Electric vehicle active power distribution network operation method driven by multi-scene dynamic pricing mechanism

Abstract

The invention discloses an electric vehicle active power distribution network operation method driven by a multi-scene dynamic pricing mechanism, and the method comprises the steps: carrying out the double-layer optimization of the scheduling between an active power distribution network operator and an electric vehicle aggregator, and building a double-layer-four-stage random robust model based on the double-layer optimization, and carrying out iterative solution on the double-layer-four-stage model through a CCC-AIS coupled dichotomy, and outputting a strategy according to a condition for terminating iteration of the double-layer-four-stage model, and outputting the strategy as an operation method. According to the multi-scene dynamic electricity price mechanism provided by the invention, the EV can be effectively guided to be charged in order according to the charging and discharging characteristics of different scenes, and actual requirements are better met. Compared with a traditional fixed TOU electricity price and disordered charging mode, the new energy consumption level is improved, the load fluctuation amplitude of the ADN is effectively reduced, and the economical efficiency of the ADN is further improved. The model combines the advantages of RO and SO, and successfully balances the robustness and economy of the system.

Description

[0001] Methodology

[0002] The present invention relates to the technical field of electric vehicle active distribution network operation, and in particular to an electric vehicle active distribution network operation method driven by a multi-scenario dynamic pricing mechanism.

[0003] Background Methods

[0004] Electric vehicles (EVs) are currently trending towards replacing fuel-powered vehicles, leading to the multi-dimensional application and development of EVs as a hot research topic. The unordered integration of large-scale EVs into the power grid will further widen the peak-to-valley load gap and increase grid load pressure. As a new type of load, large-scale EVs offer significant potential for participating in the reserve market and achieving peak-to-valley load shifting. However, uncertainty surrounding EV users' willingness to participate in aggregator regulation makes it difficult to accurately quantify the reserve capacity of EV clusters, hindering aggregators' optimal market participation decisions. Summary of the Invention

[0005] The purpose of the present invention is to provide an electric vehicle active distribution network operation method driven by a multi-scenario dynamic pricing mechanism to solve the problems raised in the above background methods.

[0006] To achieve the above object, the present invention provides the following method scheme:

[0007] The electric vehicle active distribution network operation method driven by a multi-scenario dynamic pricing mechanism includes:

[0008] A two-layer optimization is performed on the scheduling between active distribution network operators and electric vehicle aggregators. A two-layer-four-level stochastic robust model is established based on the two-layer optimization. The two-layer-four-level model is iteratively solved through the bisection method of coupled CCG-AIS. The strategy is output according to the conditions for terminating the iteration of the two-layer-four-level model, and the output strategy is used as the operation method.

[0009] Furthermore, the two-tier optimization includes dividing electric vehicle charging into reservation, real-time and emergency charging systems based on active distribution network losses, load and renewable energy matching, and using a multi-scenario dynamic pricing mechanism, guiding electric vehicles to participate in active distribution network optimization through differentiated pricing strategies.

[0010] Furthermore, the active distribution network operator is responsible for the operation and settlement of the market, and the electric vehicle aggregator is the integration of all charging piles in the active distribution network in a specified area, responsible for the charging and discharging scheduling of electric vehicles in the area;

[0011] Electric vehicle aggregators voluntarily sign contracts with electric vehicle users who are willing to participate in demand response, obtaining control over the charging and discharging of electric vehicle users during the time period when the electric vehicle users are connected to the charging piles. The electric vehicle aggregators will centrally dispatch the charging and discharging status of the electric vehicles.

[0012] The EV aggregator formulates a charging and discharging plan based on the initial power consumption, expected off-grid time, and required power consumption information uploaded by EV users through the data collection system, combined with the dispatch instructions of the active distribution network operator, and feeds it back to the active distribution network operator;

[0013] Active distribution network operators implement multi-scenario dynamic pricing strategies for different electric vehicle operation scenarios based on the current grid losses, the matching status of renewable energy and load, and the charging and discharging data provided by electric vehicle aggregators, and clarify the charging and discharging prices for electric vehicle users at different times.

[0014] Furthermore, the constraints of the active distribution network include power flow balance constraints, branch current constraints, node voltage constraints, and line current carrying capacity constraints. The specific expressions are as follows:

[0015]

[0016] (V i min ) 2 ≤U i,t ≤(V i max ) 2

[0017]

[0018] Where, P j,t , Q j,t and U j,t are the active power, reactive power and square of node voltage injected into node j at time t, Ω n is the set of branch nodes, r ij 、x ij are the resistance and reactance of branch ij, L ij is the square of the current in branch ij, is the maximum value of the branch ij current, and is the maximum and minimum value of the voltage at node i, P ij,t With Q ij,t is the active power and reactive power flowing from node i to node j in branch ij at time t, is the maximum apparent power allowed to flow through branch ij, P jk,t , Q jk,t They are respectively the active power and reactive power injected into branch jk at time t, Ui,t is the node voltage of node i at time t, L ij,t is the square of the current in branch ij at time t.

[0019] Furthermore, the electricity consumption behavior of electric vehicles is described by the initial charging period probability distribution model and the daily mileage probability model.

[0020] Furthermore, the probability distribution model of the initial charging period is expressed as follows:

[0021]

[0022] f EV (t) is the probability distribution model of the initial charging period, μ1 is the mean of the arrival and departure time of electric vehicles, and σ1 is the standard deviation of the arrival and departure time of electric vehicles.

[0023] Furthermore, the probability model of daily mileage of electric vehicles is considered to obey the normal distribution f M (M d ), its mathematical expression is:

[0024]

[0025] M d is the daily mileage of electric vehicles, σ M and μ M M d the standard deviation and mean of

[0026] Introducing electric vehicle charging time to describe the electric vehicle charging time T ch , the expression is:

[0027]

[0028] E d,100 is the power demand of an electric vehicle when it travels 100 kilometers, B c is the battery capacity of the electric vehicle, and is the rated power and charging efficiency of the electric vehicle, is the battery capacity of the electric vehicle;

[0029] The probability density function of electric vehicles is sampled using the Monte Carlo method to obtain the starting charging time and charging duration, and calculate its daily basic charging load.

[0030]

[0031] t P This is the time when electric vehicles start charging.

[0032] Furthermore, demand response is divided into interruptible load and time-shiftable load. The specific expressions are as follows:

[0033]

[0034]

[0035] For interruptible loads, is the proportion of interruptible load, is the total electrical load, is a time-shiftable load, is the maximum value of the time-shiftable load, is the minimum value of the time-shiftable load, t represents time t, and T is 24h.

[0036] Furthermore, renewable energy includes wind turbines and photovoltaic power generation. The power output of wind turbines and photovoltaic power generation is described by probability density functions, which are expressed as follows:

[0037]

[0038] where h = v N / v ci -1

[0039] The scale parameter C of the first Weibull distribution, the charging price c, the scale parameter K of the second Weibull distribution, V ci is the cut-in wind speed of the wind turbine, h is a constant, v N is the rated wind speed of the wind turbine, P N is the rated installed capacity of the wind turbine, the first shape parameter α, the second shape parameter β, P PV is the photovoltaic power generation power, is the maximum value of photovoltaic power generation, P WT is the wind turbine power, f FG (P WT ),f GG (P PV ) are the probability density functions of wind turbines and photovoltaic power generation, respectively.

[0040] Furthermore, a two-layer-four-level stochastic robust model is established based on two-layer optimization, specifically including the upper-layer active distribution network operator and the lower-layer electric vehicle aggregator. The upper-layer active distribution network operator adopts a three-level max-max-min model, and the lower-layer electric vehicle aggregator adopts a single-level min model. The multi-scenario dynamic pricing mechanism connects the three-level max-max-min model and the single-level min model respectively.

[0041] Furthermore, the multi-scenario dynamic pricing mechanism actively adjusts price differences through the distribution network to attract or restrict the charging and discharging behavior of electric vehicles. The specific expression is as follows:

[0042]

[0043] Where, The price of charging an electric car, is the electricity price for electric vehicles, The valley electricity price is the time-of-use electricity purchase price. The electricity price is based on time of use. is the time-of-use electricity price, and the network loss coefficient μ EV , P loss,t,i→j is the branch at time t ij The network loss between loss,t,j→k is the network loss between branches jk at time t, is the equivalent load, is the equivalent total power of wind and solar power, c t As the basic charging electricity price, is the total electrical load, is a time-shiftable load, are the charging and discharging power of electric vehicles, Interruptible load.

[0044] Furthermore, the first level of the three-level max-max-min model finds the probability distribution of the uncertainty variable, the second level max identifies the output of the uncertainty variable, and the third level min optimizes the operating cost. The three-level max-max-min model also includes the transaction cost between the active distribution network and the main network. Network loss cost Demand response costs and penalty costs for curtailing wind and solar power The specific expression is as follows:

[0045]

[0046] P ss =(P t WT ,P t PV )

[0047] Scenario probability σ s ∈Ω σ ,Ω σ is the set of scenario probabilities, the uncertainty variable P of renewable energy ss , s is the running scenario, N s It's a cut scene. is a set of uncertain variables, P tbuy_grid 、P t sell_grid Distribution refers to the amount of electricity purchased and sold between the active distribution network and the upper power grid at time t. They are the power purchase and sales prices of the active distribution network, and the network loss coefficient μ loss ,Ω line is the set of active distribution network branches, P i PV,SJ is the actual photovoltaic power generation output, P WT,SJ is the actual wind turbine output, μ RES is the penalty coefficient for curtailment of renewable energy, x is the first-stage decision variable, μ DR is the demand response coefficient, node i, I ij,t is the maximum value of the branch ij current at time t, r ij is the resistance of branch ij, Ω n is the set of branch nodes, P WT is the wind turbine power, P PV is the photovoltaic power generation power, P t WT 、P t PV They represent the wind turbine power and photovoltaic power generation at time t respectively.

[0048] Furthermore, a comprehensive norm constraint consisting of 1-norm and ∞-norm conditions is used to constrain the scene probability. The specific expression is as follows:

[0049]

[0050] is the initial probability, θ1 is the 1-norm confidence, θ ∞ is the infinite norm confidence, α1 is the 1 norm confidence parameter, α ∞ is the infinity norm confidence parameter, N AL It is the historical data before the reduction;

[0051] Uncertain variable P of renewable energy ss Using polyhedron uncertainty modeling, the expression is as follows:

[0052]

[0053] P t ss,exp is the expected value of the uncertain variable, P t ss+ 、P t ss- are the positive and negative deviations of the uncertainty variable, is a 0-1 variable with positive deviation, is a 0-1 variable with negative bias, Γ ss is the confidence level of the ∞-norm.

[0054] Furthermore, the single-level min model is shown as follows:

[0055]

[0056] are the charging and discharging powers of the electric vehicles of the g-th electric vehicle aggregator, G is the set of g, and T is 24h.

[0057] Furthermore, the bisection method of coupling CCG-AIS is used to iteratively solve the two-layer-four-level model. The output strategy for terminating the iteration condition of the two-layer-four-level model includes:

[0058] Identify independent optimization variables;

[0059] By fixing the independent optimization variables, a three-level max-max-min model is used as the original optimization problem;

[0060] Decouple the original optimization problem into two sub-problems, and then solve each sub-problem alternately;

[0061] When the difference between the objective function values ​​of the two subproblems is less than the set error range, the original optimization problem converges. The three-level max-max-min model takes the dynamic electricity price set by the active distribution network as the convergence target and converges through the bisection method;

[0062] The bisection method converges to determine whether it satisfies the If the convergence condition is not satisfied, then according to the constraint Resolve;

[0063] in are the electric vehicle charging prices for the λ-2, λ-1, λ and λ+1 iterations respectively, and ε is the set tolerance.

[0064] To achieve the above object, the present invention also provides the following method scheme:

[0065] The electric vehicle active distribution network operation equipment driven by a multi-scenario dynamic pricing mechanism includes:

[0066] The optimization unit and the solution unit are used. The optimization unit is used to perform two-level optimization on the scheduling between the active distribution network operator and the electric vehicle aggregator. The solution unit is used to establish a two-level-four-level stochastic robust model based on the two-level optimization, and iteratively solve the two-level-four-level model through the bisection method of coupling CCG-AIS. The strategy is output according to the conditions for terminating the iteration of the two-level-four-level model, and the output strategy is used as the operation method.

[0067] To achieve the above object, the present invention also provides the following method scheme:

[0068] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0069] To achieve the above object, the present invention also provides the following method scheme:

[0070] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any one of the methods described above.

[0071] Compared with the existing method, the present invention has the following beneficial effects:

[0072] This paper proposes a two-layer four-level collaborative optimization scheduling strategy based on a multi-scenario dynamic pricing mechanism, which is applied to the active distribution network (ADN) containing electric vehicles. First, a two-layer optimization model between ADNO and EVA is constructed, and a multi-scenario dynamic pricing mechanism is used as the key bridge connecting the two-layer model. Through differentiated pricing strategies, EVs are guided to participate in the optimized operation of ADN. Subsequently, considering the impact of renewable energy output fluctuations on trading strategies, a four-level stochastic robust model in the form of max-max-min-min is further proposed based on the two-layer model. Finally, the two-layer four-level model is iteratively solved using the dichotomy method of coupled CCG-AIS, and the following conclusions are drawn:

[0073] (1) The proposed multi-scenario dynamic electricity price mechanism can effectively guide EVs to charge in an orderly manner according to the charging and discharging characteristics of different scenarios, which is more in line with actual needs. Compared with the traditional fixed TOU electricity price and disorderly charging mode, this mechanism not only improves the level of new energy consumption, but also effectively reduces the load fluctuation amplitude of ADN, further improving the economic efficiency of ADN.

[0074] (2) The adopted stochastic robust optimization model fully considers the uncertainty of renewable energy faced by ADN during operation. The model combines the advantages of RO and SO, taking into account the range of uncertain variables and utilizing limited probability distribution data, successfully balancing the robustness and economy of the system.

[0075] (3) The CCG-AIS coupled bisection method successfully solved the complex two-layer four-level model in 31.9 seconds. Compared with traditional heuristic algorithms, this algorithm not only significantly reduces the number of iterations and solution time, but also converges to the optimal solution more quickly, greatly improving computational efficiency. In addition, even if the EVA scale is further expanded, the algorithm still shows excellent convergence and computational efficiency when solving the two-layer four-level model. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a framework diagram of the active power distribution network containing electric vehicles of the present invention.

[0077] Figure 2 This is the solution process of the double-layer-four-level model of the active distribution network containing electric vehicles of the present invention.

[0078] Figure 3 This is the improved IEEE33 node structure diagram of the present invention.

[0079] Figure 4 This is a wind power output scenario diagram of the present invention.

[0080] Figure 5 This is a photovoltaic output scene diagram of the present invention.

[0081] Figure 6 This is the EV charging demand diagram under disordered state of the present invention.

[0082] Figure 7 This is the TOU (time-of-use) electricity price diagram of the present invention.

[0083] Figure 8 This is a diagram of the dichotomy iterative process of the present invention.

[0084] Figure 9 This is a real-time electricity price chart for each EVA of the present invention.

[0085] Figure 10 This is the EVA1 reservation scheduling plan diagram of the present invention.

[0086] Figure 11 This is the EVA1 real-time scheduling plan diagram of the present invention.

[0087] Figure 12 This is the EVA1 emergency dispatch plan diagram of the present invention.

[0088] Figure 13 This is a comparison diagram of the total load power of the ADN of the present invention.

[0089] Figure 14 This is a diagram showing the PV power distribution of the renewable energy output of the present invention.

[0090] Figure 15 This is a diagram of the WT power distribution of the renewable energy output of the present invention.

[0091] Figure 16 This is a performance diagram of the different number of EVA access algorithms of the present invention.

[0092] Figure 17 This is a diagram of the equipment model of the method of the present invention.

[0093] Figure 18 This is a diagram of the internal structure of the computer device of the present invention. DETAILED DESCRIPTION

[0094] The following will clearly and completely describe the method scheme in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary method personnel in this field without making creative efforts are within the scope of protection of the present invention.

[0095] See also Figures 1 to 18 , the present invention provides a method scheme:

[0096] To address the challenges posed by renewable energy uncertainty and disorderly charging of electric vehicles (EVs) to the safe and economic operation of active distribution networks (ADNs), a two-tiered, four-tiered collaborative optimization scheduling strategy for active distribution networks (ADNs) with EVs, based on a multi-scenario dynamic pricing mechanism, was proposed. First, a two-tiered optimization model was constructed between active distribution network operators (ADNOs) and electric vehicle aggregators (EVAs). Based on the distribution network losses and the matching degree between load and renewable energy, a multi-scenario dynamic pricing mechanism was designed. EV charging was categorized into scheduled, real-time, and emergency charging modes, guiding EV participation in ADN optimization through differentiated pricing strategies. Subsequently, considering the impact of renewable energy output fluctuations on the trading strategy, a two-tiered, four-tiered stochastic robust model with the max-max-min-min formula was established based on the aforementioned model to effectively balance the economic and robustness of the ADN. Finally, the two-tiered, four-tiered model was iteratively solved using a bisection method coupled with CCG-AIS. The effectiveness of the model and algorithm was verified on a modified IEEE 33-bus system.

[0097] 1 ADN framework and modeling including electric vehicles

[0098] 1.1 ADN Framework

[0099] The proposed framework of active distribution network with electric vehicles is as follows Figure 1As shown in the figure, this framework establishes a complete market structure by introducing key roles, including active distribution network operators (ADNOs) and electric vehicle aggregators (EVAs). Within this structure, ADNOs are responsible for market operation and settlement, while EVAs, as the aggregator of all charging stations within a specific active distribution network (ADN), are responsible for scheduling EV charging and discharging within the region. Specifically, EVAs voluntarily sign contracts with EV users willing to participate in demand response, obtaining control over charging and discharging during the time period when the user is connected to a charging station. The EVAs then centrally schedule the EV's charging and discharging status. Based on information uploaded by EV users via a data acquisition system, such as initial power consumption, desired off-grid time, and required power consumption, and combined with ADNO dispatch instructions, the EVAs formulate corresponding charging and discharging plans and provide feedback to the ADNO. Subsequently, based on current grid losses, the matching of renewable energy and load, and charging and discharging data provided by the EVAs, ADNOs implement multi-scenario dynamic pricing strategies for different EV operating scenarios, defining charging and discharging prices for EV users at different times. This multi-scenario dynamic pricing mechanism not only ensures a dynamic balance of grid load but also effectively manages EV charging and discharging demand, ensuring the rationality and flexibility of charging pricing strategies.

[0100] 1.2 ADN Modeling

[0101] ADN network constraints include power flow balance constraints, branch current constraints, node voltage constraints, and line current carrying capacity constraints, which are specifically expressed as:

[0102]

[0103] Where, P j,t , Q j,t and U j,t It is the active power, reactive power and square of node voltage injected into node j (at time t). n is the set of branch nodes. ij 、x ij is the resistance and reactance of branch ij. ij is the square of the current in branch ij. is the maximum value of the branch current ij. and is the maximum and minimum voltage of node i. ij,t With Q ij,t are the active power and reactive power flowing from node i to node j in branch ij, respectively. is the maximum apparent power allowed to flow through line ij.

[0104] 1.3EV Individual Modeling

[0105] For private cars, their charging behavior is often influenced by the owner's travel habits and lacks obvious regularity. Therefore, their electricity consumption behavior can be described by a probability distribution model of the initial charging period and a probability model of the daily mileage.

[0106] The probability distribution of the initial charging period can be expressed as follows:

[0107]

[0108] μ1 is the mean of EV arrival and departure times, and σ1 is the standard deviation of EV arrival and departure times.

[0109] The daily charging load demand of electric vehicles is closely related to their daily mileage and charging duration. In general, the daily mileage of electric vehicles is considered to follow a normal distribution, and its mathematical expression is:

[0110]

[0111] M d is the daily mileage of electric vehicles, σ M and μ M M d The standard deviation and mean of .

[0112] The electric vehicle charging time is introduced to describe the charging time of electric vehicles. The expression is:

[0113]

[0114] E d,100 is the power demand of an electric vehicle when it travels 100 kilometers, B c is the battery capacity of the electric vehicle. ch Charging time for electric vehicles, and The rated power and charging efficiency of electric vehicles; is the battery capacity of the electric vehicle.

[0115] The Monte Carlo method is used to sample the probability density function of electric vehicles to obtain the starting charging time and charging duration, thereby calculating its daily basic charging load:

[0116]

[0117] P t EV is the basic charging load of electric vehicles; P The starting time for charging electric vehicles.

[0118] 1.4 Demand Response Modeling

[0119] Demand response refers to the load adjusting its daily electricity consumption pattern according to changes in price or incentive signals, thereby achieving efficient allocation of power system resources. Demand response is divided into interruptible loads and time-shiftable loads. The specific expression is as follows:

[0120]

[0121] is the size of the interruptible load; is the proportion of interruptible load; is the total electrical load; It is the basic load and can be obtained directly; is the size of the time-shiftable load; is the maximum value of the time-shiftable load.

[0122] 1.5 Renewable Energy Generation Modeling

[0123] Since the power output of wind turbines (WT) and photovoltaic power generation (PV) is random, a probability density function (PDF) is usually used to describe its uncertainty. The power output of WT can be represented by a Weibull distribution, while the power output of PV follows a Beta distribution. Their expressions are:

[0124]

[0125]

[0126] where h = v N / v ci -1.

[0127] C and K represent the scale parameters of the first and second Weibull distributions respectively. v is the actual wind speed of the wind turbine (abbreviated as wind turbine), which is a known data. ci is the cut-in wind speed, v co To cut out the wind speed is known data, v N is the rated wind speed of the fan, P N is the rated installed capacity of the wind turbine. The first shape parameter α, the second shape parameter β; is the maximum value of photovoltaic power f FG (P WT ),f GG (P PV ) are the probability density functions of wind turbines and photovoltaics, respectively.

[0128] 2. ADN Scheduling Method Including Electric Vehicles

[0129] To coordinate the scheduling of electric vehicles and active distribution networks, a two-tier, four-tier coordinated optimization scheduling approach is proposed. The upper-tier ADNO employs a three-tier max-max-min model, while the lower-tier EVA uses a single-tier min model, both aiming to minimize operating costs. A multi-scenario dynamic pricing mechanism serves as a bridge between the two.

[0130] 2.1 Multi-scenario dynamic pricing mechanism

[0131] To better meet the diverse charging and discharging needs of EVAs, these scenarios are categorized into three types: scheduled, real-time, and emergency. ADNO, incorporating EVA charging and discharging information, sets differentiated dynamic electricity prices for EVAs based on these scenarios, guiding them to optimize charging and discharging operations. This strategy not only helps EVAs plan their charging and discharging appropriately but also encourages them to actively participate in ADN's coordinated scheduling, fully leveraging EVAs' potential in flexible grid regulation.

[0132] 2.1.1 Dynamic Pricing Mechanism Model

[0133] The dynamic pricing mechanism is the core component of the two-tier, four-level model. With the widespread adoption of EVs, load fluctuations in the distribution network have increased significantly. Traditional time-of-use (TOU) electricity prices struggle to accurately reflect actual load conditions and effectively guide EV charging and discharging. To this end, a more reasonable charging pricing mechanism has been designed, combining the ADN network losses and the matching degree between load and renewable energy output. ADN adjusts price differences to attract or restrict EV charging and discharging behavior, thereby optimizing the operation of the distribution network. The specific expression is as follows:

[0134]

[0135] Where, The (known or final) price of EV charging; is the (known or final) EV electricity selling price; The valley electricity price is the time-of-use electricity purchase price; It is the time-of-use electricity purchase price; is the time-of-use electricity price; network loss coefficient μ EV ;P loss,t,i→j is the network loss between branches ij; is the equivalent load; is the equivalent total power of wind and solar power; c t As the basic charging electricity price; are the EV charging and discharging power respectively.

[0136] Through a (multi-scenario) dynamic pricing mechanism, ADNO provides EVA with price discounts, allowing EVA to adjust its EV charging and discharging needs during periods with lower electricity prices. The specific pricing process is as follows:

[0137] 1) Step 1. Determine the relationship between the ADN equivalent load distribution and the value of renewable energy generation, and formulate a unified ADN power purchase price based on the time-of-use electricity price;

[0138] 2) Step 2. Considering the differences in branch network losses when EVA is connected to different nodes, the network loss factor is converted into a price coefficient and included in the unified electricity price established in step 1. At this time, the entire microgrid presents differentiated electricity purchase prices;

[0139] 3) Step 3: The upper-layer ADN and the lower-layer EVA iteratively update the dynamic electricity price.

[0140] 2.1.2 Multi-scenario dynamic pricing mechanism model

[0141] a) Scheduled charging and discharging mechanism

[0142] The reservation charging and discharging mechanism is designed to meet the charging needs of EV users with foresight and planning. Based on the reservation information of EVA and the load forecast information of ADN, ADNO will publish the day-ahead price to all EV users in advance through a dynamic pricing mechanism and sign a day-ahead transaction agreement with EVA. In this scenario, the basic charging electricity price c t Lower, EV users can plan charging time and charging amount in advance and adjust their charging plans appropriately according to price signals to enjoy more favorable prices.

[0143] b) Real-time charging and discharging mechanism

[0144] The real-time charging and discharging mechanism is suitable for EV users to charge immediately without making reservations in advance. In this scenario, the basic charging electricity price c t This mechanism enables EV users to flexibly adjust their charging behavior according to dynamic electricity prices and actively participate in the optimized scheduling of ADN, thereby improving the overall flexibility and operational efficiency of the system.

[0145] c) Emergency charging mechanism

[0146] The emergency charging mechanism is designed to deal with emergencies, such as when the EV battery is about to run out. Under this mechanism, the basic charging price c t When set to the highest value, EVs are charged with high priority, skipping the normal process. This ensures that EV users have sufficient power at critical moments and avoid travel interruptions due to insufficient power. This mechanism is mainly aimed at EV users who only focus on electricity demand and do not care about charging costs. Therefore, these users generally do not adjust their charging behavior due to fluctuations in electricity prices.

[0147] 2.2 ADNO model

[0148] 2.2.1 Utility Function

[0149] During ADNO's operation, the scheduling process is modeled as a three-level max-max-min stochastic robust structure, taking into account the uncertainty and probabilistic characteristics of the output of renewable energy (such as wind power and photovoltaics). The first level max is used to find the most unfavorable probability distribution of the uncertainty variable, the second level max is used to identify the most unfavorable output of the uncertainty variable, and the third level min is used to minimize the operating costs, including the transaction costs between ADN and the main network. Network loss cost Demand response costs and penalty costs for curtailing wind and solar power The specific expression is as follows:

[0150]

[0151]

[0152] Scenario probability σ s ∈Ω σ Theoretically, it can be within any range. However, to ensure the economy and feasibility of the scheduling scheme, we use a comprehensive norm constraint consisting of a 1-norm condition and an ∞-norm condition to constrain the probability of each scenario:

[0153]

[0154] Among them, θ1 is the 1-norm confidence, θ ∞ is the infinite norm confidence, α1 is the 1 norm confidence parameter, α ∞ is the infinity norm confidence parameter, N AL It is the historical data of the scene after reduction;

[0155] In addition, the uncertainty variable P of renewable energy (wind and solar) ss Using polyhedron uncertainty modeling, the expression is as follows:

[0156]

[0157] Among them, P t ss,exp is the expected value P of the uncertain variable t ss+ 、P t ss- are the positive and negative deviations of the uncertainty variable are 0-1 variables with positive and negative deviations, respectively.

[0158] 2.2.2 Strategy Space (for the EVA Model in 2.2.1)

[0159] 1) Power constraints between ADN and upper-level power grid

[0160]

[0161] Where, It is the maximum amount of electricity purchased and sold between ADN and the upper-level power grid.

[0162] 2) SVC operation constraints

[0163]

[0164] Where, Reactive power injected by node i of SVC. The maximum reactive power injected into node i. SVC is a reactive power compensation device: it is indispensable in power systems. Its main function is to improve the power factor of the power supply and distribution system, thereby increasing the utilization rate of transmission and substation equipment, improving power efficiency, and reducing power costs.

[0165] 3) Renewable energy constraints

[0166]

[0167] are the maximum values ​​of renewable energy curtailment.

[0168] 4) Node power balance constraints

[0169]

[0170] Pi ,t Q i,t are the active power and reactive power injected by node i at time t, is the equivalent reactive load.

[0171] 2.3EVA model

[0172] 2.3.1 Utility Function

[0173] The EVA model takes minimizing the overall operating cost as its objective function, including the cost of electricity purchase and sale from EVA to AND, which is expressed as follows:

[0174]

[0175] 2.3.2 Strategy Space (for the EVA Model in 2.3.1)

[0176] 1) EV individual constraints

[0177] The SOC of an EV individual satisfies the following constraints:

[0178]

[0179] The individual charging and discharging of EVs meets the following constraints:

[0180]

[0181] 2) EVA Constraints

[0182] To construct an EVA dispatchability model, a Minkowski addition algorithm was used to combine the charge and discharge power boundaries of multiple EVs with the battery safety capacity boundaries to derive the upper and lower bounds of the EVA dispatchability. This method effectively reduces the dimensionality of the decision variables and simplifies the calculation process.

[0183]

[0184] are the maximum charging power and discharge power of electric vehicles, is the maximum capacity of a single electric vehicle, is the maximum value of the EV aggregator capacity, a binary variable K1 is the set of k nodes, N is the set of n nodes,

[0185] Considering the step power change that may be caused by multiple EVs entering or leaving the grid in an EVA, a method for correcting the dynamic energy storage power of the cluster is proposed. It can be calculated by the following formula:

[0186]

[0187] n is the number of electric vehicles, is the amount of electricity left by the electric car, is the amount of electricity reached by the electric vehicle.

[0188] In summary, based on Minkowski addition, the dynamic energy balance capability of multiple EVs when participating in ADN scheduling is defined as the dispatchable capability of EVA. Its model is shown in the following formula:

[0189]

[0190] η d is the discharge efficiency, η c It is the charging efficiency.

[0191] 3 Solutions

[0192] The CCG-AIS nested bisection method is used to solve the two-layer optimization model of ADN with electric vehicles. Since the upper-layer ADN problem has a complex three-level max-max-min structure and is difficult to solve directly, the CCG-AIS algorithm is introduced for conversion to simplify the solution. In addition, in this two-layer model, the dynamic electricity price set by the ADN will fluctuate with the coordinated scheduling plan of EVA and ADN. Traditional heuristic algorithms (such as genetic algorithms and ant colony algorithms) have a limited convergence range of electricity prices, resulting in slow solution speed and difficulty in ensuring convergence. To this end, a bisection method is used to track electricity price changes. When the electricity price converges, it indicates that the upper and lower layer scheduling has achieved collaborative optimization.

[0193] 3.1CCG-AIS Algorithm

[0194] The max-max-min model of ADNO in the upper model can be simplified as follows:

[0195]

[0196] This shorthand form is a three-level optimization problem, where the first constraint contains , the second constraint contains , and the third constraint contains . It can ensure the minimization of ADN costs under the worst probability distribution and the most unfavorable renewable energy output conditions. To solve such problems, the CCG-AIS method can be used: first, determine the independent optimization variables; then, by fixing different independent optimization variables, decouple the original optimization problem into multiple simplified sub-problems; finally, solve each sub-problem alternately. Where σ s and ψ are two sets of independent optimization variables, by fixing σ s and ψ, decoupling the original problem into σ s -fixed problem (SPS1) and ψ-fixed problem (SPS2). Ψ1 is the set of independent optimizations of ψ. The constant coefficient matrix b T , I, K1, J, H, Gx are constant matrices multiplied by the first stage variables, and f(ψ) is a function of the independent optimization variables.

[0197] a)SPS1

[0198]

[0199] (σ s ) * is a known probability value.

[0200] By applying KKT conditions and big-M method, the max-min problem is combined into a single-layer model as follows:

[0201]

[0202] δ is the Lagrange coefficient. The transformed model can be directly solved by the solver to obtain the optimal solution. A large constant M. Auxiliary variable ν1. Gy is a constant matrix multiplied by the second stage variables.

[0203] b)SPS2

[0204]

[0205] Despite the existence of σ s The nonlinear term of the product with x, but due to σ s Since x and x are located in two independent layers, there is no need to convert them through KKT conditions. The SP problem can be decomposed into S+1 linear programming problems.

[0206] The sth planning SP problem is as follows:

[0207]

[0208] stGx≤f(ψ)(53)

[0209] is the optimal value for each scenario. Then, Substitute into the following model to obtain the worst-case scenario probability (distribution value) σ s and the optimal solution to the SP problem

[0210]

[0211] When the difference between the objective function values ​​of the two subproblems (i.e., SPS1 and SPS2) is less than a certain error range, it can be considered that convergence has been achieved, as shown in the following formula (55).

[0212]

[0213] ξ is the Lagrange multiplier.

[0214] 3.2 Bisection Solution

[0215] After CCG-AIS conversion, both the upper-layer ADN model and the lower-layer EVA model can be solved iteratively using the GUROBI solver. The model uses the dynamic electricity price formulated by ADN as the convergence target and accelerates convergence through the bisection method. Specifically, the bisection method determines whether the convergence condition of formula (56) is met in each iteration. If not, it is re-solved according to constraint (57). By halving the upper and lower boundary intervals, the search range is narrowed and the algorithm efficiency is improved. In summary, the solution flow chart of the two-layer-four-level model of the active distribution network containing electric vehicles is as follows: Figure 2 shown.

[0216]

[0217] ε is the set tolerance.

[0218] 4 Case Analysis

[0219] The simulation is conducted using an IEEE 33-node distribution network with a base capacity of 100MW and a base voltage of 12.66kV. Bus 0 is set as the root node and the voltage value is 1.00 (unit per unit). To ensure the stability of the voltage within the ADN, the voltage range is set between 0.9 and 1.1 (unit per unit). Assume that three EVAs, SVCs, and renewable energy sources (wind and solar) are connected to the ADN. The specific access nodes are as follows: Figure 3 As shown in the figure, the reactive power compensation capacity of the SVC is set to -500 to 500 kVar. The simulation time span is set from 00:00 to 24:00, with an interval of 1 hour. Furthermore, the number of EVs participating in energy management in the ADN is set to 80. Each EV has a battery capacity of 30 kWh, a maximum charging power of 10 kW, and a charging efficiency of 0.88. The Monte Carlo method is used to sample the relevant probability distributions one by one to obtain the SOC data for each EV at the time of arrival, time of leaving the grid, and time of joining the grid. Figures 4 to 6 , shows the total charging demand of each EVA in the disordered state, and allocates the total demand to multiple scenarios such as appointment system, real-time system and emergency system in a ratio of 7:2:1. The key operating parameters of ADN are shown in Table 1. The time-of-use electricity price (TOU) is shown in Figure 7 shown.

[0220] All simulations were performed in Matlab R2018b, using the GUROBI solver for the two-layer, four-level model. The computer configuration was an Intel Core i7 processor with a main frequency of 1.8 GHz and 16 GB of memory.

[0221] Table 1 Key operating parameters of the system

[0222]

[0223] 4.1 Iterative Analysis

[0224] 4.1.1 Iterative Process Analysis

[0225] To highlight the advantages of the proposed bisection method in solving the two-layer model, a comparative analysis was conducted with the genetic algorithm and the ant colony algorithm. The three methods all used the same input data, and the results are shown in Table 2. Compared with the genetic algorithm, the bisection method reduced the iteration time by 89.47% and the number of iterations by 91.68%; compared with the ant colony algorithm, the iteration time was reduced by 92.29% and the number of iterations was reduced by 97.75%. In addition, the bisection method achieved the lowest cost for both ADNO and EVA, outperforming the other two algorithms. Figure 8The bisection method demonstrates its convergence process, showing that the system gradually stabilizes after a limited number of oscillations. At this point, the embedded C&CG-AIS algorithm averaged 2.75 iterations. Analysis results show that the bisection method not only significantly reduces the number of iterations and time, but also converges to the optimal solution more quickly, significantly improving efficiency and saving computational resources and time, demonstrating its clear advantages in solving two-layer models.

[0226] Table 2 Comparison of solution results of different optimization algorithms

[0227]

[0228] 4.1.2 Iteration Results Analysis

[0229] The EVA electricity price optimization results of the proposed two-layer-four-level model in multiple scenarios are as follows: Figure 9 The figure only shows the changing trend of real-time charging electricity prices. Prices for other scenarios are adjusted upward and downward based on this trend and are therefore not shown separately. The results show that EVA2, connected to ADN node 20, has a higher overall electricity price, while EVA1, connected to ADN node 17, has a lower overall price. This is primarily because EVA1 and WT are connected to the same node, and the WT's output changes match the node's load demand trends, resulting in lower grid losses and greater EV charging and discharging margin, which in turn reduces its charging price. In contrast, EVA2 has no power supply nearby and a higher node load, resulting in greater grid losses, limited EV charging and discharging space, and a relatively higher electricity price. Analysis shows that EVA charging prices are influenced by factors such as grid losses, node load, and renewable energy matching. The proposed multi-scenario dynamic pricing mechanism effectively accounts for these factors, rationally guiding coordinated optimization between upper and lower levels, and achieving a win-win situation for all parties.

[0230] also, Figures 10 to 12The figure shows the charge and discharge plans of EVA1 under multiple scenarios. As can be seen in the figure, in scheduled scheduling, EVA is able to fully utilize low-price periods (such as 3:00-5:00, 17:00, and midnight) for large-scale charging, as charging demand is highest and charging prices are lowest. These periods typically coincide with peak wind power generation and low load periods, and EVA has ample dispatchable capacity, maximizing renewable energy consumption through EV charging. Furthermore, scheduled scheduling also involves large-scale discharge during periods with higher electricity prices (10:00 PM) to maximize economic benefits. EVA's overall strategy is to charge when prices are low and discharge when prices are high, thereby optimizing economic returns. In real-time scheduling, while charging prices are slightly higher, greater flexibility is offered. EVA charges during multiple periods, including 3:00, 9:00, 10:00, and 16:00, and discharges at 11:00. Real-time scheduling dynamically adjusts the charge and discharge strategy based on actual electricity price fluctuations and grid demand. While it avoids discharging during peak periods, it ensures maximum efficiency and maintains a balance between supply and demand by flexibly responding to changes in the electricity market. Emergency charging has the highest price and is used only in emergency situations to ensure a stable EV charge level. Charging is scheduled at 3:00, 9:00, and 16:00, minimizing charging during periods of high electricity prices and preventing discharge. This schedule prioritizes EV operational safety, mitigating costs and risks during peak periods. Overall, these three schedules offer distinct advantages in terms of economic efficiency, flexibility, and safety. Dynamic pricing strategies have been implemented for these schedules to better meet real-world needs.

[0231] 4.2 Analysis of the impact of EV charging optimization results on the distribution network side

[0232] To verify the superiority of the multi-scenario dynamic pricing mechanism, the following three scenarios were set up for comparison under the same optimization conditions:

[0233] Scenario 1: Disorderly charging (i.e., disorderly charging is adopted in the entire area in ADN).

[0234] Scenario 2: Fixed TOU electricity price (i.e., fixed time-of-use electricity price is used in the entire ADN area for EV charging optimization).

[0235] Scenario 3: Proposed multi-scenario dynamic pricing model.

[0236] Figure 13Comparisons of total ADN load power under three scenarios are presented. As can be seen from the figure, under the proposed multi-scenario dynamic electricity pricing model, orderly EV charging significantly reduces load fluctuations compared to disordered charging in Scenario 1, effectively shifting peak loads. Compared to orderly charging under a fixed time-of-use electricity price in Scenario 2, this model further reduces load fluctuations and demonstrates a greater effectiveness in guiding EV charging. Analysis results demonstrate that the proposed model can effectively reduce load fluctuations and mitigate distribution network losses caused by the concentrated access of EV charging loads.

[0237] Furthermore, Table 3 shows that, compared to Scenario 1 (disordered EV charging), Scenario 3 (multi-scenario dynamic pricing model) reduces grid loss costs and wind and solar curtailment costs by 10.84% ​​and 30.02%, respectively. This demonstrates that the proposed multi-scenario dynamic electricity pricing model, by fully accounting for ADN grid losses and the matching degree between load and renewable energy, can significantly reduce distribution network losses. Compared to Scenario 2 (fixed electricity price), grid loss costs are further reduced, and the wind and solar power integration rate is improved. This demonstrates that the model not only reduces grid loss costs over traditional pricing mechanisms but also increases the wind and solar power integration rate. This analysis demonstrates that the proposed multi-scenario dynamic electricity pricing mechanism effectively improves the economic performance of ADN and the wind and solar power integration rate.

[0238] Table 3 Comparison of optimization results in different scenarios

[0239] Scenario Network loss cost / CNY Penalty for curtailing wind and solar power / CNY Wind and solar power consumption rate / % Scenario 1 297.72 136.82 97.55 Scenario 2 271.66 108.05 98.12 Scenario 3 265.44 95.74 98.53％

[0240] 4.3 Comparative analysis of uncertainty

[0241] The power distribution of renewable energy output in ADN obtained by optimizing the two-layer-four-level model is as follows: Figures 14 and 15 As shown in the figure, in the worst-case scenario, actual renewable energy output is mostly lower than forecasted during periods of fluctuation. Therefore, the ADN must compensate for the difference between actual and forecasted output by discharging EVs or increasing power purchases from the upstream power grid. This result is consistent with the model's requirement for optimizing operating costs under the worst-case scenario. Analysis shows that the two-layer-four-level model enhances the system's ability to withstand uncertainty by increasing operating costs, thereby improving operational robustness.

[0242] Furthermore, the ADN max-max-min stochastic robust optimization structure was compared with other optimization methods, including deterministic optimization (CO), robust optimization (RO), and stochastic optimization (SO). The results are shown in Table 4. The total system cost is the sum of the operating costs of ADNO and EVA. As shown in Table 4, although deterministic optimization has the lowest cost, this does not necessarily mean it is the optimal solution. This is because deterministic optimization ignores uncertainty, which can lead to higher hidden costs in actual operation. Among uncertain optimization methods, SO offers the best economic performance, but it is difficult to accurately obtain the probability distribution of uncertain variables in engineering applications. RO only requires the range of uncertain variables and requires less data, but its economic performance is poor. The proposed optimization method combines the advantages of RO and SO, considering the range of uncertain variables while using limited probability distribution data, successfully balancing robustness and economic performance.

[0243] Table 4 Comparison of total system costs of optimization methods

[0244] Optimization methods CO RO Proposed SO Total system cost 3398.06 3695.86 3502.28 3471.11

[0245] 4.4 Expanding the Scale of the Case

[0246] To further verify the performance of the bisection method of the coupled CCG-AIS, the algorithm convergence and computation time were analyzed under different scales of EVA access (the number of electric vehicles in each EVA was the same, and only the parameters of each probability distribution were adjusted). The results are shown in Figure 2. Figure 16 As shown in the figure, as the EVA scale increases, even in the complex case of nine EVAs simultaneously connected to the ADN, the model converges in only 354.02 seconds, far less than the scheduling period, and with a large computational time margin. Analysis shows that even with further expansion of the EVA scale, the bisection method coupled with the CCG-AIS algorithm maintains good convergence and computational efficiency when solving the two-layer four-level model.

[0247] 5 Conclusion

[0248] A two-tier, four-level collaborative optimization scheduling strategy based on a multi-scenario dynamic pricing mechanism is proposed for active distribution networks (ADNs) containing electric vehicles. First, a two-tier optimization model between ADNO and EVA is constructed, and a multi-scenario dynamic pricing mechanism is used as a key bridge connecting the two-tier model. Through a differentiated pricing strategy, EVs are guided to participate in the optimized operation of the ADN. Subsequently, considering the impact of renewable energy output fluctuations on the trading strategy, a four-level stochastic robust model in the form of max-max-min-min is further proposed based on the two-tier model. Finally, the two-tier, four-level model is iteratively solved using the bisection method of coupled CCG-AIS, and the following conclusions are drawn:

[0249] (1) The proposed multi-scenario dynamic electricity price mechanism can effectively guide EVs to charge in an orderly manner according to the charging and discharging characteristics of different scenarios, which is more in line with actual needs. Compared with the traditional fixed TOU electricity price and disorderly charging mode, this mechanism not only improves the level of new energy consumption, but also effectively reduces the load fluctuation amplitude of ADN, further improving the economic efficiency of ADN.

[0250] (2) The adopted stochastic robust optimization model fully considers the uncertainty of renewable energy faced by ADN during operation. The model combines the advantages of RO and SO, taking into account the range of uncertain variables and utilizing limited probability distribution data, successfully balancing the robustness and economy of the system.

[0251] (3) The CCG-AIS coupled bisection method successfully solved the complex two-layer four-level model in 31.9 seconds. Compared with traditional heuristic algorithms, this algorithm not only significantly reduces the number of iterations and solution time, but also converges to the optimal solution more quickly, greatly improving computational efficiency. In addition, even if the EVA scale is further expanded, the algorithm still shows excellent convergence and computational efficiency when solving the two-layer four-level model.

[0252] like Figure 17 As shown, the electric vehicle active distribution network operation equipment driven by the multi-scenario dynamic pricing mechanism includes a building module and a scheduling module;

[0253] The optimization unit and the solution unit are used. The optimization unit is used to perform two-level optimization on the scheduling between the active distribution network operator and the electric vehicle aggregator. The solution unit is used to establish a two-level-four-level stochastic robust model based on the two-level optimization, and iteratively solve the two-level-four-level model through the bisection method of coupling CCG-AIS. The strategy is output according to the conditions for terminating the iteration of the two-level-four-level model, and the output strategy is used as the operation method.

[0254] like Figure 18 As shown, the proposed multi-scenario dynamic pricing mechanism-driven EV active power distribution network operation equipment can utilize various computing devices, including desktop computers, laptops, PDAs, and cloud servers. This equipment is composed of multiple components, including a processor and memory, but is not limited to these components and may include other components or different combinations. The device examples provided here are for illustrative purposes only and are not intended to be limiting.

[0255] The processor can be any type of device, such as a CPU, DSP, ASIC, or FPGA. It can also include programmable logic devices, discrete gates, transistor logic devices, and discrete hardware components. Memory is used to store computer programs and modules. The processor runs or executes the computer program modules stored in the memory and calls the data stored in the memory, thereby realizing the various functions of the active distribution network operation of electric vehicles driven by the multi-scenario dynamic pricing mechanism.

[0256] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory may include a high-speed random access memory and may also include a non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0257] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the method for operating an active distribution network for electric vehicles driven by a multi-scenario dynamic pricing mechanism.

[0258] If the electric vehicle active distribution network operation method driven by the multi-scenario dynamic pricing mechanism is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device.

[0259] Based on this understanding, the present invention implements all or part of the processes in the above-mentioned method for operating an active electric vehicle power distribution network driven by a multi-scenario dynamic pricing mechanism, and can also be completed by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-mentioned method for operating an active electric vehicle power distribution network driven by a multi-scenario dynamic pricing mechanism. The computer program includes computer program code, which can be in source code form, object code form, executable file, or preset intermediate form.

[0260] Computer-readable storage media may refer to any entity or device that can carry computer program code, including recording media, USB flash drives, mobile hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), software distribution media, etc.

[0261] It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable storage media do not include electrical carrier signals and telecommunication signals.

[0262] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An electric vehicle active distribution network operation method driven by a multi-scenario dynamic pricing mechanism, characterized in that: include: A two-layer optimization is performed on the scheduling between active distribution network operators and electric vehicle aggregators. A two-layer-four-level stochastic robust model is established based on the two-layer optimization. The two-layer-four-level model is iteratively solved through the bisection method of coupled CCG-AIS. The strategy is output according to the conditions for terminating the iteration of the two-layer-four-level model, and the output strategy is used as the operation method.

2. The method according to claim 1, wherein The two-tier optimization includes dividing electric vehicle charging into reservation, real-time and emergency charging systems based on active distribution network losses, load and renewable energy matching, and using a multi-scenario dynamic pricing mechanism. Through differentiated pricing strategies, electric vehicles are guided to participate in active distribution network optimization.

3. The method according to claim 2, wherein Active distribution network operators are responsible for market operation and settlement. Electric vehicle aggregators are the integration of all charging piles within the active distribution network in a specified area and are responsible for the charging and discharging scheduling of electric vehicles in the area. Electric vehicle aggregators voluntarily sign contracts with electric vehicle users who are willing to participate in demand response, obtaining control over the charging and discharging of electric vehicle users during the time period when the electric vehicle users are connected to the charging piles. The electric vehicle aggregators will centrally dispatch the charging and discharging status of the electric vehicles. The EV aggregator formulates a charging and discharging plan based on the initial power consumption, expected off-grid time, and required power consumption information uploaded by EV users through the data collection system, combined with the dispatch instructions of the active distribution network operator, and feeds it back to the active distribution network operator; Active distribution network operators implement multi-scenario dynamic pricing strategies for different electric vehicle operation scenarios based on the current grid losses, the matching status of renewable energy and load, and the charging and discharging data provided by electric vehicle aggregators, and clarify the charging and discharging prices for electric vehicle users at different times.

4. The method according to claim 3, wherein The constraints of the active distribution network include power flow balance constraints, branch current constraints, node voltage constraints, and line current carrying capacity constraints. The specific expressions are as follows: (V i min ) 2 ≤U i,t ≤(V i max ) 2 Where, P j,t , Q j,t and U j,t are the active power, reactive power and square of node voltage injected into node j at time t, Ω n is the set of branch nodes, r ij 、x ij are the resistance and reactance of branch ij, L ij is the square of the current in branch ij, is the maximum value of the branch ij current, V i max With V i min is the maximum and minimum value of the voltage at node i, P ij,t With Q ij,t is the active power and reactive power flowing from node i to node j in branch ij at time t, is the maximum apparent power allowed to flow through branch ij, P jk,t , Q jk,t They are respectively the active power and reactive power injected into branch jk at time t, U i,t is the node voltage of node i at time t, L ij,t is the square of the current in branch ij at time t.

5. The method according to claim 3, wherein The electricity consumption behavior of electric vehicles is described by the initial charging period probability distribution model and the daily mileage probability model.

6. The method according to claim 5, wherein The probability distribution model of the initial charging period is expressed as follows: f EV (t) is the probability distribution model of the initial charging period, μ1 is the mean of the arrival and departure time of electric vehicles, and σ1 is the standard deviation of the arrival and departure time of electric vehicles.

7. The method according to claim 5, wherein The probability model of daily mileage of electric vehicles is assumed to follow a normal distribution f M (M d ), the mathematical expression is: M d is the daily mileage of electric vehicles, σ M and μ M M d the standard deviation and mean of Introducing electric vehicle charging time to describe the electric vehicle charging time T ch , the expression is: E d,100 is the power demand of an electric vehicle when it travels 100 kilometers, B c is the battery capacity of the electric vehicle, and is the rated power and charging efficiency of the electric vehicle, is the battery capacity of the electric vehicle; The probability density function of electric vehicles is sampled by the Monte Carlo method to obtain the starting charging time and charging duration, and the daily basic charging load P is calculated. t EV : t P This is the time when electric vehicles start charging.

8. The method according to claim 3, wherein Demand response is divided into interruptible load and time-shiftable load. The specific expressions are as follows: For interruptible loads, is the proportion of interruptible load, is the total electrical load, is a time-shiftable load, is the maximum value of the time-shiftable load, is the minimum value of the time-shiftable load, t represents time t, and T is 24h.

9. The method according to claim 3, wherein Renewable energy sources include wind turbines and photovoltaic power generation. The power output of wind turbines and photovoltaic power generation is described by probability density functions, which are expressed as follows: where h = v N / v ci -1 The scale parameter C of the first Weibull distribution, the charging price c, the scale parameter K of the second Weibull distribution, V ci is the cut-in wind speed of the wind turbine, h is a constant, v N is the rated wind speed of the wind turbine, P N is the rated installed capacity of the wind turbine, the first shape parameter α, the second shape parameter β, P PV is the photovoltaic power generation power, is the maximum value of photovoltaic power generation, P WT is the wind turbine power, f FG (P WT ),f GG (P PV ) are the probability density functions of wind turbines and photovoltaic power generation, respectively.

10. The method according to claim 2, wherein Based on two-layer optimization, a two-layer-four-level stochastic robust model is established, which specifically includes the upper-layer active distribution network operator and the lower-layer electric vehicle aggregator. The upper-layer active distribution network operator adopts a three-level max-max-min model, and the lower-layer electric vehicle aggregator adopts a single-level min model. The multi-scenario dynamic pricing mechanism connects the three-level max-max-min model and the single-level min model respectively.

11. The method according to claim 10, wherein The multi-scenario dynamic pricing mechanism actively adjusts price differences through the distribution network to attract or restrict the charging and discharging behavior of electric vehicles. The specific expression is as follows: Where, The price of charging an electric car, is the electricity price for electric vehicles, The valley electricity price is the time-of-use electricity purchase price. The electricity price is based on time of use. is the time-of-use electricity price, μ EV is the network loss coefficient, P loss,t,i→j is the network loss between branches ij at time t, P loss,t,j→k is the network loss between branches jk at time t, is the equivalent load, is the total equivalent power of wind and solar power, c t As the basic charging electricity price, is the total electrical load, is a time-shiftable load, are the charging and discharging power of electric vehicles, Interruptible load.

12. The method according to claim 11, wherein The first level of the three-level max-max-min model finds the probability distribution of uncertain variables, the second level max identifies the output of uncertain variables, and the third level min optimizes the operating cost. The three-level max-max-min model includes the transaction costs between the active distribution network and the main network. Network loss cost Demand response costs and penalty costs for curtailing wind and solar power The specific expression is as follows: P ss =(P t WT ,P t PV ) Scenario probability σ s ∈Ω σ ,Ω σ is the set of scenario probabilities, the uncertainty variable P of renewable energy ss , s is the running scenario, N s It's a cut scene. is a set of uncertain variables, P t buy_grid 、P t sell_grid Distribution refers to the amount of electricity purchased and sold between the active distribution network and the upper power grid at time t. They are the purchase and sale prices of electricity from the active distribution network, μ loss is the network loss coefficient, Ω line is the set of active distribution network branches, P i PV,SJ is the actual photovoltaic power generation output, P WT,SJ is the actual wind turbine output, μ RES is the penalty coefficient for curtailment of renewable energy, x is the first-stage decision variable, μ DR is the demand response coefficient, node i, I ij,t is the maximum value of the branch ij current at time t, r ij is the resistance of branch ij, Ω n is the set of branch nodes, P WT is the wind turbine power, P PV is the photovoltaic power generation power, P t WT 、P t PV They represent the wind turbine power and photovoltaic power generation at time t respectively.

13. The method according to claim 12, wherein: The scene probability is constrained by a comprehensive norm constraint consisting of 1-norm and ∞-norm conditions. The specific expression is as follows: is the initial probability, θ1 is the 1-norm confidence, θ ∞ is the infinite norm confidence, α1 is the 1 norm confidence parameter, α ∞ is the infinity norm confidence parameter, N AL It is the historical data before the reduction; Uncertain variable P of renewable energy ss Using polyhedron uncertainty modeling, the expression is as follows: P t ss,exp is the expected value of the uncertain variable, P t ss+ 、P t ss- are the positive and negative deviations of the uncertainty variable, is a 0-1 variable with positive deviation, is a 0-1 variable with negative bias, Γ ss is the confidence level of the ∞-norm.

14. The method according to claim 13, wherein The single-level min model is shown as follows: are the charging and discharging powers of the electric vehicles of the g-th electric vehicle aggregator, G is the set of g, and T is 24h.

15. The method according to claim 14, wherein The two-layer-four-level model is iteratively solved by coupling the CCG-AIS dichotomy method. The output strategy for terminating the iteration according to the two-layer-four-level model includes: Identify independent optimization variables; By fixing the independent optimization variables, a three-level max-max-min model is used as the original optimization problem; Decouple the original optimization problem into two sub-problems, and then solve each sub-problem alternately; When the difference between the objective function values ​​of the two subproblems is less than the set error range, the original optimization problem converges. The three-level max-max-min model takes the dynamic electricity price set by the active distribution network as the convergence target and converges through the bisection method; The bisection method converges to determine whether it satisfies the If the convergence condition is not satisfied, then according to the constraint Resolve; in are the electric vehicle charging prices for the λ-2, λ-1, λ and λ+1 iterations respectively, and ε is the set tolerance.

16. Electric vehicle active distribution network operation equipment driven by a multi-scenario dynamic pricing mechanism, characterized by: include: The optimization unit and the solution unit are used. The optimization unit is used to perform two-level optimization on the scheduling between the active distribution network operator and the electric vehicle aggregator. The solution unit is used to establish a two-level-four-level stochastic robust model based on the two-level optimization, and iteratively solve the two-level-four-level model through the bisection method of coupling CCG-AIS. The strategy is output according to the conditions for terminating the iteration of the two-level-four-level model, and the output strategy is used as the operation method.

17. The device according to claim 16, characterized in that The two-tier optimization includes dividing electric vehicle charging into reservation, real-time and emergency charging systems based on active distribution network losses, load and renewable energy matching, and using a multi-scenario dynamic pricing mechanism. Through differentiated pricing strategies, electric vehicles are guided to participate in active distribution network optimization.

18. The device according to claim 17, wherein Active distribution network operators are responsible for market operation and settlement. Electric vehicle aggregators are the integration of all charging piles within the active distribution network in a specified area and are responsible for the charging and discharging scheduling of electric vehicles in the area. Electric vehicle aggregators voluntarily sign contracts with electric vehicle users who are willing to participate in demand response, and obtain the charging and discharging control rights of the electric vehicle users during the period of access to the charging pile. Unified scheduling of charging and discharging status of electric vehicles; The EV aggregator formulates a charging and discharging plan based on the initial power consumption, expected off-grid time, and required power consumption information uploaded by EV users through the data collection system, combined with the dispatch instructions of the active distribution network operator, and feeds it back to the active distribution network operator; Active distribution network operators implement multi-scenario dynamic pricing strategies for different electric vehicle operation scenarios based on the current grid losses, the matching status of renewable energy and load, and the charging and discharging data provided by electric vehicle aggregators, and clarify the charging and discharging prices for electric vehicle users at different times.

19. The device according to claim 18, characterized in that The constraints of the active distribution network include power flow balance constraints, branch current constraints, node voltage constraints, and line current carrying capacity constraints. The specific expressions are as follows: (V i min ) 2 ≤U i,t ≤(V i max ) 2 Where, P j,t , Q j,t and U j,t are the active power, reactive power and square of node voltage injected into node j at time t, Ω n is the set of branch nodes, r ij 、x ij are the resistance and reactance of branch ij, L ij is the square of the current in branch ij, is the maximum value of the branch ij current, V i max With V i min is the maximum and minimum value of the voltage at node i, P ij,t With Q ij,t is the active power and reactive power flowing from node i to node j in branch ij at time t, is the maximum apparent power allowed to flow through branch ij, P jk,t , Q jk,t They are respectively the active power and reactive power injected into branch jk at time t, U i,t is the node voltage of node i at time t, L ij,t is the square of the current in branch ij at time t.

20. The apparatus of claim 18, wherein The electricity consumption behavior of electric vehicles is described by the initial charging period probability distribution model and the daily mileage probability model.

21. The device according to claim 20, characterized in that The probability distribution model of the initial charging period is expressed as follows: f EV (t) is the probability distribution model of the initial charging period, μ1 is the mean of the arrival and departure time of electric vehicles, and σ1 is the standard deviation of the arrival and departure time of electric vehicles.

22. The device according to claim 20, wherein The probability model of daily mileage of electric vehicles is assumed to follow a normal distribution f M (M d ), the mathematical expression is: M d is the daily mileage of electric vehicles, σ M and μ M M d the standard deviation and mean of Introducing electric vehicle charging time to describe the electric vehicle charging time T ch , the expression is: E d,100 is the power demand of an electric vehicle when it travels 100 kilometers, B c is the battery capacity of the electric vehicle, and is the rated power and charging efficiency of the electric vehicle, is the battery capacity of the electric vehicle; The probability density function of electric vehicles is sampled by the Monte Carlo method to obtain the starting charging time and charging duration, and the daily basic charging load P is calculated. t EV : t P This is the time when electric vehicles start charging.

23. The apparatus of claim 18, wherein Demand response is divided into interruptible load and time-shiftable load. The specific expressions are as follows: For interruptible loads, is the proportion of interruptible load, is the total electrical load, is a time-shiftable load, is the maximum value of the time-shiftable load, is the minimum value of the time-shiftable load, t represents time t, and T is 24h.

24. The apparatus of claim 18, wherein Renewable energy sources include wind turbines and photovoltaic power generation. The power output of wind turbines and photovoltaic power generation is described by probability density functions, which are expressed as follows: where h = v N / v ci -1 The scale parameter C of the first Weibull distribution, the charging price c, the scale parameter K of the second Weibull distribution, V ci is the cut-in wind speed of the wind turbine, h is a constant, v N is the rated wind speed of the wind turbine, P N is the rated installed capacity of the wind turbine, the first shape parameter α, the second shape parameter β, P PV is the photovoltaic power generation power, is the maximum value of photovoltaic power generation, P WT is the wind turbine power, f FG (P WT ),f GG (P PV ) are the probability density functions of wind turbines and photovoltaic power generation, respectively.

25. The apparatus of claim 17, wherein Based on two-layer optimization, a two-layer-four-level stochastic robust model is established, which specifically includes the upper-layer active distribution network operator and the lower-layer electric vehicle aggregator. The upper-layer active distribution network operator adopts a three-level max-max-min model, and the lower-layer electric vehicle aggregator adopts a single-level min model. The multi-scenario dynamic pricing mechanism connects the three-level max-max-min model and the single-level min model respectively.

26. The device according to claim 25, characterized in that The multi-scenario dynamic pricing mechanism actively adjusts price differences through the distribution network to attract or restrict the charging and discharging behavior of electric vehicles. The specific expression is as follows: Where, The price of charging an electric car, is the electricity price for electric vehicles, The valley electricity price is the time-of-use electricity purchase price. The electricity price is based on time of use. is the time-of-use electricity price, and the network loss coefficient μ EV , P loss,t,i→j is the network loss between branches ij at time t, P loss,t,j→k is the network loss between branches jk at time t, is the equivalent load, is the total equivalent power of wind and solar power, c t As the basic charging electricity price, is the total electrical load, is a time-shiftable load, are the charging and discharging power of electric vehicles, Interruptible load.

27. The device according to claim 26, characterized in that The first level of the three-level max-max-min model finds the probability distribution of uncertain variables, the second level max identifies the output of uncertain variables, and the third level min optimizes the operating cost. The three-level max-max-min model includes the transaction costs between the active distribution network and the main network. Network loss cost Demand response costs and penalty costs for curtailing wind and solar power The specific expression is as follows: P ss =(P t WT ,P t PV ) Scenario probability σ s ∈Ω σ ,Ω σ is the set of scenario probabilities, the uncertainty variable P of renewable energy ss , s is the running scenario, N s It's a cut scene. is a set of uncertain variables, P t buy_grid 、P t sell_grid Distribution refers to the amount of electricity purchased and sold between the active distribution network and the upper power grid at time t. They are the power purchase and sales prices of the active distribution network, and the network loss coefficient μ loss ,Ω line is the set of active distribution network branches, P i PV,SJ is the actual photovoltaic power generation output, P WT,SJ is the actual wind turbine output, μ RES is the penalty coefficient for curtailment of renewable energy, x is the first-stage decision variable, μ DR is the demand response coefficient, node i, I ij,t is the maximum value of the branch ij current at time t, r ij is the resistance of branch ij, Ω n is the set of branch nodes, P WT is the wind turbine power, P PV is the photovoltaic power generation power, P t WT 、P t PV They represent the wind turbine power and photovoltaic power generation at time t respectively.

28. The apparatus of claim 27, wherein The scene probability is constrained by a comprehensive norm constraint consisting of 1-norm and ∞-norm conditions. The specific expression is as follows: is the initial probability, θ1 is the 1-norm confidence, θ ∞ is the infinite norm confidence, α1 is the 1 norm confidence parameter, α ∞ is the infinity norm confidence parameter, N AL It is the historical data before the reduction; Uncertain variable P of renewable energy ss Using polyhedron uncertainty modeling, the expression is as follows: P t ss,exp is the expected value of the uncertain variable, P t ss+ 、P t ss- are the positive and negative deviations of the uncertainty variable, is a 0-1 variable with positive deviation, is a 0-1 variable with negative bias, Γ ss is the confidence level of the ∞-norm.

29. The apparatus of claim 28, wherein The single-level min model is shown as follows: are the charging and discharging powers of the electric vehicles of the g-th electric vehicle aggregator, G is the set of g, and T is 24h.

30. The apparatus of claim 29, wherein The two-layer-four-level model is iteratively solved by coupling the CCG-AIS dichotomy method. The output strategy for terminating the iteration according to the two-layer-four-level model includes: Identify independent optimization variables; By fixing the independent optimization variables, a three-level max-max-min model is used as the original optimization problem; Decouple the original optimization problem into two sub-problems, and then solve each sub-problem alternately; When the difference between the objective function values ​​of the two subproblems is less than the set error range, the original optimization problem converges. The three-level max-max-min model takes the dynamic electricity price set by the active distribution network as the convergence target and converges through the bisection method; The bisection method converges to determine whether it satisfies the If the convergence condition is not satisfied, then according to the constraint Resolve; in are the electric vehicle charging prices for the λ-2, λ-1, λ and λ+1 iterations respectively, and ε is the set tolerance.

Images