Large-scale EV layered charging method considering cooperation of multiple charging modes

Abstract

A large-scale EV layered charging method considering cooperation of multiple charging modes comprises the steps that distribution network power load data is collected to be used for describing power load user portraits of all communities, then data characteristic quantities are constructed, clustering is carried out through regional power load clustering after data standardization processing, and a typical power load curve of the region is obtained; and matching the user with the charging mode based on a particle swarm algorithm according to the user characteristics and the charging mode cost. Characteristics of different charging modes are utilized, charging requirements of users with different characteristics in different time periods are met, advantage complementation of the charging modes can be achieved through cooperation of the different charging modes, the charging requirements of the EV users can be met continuously and economically, a basis is provided for a power grid dispatching center to make a dispatching plan, and the operation efficiency of a power grid is optimized.

Description

Technical Field

[0001] This invention relates to a technology in the field of power resource allocation, specifically a tiered charging method for large-scale electric vehicles (EVs) that takes into account different load user profiles. Background Technology

[0002] Most existing planning techniques for large-scale electric vehicle charging mode coordination are based on master-slave game theory. Master-slave game theory considers the game relationship between upper and lower layers, and can coordinate the charging and discharging behavior of electric vehicles while ensuring grid stability. However, few optimization methods consider achieving complementary advantages through the coordination of different charging modes when multiple charging modes coexist. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes a large-scale tiered charging method for EVs that considers the synergy of multiple charging methods. By leveraging the characteristics of different charging methods, it meets the charging needs of users with different characteristics at different times. By coordinating different charging methods, the advantages of each method can be complemented, sustainably and economically meeting the charging needs of EV users. This provides a basis for the power grid dispatch center to formulate dispatch plans and optimizes the efficiency of power grid operation.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a large-scale tiered charging method for EVs that considers user profiles with different loads. After collecting power load data from the distribution network to characterize the power load user profiles of each community, data feature quantities are constructed. After data standardization processing, clustering is performed through regional power load clustering to obtain the typical power load curve of the region. Based on user characteristics and charging method costs, users are matched with charging methods using a particle swarm optimization algorithm.

[0006] This invention relates to a system for implementing the above method, comprising: a clustering unit and a matching unit, wherein: the clustering unit performs K-means and elbow rule-based clustering based on the power load curve of a certain region and considering the local renewable energy abundance information to obtain a typical load profile of the local area; the matching unit performs two-layer particle swarm optimization algorithm power matching based on vehicle networking and battery swapping characteristics information to obtain a result that the advantages of different charging methods can be complemented by coordinating different charging methods. Attached Figure Description

[0007] Figure 1 This is a parameter view of an example from the present invention;

[0008] The graph shows: (a) the number of electric vehicles and renewable energy generation; (b) changes in electricity prices.

[0009] Figure 2 This is a flowchart of the present invention;

[0010] Figure 3 The image shown is a rendering of an example.

[0011] In the figure: (a) Typical power load curves of Inner Mongolia after clustering; (b) Comparison of energy exchange methods and collaborative approaches. Detailed Implementation

[0012] like Figure 2 As shown in the figure, this embodiment involves a large-scale EV hierarchical charging method that considers user profiles with different loads. After collecting power load data from the distribution network to characterize the power load user profiles of each community, data feature quantities are constructed. After data standardization processing, clustering is performed through regional power load clustering to obtain the typical power load curve of the region. Based on user characteristics and charging method costs, users and charging methods are matched based on particle swarm optimization algorithm.

[0013] The data features include:

[0014] ① Maximum and minimum loads reflecting the power load level: P l，max =max{P l (0), P l (Δt), L, P l (t), L, P l (T)},P l，min =min{P l (0), P l (Δt), L, P i (t), L, P l (T)}, where: P l，max and P l，min These represent the maximum and minimum electrical loads within region l during time period T; P l The electrical load of region l; D t The time interval for collecting charging load data; T is the time period of the power load.

[0015] ② Average power load, reflecting the overall level of power load in the area: Where: P l，aver Let be the average power load within region l over time period T. Using this characteristic, the acceptable overall EV charging load for this region can be determined.

[0016] ③ Power load variance, reflecting the dispersion of power load and the fluctuation of the power load curve in the region. Wherein: S l Let be the variance of the electricity load in region l. Utilizing this characteristic value is beneficial for developing cost-effective charging schemes and improving the local consumption level of distributed renewable energy.

[0017] ④ Distributed renewable energy absorption capacity Where: P l，w and P l，s These represent the distributed wind and solar power absorption capacity of region l, respectively; P w and P s These are distributed wind power and photovoltaic power generation outputs, respectively. Considering the capacity for distributed renewable energy absorption as a characteristic quantity is beneficial for improving the local absorption level of distributed renewable energy in some regions.

[0018] The data standardization process specifically involves using the Z-Score standardization method to ensure comparability of results. in: Z represents the d-th feature of the Z-Score-normalized regional load data. aver S and Z represent the mean and standard deviation of the population characteristic data, respectively; l，d Let d be the d-th feature data of region l; D is the number of feature data in each region.

[0019] The aforementioned regional power load clustering specifically involves clustering standardized regional power load characteristics using the K-means algorithm to obtain typical power load curves for the region. To overcome the issue of the number of cluster centers, the elbow rule is used to select the number of cluster centers. Where: E represents the total distortion level; K means，d (k) is the d-th feature of the k-th center point; K is the number of center points.

[0020] The V2G station and BSS charging model mentioned above include:

[0021] a) V2G station charging model: This model aims to optimize the total cost of EVs and formulates the optimal energy exchange scheme, satisfying the following: P min ≤P n (t)≤P max , SOC min ≤SOC n，ini ≤SOC n，obj ≤SOC max SOC min ≤SOC n (t)≤SOC max , t0≤T ini，n <T n ≤T, N≥max(N) EV (t0), N EV (t0+Δt), L, NEV (T)), where: C V2a and C V2a，n These represent the total cost of the EV under V2G energy exchange and the cost of the nth EV, respectively; T n P represents the network outage time for electric vehicle n; n For EVn's V2G power; a n V2G pricing for EVn; BDC n The battery loss cost is EVn; a1, a2, β1, and β2 are all battery loss model parameters; P max and P min These are the maximum and minimum power outputs for V2G in electric vehicles, respectively; a Grid The electricity price sold by the power grid; SOC n，obj and SOC n，ini These represent the expected and initial state of charge (SOC) values ​​of the battery for electric vehicle n, respectively; Q n Rated capacity of electric vehicle battery; SOC max and SOC min These represent the battery's maximum and minimum nuclear states; SOC n (t) represents the real-time state of charge of the battery for EVn; T n，ini P is the initial time for EVn to connect to the network. re and P′ re These represent the distributed new energy power output values ​​before and after electric vehicle consumption; N is the number of charging piles in the V2G charging station, and N is the maximum number of EVs that can be charged in the station; N EV The number of EVs matched to V2G charging stations.

[0022] b) BS Capacity Exchange Model: The optimal energy exchange scheme is formulated with the goal of minimizing the total cost of the BSS, satisfying the following: P min ≤P m (t)≤P max , SOC min ≤SOC m，ini ≤SOC m，obj ≤SOC max SOC min ≤SOC m (t)≤SOC max , t0≤T ini，m <T m ≤T, M B ≥max(M EV (t0), M EV (t0+Δt), L, M EV (T)), Where: C BS and C BS，mThese represent the total cost of the EV and the cost of the nth EV under the BS method, respectively; T m，ini and T m P represents the start and end times of battery charging at position m in the BSS. The start time is when the battery at that position, representing the desired SOC value, has just completed its battery exchange with the EV; the SOC value at that position is the EV's SOC value before the exchange. The end time is the time it takes for the battery at the corresponding position in the BSS to complete the battery exchange with the EV, or the time it takes for the battery at that position to reach the desired SOC value. m Q represents the battery charging power at position m in the BSS. m This refers to the rated capacity of the battery. MB represents the number of batteries within the BSS.

[0023] The two-layer capacity exchange model includes: an upper-layer capacity exchange model minC, which uses the number of EVs randomly matched to V2G charging stations and BSSs in a single time period as the decision variable. BS (t)+C V2G (t), and the lower-level capacity exchange model minC with the individual user cost in the process of individual electric vehicles supplementing electricity in V2G or BS as the optimization objective. EV.n (t), where: C EV，n Let n be the charging cost for an individual electric vehicle.

[0024] The aforementioned particle swarm optimization algorithm specifically includes:

[0025] Step 1) Data collection for cluster analysis of typical load curves, serving as parameters for bi-level optimization; individual optimality and global optimality are designed to be infinite.

[0026] Step 2) Randomly allocate the number of EVs, update the EV allocation array, and simultaneously perform lower-level capability exchange between V2G charging stations and battery swapping stations.

[0027] Step 3) If the cost of the array after the swap in Step 2) is lower, then determine whether it is the historical best; otherwise, update the particles to the swapped array.

[0028] Step 4) Iteratively update the individual optimal particles until the global optimum is reached;

[0029] Step 5) Update the global optimal particle until the termination condition is met; otherwise, update the EV allocation array again and return to step 2.

[0030] Through specific practical experiments, Figure 1Under the specific environmental settings of the number of charging electric vehicles and renewable energy power generation parameters shown, the proposed method was compared with the methods considering only V2G charging (Method 1) and only battery swapping (Method 2) in simulations. The tests were conducted on a single PC with a 2.20GHz CPU, 16GB of memory, and a 64-bit operating system. Specific parameters are shown in Table 1.

[0031] Table 1 Parameter name numerical values Parameter name numerical values T (hours) 24 P max (kilowatt) 70 Δt (hours) 1 P min (hours) 0 t0 (hours) 0 SOC n,ini ]]> 0.4 <![CDATA[N EV (individual)]]> 100 <![CDATA[SOC n,obj ]]> 0.8 <![CDATA[M B (individual)]]> 100 <![CDATA[Q n (kilowatt-hours) 70 <![CDATA[N EV2 (individual)]]> 250 SOCmax 0.8 <![CDATA[M B2 (individual)]]> 200 SOCmin 0.2 α1 ($ / kW) 0.2281×10-6 <![CDATA[SOC m,obj ]]> 0.8 α2 ($ / kW²) 0.8442×10-6 <![CDATA[SOC m,ini ]]> 0.4 β1 5 β2 2

[0032] Table 2 Electricity costs Battery loss cost Total cost Method 1 2.5148 0.6550 3.1699 Method 2 2.4186 0.2217 2.6403 The proposed method 2.2605 0.1069 2.3674

[0033] Table 3

[0034] Compared to existing technologies, this method achieves the lowest charging power during peak electricity price periods and the highest charging power during off-peak electricity price periods. This is because the proposed method combines the advantages of V2G (Vehicle-to-Government) methods, which consider user downtime, with the long flexible charging time of BS (Base Station) methods, matching the charging mode based on user downtime and cost. This method's prediction technology considers load user profiles and incorporates the elbow method.

[0035] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A method for hierarchical charging of large-scale EVs considering different load user profiles, characterized by, The data feature quantity is constructed after collecting power load data of a distribution network for depicting user portraits of power load of each cell, and the typical power load curve of the region is obtained through clustering of regional power load after data standardization processing; and the user and the charging mode are matched based on the particle swarm algorithm according to the user features and the charging mode cost.

2. The method of claim 1, wherein the method is characterized by, The data feature quantity includes: l，max = max{P l (0), P l (Δt), L, P l (t), L, P l (T)}, P l，min = min{P l (0), P l (Δt), L, P l (t), L, P l (T)}, where: P l，max and P l，min are the maximum and minimum power loads in the time period T of the region l; P l is the power load of the region l; D t is the time interval of the data collection of the charging load; and T is the time period of the power load.​ ② Reflects the overall level of the average power load of the region: Wherein: P l，aver The average load of the power load in the region l time period T, using the characteristic quantity, the region can accept the EV total charging load situation; ③ Power load variance reflecting the dispersion degree of power load in the region and the fluctuation of power load curve wherein: S l is the power load variance value of the region I, using this characteristic quantity, it is beneficial to reasonably formulate a high-cost benefit charging scheme and improve the local consumption level of distributed renewable energy. (4) Distributed new energy consumption capacity Wherein: P l，w and P l，s are the distributed wind power and photovoltaic power consumption capacity of region I respectively; P w and P s are the distributed wind power and photovoltaic power output respectively. Considering the distributed new energy consumption capacity as a characteristic quantity is conducive to improving the local consumption level of distributed new energy in some areas.

3. The method of claim 1, wherein the method is characterized by, The data standardization processing is specifically: adopting a Z-Score standardization method, ensuring comparability of results by data standardization processing, Wherein: is the dth feature quantity of the Z-Score standardized regional l load data; Z aver and S are the average value and the standard deviation of the overall feature data, respectively; Z l，d is the dth feature data of the regional l; and D is the number of feature data of each region.

4. The method of claim 1, wherein the method is characterized by, The regional power load clustering comprises: clustering the normalized regional power load characteristic quantity based on a K-means algorithm to obtain a typical power load curve of the region, and the elbow rule is used to select the number of cluster center points to overcome the problem of the number of cluster center points, Wherein: E is the total distortion degree; K means，d (k) is the dth characteristic quantity of the kth center point; and K is the number of center points.

5. The method of claim 1, wherein the method is characterized by, The V2G station and the BSS charging model include: a) V2G station charging model: This model aims to develop an optimal energy exchange scheme with the goal of optimizing the total cost of EVs, satisfying: P min ≤P n (t)≤P max , SOC min ≤SOC n，ini ≤SOC n，obj ≤SOC max , SOC min ≤SOC n (t)≤SOC max , t0≤T ini，n <T b ≤T, N≥max(N EV (t0)N EV (t0+Δt), L, N EV (T)), where: C V2G and C V2G，n are the total cost of EV and the cost of the nth EV under V2G energy exchange, respectively; T n is the off-grid time of the nth electric vehicle; P n is the V2G power of EVn; a n is the V2G price of EVn; BDC n is the battery loss cost of EVn; a1, a2, β1, and β2 are battery loss model parameters; P max and P min are the maximum and minimum V2G power of the electric vehicle, respectively; a Grid is the grid electricity selling price; SOC n，obj and SOC n，ini are the desired and initial state of charge values of the nth electric vehicle, respectively; Q n is the rated capacity of the electric vehicle battery; SOC max and SOC min are the maximum and minimum state of charge of the battery, respectively; SOC n (t) is the real-time state of charge of the battery of EVn; T n，ini is the initial time of EVn networking; P re and P′ re are the distributed new energy output values before and after the electric vehicle is accommodated, respectively; N is the number of charging piles in the V2G charging station, which is the maximum number of EVs in the station; N EV is the number of EVs matched to the V2G charging station. b) BS capability exchange model: to develop the optimal energy exchange scheme with the goal of minimizing the total cost of BSS, to meet: P min ≤P m (t)≤P max , SOc min ≤SOC m，ini ≤SOC m，obj ≤SOC max , SOC min ≤SOC m (t)≤SOC max , t0≤T ini，m <T m ≤t, M B ≥max(M EV (t0), M EV (t0+Δt), L, M EV (T)), Where: C BS and C BS，m are the total cost of EV and the cost of the nth EV under BS mode; T m，ini and T m are the battery charging start time and end time of position m in BSS, the charging start time here is the battery of the expected SOC value of the position and the EV just completed the battery exchange, the SOC value of the battery of the position is the battery SOC value before the EV battery exchange, the charging end time is the time when the battery of the corresponding position in BSS exchanges with the EV, or the time when the SOC value of the battery of the position reaches the expected SOC value, P m is the battery charging power of position m in BSS, Q m is the rated capacity of the battery, and MB is the number of batteries in BSS.

6. The method of claim 1, wherein the method further comprises: The double-layer capacity exchange model includes: an upper-layer energy exchange model minC BS (t) + C V2G (t), and a lower-layer capacity exchange model minC EV.n (t) taking individual user cost of individual electric vehicles in the power supplement process in V2G or BS as an optimization target, wherein: C EV,n is the charging cost of the electric vehicle individual n.

7. The method of claim 1, wherein the method further comprises: The particle swarm algorithm specifically includes: Step 1) data collection, used for clustering analysis of typical load curve, as parameters of double-layer optimization; individual optimal and global optimal are designed as infinity; Step 2) randomly allocating EV quantity, updating EV allocation array, and meanwhile, performing lower-layer capacity exchange of V2G charging station and battery swap station; Step 3) judging whether it is the historical optimal when the array cost is lower after the exchange in step 2), otherwise updating the particle as the exchanged array; Step 4) iteratively updating individual optimal particle until global optimal; Step 5) updating global optimal particle until satisfying the termination condition, otherwise re-updating EV allocation array and returning to step 2).

8. A scalable EV hierarchical charging system implementing the method of any one of claims 1-7, considering different load user profiles, characterized by, It includes: The clustering unit and the matching unit, wherein: the clustering unit performs clustering processing based on K-means and elbow rule according to the power load curve of a region and considering local renewable energy rich feature information, and obtains the result of local typical load portrait; the matching unit performs power matching processing based on double-layer particle swarm algorithm according to the characteristics information of Internet of Vehicles and battery swap, and obtains the result of complementary advantages of charging modes through cooperation of different charging modes.

Images