A method for controlling the daily rolling orderly charging of electric vehicles under a hierarchical architecture

Abstract

The present invention discloses a method for controlling the daily rolling orderly charging of electric vehicles under a hierarchical architecture, comprising the following steps: 1. Determine the overall operation architecture and time scale of the system, including establishing a hierarchical control structure system of the distribution network layer and the charging station layer, and adopt the form of forward rolling optimization on the time scale; 2. At the charging station layer, analyze and realize the convergence of the charging demand of electric vehicles, and finally form the overall charging demand of each charging station in the distribution network system; 3. According to the overall charging demand of the charging station obtained in step 2 and the safety constraints of the distribution network layer itself, complete the coordinated operation optimization of the distribution network and the charging station, and formulate the charging plan of the charging station; 4. Based on the fuzzy power allocation algorithm, each charging station will complete the charging power allocation of the electric vehicles in the station according to the charging plan provided by the distribution network layer. The present invention can achieve the goals of peak shaving and valley filling of the distribution network and improving the economic efficiency of the charging station operation through the application of the orderly charging strategy.

Description

Technical Field

[0001] The present invention belongs to the technical field of integrated energy management, and relates to a method for controlling the daily rolling orderly charging of an electric vehicle under a layered architecture. Background Art

[0002] In recent years, due to technological progress and the expansion of environmental protection needs, the number of electric vehicles (EVs) in my country is growing rapidly, and their large-scale access to the grid has become a new trend in the development of distribution networks. EV access to the grid is highly random. If it is not managed for charging, a new peak load will be generated outside the original load distribution. At the same time, the long-term residence characteristics of EVs enable them to participate in the economic operation and auxiliary services of the power grid as a transferable load or mobile energy storage. In order to achieve the economic and safe operation of the distribution network after EV access, the charging strategy of EVs has received widespread attention from researchers.

[0003] From the perspective of control structure, solutions for EV charging scheduling can be mainly divided into three types: centralized control, distributed control, and hierarchical control. Centralized control refers to the collection of all EV data in the system and the formulation of charging tasks by a unified controller. It has extremely high requirements for the reliability of the controller's calculation and communication. It is not suitable for scenarios where large-scale EVs are connected to the power grid and is difficult to implement widely. Distributed control usually coordinates control in the form of multiple agents, distributing the amount of calculation required for the original centralized control to multiple agents. This method has high scheduling efficiency and problem-solving capabilities, and is suitable for energy management of EV access to a single charging station or microgrid. Hierarchical control is manifested in the form of "indirect" management of EVs by the power grid, introducing charging station operators to coordinate grid-side operations and EV-side charging.

[0004] From the perspective of the time scale of system operation, the scheduling of EVs can be mainly divided into three types: day-ahead, intraday, and multi-stage. Day-ahead optimization requires obtaining the charging parameters of all connected EVs within a day based on the day-ahead declaration mechanism combined with historical travel patterns and predictions. The literature (He Y, Venkatesh B, Guan L. Optimal Scheduling for Charging and Discharging of Electric Vehicles [J]. IEEE Transactions on Smart Grid, 2012, 3 (3): 1095-1105.) points out that the global optimal scheduling scheme requires parameter information of EVs within the next day, which is difficult to implement in practice. Therefore, studying scheduling schemes within the day or on multiple time scales is of great significance for promoting the orderly connection of EVs.

[0005] Reference (Zhang Liang, Yan Zheng, Feng Donghan, et al. Orderly charging strategy in electric vehicle charging stations using two-stage optimization model [J]. Power System Technology, 2014, 38 (004): 967-973.) established a two-stage centralized operation optimization model for charging stations, which considered the maximization of charging revenue of charging stations and the minimization of system peak-to-valley difference respectively. However, the centralized solution has high requirements on the controller and is prone to "dimensionality" problems when EVs are widely connected. Reference (Bahrami S, Wong VW SA potentialgame framework for charging PHEVs in smart grid [C] / / Communications, Computers & Signal Processing. IEEE, 2015: 28-33.) established the EV charging problem as a potential game problem between EV agents and solved it based on Nash equilibrium. This distributed optimization model can effectively reduce the peak load caused by EV charging, but ignores the actual network model of EV charging. At the same time, this distributed solution is mainly suitable for small-scale EV access scenarios. Reference (Huang Yinuo, Guo Chuangxin, Wang Licheng, et al. Electric vehicle clustering dispatch strategy considering user satisfaction [J]. Automation of Electric Power Systems, 2015, 39(17): 183-191.) uses EV individual parameters as discriminants to cluster and complete energy demand aggregation and subsequent EV group charging optimization. This method is more suitable for model building in the day-ahead optimization stage, and all EVs in the station need to be regrouped before energy demand aggregation to reduce the scale of the problem. Reference (Pan Zhenning, Zhang Xiaoshun, Yu Tao, et al. Hierarchical real-time optimization dispatch of large-scale electric vehicle clusters [J]. Automation of Electric Power Systems, 2017(16): 102-110.) also regroups cluster electric vehicles in the charging station and completes real-time power allocation in the charging station based on the energy consistency algorithm, but this allocation mechanism has the problem of errors between the planned and actual charging power.

[0006] At present, EV charging management needs to solve the following problems: First, to ensure the effectiveness of EV charging strategies, the structure of the actual distribution network and its operating constraints must be taken into account. Secondly, with the increase in the number of EVs connected to the grid, if a single EV is directly scheduled as a controllable resource independently, the optimization calculation will be complicated and the algorithm dimension will be too large. In addition, there is uncertainty in the time of EV access to the grid, and the operation of commercial charging stations may result in unreliable reporting of charging information, making it difficult to implement day-ahead optimization based on global information; for the power allocation problem within a single charging station, an accurate, effective and convenient charging power allocation mechanism needs to be implemented.

[0007] In response to the above problems, the present invention develops a hierarchical architecture of a distribution network including charging stations to meet the needs of large-scale EV scheduling; develops and solves a daily rolling optimization problem that takes into account the interests of both charging stations and distribution networks based on a distributed algorithm, and develops a short-term scheduling plan within the day. Based on known EV information, the present invention combines short-term load information and predicted information of vehicles entering the network to build an EV cluster model within a rolling window to adapt to the uncertainty of EV access to the network and the privacy requirements of charging station operators; and designs a fuzzy power allocation mechanism to quickly determine the charging power of each EV in the EVA. Summary of the invention

[0008] In view of the above problems, the present invention proposes a method for orderly rolling charging of electric vehicles within a day under a hierarchical architecture, formulates a hierarchical architecture including a distribution network of charging stations to meet the needs of large-scale EV scheduling; formulates and solves the intraday rolling optimization problem that takes into account the interests of both charging stations and distribution networks based on a distributed algorithm, formulates an intraday short-term scheduling plan, and constructs an EV cluster model within a rolling window based on known EV information, combined with short-term load information and predicted information of vehicles entering the network, to adapt to the uncertainty of EV access to the network and the privacy requirements of charging station operators; and designs a fuzzy power allocation mechanism to quickly formulate the charging power of each EV in the EVA.

[0009] The present invention is achieved by at least one of the following technical solutions.

[0010] A method for controlling the daily rolling orderly charging of an electric vehicle under a layered architecture comprises the following steps:

[0011] Step 1: Determine the overall operation architecture and time scale of the system, including establishing a hierarchical control structure system of the distribution network layer and the charging station layer, and adopting the form of forward rolling optimization on the operation time scale;

[0012] Step 2: At the charging station level, analyze and realize the aggregation of charging demands of electric vehicles, and finally form the overall charging demands of each charging station in the distribution network system;

[0013] Step 3: Based on the overall charging demand of the charging station obtained in step 2 and the safety constraints of the distribution network layer itself, complete the coordinated operation optimization of the distribution network and the charging station, and formulate a charging plan for the charging station;

[0014] Step 4: Based on the fuzzy power allocation algorithm, each charging station will complete the charging power allocation of electric vehicles in the station according to the charging plan provided by the distribution network layer.

[0015] Furthermore, in step 1, the system is divided into a distribution network layer and a charging station layer. The distribution network layer is based on the actual distribution network line model and is equipped with a distribution network controller with communication and computing functions; the charging station layer includes charging stations located at different nodes in the distribution network, and each charging station is equipped with a controller for controlling the charging power distribution of electric vehicles (Electrical vehicle EV) in the station and communicating and interacting with the upper distribution network.

[0016] Furthermore, the system's optimization form operates in a continuously rolling intraday rolling optimization scheduling form: first, based on the current EV data and short-term predicted charging load and other load forecast data, the optimal power flow problem of the distribution network-charging station in the corresponding sliding window time domain is solved, and only the charging strategy of the charging station and the electric vehicles in the station at the current moment is implemented; as time moves forward, the load forecast information and electric vehicle data information are updated, and then the optimization problem of the corresponding sliding window at the next moment is reformulated.

[0017] Furthermore, in step 2, when the electric vehicle is connected to the charging station, the basic charging information of the electric vehicle includes the total battery capacity B i , Start access time t i in , expected off-grid period t i dep , current state of charge soc i in , expected state of charge soc i d , maximum charging power p i max are all determined; and the electric vehicle charging boundary information is generated based on the above basic information, including the upper and lower limits of the charging power and the upper and lower limits of the accumulated energy.

[0018] Furthermore, the overall charging demand of the charging station includes the total charging demand of all real electric vehicles currently in the station and the total charging demand of all predicted electric vehicles that are expected to enter the station within the current optimization window.

[0019] Furthermore, the total charging demand of all real electric vehicles in the current station is calculated as follows:

[0020]

[0021] Where: L represents the size of the rolling optimization window, t represents the current optimization period, and k represents the specific time within the period. and are the lower and upper bounds of the actual EVA power at node i in each time period within the window, P EVA,i (t+k) represents the actual charging station power at time t+k, P EVA,iis the actual charging station power at node i, and They represent the lower and upper bounds of the accumulated energy of the actual EVA at node i in each period within the sliding window, is the expected energy of the actual EVA in the entire sliding window;

[0022] The total charging demand of all predicted EVs that are expected to enter the station within the current optimization window is calculated as follows:

[0023]

[0024] Where: and are the lower and upper bounds of the power of EVA in each time period within the window, represents the total charging power of EVA predicted at time t+k, and They are respectively the lower and upper bounds of the cumulative energy of EVA in each period within the sliding window.

[0025] Furthermore, the optimization scheduling goal of step 3 is to minimize the network loss and charging cost:

[0026]

[0027] Where: L is the sliding window length, N ev For the electric car station assembly, P i (t) and P i pre (t) are the actual EVA and predicted EVA of the charging station at time t, Π(t) is the electricity price information at that time, r j,k with i j,k are the impedance and current of the distribution network branch (j, k), respectively, and E represents the set of all distribution network branches. is the total loss of the distribution network circuit, and γ is the weight value.

[0028] Furthermore, step 3 includes the following steps: first, each charging station completes the analysis of the total charging demand within the station, takes the charging station operation model as a constraint, responds to the electricity price during the period, and uses the sub-controller to solve the sub-problem of minimizing the charging cost; then, the distribution network operator uses the distribution network main controller to use the system flow constraint as the optimization constraint to solve the sub-problem with the goal of minimizing network loss; the distribution network controller and the charging station controller iteratively solve each sub-problem until convergence; the solution of each sub-problem is completed locally, mainly relying on the power decision made by the communication transmission between controllers.

[0029] Furthermore, step 4 allocation is to realize the energy allocation mechanism of charging station-EV based on fuzzy rules. The allocation mechanism is: for each EV, the power allocation factor used to measure the power allocation weight is determined according to its own charging urgency and charge state, that is, the input of the fuzzy controller is the charging urgency and charge state of the EV, and the output is the power allocation factor; the input variables and output variables are divided into three levels: low (L), medium (M), and high (H).

[0030] Furthermore, in step 4, the charging power allocation of electric vehicles in the station is completed based on the fuzzy power allocation algorithm, and the steps are as follows:

[0031] (1) The node charging station receives the charging power at time t set by the distribution network layer optimization dispatch;

[0032] (2) Power is allocated to EVs that cannot be dispatched in the station or are about to leave the station at the next moment according to their rated power or expected energy;

[0033] (3) Except for the EVs that have completed power allocation, the remaining EVs in the station constitute the set N res , the EV parameter information in the set is input to the fuzzy controller, and the output power allocation factor of the jth EV is df j , and press df j The pre-allocated charging power p of the jth EV is obtained by numerical calculation j (t);

[0034] (4) The pre-allocated power p of each EV in step (3) j (t) Perform safety inspection and formulate the following rules: If the set N res If all EVs in meet the EV power and energy boundary conditions, the power allocation of the charging station is completed at time t; otherwise, the power of the individuals that do not meet the constraints is corrected;

[0035] (5) From the set N res Eliminate the EVs that have completed power correction, update the remaining total power, and return to step (3) to redistribute power;

[0036] The power allocated to the EV meets the safety inspection rules, and the power allocation of the charging station is completed at time t.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] (1) The present invention includes a complete electric vehicle charging architecture and charging strategy. To ensure the effectiveness of the EV charging strategy, the structure of the actual distribution network and its operating constraints are taken into account.

[0039] (2) The electric vehicles of the present invention are first connected to the power grid in a distributed manner, and then participate in the economic operation of the distribution network in the form of clusters to reduce the scale of the problem. Hierarchical scheduling usually requires cluster division based on the characteristic parameters of EVs. The present invention directly establishes an actual EVA model of energy aggregation within the station in the rolling optimization time domain from the perspective of real-time operation of the electric vehicle station.

[0040] (3) In order to address the uncertainty of EV access time, a short-term scheduling plan is formulated. This plan is based on known EV information, short-term load information and predicted information of vehicles that will be connected to the grid to build an electric vehicle cluster model within a rolling window, and implement rolling scheduling for the economic operation of the distribution network, which is more suitable for the future form of the distribution network.

[0041] (4) The fuzzy power allocation mechanism of the electric vehicle charging station included in the present invention can meet the charging needs of EV users and facilitate the implementation of practical projects.

[0042] (5) The present invention formulates a rolling optimization problem that takes into account the interests of both charging stations and distribution networks, and solves it through a distributed algorithm to adapt to the uncertainty of EV access to the network and the privacy requirements of charging station operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a system control layered architecture diagram of an embodiment of the present invention;

[0044] Figure 2 is a diagram of a system operation rolling optimization process according to an embodiment of the present invention;

[0045] Figure 3a is the electric vehicle charging power boundary diagram;

[0046] Figure 3b is the cumulative energy boundary map of off-grid electric vehicles within the sliding window;

[0047] Figure 3c It is the cumulative energy boundary diagram of the non-off-grid electric vehicles within the sliding window;

[0048] Figure 4 is a diagram of the implementation mechanism of the fuzzy controller used in the electric vehicle power distribution involved in the embodiment of the present invention;

[0049] Figure 5a It is the membership function curve of the input (charge state) of the fuzzy controller;

[0050] Figure 5b It is the membership function curve of the input (charging urgency) of the fuzzy controller;

[0051] Figure 5c It is the membership function curve of the output (power allocation factor) of the fuzzy controller. DETAILED DESCRIPTION

[0052] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings, but the implementation of the present invention is not limited thereto.

[0053] A method for controlling the daily rolling orderly charging of an electric vehicle under a layered architecture comprises the following steps:

[0054] Step 1, determine the overall operation architecture and time scale of the system, including establishing a hierarchical control structure system of the distribution network layer-charging station layer, and adopting the form of rolling optimization towards the previous day on the operation time scale; specifically, the system is a two-layer structure including a distribution network layer and a charging station layer, the distribution network layer is based on the actual distribution network line model, and is equipped with a distribution network controller with communication and computing functions; the charging station layer is composed of charging stations located at different locations in the distribution network, and each is equipped with a controller for controlling the charging power distribution of all EVs in the station and the communication interaction with the upper power grid.

[0055] Step 2: At the charging station level, analyze and realize the charging demand aggregation of electric vehicles, and finally form the overall charging demand of each charging station in the distribution network system, realize the charging demand aggregation of electric vehicles, and form the overall charging demand of charging stations. When an electric vehicle is connected to a charging station, the basic charging information of the electric vehicle includes the total battery capacity B i , Start access time t i in , expected off-grid period t i dep , current state of charge soc i in , expected state of charge soc i d , maximum charging power p i max are all determined; and the electric vehicle charging boundary information is generated based on the above basic information, including the upper and lower limits of the charging power and the upper and lower limits of the accumulated energy.

[0056] Step 3: Based on the overall charging demand of the charging station obtained in step 2 and the safety constraints of the distribution network layer itself, complete the coordinated operation optimization of the distribution network and the charging station, and formulate a charging plan for the charging station;

[0057] Step 4: Based on the fuzzy power allocation algorithm, each charging station will complete the charging power allocation of electric vehicles in the station according to the charging plan provided by the distribution network layer. The steps are as follows:

[0058] (1) The node charging station receives the charging power at time t set by the distribution network layer optimization dispatch;

[0059] (2) Power is allocated to EVs that cannot be dispatched in the station or are about to leave the station at the next moment according to their rated power or expected energy;

[0060] (3) Except for the EVs that have completed power allocation, the remaining EVs in the station constitute the set N res , the EV parameter information in the set is input to the fuzzy controller, and the output power allocation factor of the jth EV is df j , and press df j The pre-allocated charging power p of the jth EV is obtained by numerical calculation j (t);

[0061] (4) The pre-allocated power p of each EV in step (3) j (t) Perform safety inspection and formulate the following rules: If the set N res If all EVs in meet the EV power and energy boundary conditions, the power allocation of the charging station is completed at time t; otherwise, the power of the individuals that do not meet the constraints is corrected;

[0062] (5) From the set N res Eliminate the EVs that have completed power correction, update the remaining total power, and return to step (3) to redistribute power; the power allocated to the EVs meets the safety inspection rules, and the power allocation of the charging station is completed at time t.

[0063] like Figure 1 The diagram shows a hierarchical architecture of system control formed by an embodiment of the present invention, and the orderly charging control strategy corresponding to the present invention is applicable to the hierarchical architecture. The architecture model can be mainly divided into a distribution network layer and a charging station layer. The distribution network layer is based on the actual distribution network line model and is equipped with a distribution network controller with communication and computing functions; the charging station layer is composed of charging stations located at different locations in the distribution network, and each is equipped with a controller for controlling the charging power distribution of all EVs in the station and the communication interaction with the upper power grid. In the hierarchical control model, first, each charging station controller in the lower charging station layer will obtain the battery information and charging plan of the EV in the station, obtain the charging demand information of the EV through calculation, and aggregate the charging demand of all EVs into the overall charging demand of each charging station, which is uploaded to the distribution network controller; the upper distribution network layer can formulate the charging plan of each node charging station based on the total energy demand information uploaded by the charging station layer controller and the security information of the distribution network itself; each charging station in the lower charging station layer finally completes the charging power distribution of EVs in the station according to the total charging plan formulated by the distribution network layer.

[0064] Consider a radial distribution network with N nodes. The distribution network model is described as a connected graph G = {N, E}, where N represents the set of all nodes and E represents the set of branches consisting of all connected nodes. Introduce auxiliary variables v j 、i j,k :

[0065]

[0066] The flow calculation is based on the Distflow equation. The flow equation is shown in formula (2):

[0067]

[0068] Where: P j,k , Q j,k are the active power and reactive power flowing through branch (j, k) respectively; P k,h , Q k,h are the active power and reactive power flowing through the branch (k, h) respectively; P k , Q k are the active power and reactive power injected into node k respectively; V j With V k Represent the voltages of node j and node k respectively, I j,k is the current on branch (j, k); v j 、v k 、i j,k V j 、V k ,I j,k The corresponding auxiliary variable; r j,k With x j,k are the resistance and reactance values ​​on branch (j, k) respectively.

[0069] The quadratic non-convex constraint in formula (2) can be transformed into a standard second-order cone relaxation form:

[0070]

[0071] Add the following voltage safety constraints to each node:

[0072]

[0073] Where: v max and v min The bus voltage at the root node is set to a constant, whose value is v ref , when power is transmitted from a single bus to a radial distribution network, the active power P1 and reactive power Q1 at the root node satisfy:

[0074]

[0075] Where: P 1,j , Q 1,j They represent the active power and reactive power flowing through branch (1, j) respectively. Branch (1, j) represents the branch connected to the root node.

[0076] All nodes except the root node are PQ nodes, which meet the following requirements:

[0077]

[0078] Where: P min With P max are the lower and upper limits of active power, Q min With Q max are the lower and upper limits of reactive power respectively; P L,j , Q L,j is the fixed active and reactive loads at node j, P EVA,j Represents the power of the charging station at node j. If there is no charging station at the node, the value is 0.

[0079] like Figure 2 The diagram shows the rolling optimization of the system operation, considering the form of optimal scheduling that rolls forward over time: first, based on the EV data at the current moment and the short-term predicted charging load and other load forecast data, solve the optimization problem in the corresponding sliding window time domain, and only implement the charging strategy of the charging station and the EV in the station at the current moment; as time moves forward, update the load forecast information and EV data information, and then re-formulate the optimization problem of the corresponding sliding window at the next moment. The advantages of implementing intraday rolling optimization scheduling are: first, rolling optimization can effectively exert the accuracy of short-term forecast information in the distribution network; second, the rolling forward form is more suitable for the characteristics of random EV entry.

[0080] When an EV is connected to a charging station at a certain node at a certain time, its basic information includes the total battery capacity B i , Start access time t i in , expected off-grid period t i dep , current state of charge soc i in , off-grid expected state of charge soc i dep , maximum charging power p i max Upload to the charging station controller.

[0081] At time t, the charge and energy information of the EV is first updated according to equations (7)-(8).

[0082]

[0083] Where: represents the initial access period of EV, p i (t-1) represents the charging power at time t-1, represents the expected off-grid state of charge, η is the efficiency of the charging device, soc i in(t) is the initial charge of EV at time t, E i n (t) is the total energy demand of EV at time t, E i (t) represents the dispatchability of EV. When the energy that EV can accumulate when charging at the maximum charging power is greater than its total required energy, it indicates that EV has the flexibility to adjust the charging power, that is, E i (t) greater than 0 indicates that the EV is dispatchable, otherwise the EV is not dispatchable.

[0084] For non-dispatchable EVs, consider them as having a power of p i (t) fixed load, where p i (t)Satisfy:

[0085]

[0086] For dispatchable EVs, their charging flexibility is determined by the upper and lower bounds of the charging power and the upper and lower bounds of the accumulated energy. In order to ensure that the EV can complete the charging task within the residence time, the expected energy within the sliding window is defined as the charging target of the EV in the rolling time domain. For individual EVs, it is necessary to achieve the required energy through power combinations at different times within their controllable time period. Equations (10) and (11) specify the upper and lower bounds of the power at each moment within the controllable time period, and their size depends on the EV charging power characteristics. In this paper, the default power lower bound is the state where the EV is not charging, such as Figure 3a The figure shows the boundary curve of EV charging power. Equations (12) and (13) define the boundary of cumulative energy. The lower bound of cumulative energy corresponds to the EV charging at the latest time to the time of leaving the grid to complete the charging task; the upper bound of cumulative energy corresponds to the EV starting to charge at maximum power at time t and completing the charging task ahead of schedule.

[0087] Charging power lower limit for:

[0088]

[0089] Charging power upper limit for:

[0090]

[0091] Cumulative energy upper bound for:

[0092]

[0093] Cumulative energy lower bound for:

[0094]

[0095] Expected Energy calculate:

[0096]

[0097] Where: η is the efficiency of the charging device, t i is the remaining controllable period of EV, satisfying t i =(t i dep -t) / Δt, L is the length of the sliding window, E i n is the total energy demand of EV, and They represent the lower and upper bounds of the charging power, respectively. and Respectively represent the lower and upper bounds of the cumulative energy, and Δt represents the unit time interval. Figure 3b As shown in , if the EV leaves within the sliding window corresponding to the current optimization moment, its expected energy is the demand energy updated at the current moment; Figure 3c As shown in the figure, if the EV does not leave the grid within the current sliding window, the expected energy of the electric vehicle within the sliding window is specified as The energy that can be achieved by charging with an average power strategy.

[0098] Hierarchical scheduling usually requires clustering based on the characteristic parameters of EVs. The present invention directly establishes an actual EVA model of energy aggregation in the station within the rolling optimization time domain from the perspective of real-time operation of the electric vehicle station. The constraint model of the electric vehicle station at node i is mainly determined by the model of the EVs that actually enter the station during the current control period, that is, the power and energy operation constraints of all EVs in the station shown in equations (10) to (14) are taken into account to form the power and energy demand constraints at the station level:

[0099]

[0100] Where: and represents the lower and upper bounds of the charging power of the jth EV, represents the set of all EVs at node i, represents the expected energy of the jth EV, and They represent the lower and upper bounds of the charging power of the charging station at node i, respectively. and They represent the lower and upper bounds of the accumulated energy of the charging station at node i respectively.

[0101] In the t-th control period, within the sliding window of length L, the charging station operation constraint model at node i should include the power constraint at each moment in the sliding window, the cumulative energy constraint at each moment in the sliding window, and the third term represents the overall expected energy constraint of the sliding window. Specifically, as shown in formula (16):

[0102]

[0103] Where: and are the lower and upper bounds of the actual EVA power in each time period within the window, P EVA,i (t+k) represents the actual EVA power at time t+k, and are the lower and upper bounds of the actual EVA accumulated energy in each period within the sliding window, is the expected energy of the actual EVA in the entire sliding window.

[0104] In addition, the intraday charging load forecast can continuously adjust the forecast value by combining the historical rules of vehicle charging with the real-time traffic network and vehicle operation status, which is easier to have higher prediction accuracy than the day-ahead forecast. Assume that at a certain moment, the charging station can use a certain prediction method to obtain the EVA charging demand information formed by the EV that is about to enter the station in the current rolling window. The charging load prediction model in the sliding window of length L in the tth control period is obtained:

[0105]

[0106] Where: and are the lower and upper bounds of the power of EVA in each time period within the window, represents the predicted EVA power at time t+k, and They are respectively the lower and upper bounds of the cumulative energy of EVA in each period within the sliding window.

[0107] Step 1 determines that the time scale of system operation is real-time scheduling within the day, so the optimization problem solved in step 3 is actually a rolling optimization problem. In each time period, firstly, based on the EV data at the current moment and the short-term predicted charging load and other load forecast data, the optimization problem in the current sliding window time domain is solved, and only the charging strategy of the charging station and the EV in the station at the current moment is implemented; then, as time moves forward, the load forecast information and EV data information are updated, and the optimization problem of the corresponding sliding window at the next moment is re-formulated. The advantages of implementing intraday rolling optimization scheduling are that, first, rolling optimization can effectively exert the accuracy of short-term forecast information in the distribution network; second, the rolling forward form is more suitable for the characteristics of random EV entry.

[0108] A distributed solution to this optimization scheduling problem. This solution is based on the alternating direction multiplier method and is a distributed solution for distribution network-multi-node charging stations. First, each charging station completes the total charging demand analysis within the station, takes the charging station operation model as a constraint, responds to the electricity price during the period, and uses the sub-controller to solve the sub-problem of minimizing the charging cost; then, the distribution network operator uses the distribution network main controller to use the system flow constraint as the optimization constraint to solve the sub-problem with the goal of minimizing network loss. The distribution network controller (master) and the charging station (sub) controller iteratively solve each sub-problem until convergence. The solution to each sub-problem is completed locally, mainly relying on the power decision made by the communication transmission between controllers. The optimization goal of the distribution network layer is to minimize the operating network loss, and the optimization goal of each charging station at the charging station layer is to minimize the charging cost. In summary, the system objective function expressed in the form of rolling optimization is:

[0109]

[0110] Where: L is the sliding window length, N ev For the electric car station assembly, P i (t) and P i pre (t) are the actual EVA and predicted EVA of the charging station at time t, Π(t) is the electricity price information at that time, is the total circuit loss of the distribution network, γ is the weight value, γ>0, and the larger the value is, the more the optimization objective focuses on reducing power loss rather than maximizing the utility of the charging station.

[0111] The solution to (18) is based on a hierarchical architecture and combines the ADMM algorithm to implement a distributed solution for the distribution network center-multi-node charging station. This can not only alleviate the computing pressure faced by the master controller, but also ensure the privacy of the charging station operator to a certain extent. The power and energy information data collection of EVs is completed in the charging station. After completing the information collection work, each charging station uses its own operating constraint model as the optimization constraint, responds to the electricity price during the period, and uses the sub-controller to solve the sub-problem of minimizing the charging cost; the distribution network operator uses the distribution network main controller to use the system flow constraint as the optimization constraint to solve the sub-problem with the goal of minimizing network loss. The sub-problem is iteratively solved by the distribution network controller (master) and the charging station (sub) controller until convergence. In the process of solving the two sub-problems, the constraints and optimization objectives are solved locally, and the optimization strategies formulated by each are transmitted in mutual communication.

[0112] Definition: P EVA,i and are the decision powers of the actual EVA and predicted EVA solved by the sub-controller, Q EVA,i The reactive power solved for the subcontrol. ik and is the decision power of the main controller at the kth iteration. and are the corresponding residuals respectively. The iterative solution steps based on the ADMM algorithm are as follows:

[0113] 1) For the kth iteration, each charging station sub-controller i first solves the optimization problem of the charging station node:

[0114]

[0115] Where: p and ρ q They represent the coefficients corresponding to the active and reactive powers to be solved, respectively. and Respectively represent the residuals corresponding to the active and reactive powers solved, P EVA,i (t) k , They represent the actual EVA charging power and the predicted EVA charging power of the charging station at node i at the kth iteration, respectively. EVA,i (t) k It represents the reactive power at node i (node ​​i is the charging station node) at the kth iteration.

[0116] After solving, update the strategy information:

[0117]

[0118] Each sub-controller completes a communication with the main controller and sends To the main controller;

[0119] 2) After receiving the strategy information from all sub-controllers, the distribution network master controller solves the optimization problem:

[0120]

[0121] After solving, update the strategy information:

[0122]

[0123] Where: P i (t) k+1 , Q i (t) k+1 They represent the active power and reactive power set by the main controller at the k+1th iteration, P i (t), Q i (t) is the solution of the optimization problem in 2), which serves as the active and reactive power of the main controller.

[0124] 3) Residual update:

[0125]

[0126] If v j k+1 with u j k+1 If the convergence criterion is met, the iterative loop is exited and the solution is completed; otherwise, the main controller will (P i (t) k+1 ,Q i (t) k+1 ) is sent to each sub-controller, the number of iterations + 1, and returns to step 1).

[0127] Step 1) corresponds to uploading the power demand of each charging station in the hierarchical architecture, and step 2) corresponds to the charging power plan formulation in the hierarchical architecture. Therefore, demand uploading and plan formulation are actually an iterative convergence process. The distribution network decision power at the time of convergence is the overall charging power of the charging station.

[0128] For individual EVs, it is necessary to combine the power at different times within its controllable period to achieve the required energy. After determining the charging boundary information (including prediction data) of all EVs in the charging station in the current period, combine them in time sequence to finally obtain the overall energy demand of the charging station, completing step 3.

[0129] like Figure 4 The figure shows the implementation mechanism of the fuzzy controller used in the electric vehicle power distribution involved in the embodiment of the present invention. Step 4 completes the charging power distribution of the charging station based on the fuzzy power distribution algorithm. The core of the fuzzy controller is the design of the fuzzy controller. For each EV, according to its own charging urgency U i n (t) and soc i (t) Obtain their respective "power allocation factors", which are used to measure the power allocation weight and affect the distribution of the total charging power in the station. The fuzzy controller can perform fuzzification after the parameter input, and obtain the output parameter "power allocation factor" after the fuzzy reasoning and defuzzification process based on the designed fuzzy rules. Finally, the power of the charging station is "fairly" allocated according to the size of the power allocation factor of each EV.

[0130] The original intention of EV users to join the network is to complete the charging task "as soon as possible" to meet travel needs. The present invention realizes real-time allocation of charging power based on the charging urgency of EV and the current charge amount, where the charging urgency of EV at time t is:

[0131]

[0132] It is expressed as the ratio of EV required energy to rechargeable energy. The closer the value is to 1, the higher the charging urgency is, and the closer the value is to 0, the lower the charging urgency is.

[0133] The power allocation algorithm steps for a single charging station-EV at time t are as follows:

[0134] 1) Receive the active power P of the node charging station at time t set by the distribution network optimization dispatch EVA,i (t);

[0135] 2) For non-dispatchable EVs, the rated power p j Allocation, for EVs off-grid at time t+1, the expected energy The remaining total power is obtained as shown in formula (25), where represents the set of all EVs in the charging station at node i; E j Indicates the dispatchability of the j-th EV. When the value is not greater than 0, it means that it is not dispatchable. represents the off-grid time of the jth vehicle;

[0136]

[0137] 3) Except for the EVs that have completed power allocation, the remaining EVs in the station constitute a set N res The power allocation factor of the jth EV in the set is inferred by the fuzzy controller as df j , the pre-allocated charging power of the jth EV is p j (t), satisfying:

[0138]

[0139] 4) The charging power of all dispatchable EVs at time t must meet the constraints of power and accumulated energy:

[0140]

[0141] Where: and They represent the lower and upper limits of EV charging power, respectively. and They represent the lower and upper limits of the accumulated energy of EV respectively.

[0142] The pre-allocated power p of each EV in step 3) j (t) Perform safety inspection and formulate the following rules: If the set N res If all EVs satisfy equation (26), the power allocation of the charging station is completed at time t; otherwise, the power of the individuals that do not meet the constraints is corrected: if there is a pre-allocated power p of the EV individual j Less than Then pj Corrected to If p j Greater than Then p j Corrected to Finally, the modified power is allocated to the EVs that do not meet the charging power constraint; res Eliminate the EV individuals that have completed power correction, update the remaining total power, and return to step 3) to redistribute power;

[0143] 5) The power allocated to the EV meets the safety inspection rules, and the power allocation of the charging station is completed at time t.

[0144] The input variables and output variables of the fuzzy logic controller are divided into three levels: low (L), medium (M), and high (H). The fuzzy membership function determines that when the EV charge state is at a low level or the charging urgency is at a high level, the power allocation factor output by the fuzzy controller is not at a low level. The specific fuzzy rules are shown in Table 1. When the current charge of the EV is at a low level or the charging urgency is at a high level, the power allocation factor is not low. The input variables of the controller are described by the Gaussian membership function, and the output variables are described by the triangular membership function, such as Figure 5a , Figure 5b , Figure 5c Shown is the membership function curve of the input and output variables of the fuzzy controller.

[0145] Table 1 Fuzzy rules

[0146]

[0147]

[0148] The present invention can be applied to the daily operation scenarios of the distribution network including charging stations, and can effectively achieve the goals of peak shaving and valley filling of the distribution network and improve the economic efficiency of charging station operation. The present invention establishes a hierarchical architecture model of EV orderly charging and an energy demand model of electric vehicle clusters in a rolling window, and proposes an EV charging strategy corresponding to the architecture. At the distribution network layer, based on the second-order cone relaxation technology, a distribution network operation optimization model is established with the optimal power flow operation of the distribution network and the minimum charging cost of the charging station as the goal, and the Alternating Direction Method of Multipliers ADMM algorithm is used to solve the optimization model; at the charging station layer, the charging power allocation of individual EVs is completed based on the fuzzy allocation mechanism, which can effectively reduce the maximum peak-to-valley difference of the system while improving the economic efficiency of the charging station operation. The charging station power allocation mechanism can meet the charging needs of EVs.

[0149] The above contents are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above examples. Any other modifications, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principles of the present invention shall be equivalent replacement modes and shall be included in the protection scope of the present invention.

Claims

1. A method for controlling the daily rolling orderly charging of electric vehicles under a layered architecture, characterized in that: The following steps are involved: Step 1: Determine the overall operation architecture and time scale of the system, including establishing a hierarchical control structure system of the distribution network layer and the charging station layer, and adopting the form of forward rolling optimization on the operation time scale; Step 2: At the charging station level, analyze and realize the aggregation of charging demands of electric vehicles, and finally form the overall charging demands of each charging station in the distribution network system; The overall charging demand of the charging station includes the total charging demand of all real electric vehicles currently in the station and the total charging demand of all predicted electric vehicles that are expected to enter the station within the current optimization window; The total charging demand of all real electric vehicles currently in the station is calculated as follows: Where: L represents the size of the rolling optimization window, t represents the current optimization period, and k represents the specific time within the period. and are the lower and upper bounds of the actual EVA power at node i in each time period within the window, P EVA,i (t+k) represents the actual charging station power at node i at time t+k, and They represent the lower and upper bounds of the accumulated energy of the actual EVA at node i in each period within the sliding window, is the expected energy of the actual EVA in the entire sliding window; The total charging demand of all predicted EVs that are expected to enter the station within the current optimization window is calculated as follows: Where: and are the lower and upper bounds of the power of EVA in each time period within the window, It represents the total charging power of EVA predicted at time t+k; Step 3: Based on the overall charging demand of the charging station obtained in step 2 and the safety constraints of the distribution network layer itself, complete the coordinated operation optimization of the distribution network and the charging station, and formulate a charging plan for the charging station; Step 4: Based on the fuzzy power allocation algorithm, each charging station will complete the charging power allocation of electric vehicles in the station according to the charging plan provided by the distribution network layer.

2. According to the hierarchical architecture of claim 1, the method for controlling the daily rolling orderly charging of electric vehicles is characterized by: In step 1, the system is divided into a distribution network layer and a charging station layer. The distribution network layer is based on the actual distribution network line model and is equipped with a distribution network controller with communication and computing functions; the charging station layer includes charging stations located at different nodes in the distribution network, and each charging station is equipped with a controller for controlling the charging power distribution of electric vehicles (Electrical vehicle EV) in the station and communicating with the upper distribution network.

3. The method for controlling the daily rolling orderly charging of electric vehicles under a layered architecture according to claim 2 is characterized by: The system's optimization form is a daily rolling optimization scheduling form that keeps rolling forward: first, based on the current EV data and short-term predicted charging load and other load forecast data, the optimal power flow problem of the distribution network-charging station in the corresponding sliding window time domain is solved, and only the charging strategy of the charging station and the electric vehicles in the station at the current moment is implemented; As time goes forward, the load forecast information and electric vehicle data information are updated, and the optimization problem of the sliding window corresponding to the next moment is reformulated.

4. The method for controlling the daily rolling orderly charging of electric vehicles under a layered architecture according to claim 1 is characterized in that: In step 2, when the electric vehicle is connected to the charging station, the basic charging information of the electric vehicle includes the total battery capacity B i , Start access time t i in , expected off-grid period t i dep , current state of charge soc i in , expected state of charge soc i d , maximum charging power p i max are all determined; and the electric vehicle charging boundary information is generated based on the above basic information, including the upper and lower limits of the charging power and the upper and lower limits of the accumulated energy.

5. The method for controlling the daily rolling orderly charging of electric vehicles under a layered architecture according to claim 1 is characterized in that: The optimization scheduling objectives of step 3 are to minimize network losses and charging costs: Where: N ev For the electric car station assembly, P i (t) and P i pre (t) are the actual EVA and predicted EVA of the charging station at time t, Π(t) is the electricity price information at that time, r j,k with i j,k are the impedance and current of the distribution network branch (j, k), respectively, and E represents the set of all distribution network branches. is the total loss of the distribution network circuit, and γ is the weight value.

6. The method for controlling the daily rolling orderly charging of electric vehicles under a layered architecture according to claim 1 is characterized by: Step 3 includes the following steps: first, each charging station completes the analysis of the total charging demand within the station, takes the charging station operation model as a constraint, responds to the electricity price during the period, and uses the sub-controller to solve the sub-problem of minimizing the charging cost; then, the distribution network operator uses the distribution network main controller to use the system flow constraint as the optimization constraint to solve the sub-problem with the goal of minimizing network loss; the distribution network controller and the charging station controller iteratively solve each sub-problem until convergence; the solution of each sub-problem is completed locally, and it is a power decision made by relying on communication transmission between controllers.

7. The method for controlling the daily rolling orderly charging of electric vehicles under a layered architecture according to claim 1 is characterized in that: Step 4 allocation is to realize the energy allocation mechanism of charging station-EV based on fuzzy rules. The allocation mechanism is: for each EV, the power allocation factor used to measure the power allocation weight is determined according to its own charging urgency and charge state. That is, the input of the fuzzy controller is the charging urgency and charge state of the EV, and the output is the power allocation factor; the input variables and output variables are divided into three levels: low (L), medium (M), and high (H).

8. A method for controlling the daily rolling orderly charging of an electric vehicle under a layered architecture according to any one of claims 1 to 7, characterized in that: In step 4, the charging power allocation of electric vehicles in the station is completed based on the fuzzy power allocation algorithm. The steps are as follows: (1) The node charging station receives the charging power at time t set by the distribution network layer optimization dispatch; (2) Power is allocated to EVs that cannot be dispatched in the station or are about to leave the station at the next moment according to their rated power or expected energy; (3) Except for the EVs that have completed power allocation, the remaining EVs in the station constitute the set N res , the EV parameter information in the set is input to the fuzzy controller, and the output power allocation factor of the jth EV is df j , and press df j The pre-allocated charging power p of the jth EV is obtained by numerical calculation j (t); (4) The pre-allocated power p of each EV in step (3) j (t) Perform safety inspection and formulate the following rules: If the set N res If all EVs in meet the EV power and energy boundary conditions, the power allocation of the charging station is completed at time t; otherwise, the power of the individuals that do not meet the constraints is corrected; (5) From the set N res Eliminate the EVs that have completed power correction, update the remaining total power, and return to step (3) to redistribute power; (6) The power allocated to the EV meets the safety inspection rules, and the power allocation of the charging station is completed at time t.

Images