] Step S3: Based on the column and constraint generation algorithm, solve the problem-driven scene reduction model of the sub-bar to obtain the representative scene set, nominal probability and corresponding probability offset boundary, which are used as the scene set input for the subsequent optical storage and charging system optimization configuration model.

[0067] Step S4: Establish a capacity optimization configuration model for the photovoltaic storage and charging system.

[0068] Step S5: Solve the photovoltaic-storage-charging system distributed rod capacity optimization configuration model based on column and constraint generation algorithm to obtain an optimized configuration scheme including photovoltaic installed capacity, energy storage rated power, energy storage rated capacity, electric vehicle charging station capacity and number of charging piles.

[0069] Step S6: Substitute the system capacity configuration scheme obtained in step S5 back into the original scenario and run it. Calculate the total system cost under the nominal probability and the worst-case probability distribution respectively to evaluate the economy and robustness of the representative scenario configuration scheme in the original scenario.

[0070] The present invention will be further illustrated below using a photovoltaic, energy storage, and charging system in a highway service area as an example.

[0071] The structure of the photovoltaic-storage-charging system in highway service areas is as follows: Figure 2 As shown, it includes a photovoltaic system, an energy storage system, and an EV charging station.

[0072] Step S1: Define the original source-load scene set for the highway service area:

[0073]

[0074] In the formula, N represents the original number of scenes; each scene Includes unit photovoltaic output for T time periods within a day EV charging load Service area basic load ; Let be the nominal probability set of the original scene.

[0075] The capacity configuration variables for a photovoltaic-storage-charging system include photovoltaic installed capacity, rated energy storage power, rated energy storage capacity, EV charging station configuration capacity, and the number of charging piles, defined as follows:

[0076]

[0077] For any original scene The corresponding single-scenario capacity configuration optimization problem is defined as follows:

[0078]

[0079] In the formula, Represents the set of capacity configuration constraints; Indicating in capacity configuration scheme and scene The total system cost includes investment costs and operating costs for this scenario.

[0080] Then the scene The optimal objective function value for a single scenario is:

[0081]

[0082] Scene Substituting the optimal capacity scheme obtained below back into another scenario We can obtain:

[0083]

[0084] Therefore, the opportunity loss matrix can be obtained. :

[0085]

[0086] In the formula, Indicates the use case Representative scenarios This results in a loss of capacity configuration performance. The larger the value, the more likely it is to be a scene. For the scene The weaker the representativeness of the decision-making process; when When it is close to 0, it indicates a scenario For the scene It has good representativeness in configuration decision-making.

[0087] To obtain a symmetric problem-driven metric, define the scenario. With Scene The problem-driven loss distance between them is:

[0088]

[0089] In the formula, Indicates the use case Optimal configuration alternative scenarios When determining the optimal configuration scheme, in the scenario Opportunity losses resulting from this; Indicates the use case Optimal configuration alternative scenarios When determining the optimal configuration scheme, in the scenario The opportunity loss incurred. Therefore, Able to reflect the scene With Scene The bidirectional decision-making differences between the two scenarios effectively quantify their mutual decision-making adaptability. Furthermore, a smaller problem-driven loss distance indicates that the two scenarios play similar roles in capacity configuration; a larger problem-driven loss distance indicates a more significant difference in the impact of the two scenarios on the system configuration outcome.

[0090] Step S2, in the scene reduction stage of the original scene, this invention introduces a comprehensive norm probability ambiguity set:

[0091]

[0092] In the formula, For probability ambiguity set, Let be the scene probability distribution vector. For the scene The probability, The nominal probability distribution vector; , They are respectively the 1-norm radius and - Norm radius, used to constrain the overall probability shift and the maximum shift of individual scene probabilities;

[0093] in,

[0094]

[0095] In the formula, , They are respectively 1-norm and - The confidence level parameter corresponding to the norm probability offset bound; The scaling factor for the radius of the ambiguity set; This represents the number of valid samples used to estimate the probability distribution of the scene.

[0096] Define a binary variable representing the scenario selection. and scene assignment variables :

[0097]

[0098] When the scene Assigned to representative scene At that time, the problem-driven loss distance is Therefore, the scenario The representative error is:

[0099]

[0100] Considering the above probability distribution shift, the following problem-driven scene reduction model for the sub-bar is constructed:

[0101]

[0102] In the formula, the first term is the problem-driven loss under the worst probability distribution, and the second term is the penalty term representing the number of scenarios; To weigh factors, constraints Guarantee Scenarios It can only be assigned to a cluster representing a specific scenario; Represents the self-assigned state of candidate scenario j, with constraints Then strictly limited, if the scene Once selected as a representative scenario, it must belong to the cluster it represents; constraint Ensure that each scenario has a unique affiliation, meaning it can only be assigned to a single cluster; constraints This limits the range of representative scenarios.

[0103] Step S3, based on the Column and Constraint Generation (C&CG) algorithm, solve the problem-driven scenario reduction model of the sub-bar as follows:

[0104] The original problem is decomposed into a first-stage master problem (MP) and a second-stage subproblem (SP). Given a scenario variable P, MP1 is solved to obtain a lower bound LB1 for the optimal value of the original problem. The first-stage variables u and v obtained from MP1 are substituted into SP1 to obtain an upper bound UB1 for the optimal value of the original problem. The scenario variable P obtained from SP1 is returned to MP1, and new robust cut constraints are added to MP1 accordingly. The process of first relaxing and then pursuing is repeated, alternating between solving MP1 and SP1, thus forming a C&CG loop. The specific form is as follows:

[0105]

[0106]

[0107] In the formula, The auxiliary variable introduced represents the optimal value of the objective function in the second stage; This represents the worst probability distribution obtained from the subproblems in the l1st iteration; Solving the main problem The resulting representative error vector; is an auxiliary variable used to linearize the 1-norm constraint, representing the absolute offset of the probability of scenario i relative to the nominal probability.

[0108] As the C&CG algorithm iterates, the worst-case scenario probability distribution generated after solving SP1 is retained and passed to the current MP1. The number of variables, constraints, and cutting planes in MP1 increases continuously, so LB1 will become larger and larger. At the same time, the scenario reduction variables provided by MP1 to SP1 also change, so UB1 will become smaller and smaller. If the convergence criterion (15) is satisfied, the loop exits and the optimal solution is returned.

[0109]

[0110] In the formula, Let be a very small positive real number, representing the convergence gap.

[0111] Therefore, we can obtain the representative set of scenarios and their nominal probabilities:

[0112]

[0113] In the formula, This represents the Kth representative scene obtained after the r-th iteration. The nominal probability of the Kth representative scene is obtained by summing the original scene probabilities of its corresponding cluster.

[0114] Since the original scene probabilities have upper and lower offset boundaries, to ensure the consistency of scene probability offsets, the upper and lower offset boundaries representing the scene are obtained by mapping the original source load scene probability boundary to the representative scene probability space:

[0115]

[0116] The above-aggregated representative scenarios and their probabilities will serve as inputs for the subsequent sub-Bruker optimization configuration model.

[0117] Step S4: Establish the optimal configuration model of the distributed blobs for the photovoltaic-storage-charging system, as follows:

[0118] The objective function for optimizing the configuration is to minimize the average daily total cost over the entire lifecycle of the photovoltaic-storage-charging system. Investment cost based on daily value Equal daily operating costs under the worst probability distribution Composition. The specific form is as follows:

[0119]

[0120]

[0121]

[0122] In the formula, and Let $\frac{k}{k}$ be the running cost of the k-th representative scenario and its corresponding scenario probability after offset. , and These are the daily investment costs for energy storage systems, photovoltaic systems, and EV charging stations, respectively. The discount rate; , and These refer to the investment periods for energy storage systems, photovoltaic systems, and EV charging stations, respectively. , and These are the investment cost coefficients for energy storage systems, photovoltaic systems, and EV charging stations, respectively. , , , and These include the operation and maintenance costs of each unit, operation and maintenance penalty costs, distribution network interaction costs, tie line power fluctuation penalty costs, and energy storage cycle aging costs. , and These are the operation and maintenance cost coefficients for energy storage systems, photovoltaic systems, and EV charging stations, respectively. , , , , These are the unit power penalty cost coefficients for EV charging queues, EV unserved demand, service rate below target value, base load reduction, and curtailment of solar power, respectively. and The transaction price between the service area and the distribution network; The unit power penalty cost coefficient for power fluctuations in the tie line between the service area and the distribution network; The number of cycles for energy storage (lithium battery) to reach the end of its lifespan by 100% deep discharge.

[0123] To fully utilize the photovoltaic resources in the service area, the upper limit of the distributed photovoltaic configuration capacity is limited by conditions such as the installation site in the service area, as shown in the following expression:

[0124]

[0125] In the formula, This represents the maximum allowable installation capacity for distributed photovoltaic systems.

[0126] Considering the increasing number of EVs on the road, the upper limit of EV charging station configuration capacity is limited by the installation scenario and grid-connected power, as shown in the following expression:

[0127]

[0128] In the formula, The maximum allowable construction capacity for EV charging stations Rated power for a single charging station.

[0129] Considering the regulation needs of the energy storage system, site conditions, grid connection conditions, and renewable energy consumption needs, a lower limit for the energy storage configuration power or a configuration ratio constraint between energy storage and photovoltaic installed capacity can be set according to specific project requirements. In this embodiment, the configuration ratio is set to 10%, as shown in the following expression:

[0130]

[0131] In the formula, Configure power for the energy storage system; , These are the upper and lower limits of the configured power for the energy storage system, respectively. This refers to the rated capacity of the energy storage system. The duration of continuous energy storage by the energy storage system at its rated power.

[0132] For photovoltaic systems, considering the physical absorption boundary and economic trade-offs in actual operation, the system should satisfy photovoltaic output balance and curtailment constraints, and quantify its penalty cost, as shown in the following expression:

[0133]

[0134] In the formula, , and These are the theoretical power, actual utilized power, and curtailed power of the photovoltaic system, respectively. This represents the total system waste-light penalty cost during the operating cycle.

[0135] During peak holiday travel periods or under extreme operating conditions such as traffic congestion, EV charging demand is highly concentrated, potentially causing the charging load to exceed the configured capacity limits of charging stations. Therefore, an EV charging model is established to characterize the EV charging load queuing backlog caused by insufficient charging station capacity, and the maximum queuing time is limited to 1 hour to meet the demand for rapid service.

[0136] Let the energy required for EV charging during time period t be... It includes instant service capabilities. and delay requirements :

[0137]

[0138] Delayed requests must be served in the next time slot; otherwise, they will be recorded as unserved requests. :

[0139]

[0140] In the formula, The service volume in time period t represents the delayed demand from the previous time period. For charging needs that are still not met after waiting for 1 hour.

[0141] Therefore, the actual charging power of the EV charging station can be obtained as follows:

[0142]

[0143] To ensure the quality of EV charging station services, a daily charging service rate is introduced. EV charging queue penalty costs EV unservice demand penalty cost and the cost of penalties for exceeding service limits :

[0144]

[0145] In the formula, For Japan's total EV demand; For the final unserved demand; It is a non-negative auxiliary variable; Target service rate for Japanese EV charging stations.

[0146] The base load covers critical public service facilities such as lighting, ventilation, fire protection, and commercial facilities. To ensure the system can still operate under extreme conditions or severe source-load mismatch, this invention introduces a base load reduction variable and a corresponding penalty mechanism, expressed as follows:

[0147]

[0148] In the formula, and These are the original base load forecast power and the actual load reduction power (load shedding) of the service area, respectively. This reduces penalty costs associated with reducing the system's base load during the operating cycle. It should be noted that... Typically, the constraint is set with reference to the value of load loss, the importance level of the service area load, or the emergency supply guarantee cost under extreme conditions, to form an absolute non-economic penalty. Therefore, the constraint is strictly defined. It must be much greater than the maximum cost of purchasing electricity.

[0149] The operational constraints of energy storage systems include charge / discharge power constraints, state of charge constraints, and charge / discharge balance constraints, as shown in the following expressions:

[0150]

[0151] In the formula, , These refer to the charging and discharging power of the energy storage system, respectively. , These are the charging and discharging state variables of the energy storage system, respectively. When it is 1, A value of 0 indicates a charging state. When When it is 0, A value of 1 indicates a discharge state; , , , , The initial state of charge, upper and lower limits of the state of charge, and charging and discharging efficiency of the energy storage system are respectively defined.

[0152] In this embodiment, the energy storage unit is a lithium battery. During operation, the greater the depth of discharge, the fewer the number of cycles the lithium battery can be used, and the shorter its cycle life.

[0153] Energy storage in cycle depth The cycle life is as follows:

[0154]

[0155] In the formula, The number of charge-discharge cycles to the end of the lifespan of a lithium battery at 100% depth of discharge. These are the fitting parameters.

[0156] The equivalent number of cycles at 100% depth of discharge is:

[0157]

[0158] Define loop variable When energy storage changes from a discharging state to a charging state:

[0159]

[0160] set up If the accumulated battery-side discharge energy is the current continuous discharge phase, then during the discharge phase:

[0161]

[0162] When a cyclic event occurs, the accumulated discharge energy from the previous moment is extracted:

[0163]

[0164] Therefore, the loop depth is defined. for:

[0165]

[0166] Ultimately, the energy storage cycle aging cost during the operating cycle for:

[0167]

[0168] Power grid interaction constraints mainly refer to the constraints that the service area and the distribution network should meet, namely, tie-line power fluctuation constraints and power interaction constraints, as shown in the following expressions:

[0169]

[0170] In the formula, , These refer to the power purchased and sold between the service area and the power distribution network, respectively. , These are the state variables for electricity purchase and electricity sale, respectively. A value of 1 indicates the electricity purchase status. When A value of 0 indicates the status of electricity sales; For tie-line crossover power limits; Net interactive power of the tie line; This represents the absolute value of the power fluctuation in the tie line. This represents the maximum gradeability.

[0171] The system power balance constraint ensures that the power on both the power supply side and the load side remains balanced in each representative scenario, as shown in the following expression:

[0172]

[0173] Step S5: Based on the column and constraint generation algorithm, solve the optimal configuration model of the optical storage and charging system using the distributed bar configuration, as follows:

[0174] The original problem is decomposed into a first-stage master problem (MP) and a second-stage subproblem (SP). Given scenario variables... Solve MP2 to obtain the lower bound LB2 of the optimal value of the original problem; substitute the first-stage variable x obtained from solving MP2 into solving SP2 to obtain the upper bound UB2 of the optimal value of the original problem; solve SP2 for the scenario variable... Return to MP2, and add new robust cut constraints to MP2 accordingly; first relax, then pursue, solving MP2 and SP2 alternately, thus forming a C&CG loop. The specific form is as follows:

[0175]

[0176]

[0177] In the formula, Configure variables for capacity; The auxiliary variable introduced represents the optimal value of the objective function in the second stage; This represents the worst-case probability distribution obtained from the subproblems in the l2nd iteration. These are auxiliary variables used to linearize the 1-norm constraint.

[0178] As the C&CG algorithm iterates, the worst-case scenario probability distribution generated after solving SP2 is retained and passed to the current MP2. The number of variables, constraints, and cutting planes in MP2 increases continuously, so LB2 will become larger and larger. At the same time, the scenario reduction variables provided by MP2 to SP2 are also changing, so UB2 will become smaller and smaller. If the convergence criterion (35) is satisfied, the loop exits and the optimal solution is returned.

[0179]

[0180] In the formula, Let be a very small positive real number, representing the convergence gap.

[0181] Step S6: Substitute the system configuration results back to the original scenario for evaluation, as follows:

[0182] To ensure that the system operation results conform to the actual situation and to compare and analyze them with traditional methods under the same standard, this invention substitutes the final capacity configuration scheme back into the original scenario to obtain the total system cost under the nominal probability and the total system cost under the worst probability distribution.

[0183] Therefore, the total system cost under the nominal probability is:

[0184]

[0185] In the formula, This is the final capacity configuration scheme.

[0186] The total system cost under the worst probability distribution is:

[0187]

[0188] The above results can test the economy and robustness of the proposed method under the original uncertain set, and can be compared and evaluated with methods such as statistical distance clustering, random scene selection or direct bloat optimization under the same standard.

[0189] The problem-driven distributed rod capacity configuration method for the photovoltaic-storage-charging system in highway service areas described in this embodiment sets the model parameters as follows:

[0190] 1. Set model parameters

[0191] Setting a comprehensive norm ambiguity set 0.50 0.99 It is 0.15.

[0192] Other relevant parameters for the service area are shown in Table 1.

[0193] Table 1 Service Area Parameters

[0194]

[0195] Table 2 represents the results of scenario optimization configuration.

[0196] 2. System performance throughout the year

[0197] To verify the rationality and effectiveness of the model established in this invention, the system capacity configuration scheme obtained based on the representative scenario is substituted into the annual scenario operation to analyze the operating effect of the system capacity configuration scheme under nominal probability and worst probability.

[0198] Table 3. Results of Scene Back-up Generation Throughout the Year

[0199]

[0200] To evaluate the sample generalization ability of the reduced scenario configuration scheme, the capacity scheme obtained based on the representative scenario was fixed and back-substituted into the original scenario throughout the year. The nominal total cost of the representative scenario was RMB 9550.45 / day, and the nominal back-substituted total cost for the whole year was RMB 9022.14 / day, with a deviation of 5.86%; the corresponding worst-case total costs were RMB 10736.75 / day and RMB 10494.11 / day, with a relative deviation of 2.31%. This indicates that the reduced representative scenario has a good representation ability of the original scenario throughout the year, especially under the worst-case scenario probability. There were only 7 days of unserved demand throughout the year, and there was no constraint on the EV charging service rate being below 90% on any day, with a minimum service rate of 97.380%. The results show that the 1260kW charging facility combined with a one-hour demand extension can meet the vast majority of traffic charging needs, while retaining a small amount of economic flexibility to cope with extreme demand. The number of days with grid-connected power limit was 13 days throughout the year, accounting for only 3.55% of the total annual capacity. Furthermore, there was no reduction in base load or system power deficit, indicating that the resulting capacity configuration has high power supply reliability.

Claims

1. A problem-driven method for configuring the capacity of a distributed energy storage and charging system, characterized in that, Including the following steps: S1 obtains the original source-load scenario set of the optical storage-charging system, uses the single-scenario capacity optimization configuration model to obtain the capacity configuration scheme for each scenario in the original source-load scenario set, and uses the inter-scenario back-substitution method to calculate the problem-driven loss distance between scenarios. S2 constructs a scenario reduction model based on the comprehensive norm probability ambiguity set and the problem-driven loss distance; S3 solves the scene reduction model to obtain the representative scene set and nominal probability, and maps the original source load scene probability boundary to the representative scene probability space to obtain the representative scene probability offset boundary; S4 constructs a multi-bar capacity optimization configuration model for the optical storage and charging system; S5 inputs the representative scenario set, nominal probability, and representative scenario probability offset boundary into the sub-Blu-ray rod capacity optimization configuration model, solves the sub-Blu-ray rod capacity optimization configuration model, and obtains the capacity configuration scheme of the optical storage and charging system.

2. The capacity configuration method as described in claim 1, characterized in that, The original source load scene set is as follows: In the formula, N represents the number of original source load scenes, and the th... A scenario Includes unit photovoltaic output for T time periods within a day EV charging load and base load ; The nominal probability set of the original source-load scenario; ; The single-scenario capacity optimization configuration model is as follows: In the formula, Indicates in the scene The following capacity configuration scheme, Represents the set of capacity configuration constraints. Indicating in capacity configuration scheme and scene The total system cost is as follows , Indicates photovoltaic installed capacity, Indicates the rated power of energy storage, Indicates the rated capacity of energy storage, Indicates the configuration capacity of EV charging stations, Indicates the number of charging stations; The method employing inter-scene back-substitution to calculate the problem-driven loss distance between scenes is as follows: Scene Capacity configuration scheme Return to scene ,get In the scene Total system cost ; to the scene Capacity configuration scheme Return to scene ,get In the scene Total system cost Scene With Scene The problem of distance driving loss In the formula, , Scenes Scene Capacity configuration scheme , The total system cost.

3. The capacity configuration method as described in claim 1, characterized in that, The comprehensive norm probability ambiguity set is In the formula, For probability ambiguity set, Let be the scene probability distribution vector. For the scene The probability, The nominal probability distribution vector; , They are respectively the 1-norm radius and - Norm radius, used to constrain the overall probability shift and the maximum shift of a single scene probability; N is the number of original source load scenes; in, In the formula, , They are respectively 1-norm and - The confidence level parameter corresponding to the norm probability offset bound; The scaling factor for the radius of the ambiguity set; This is the number of effective samples used to estimate the probability distribution of the scene; The problem-driven scenario reduction model for the sub-bar is as follows: In the formula, The binary variable chosen to represent the scenario. Assign variables to the scene; ； For the scene The representative error is, , For the scene Assigned to scene The problem of time-driven loss distance; As a weighting factor, Representing candidate scenarios The self-allocation state, , These represent the minimum and maximum number of scenarios.

4. The capacity configuration method as described in claim 1, characterized in that, The objective function of the distributed bar capacity optimization configuration model for the photovoltaic energy storage and charging system is: In the formula, This represents the average daily total cost of the photovoltaic energy storage and charging system. This refers to the daily value of investment costs, including the daily value of investment costs for energy storage systems, photovoltaic systems, and EV charging stations. , , ; The operating cost is the daily value under the worst-case probability distribution. and Let $\mathbf{k}$ be the running cost of the k-th representative scenario and its corresponding scenario probability after offset. This indicates the number of scenes. Represents the set of probability constraints for the representative scenario; the operating cost of the k-th representative scenario. This includes the operation and maintenance costs of each unit. Operation and maintenance penalty costs Distribution network interaction costs and the cost of power fluctuations in the tie line ; The constraints of the distributed bar capacity optimization configuration model of the photovoltaic energy storage and charging system include: Configuration constraints of energy storage systems, photovoltaic systems and EV charging stations; operation constraints of energy storage systems, photovoltaic systems and EV charging stations; power purchase and sales constraints; tie line power fluctuation constraints; and system operation power balance constraints.

5. The capacity configuration method as described in claim 4, characterized in that, The aforementioned maintenance penalty cost Including the total cost of system light curtailment penalty during the operating cycle and system base load reduction penalty costs It also includes the penalty costs for ensuring the quality of EV charging station services: EV charging queuing penalty costs. EV unservice demand penalty cost and the cost of penalties for exceeding service limits ; The penalty cost for ensuring the service quality of EV charging stations is determined according to the following method: S1 establishes an EV charging model, which characterizes the EV charging load queuing backlog caused by insufficient charging station capacity and limits the maximum queuing charging time: Let the energy required for EV charging during time period t be... , For immediate service energy, To meet the delay requirements; , This represents the service volume in time period t for the delayed demand from the previous time period. For charging demands that remain unmet even after waiting for the maximum queuing time, the actual charging power of the EV charging station is... , , For EV charging station capacity; S2 determines the penalty cost for poor EV charging station service quality. ； In the formula, For Japan's total EV demand; For the final unserved demand; It is a non-negative auxiliary variable; To achieve the target service rate for Japanese EV charging stations, , , These are the additional penalty cost coefficients for EV charging queues, EV unserved demand, and service rates below the target value.

6. The capacity configuration method as described in claim 4, characterized in that, The energy storage unit of the energy storage system is a lithium battery; the operating cost of the k-th representative scenario It also includes the cost of energy storage cycle aging. The energy storage cycle aging cost It is determined by the following method: Energy storage in cycle depth The cycle life is as follows: In the formula, The number of charge-discharge cycles (CDRs) required for a lithium battery to reach 100% depth of discharge and reach the end of its lifespan. These are the fitting parameters; The equivalent number of cycles at 100% depth of discharge is: When energy storage changes from a discharging state to a charging state, the cyclic variable for: set up If the accumulated battery-side discharge energy is the current continuous discharge phase, then during the discharge phase: When a cyclic event occurs, the accumulated discharge energy from the previous moment is extracted: Therefore, the loop depth for: Determine the energy storage cycle aging cost during the operating cycle. for: 。

Images