Full power spectrum electric vehicle aggregated feasible region high-dimensional projection simplification method

Abstract

This invention provides a simplified high-dimensional projection method for the feasible region of full-power spectrum electric vehicle aggregation. First, a simplified model containing the aggregated power vector of the electric vehicle cluster is constructed. The model is defined by multiple linear inequality constraints, each constraint being in the form that the dot product of the constraint coefficient row vector and the aggregated power vector lies between upper and lower bounds. The constraint coefficient row vectors are generated using binary encoding, and the index set consists of constraints satisfying "number of non-zero segments + number of zero segments ≤ 3", resulting in a total of only 2T constraints. 2 The upper and lower bounds of the constraints are obtained by partitioning non-zero segments, recursively calculating auxiliary parameters, and accumulating the contributions of all electric vehicles. This invention does not require symmetric or constant assumptions about charging and discharging power and is applicable to electric vehicles with the full power spectrum. Compared with existing technologies, this model greatly reduces the computational burden while maintaining high accuracy, enabling second-level scheduling decisions and meeting the day-ahead, intraday, and real-time scheduling needs of large-scale electric vehicle clusters.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and specifically to a simplified method for high-dimensional projection of the feasible region of full-power spectrum electric vehicle aggregation. Background Technology

[0002] With the rapid growth of electric vehicle ownership, vehicle-to-grid (V2G) technology enables electric vehicle clusters to discharge during peak load periods and charge during off-peak periods, thereby improving grid economy and reliability. A prerequisite for achieving this goal is the ability to accurately and efficiently describe the aggregated power feasible region (AFR) of the electric vehicle cluster, which is the set of total charging and discharging power that all electric vehicles can provide at different time points.

[0003] Existing feasible region modeling methods can be mainly divided into three categories. The first category is the precise projection method, which is based on the high-dimensional polyhedron projection theory. Starting from the individual constraints of each electric vehicle, it obtains the precise feasible region at the cluster level through mathematical projection. For example, patent document CN 119272532 A discloses a method for feasible region aggregation of 5G base station backup energy storage based on constrained Chino polyhedra. It uses Minkowski summation to achieve rapid aggregation of feasible regions and utilizes the geometric properties of Chino polyhedra to reduce the computational complexity of projection.

[0004] However, the number of constraints in this type of method is strongly coupled with the number of resources and time nodes. When the number of resources exceeds several hundred, the number of constraints increases dramatically (for example, the number of constraints can reach O(m)). T (where m is the number of vehicles and T is the number of time points), which cannot be applied to the scheduling scenarios of large-scale electric vehicle clusters in practice.

[0005] The second category is approximation methods, which employ external or internal approximation strategies to approximate the true feasible region with a few constraints. For example, patent document CN 121835155 A discloses a grid flexibility resource aggregation method that introduces an internal approximation flexibility aggregation method to aggregate user-side distributed flexibility resources. Although this scheme considers grid topology and operational constraints and achieves node flexibility aggregation of multiple types of resources, the conservatism of the internal approximation method leads to a smaller aggregated feasible region, and it is not specifically optimized for fast-charging electric vehicles with dynamically changing power, still suffering from large approximation errors and inapplicability to full power spectrum scenarios. Meanwhile, the feasible region constructed by the external approximation method includes the true feasible region but carries the risk of infeasible scheduling commands; the internal approximation method guarantees the feasibility of scheduling commands but is too conservative, sacrificing scheduling flexibility. Approximation methods are usually simplified based on power symmetry assumptions or constant power assumptions, failing to accurately describe the highly asymmetric charging and discharging capabilities of fast-charging electric vehicles and the dynamic impact of state of charge on the power boundary.

[0006] The third category is a simplified method based on clustering or scenarios, which constructs a typical model after clustering electric vehicles according to parameters. During the clustering process, vehicles with different power levels are grouped into the same typical model, leading to an underestimation of the scheduling capability of high-power fast-charging vehicles. When the power distribution of electric vehicles spans nearly 100 times from 3.5 kW to 350 kW, the clustering method struggles to accurately characterize the full power spectrum characteristics (i.e., vehicles with different power levels) using a limited number of typical models.

[0007] In summary, the existing technology has the following core defects:

[0008] 1. The strong coupling between the number of constraints and the number of resources leads to excessively high computational complexity during large-scale aggregation;

[0009] 2. Simplified assumptions based on power symmetry or constant constraints lead to insufficient model accuracy, making them unsuitable for energy storage resources with dynamically changing power.

[0010] 3. Clustering methods lose accuracy due to parameter averaging and are difficult to cover the full power spectrum.

[0011] Therefore, there is a need for a simplified aggregation feasible region method that decouples the number of constraints from the number of resources, does not require power limit assumptions, and is applicable to electric vehicles with the full power spectrum. Summary of the Invention

[0012] To address the shortcomings of existing technologies, the purpose of this invention is to provide a simplified high-dimensional projection method for the aggregated feasible region of electric vehicles across the full power spectrum. This method utilizes structured constraints, reducing the number of constraints to a mere fraction. It is independent of the number of electric vehicles and has no power limitation assumptions, making it applicable to electric vehicles with a full power spectrum, such as those using fast charging.

[0013] To achieve the above objectives, the present invention provides a simplified high-dimensional projection method for the aggregated feasible region of full-power spectrum electric vehicles, comprising the following steps:

[0014] A simplified model of the aggregated feasible region of an electric vehicle cluster is constructed. The simplified model is expressed as follows: the aggregated power vector is located in a feasible set, which is defined by multiple linear inequality constraints. Each constraint is in the form that the dot product of the constraint coefficient row vector and the aggregated power vector is between the lower and upper bounds of the constraint, where the constraint index belongs to a specific index set.

[0015] The specific index set consists of the indices of all constraint coefficient row vectors that satisfy the condition that the sum of the number of non-zero segments and the number of zero segments is less than or equal to 3; the length of each constraint coefficient row vector is T, and each element takes the value 0 or 1;

[0016] The constraint coefficient row vector is generated according to the binary encoding rule: for any two time nodes a and b, if a≤b, the aggregate power in the corresponding continuous time period [a,b] is summed, and the index is equal to the sum of the binary bit weights of each time node in that time period; if a>b, the aggregate power in the first time period [1,b] and the last time period [a,T] is summed, and the index is equal to the sum of the binary bit weights of the two time periods.

[0017] The upper and lower bounds of the constraints are as follows: Each constraint coefficient row vector is divided into several non-zero segments; for each electric vehicle, the upper and lower bound auxiliary parameters of each segment are recursively calculated; the contributions of all electric vehicles are summed to obtain the upper and lower bounds of the constraints at the cluster level; the total number of constraints in the simplified model is... .

[0018] As a second aspect of the present invention, an electric vehicle cluster scheduling system is proposed, comprising: a data acquisition module for acquiring the rated charging and discharging power, battery capacity, charging and discharging efficiency, arrival and departure time, initial state of charge, and expected state of charge of electric vehicles; a constraint upper and lower bound calculation module for calculating the upper and lower bounds of each constraint based on the acquired data and according to the recursive accumulation method; a scheduling optimization module for solving the optimal aggregated power vector by using the grid-side or user-side objective as the objective function and combining the simplified model; and an instruction issuance module for decomposing the optimal aggregated power into charging and discharging instructions for individual electric vehicles and issuing them.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0020] 1. This invention introduces a constraint screening criterion of "number of non-zero fragments + number of zero fragments ≤ 3" to eliminate a large number of fine-grained constraints that contribute little to the feasible region boundary, retaining only the core constraints required to describe the aggregated power capability of electric vehicle clusters. This criterion ensures that the total number of constraints is precisely [value missing]. This method is completely decoupled from the number of electric vehicles. Furthermore, based on a recursive accumulation method using a non-zero fragment index set, the energy coupling constraint of each electric vehicle is precisely converted into upper and lower bounds of the linear combination of aggregated power, avoiding errors introduced by traditional approximation methods due to power limitation assumptions or symmetric simplifications.

[0021] 2. By using a recursive accumulation method based on a non-zero fragment index set, this invention accurately converts the energy constraint of each resource into the upper and lower bounds of the linear combination of aggregated power, avoiding the errors introduced by the power constraint assumption or symmetric simplification in traditional approximation methods, and ensuring model accuracy. Attached Figure Description

[0022] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0023] Figure 1 This is a schematic diagram of the overall processing flow for simplifying the high-dimensional projection of the feasible domain of full-power spectrum electric vehicle aggregation, as proposed in one embodiment of the present invention.

[0024] Figure 2 This is a flowchart of the constraint coefficient vector generation rules proposed in one embodiment of the present invention for illustrating the binary encoding rules and filtering process of continuous time period constraints and cross-time period constraints.

[0025] Figure 3 This is a flowchart illustrating the recursive calculation of auxiliary parameters for a single electric vehicle, as proposed in one embodiment of the present invention. Detailed Implementation

[0026] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0027] The present invention will be further described in detail below with reference to the accompanying drawings, but this is not intended to limit the scope of the invention.

[0028] like Figures 1-3 As shown, a simplified high-dimensional projection method for the aggregation feasible region of electric vehicles with full power spectrum is proposed, which includes the following steps:

[0029] I. Constructing a simplified model of the aggregated feasible region for electric vehicle clusters

[0030] The aggregated feasible region model is expressed as follows: the aggregated power vector lies within a feasible set, which is defined by multiple linear inequality constraints. Each constraint is in the form that the dot product of the constraint coefficient row vector and the aggregated power vector lies between the lower and upper bounds of the constraint, where the constraint index belongs to a specific index set.

[0031]

[0032] Wherein, the aggregate power vector is denoted as , its origin The aggregated power at each time point constitutes the power of the system. : Represents the aggregated power vector of the electric vehicle cluster across the entire time scale. Let T be the aggregate power at the t-th time point, in kW, with the superscript T indicating vector transpose; This represents the total number of time intervals within the scheduling period. For example, if the entire day is scheduled at 15-minute intervals, then... =96.

[0033] H2 is a specific index set consisting of all constraint coefficient vectors that satisfy the condition "Number of non-zero fragments + Number of zero fragments ≤ 3". The index is composed of H2, and the size of H2 is ; For the first The coefficient row vector corresponding to each constraint has a length of . Each element is either 0 or 1, and its function is to select the power of several time nodes from the aggregated power vector p and perform a linear combination. , The first The constraints have lower and upper bounds, in kW. Both represent the number of electric vehicles. The function is obtained by accumulating the energy constraint contribution of each vehicle.

[0034] The number of constraints obtained through screening in this model is: With the number of electric vehicles It is irrelevant, and each constraint is a linear combination of aggregate power with coefficients of only 0 or 1, which facilitates binary storage and fast calculation.

[0035] II. Structured Generation of Constraint Coefficient Vectors

[0036] The constraint coefficient row vector is generated according to the binary encoding rule: for any two time node numbers a and b, if a≤b, the aggregate power in the corresponding continuous time period [a,b] is summed, and the index is equal to the sum of the binary bit weights of each time node in that time period; if a>b, the aggregate power in the first time period [1,b] and the last time period [a,T] is summed, and the index is equal to the sum of the binary bit weights of the two time periods.

[0037] Specifically, first, define the set of time nodes. For any integer constraint coefficient vector The corresponding linear combinations fall into the following two categories.

[0038] The first constraint is a continuous time period (a≤b), which represents the sum of the aggregation power over a continuous time period from the a-th time node to the b-th time node:

[0039]

[0040] To facilitate computer storage and indexing, each constraint is assigned a binary-coded index. For continuous time period constraints, the index... The calculation formula is:

[0041]

[0042] The second constraint is the cross-time period constraint (a>b), which is used to handle scheduling scenarios that span across days or the next day. It represents the sum of the aggregated power in the first time period [1,b] and the last time period [a,T].

[0043]

[0044] Correspondingly, indexes with cross-time constraints The result is obtained by adding the two binary codes:

[0045]

[0046] It should be noted that: a and b are time node indices, with values ​​ranging from 1 to... When a≤b, the constraint covers the continuous time interval [a,b]; when a>b, the constraint covers two separate time intervals: [1,b] and [a,T]. The time node number, with a value range from 1 to... . The weights of the binary bits are used to generate a unique index. Since each time point corresponds to a bit, different combinations of (a,b) generate different indices, making it easier for the program to automatically generate all constraints.

[0047] Each constraint corresponds to a length of... 0-1 vector The position of '1' indicates the time node involved in its constraint summation. Index The binary representation uniquely determines the set of time nodes involved in the summation, enabling the constraint generation process to be fully automated.

[0048] Furthermore, a constraint filtering condition is introduced: "Number of non-zero fragments + Number of zero fragments ≤ 3". Here, a non-zero fragment is defined as... A subsequence consisting of consecutive 1s is defined as a subsequence consisting of consecutive 0s. The above condition is equivalent to the following: the time nodes involved in the power summation corresponding to the constraint form at most two consecutive intervals on the time axis. If a single consecutive interval is formed, it can be located at any position; if two consecutive intervals are formed, these two intervals must be located at the beginning and end of the time axis, respectively, with a discontinuity in between.

[0049] The rationale behind the above selection is that the boundary of the feasible region for electric vehicle cluster aggregation is primarily determined by the energy accumulation constraint within continuous charging and discharging periods. Overly fragmented 0-1 alternations (such as "1,0,1,0,1") contribute very little to the feasible region boundary and their number grows exponentially, significantly increasing the model size without improving accuracy. By limiting the sum of non-zero segments and zero segments to no more than 3, the following three types of constraints are precisely preserved: a single continuous period, two separate periods (one at the beginning and one at the end), and a continuous period with zeros at the beginning and end. These three periods can fully describe the upper and lower bounds of energy within the access period and the energy coupling relationship across periods (such as overnight).

[0050] Through derivation and experimental verification, all constraint coefficient vectors satisfying the above conditions have exactly the following: The number of constraints is quadratically related to the number of time nodes T, and completely decoupled from the number of electric vehicles m. If this condition is relaxed, the number of constraints will increase sharply to O(T³) or even higher, losing the simplification meaning; if it is tightened further, cross-time period constraints will be missed, resulting in an overly conservative feasible region.

[0051] III. Calculate the upper and lower bounds of the constraints

[0052] The upper and lower bounds of the constraints are obtained as follows: Each constraint coefficient row vector is divided into several non-zero segments; for each electric vehicle, the upper and lower bound auxiliary parameters of each segment are recursively calculated; then, the contributions of all electric vehicles are summed to obtain the cluster-level constraint upper and lower bounds; the total number of constraints in the simplified model is... .

[0053] Specifically, for arbitrary constraint indexes To calculate the upper and lower bounds of the total capacity of all electric vehicles in the cluster on this linear combination, a recursive accumulation method based on the non-zero fragment index set is proposed, which can accurately convert the energy coupling constraint of each electric vehicle into the linear constraint boundary of the aggregated power.

[0054] For any constraint, divide adjacent non-zero elements in its coefficient row vector into a non-zero segment, record the minimum and maximum indices of each segment, and count the total number of non-zero segments.

[0055] Specifically: A segment is formed by all adjacent non-zero elements in the array. Let the set of indices contained in the r-th non-zero segment be denoted as . The minimum index of this segment is The maximum index is The total number of non-zero segments is denoted as... .

[0056] The following concrete example illustrates the process of partitioning non-zero segments:

[0057] Set the total number of time nodes The coefficient row vector of a certain constraint The vector has a length of 5, and each position corresponds to a time node. The specific steps for the non-zero pollen fragment are as follows:

[0058] The first element is 1, and the second element is also 1. Since they are consecutive, they are grouped into the same non-zero segment, which contains the following set of indices: The minimum index is 1 and the maximum index is 2;

[0059] The third element is 0, which indicates the end of the current segment and the search for the next non-zero segment begins;

[0060] The fourth element is 1, appearing alone (the fifth element is 0), thus forming the second non-zero segment, which contains the following set of indices. Both the minimum and maximum indexes are 4;

[0061] The 5th element is 0, and there are no subsequent non-zero elements.

[0062] After the above division, two non-zero segments were obtained, therefore the total number of non-zero segments is... .

[0063] In this embodiment, for the first For each electric vehicle (k=1,…,m), define upper and lower bound auxiliary parameters, and recursively calculate them according to the order of non-zero segments. The inputs include: time intervals. (Given constants); Each vehicle at each time point upper and lower power limits , Each vehicle at each time point Energy upper and lower bounds at the end , .

[0064] First, the recursive formula for the upper bound auxiliary parameters: For the first non-zero segment (h=1), the upper bound starting parameter is the minimum index minus the lower bound of the energy corresponding to the first index divided by the time interval length. The upper bound ending parameter is the smaller value between the upper bound starting parameter plus the total chargeable / dischargeable capacity of this segment according to the power upper bound and the upper bound of the energy corresponding to the maximum index of this segment divided by the time interval length.

[0065]

[0066]

[0067] For subsequent non-zero segments (h=2,…,Zn), the upper bound start parameter is the larger of the following: the upper bound end parameter of the previous segment plus the lower bound of the power-based total chargeable / dischargeable capacity of the interval period, and the lower bound of the energy corresponding to the minimum index of the current segment minus one, divided by the time interval length. The upper bound end parameter is the smaller of the following: the current upper bound start parameter plus the lower bound of the power-based total chargeable / dischargeable capacity of this segment, and the upper bound of the energy corresponding to the maximum index of the current segment, divided by the time interval length.

[0068]

[0069]

[0071] Similarly, the recursive formula for the lower bound auxiliary parameter, for the first non-zero segment (h=1), is as follows: the lower bound starting parameter is the minimum index minus the energy upper bound corresponding to the first index divided by the time interval length; the lower bound ending parameter is the larger of the lower bound starting parameter plus the total chargeable / dischargeable capacity of this segment according to the power lower bound and the energy lower bound corresponding to the maximum index of this segment divided by the time interval length.

[0072]

[0073]

[0074] For subsequent non-zero segments (h=2,…,Zn), the lower bound start parameter is the smaller of the following: the lower bound end parameter of the previous segment plus the total chargeable / dischargeable capacity according to the power upper bound of the interval period, and the energy upper bound corresponding to the minimum index of the current segment minus one, divided by the time interval length. The lower bound end parameter is the larger of the following: the current lower bound start parameter plus the larger of the total chargeable / dischargeable capacity according to the power lower bound of this segment and the energy lower bound corresponding to the maximum index of the current segment, divided by the time interval length.

[0075]

[0076]

[0077] In the formula, To convert the lower bound of energy before the start of the first non-zero segment into an equivalent power contribution; This represents the total chargeable / dischargeable capacity within the h-th non-zero segment, based on the upper power bound. and The operation reflects the combined constraints of the energy upper bound and the power upper bound; the total achievable power cannot exceed the power value after the energy boundary transformation; for the lower bound recursion, and The roles are reversed because the lower bound is determined by the lower bound of energy and the lower bound of power (which are usually negative).

[0078] After obtaining the auxiliary parameters for each vehicle, the upper and lower bounds of the cluster-level constraints can be obtained by summing them. The upper bound of the cluster-level constraint n. The sum of contributions to the upper bound of all electric vehicles:

[0079]

[0080] Similarly, the lower bound The sum of contributions to the lower bound of all electric vehicles:

[0081]

[0082] It should be noted that each non-zero segment For a continuous time window, the aggregated power within the window is constrained by the energy state before the window begins, the charge / discharge power limit within the window, and the energy boundary at the end of the window. The total computational cost of calculating all constraint upper and lower bounds is approximately [missing information]. That is, the number of electric vehicles Proportional to the number of time points The computational load is proportional to the square of the value. For example, as the number of vehicles or the time resolution increases, the computational load will increase accordingly. However, since the auxiliary parameter calculations for each electric vehicle are independent, multi-threaded or distributed parallel computing is used in practical engineering to distribute the computational tasks across multiple processing units, thereby significantly reducing the total time consumption and meeting real-time scheduling requirements.

[0083] Regarding the constraint size, the total number of constraints is: .For example:

[0084] like =24 (hourly scheduling) =1 hour), the constraint number is 2×24 2 =1152;

[0085] like =96 (15-minute interval), constraint number is 2×96 2 =18432.

[0086] Therefore, linear constraints can be solved in seconds using commercial solvers (such as Gurobi and CPLEX), fully meeting real-time scheduling requirements. More importantly, this model requires no power symmetry assumptions or power constraint simplifications, making it applicable to electric vehicles across the entire power spectrum, including fast-charging vehicles with rated power up to hundreds of kilowatts.

[0087] The following is a detailed explanation using the daytime dispatching of an electric vehicle cluster in a certain city as an example:

[0088] I. Setting Scene Parameters

[0089] Scheduling timescale: one day, time interval , Hour (15 minutes).

[0090] Number of electric vehicles: Vehicles.

[0091] Vehicle parameters: Rated charging and discharging power is as follows: Battery capacity Charge and discharge efficiency .

[0092] Travel patterns: 50% of vehicles arrive early and leave early (arrival time 8:00 AM, departure time 5:00 PM), and 50% arrive late and leave the next day (arrival time 6:00 PM, departure time 8:00 AM the next day). Initial SOCs all comply with... Cut off to [0.2, 0.8], with an expected SOC of 0.9.

[0093] The grid-side objective is to minimize the net load peak-to-valley difference, which is equivalent to minimizing the load variance.

[0094] II. Model Construction and Solution Steps

[0095] Step 1: Generate a set of constraint coefficient vectors ;

[0096] right Generate all that satisfy Continuous time period constraints and Cross-time constraints, total There are several combinations. Each combination corresponds to one... And calculate the index h according to the aforementioned binary encoding formula. The set of all combinations is... The total number of constraints is .

[0097] Step 2: Calculate the individual parameter sequence for each vehicle;

[0098] Take a typical "early arrival, early departure" vehicle as an example (time index) 8:00 corresponds to the index 17:00 corresponding index ):

[0099] Before the arrival time ( The vehicle is not connected to the power grid, and its power limits are both 0.

[0100] During access ( The upper limit of power is The lower bound of power is ;

[0101] After the departure time ( The upper and lower bounds of the power are both 0.

[0102] Specifically, the calculation of the upper and lower bounds of energy: , Furthermore, the charging and discharging efficiency needs to be considered for recursive updates.

[0103] Step 3: For each constraint n, perform non-zero segmentation and calculate auxiliary parameters;

[0104] With constraint n corresponding For example, if a=1 and b=2:

[0105] There is only one non-zero segment: .

[0106] Upper bound auxiliary parameter calculation:

[0107]

[0108]

[0109] If the vehicle is Not yet connected to the grid Substituting, we get:

[0110]

[0111] Similarly, the lower bound auxiliary parameter can be obtained: .

[0112] Therefore, the vehicle's contribution to the constraints is 0, and it does not change the boundaries of the constraints.

[0113] Then, corresponding to constraint n (Right now For example, if the vehicle is in Access at any time will result in non-zero fragment coverage. At this point, the auxiliary parameters calculated have a positive contribution.

[0114] Step 4: Sum the contributions of all vehicles to obtain the upper and lower bounds of the cluster;

[0115] For each constraint All The car (like Then sum the individual segments to get the result. Similarly, we can obtain .

[0116] Step 5: Invoke the optimization solver;

[0117] Using minimizing the load variance as the objective function, the feasible region of 18432 linear constraints and aggregate power is calculated. Input the solution into a commercial solver (such as Gurobi or CPLEX) to obtain the optimal aggregate power sequence. .

[0118] Step 6: Instruction decomposition and issuance;

[0119] Using the energy margin ratio allocation method, each The charging and discharging power of each electric vehicle is broken down and transmitted to each vehicle via a communication network.

[0120] In actual operation, this model demonstrates excellent computational efficiency. On a typical workstation configured with an Intel i7 processor and 16GB of memory, it takes only about 0.3 seconds from inputting the original data to constructing all 18,432 constraints; subsequently, solving for the optimal aggregated power sequence using a commercial optimization solver takes only about 2.1 seconds. Therefore, day-ahead scheduling decisions can be completed within 3 seconds, fully meeting the real-time requirements of the actual power grid for scheduling response.

[0121] This model can be directly used for the scheduling and control of electric vehicle clusters. As a specific embodiment, this invention provides an electric vehicle cluster scheduling system, which includes a data acquisition module, a constraint upper and lower bound calculation module, a scheduling optimization module, and an instruction issuance module.

[0122] The data acquisition module is responsible for acquiring the rated charging and discharging power, battery capacity, charging and discharging efficiency, arrival and departure time, initial state of charge, and expected state of charge for each electric vehicle. The constraint upper and lower bound calculation module calculates the upper and lower bounds of each constraint based on the acquired data, following the aforementioned recursive accumulation method. The scheduling optimization module uses grid-side objectives (such as minimizing load variance) or user-side objectives as the objective function, and solves for the optimal aggregated power vector using a simplified model. The command issuance module decomposes the optimal aggregated power into charging and discharging commands for individual electric vehicles and issues them to each vehicle. This system can be deployed in a grid energy management system or a virtual power plant aggregation platform to achieve day-ahead, intraday, and real-time scheduling.

[0123] Furthermore, this model is not only applicable to electric vehicle clusters but can also be easily transferred to other battery-based distributed energy sources. For example, for stationary energy storage systems and integrated photovoltaic energy storage devices, simply replace the individual parameters in the model with the parameters of the corresponding devices: replace the upper and lower bounds of charging and discharging power with the power boundary of the device, replace the upper and lower bounds of energy with the energy storage boundary of the device (determined by capacity and SOC limitations), and replace the arrival and departure times of vehicles with the available time window of the device. After completing the above replacements, the aggregate feasible region of this type of resource can be directly constructed without modifying the constraint form or calculation method.

[0124] For example, suppose there is an energy storage power station consisting of 100 stationary energy storage units, and the parameters of each unit are: power... ,capacity Available all day (i.e.) (All are connected). During migration, simply set the upper limit of power to 100 kW and the lower limit to -100 kW, and the upper limit of energy to 200 kWh and the lower limit to 0 kWh (assuming the lower limit of SOC is 0), keep other formulas unchanged, and construct the upper and lower limits of constraints in the same steps to obtain the aggregated feasible region of the energy storage power station, which can be used for ancillary services such as grid frequency regulation.

[0125] The simplified model proposed in this invention can be directly embedded into existing power grid dispatching systems (such as energy management systems (EMS) and aggregator dispatching platforms), and is suitable for day-ahead, intraday, and real-time dispatching of large-scale electric vehicle clusters. Furthermore, this model can be used for the aggregated characterization of battery-powered distributed energy resources in virtual power plants, and has broad industrial application prospects.

[0126] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A simplified high-dimensional projection method for the aggregated feasible region of full-power spectrum electric vehicles, characterized in that, Including the following steps: A simplified model of the aggregated feasible region of an electric vehicle cluster is constructed. The simplified model is expressed as follows: the aggregated power vector is located in a feasible set, which is defined by multiple linear inequality constraints. Each constraint is in the form that the dot product of the constraint coefficient row vector and the aggregated power vector is between the lower and upper bounds of the constraint, where the constraint index belongs to a specific index set. The specific index set consists of the indices of all constraint coefficient row vectors that satisfy the condition that the sum of the number of non-zero segments and the number of zero segments is less than or equal to 3; the length of each constraint coefficient row vector is T, where T is the total number of time nodes, and each element takes the value 0 or 1; The constraint coefficient row vector is generated according to the binary encoding rule: for any two time nodes a and b, if a≤b, the aggregate power in the corresponding continuous time period [a,b] is summed, and the index is equal to the sum of the binary bit weights of each time node in that time period; if a>b, the aggregate power in the first time period [1,b] and the last time period [a,T] is summed, and the index is equal to the sum of the binary bit weights of the two time periods. The upper and lower bounds of the constraints are as follows: Each constraint coefficient row vector is divided into several non-zero segments. For each electric vehicle, the upper and lower bound auxiliary parameters of each segment are recursively calculated. Then, the contributions of all electric vehicles are summed to obtain the upper and lower bounds of the constraints at the cluster level. The total number of constraints in the simplified model is... .

2. The method according to claim 1, characterized in that, In the filtering condition that the number of non-zero segments plus the number of zero segments is less than or equal to 3: non-zero segments are defined as subsequences with consecutive values ​​of 1 in the constraint coefficient row vector, and zero segments are defined as subsequences with consecutive values ​​of 0; the condition is equivalent to the power summation window corresponding to the constraint forming at most two consecutive segments on the time axis, and when two consecutive segments are formed, they are located at the beginning and end of the time axis respectively, with exactly one zero segment interruption in between.

3. The method according to claim 1, characterized in that, The method for dividing non-zero segments is as follows: for any constraint, divide adjacent non-zero elements in its coefficient row vector into a non-zero segment, record the minimum index and maximum index of each segment, and count the total number of non-zero segments.

4. The method according to claim 3, characterized in that, The recursive calculation method for the upper bound auxiliary parameter is as follows: For the first non-zero segment, the upper bound start parameter is the minimum index minus the lower bound of the energy corresponding to the first index divided by the time interval length. The upper bound end parameter is the smaller of the upper bound start parameter plus the total chargeable / dischargeable capacity of this segment by the upper bound of the power limit and the upper bound of the energy corresponding to the maximum index of this segment divided by the time interval length. For subsequent non-zero segments, the upper bound start parameter is the larger of the previous segment's upper bound end parameter plus the lower bound of the power limit of the interval period divided by the lower bound of the energy corresponding to the minimum index of the current segment minus the first index divided by the time interval length. The upper bound end parameter is the smaller of the current upper bound start parameter plus the total chargeable / dischargeable capacity of this segment by the upper bound of the power limit and the upper bound of the energy corresponding to the maximum index of the current segment divided by the time interval length.

5. The method according to claim 4, characterized in that, The recursive calculation method for the lower bound auxiliary parameter is as follows: For the first non-zero segment, the lower bound start parameter is the minimum index minus the upper bound of the energy corresponding to the first index divided by the time interval length. The lower bound end parameter is the larger of the lower bound start parameter plus the total chargeable / dischargeable capacity of this segment by the lower bound of the power index and the lower bound of the energy corresponding to the maximum index of this segment divided by the time interval length. For subsequent non-zero segments, the lower bound start parameter is the smaller of the lower bound end parameter of the previous segment plus the total chargeable / dischargeable capacity of the interval by the upper bound of the power index and the lower bound of the energy corresponding to the minimum index of the current segment minus the first index divided by the time interval length. The lower bound end parameter is the larger of the current lower bound start parameter plus the total chargeable / dischargeable capacity of this segment by the lower bound of the power index and the lower bound of the energy corresponding to the maximum index of the current segment divided by the time interval length.

6. The method according to claim 4 or 5, characterized in that, The upper and lower bounds of the constraint are obtained by summing the contributions of all electric vehicles: the upper bound of the constraint is equal to the sum of the upper bound termination parameters of all non-zero segments of all electric vehicles minus the upper bound start parameters; the lower bound of the constraint is equal to the sum of the lower bound termination parameters of all non-zero segments of all electric vehicles minus the lower bound start parameters.

7. The method according to claim 6, characterized in that, The number of constraints in the simplified model is independent of the number of electric vehicles, and the complexity of calculating the upper and lower bounds of the constraints is O(n). Furthermore, the auxiliary parameters of each vehicle are calculated independently, supporting parallel acceleration.

8. The method according to claim 1, characterized in that, Each aggregated power in the aggregated power vector takes a positive value during charging and a negative value during discharging, or adopts a bidirectional power definition according to the actual scheduling scenario.

9. The method according to claim 1, characterized in that, The simplified model is embedded in the power grid energy management system for day-ahead scheduling, intraday rolling scheduling, or real-time scheduling, wherein the total number of time nodes is determined according to the scheduling interval.

10. An electric vehicle cluster scheduling system employing the method according to any one of claims 1 to 9, characterized in that, include: Data acquisition module: acquires the rated charging and discharging power, battery capacity, charging and discharging efficiency, arrival and departure time, initial state of charge and expected state of charge of electric vehicles; Constraint upper and lower bound calculation module: Based on the collected data, calculate the upper and lower bounds of each constraint according to the recursive accumulation method; The scheduling optimization module uses the grid-side or user-side target as the objective function and combines the simplified model to solve for the optimal aggregated power vector. Command issuance module: Decomposes the optimal aggregated power into charging and discharging commands for individual electric vehicles and issues them.

Images