Automobile charging scheduling method and system based on genetic programming and capacity guardrail layer

Abstract

The present application relates to a kind of genetic programming and capacity guardrail layer-based car charging scheduling method and system, method includes: obtaining vehicle side state variable, resident basic load, distribution transformer parameter and preset safety control parameter;Vehicle side state variable is input into genetic programming priority rule tree, and priority score is calculated and priority sequence is generated;From capacity guardrail layer, transformer safe active power limit and active power budget are calculated, in combination with priority sequence, it is converted into the charging power instruction of each scheduling period;Charging power instruction is issued to charging pile execution, and charging scheduling control is completed;Finally, according to the total active load trajectory of transformer of basic load and the charging power instruction issued, offline execution heat diagnosis and node voltage screening, assess deployment feasibility.Compared with prior art, the present application realizes the decoupling of adaptive interpretable ordering and deterministic capacity protection, guarantees transformer safety and power grid smooth operation.

Description

Technical Field

[0001] This invention relates to the technical field of charging scheduling, and in particular to a vehicle charging scheduling method and system based on genetic programming and capacity guardrail layer. Background Technology

[0002] With the rapid growth in the number of electric vehicles, the concentrated charging load in residential areas at night is putting new capacity pressure on low-voltage power distribution networks and distribution transformers. Many old residential areas have distribution transformers, low-voltage cables, and parking space wiring mainly configured according to traditional residential loads, without reserving large-scale AC charging capacity for electric vehicles.

[0003] During typical evening peak hours, both residential load and electric vehicle charging load may reach high levels simultaneously. If uncontrolled charging is used, the electric vehicle load will be directly superimposed on the residential load, which may lead to excessively high load rate of distribution transformers, increased hot spot temperature, accelerated insulation aging, and insufficient voltage margin at the feeder end.

[0004] Existing electric vehicle charging scheduling methods mainly include mathematical optimization methods, heuristic priority rule methods, and data-driven control methods. Mathematical optimization methods can achieve better results when the model is complete and the prediction is accurate, but they often face problems in older residential areas due to incomplete data on load forecasting, user behavior, feeder parameters, and phase allocation. Heuristic priority rules are transparent and interpretable, but they struggle to balance average waiting time and fairness among users when capacity is scarce. Data-driven control methods are adaptable, but without an auditable execution layer, they cannot meet the interpretability requirements of protection proximity applications.

[0005] Genetic programming can evolve symbolic priority rules, facilitating the mapping of vehicle state variables to auditable charging priority scores. However, relying solely on learned priority rules cannot guarantee the safe operation of transformers. Especially under conditions of rapid changes in residential base load, fluctuations in the number of vehicles connected, differences in maximum charging power of vehicles, and the need for smooth issuance of scheduling instructions, a deterministic safety execution layer is still required to convert the priority sequence into power instructions that meet capacity constraints.

[0006] Furthermore, transformer capacity constraints are not equivalent to thermal safety and voltage safety. The IEC 60076-7 thermal model shows that load trajectory and ambient temperature affect hot spot temperature and insulation aging rate; low-voltage feeder node voltage is also affected by feeder length, line impedance, phase imbalance, and charging pile location.

[0007] Patent application CN120634174A discloses a method and system for orderly charging scheduling of electric vehicles in old communities. The method includes: acquiring charging demand data of electric vehicles in charging stations; calculating and classifying charging priorities based on the charging demand data; and constructing a charging prediction model; collecting cable temperature data and determining if there is an overheating risk; if an overheating risk exists, charging is suspended; if no overheating risk exists, the expected total charging power of the charging station is calculated based on the charging prediction model, and the transformer load rate of the power grid where the charging station is located is obtained based on the expected total charging power; and charging decisions are made based on user priority levels and the transformer load rate of the power grid. However, this method relies on online rules based on cable temperature and transformer load rate for coarse-grained stage-by-stage disconnection or power reduction, lacking an adaptive and interpretable sorting mechanism based on data-driven decoupling from physical constraints. Furthermore, it cannot achieve smooth closed-loop control in online scheduling or offline assessment of long-term transformer thermal aging and end-voltage over-limit risks. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a vehicle charging scheduling method and system based on genetic programming and capacity barrier layer. This method and system achieves adaptive and interpretable sorting of vehicle charging priorities through genetic programming rule trees, combines a capacity barrier layer to deterministically limit charging power and issue closed-loop control, and utilizes offline thermal and voltage diagnostics to prevent transformer overheating and aging and the risk of terminal voltage exceeding the limit.

[0009] The objective of this invention can be achieved through the following technical solutions: A vehicle charging scheduling method based on genetic programming and capacity barrier layers includes the following steps: Acquire vehicle-side state variables, residential basic load, power distribution transformer parameters, and preset safety control parameters for charging piles connected to residential areas; The vehicle-side state variables are input into a pre-trained genetic programming priority rule tree, and the priority score of each access vehicle is output. The access vehicles are then sorted according to the priority scores to obtain a priority sequence. The capacity barrier layer calculates the safe active power limit of the distribution transformer and the active power budget for the current scheduling period based on the parameters of the distribution transformer and the preset safety control parameters. Combined with the priority sequence, it converts the priority sequence of each access vehicle into a charging power instruction for each scheduling period. The capacity barrier layer is a deterministic safety execution unit independent of the genetic programming priority rule tree. The priority score is only used for sorting. The final charging power of each access vehicle is determined by the capacity barrier layer within the active power budget. The charging power instructions for each scheduling period are sent to the charging piles corresponding to each connected vehicle for execution, thereby completing the charging scheduling control of electric vehicles.

[0010] Furthermore, the vehicle-side state variables include connection duration, current battery level, state of charge, average charging power achieved, maximum AC charging power, and state of charge when the vehicle arrives.

[0011] Furthermore, the process by which the capacity guardrail layer calculates the safe active power limit and the active power budget of the distribution transformer based on the parameters of the distribution transformer is as follows: The rated capacity of the transformer in the distribution transformer parameters is multiplied by the static derating factor, safe load rate factor and primary reserve margin factor in the preset safety control parameters to obtain the safe active power limit of the distribution transformer. Calculate the difference between the safe active power limit of the distribution transformer and the current residential basic load. If the difference is greater than zero, the difference is determined as the active power budget for electric vehicle charging during the current scheduling period; if the difference is less than or equal to zero, the active power budget is zero.

[0012] Furthermore, the step of converting the priority sequence of each access vehicle into charging power instructions for each scheduling period also includes a rank-sharing allocation step: The ranking weight of each vehicle is calculated based on the total number of currently connected vehicles and the ranking of each vehicle in the priority sequence, with the highest priority ranking corresponding to the largest weight. The ranking weights of each vehicle are normalized to obtain the normalized budget weights; Calculate the product of the normalized budget weight and the active power budget for each vehicle, and take the smaller of the product and the maximum AC charging power allowed by the vehicle or charging pile as the initial power allocation for the corresponding vehicle.

[0013] Furthermore, the step of converting the priority sequence of each access vehicle into charging power commands for each scheduling period also includes a trickle processing step: Determine whether the state of charge of the connected vehicle has reached the preset trickle threshold; When the state of charge of the connected vehicle reaches the preset trickle threshold, the charging power command of the vehicle is switched to the preset trickle power.

[0014] Furthermore, the step of converting the priority sequence of each access vehicle into charging power instructions for each scheduling period also includes a reserved buffer pruning and backfilling step: The reserved buffer threshold is obtained by multiplying the secondary reserve margin coefficient in the preset safety control parameters with the safe active power limit of the distribution transformer. When the sum of the resident basic load and the current allocated power of each access vehicle exceeds the reserved buffer threshold, the allocated power of non-trickle-down vehicles is reduced in order of priority from low to high. The remaining available capacity is obtained by calculating the difference between the active power budget and the sum of the current allocated power of each connected vehicle. When the remaining available capacity is greater than zero, the remaining available capacity is allocated to vehicles that have not yet reached the maximum AC charging power in descending order of priority.

[0015] Furthermore, the step of converting the priority sequence of each access vehicle into charging power commands for each scheduling period also includes hard truncation, ramp limiting, smoothing, and feasible projection steps: When the sum of the current allocated power of each connected vehicle exceeds the active power budget, the active power budget is divided by the sum of the current allocated power of each connected vehicle to obtain the hard cutoff ratio, and the charging power of vehicles with a current allocated power greater than zero is reduced proportionally according to the hard cutoff ratio. Limit the change in allocated power of each access vehicle in adjacent scheduling periods to no more than the set single-step change limit; The average power of the vehicle distribution is calculated using a three-step moving average to obtain the smoothed power. When the sum of the smoothed power of each access vehicle does not exceed the active power budget, the smoothed power of each access vehicle is used as the charging power instruction for each scheduling period; when the sum of the smoothed power of each access vehicle exceeds the active power budget and the total power is greater than zero, the smoothed power of each access vehicle is scaled according to the ratio of the active power budget to the sum of the smoothed power to obtain the final charging power instruction for each scheduling period.

[0016] Furthermore, after completing the charging scheduling and control steps for the electric vehicle, the method further includes: The total active load trajectory of the transformer is synthesized based on the basic load of residents and the charging power instructions issued for each scheduling period. Based on the total active load trajectory of the transformer, the distribution transformer thermal diagnosis and feeder node voltage screening are performed offline to assess the safety feasibility after the charging power instructions for each scheduling period are issued and executed.

[0017] Furthermore, the offline execution of distribution transformer thermal diagnosis and feeder node voltage screening based on the total active load trajectory of the transformer specifically includes: The load factor is calculated by dividing the total active load trajectory of the transformer by the product of the power factor, the rated capacity of the transformer, and the static derating factor. The steady-state top-level oil temperature rise is calculated based on the load factor, and the current top-level oil temperature is updated by combining the ambient temperature, oil time constant and scheduling time step. The hot spot temperature is calculated based on the top oil temperature, rated hot spot temperature rise, load factor, and winding index. The instantaneous aging acceleration factor is calculated based on the hot spot temperature, and the equivalent aging factor is obtained by averaging the instantaneous aging acceleration factor over a period of time within the evaluation period. Based on the three-phase forward-backward power flow calculation method, the phase voltage change is calculated by multiplying the voltage sensitivity matrix with the electric vehicle charging power vector. Calculate the voltage of each phase at each node in each time period, and subtract the lower limit threshold of the low-voltage node phase voltage to obtain the voltage margin; The period when the voltage margin is less than zero is statistically analyzed, and the statistical results of the lowest node phase voltage and the over-limit time step are obtained.

[0018] A car charging scheduling system based on genetic programming and capacity guardrail layer includes: The status acquisition module is used to collect feature vectors of vehicles connected to charging piles in residential areas, monitor the basic load of residents, and read the parameters of the distribution transformer and preset safety control parameters. The genetic programming priority decision module is used to receive the feature vector output by the state acquisition module, input it into a pre-trained symbolic expression to calculate the priority score, and generate a priority sequence in descending order. The capacity guardrail layer execution module is used to receive the priority sequence output by the genetic programming priority decision module, and generate charging power instructions for each scheduling period by multi-stage safety allocation execution capacity constraint guarantee and instruction smoothing post-processing based on the distribution transformer parameters and preset safety control parameters. The instruction issuing module is used to issue the charging instruction to the charging pile corresponding to each connected vehicle through the charging pile communication interface, and to record the execution log. The thermal diagnostic module is used to calculate the hot spot temperature and assess the equivalent aging based on the transformer thermal model according to the total active load trajectory of the transformer output by the instruction issuing module, and to feed back the generated thermal status report to the genetic programming priority decision module. The voltage screening module receives the charging command output by the thermal diagnostic module, performs voltage over-limit detection based on the sensitivity matrix of the power flow calculation method, and conducts bidirectional data interaction with the capacity guardrail layer execution module to complete the deployment feasibility assessment.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. Improve the adaptability and fairness of vehicle sorting in scenarios with scarce capacity: This invention obtains the vehicle-side state variables and inputs a pre-trained genetic programming priority rule tree to calculate the priority score of each access vehicle and generate a priority sequence, thereby realizing dynamic adaptive sorting of charging priority.

[0020] 2. Reduce transformer overload risk and achieve decoupling of intelligent sequencing and safe execution: This invention calculates the safe active power limit and active power budget based on transformer parameters and preset safety control parameters through the capacity guardrail layer, and converts them into charging power instructions for each scheduling period in combination with the priority sequence, so as to strictly constrain the power allocation within the transformer safety budget and reduce the risk of transformer overload operation.

[0021] 3. Ensure closed-loop scheduling control and improve the smoothness of instruction issuance and equipment stability: This invention completes the smooth closed-loop control of the electric vehicle charging process by issuing the charging power instructions generated by the capacity guardrail layer for each scheduling period to the charging piles corresponding to each connected vehicle.

[0022] 4. Clearly define the online and offline boundaries to prevent transformer overheating and aging risks and voltage exceeding limits: This invention synthesizes the total active load trajectory of the transformer based on the basic load of residents and the charging power instructions issued for each scheduling period after the completion of charging scheduling control, and performs offline thermal diagnosis of the distribution transformer and voltage screening of feeder nodes based on this trajectory, thereby realizing a system assessment of the safety feasibility after the issuance and execution of scheduling instructions. Attached Figure Description

[0023] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the scheduling, diagnosis, and deployment evaluation process of this invention. Figure 3 Flowchart of thermal replay of transformer thermal model; Figure 4 This is a system module structure diagram of the present invention; Figure 5 A scenario diagram showing the basic load of residents, electric vehicle charging demand, and transformer capacity constraints. Figure 6 A comparison of the Pareto front of a scheduling method combining genetic programming and capacity guardrail layers with a benchmark scheduler; Figure 7 A comparison plot of priority response surfaces for genetic programming rules and composite rules; Figure 8 Comparison chart of service time and aging between capacity guardrail layer protection scheduling and uncontrolled charging; Figure 9 A schematic diagram illustrating the convergence and Pareto scale dynamics of genetic programming training. Figure 10 A comparison chart of operational boundary robustness; Figure 11 A comparison chart of single evaluation times for six-feature and fifteen-feature configurations; Figure 12 Comparison of hypervolume distribution configurations for six-feature and fifteen-feature configurations; Figure 13 This is a graph showing the voltage screening results of the three-phase forward-backward substitution method. Detailed Implementation

[0024] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0025] Example 1 This embodiment provides a vehicle charging scheduling method based on genetic programming and capacity barrier layers, such as... Figure 1 As shown, the specific process of this method is as follows: The system acquires vehicle-side state variables, residential base load, distribution transformer parameters, and preset safety control parameters for charging piles connected to residential areas. Vehicle-side state variables include connection duration, current battery level, state of charge (SOC), average received charging power, maximum AC charging power, and SOC reached. These six variables can be obtained from the charging pile, vehicle session management system, or community charging management platform; online acquisition of feeder power flow or hotspot temperature is not required.

[0026] Assuming the scheduling time step is 15 minutes, the current time is The access vehicle set is ,vehicle Six-dimensional vehicle side feature vector for: in, For vehicles The duration of the connection. This is the current battery level. The current state of charge, For the average charging power obtained, The maximum AC charging power allowed for the vehicle or charging station. This represents the state of charge of the vehicle upon arrival.

[0027] The vehicle-side state variables are input into a pre-trained genetic programming priority rule tree, which outputs a priority score for each access vehicle. The access vehicles are then sorted according to these priority scores to obtain a priority sequence. This priority score is only used for sorting and does not directly determine the final power. The final power is determined by the capacity barrier layer within the transformer's available power budget. This structure functionally separates the learning layer and the safety execution layer, facilitating maintenance personnel's auditing of the reasons for each power intervention.

[0028] The formula for calculating vehicle priority is: The symbolic expression tree obtained from genetic programming, This is a six-dimensional vehicle-side feature vector. For vehicles At any moment Priority score. It is only used for sorting and is not directly used as the final charging power.

[0029] Average external waiting time target The calculation formula is: in For the training event set, Gathering for training day, For the first A collection of daily charging events. For the event The additional external waiting time compared to unrestricted charging.

[0030] Maximum average user wait time target The calculation formula is: in For user collection, For users Average additional external wait time during the training cycle.

[0031] The capacity guardrail layer calculates the safe active power limit of the distribution transformer and the active power budget for the current scheduling period based on the parameters of the distribution transformer and the preset safety control parameters, and converts the priority sequence of each access vehicle into the final charging power command in conjunction with the priority sequence. like Figure 2 As shown, the various stages of the capacity barrier layer are executed in a fixed sequence, including dynamic budget calculation, rank-sharing allocation, trickle processing, reserved buffer trimming, backfilling, hard truncation, slope and smoothing, and feasible projection. Each stage employs deterministic rules and can record operation logs for budget reduction, trickle switching, trimming, backfilling, truncation, slope constraints, and smoothing projection.

[0032] The capacity guardrail layer first determines the safe active power limit of the distribution transformer. : in This refers to the rated capacity of the transformer. This is the static reduction coefficient. This is the safety load factor. This is a margin factor reserved for one time.

[0033] Active power budget available for electric vehicle charging during the current scheduling period The calculation formula is: When the basic load of residents Has reached or exceeded hour, When the capacity is 0, the guardrail layer will no longer allocate additional power to ordinary charging vehicles.

[0034] The formula for calculating the rank-shared allocation weight is: in This represents the current number of connected vehicles. For vehicles exist The ranking in =1 indicates the highest priority. For ranking The corresponding weights For vehicles The obtained normalized budget weights.

[0035] vehicle Initial power allocation The calculation formula is: Not exceeding the maximum permissible AC charging power of the vehicle or charging station It also does not exceed the priority weight. Budget share received .

[0036] The trickle-down treatment rules are as follows: When the vehicle State of charge Reaching the preset trickle threshold At that time, its power command Switch to trickle power This is to reduce the use of scarce capacity by vehicles in a high-charge state.

[0037] Reserved buffer threshold The definition of is: in This is a secondary reserve margin coefficient. Below This is used to trigger the trimming of low-priority vehicles before the safety limit is reached.

[0038] The target conditions for the reserved buffer pruning phase are: This condition requires the basic load of residents. With the power of each vehicle The sum does not exceed When exceeding At that time, the capacity barrier layer is reduced starting from low-priority non-trickle-down vehicles. .

[0039] Remaining available capacity during the backfill phase for: If trimming or Constraints lead to budget If not fully utilized, then prioritize them from highest to lowest. Allocated to those not yet reached The vehicles.

[0040] Hard cutoff ratio The calculation formula is: After the aforementioned processing, the power of each vehicle The sum still exceeds the budget. When, for non-zero proportionally Cutting down to bring the total charging power of electric vehicles back to Within.

[0041] The slope is restricted to: in Indicates vehicle The power variation between adjacent scheduling periods. This embodiment uses a maximum variation range of 3.5kW every 15-minute time step to avoid... Excessive abrupt changes.

[0042] The three-step moving average smoothing formula is: in Depend on , and Calculated. Smoothing can be used to reduce abrupt changes between adjacent time steps, but smoothing may introduce a slight budget overrun, thus requiring a subsequent feasible projection.

[0043] The final feasible projection formula is: If the smoothed total charging power of electric vehicles does not exceed the current available power budget If the smoothed total charging power exceeds [a certain threshold], then the smoothed power command remains unchanged; if the smoothed total charging power exceeds [a certain threshold], then [the power command will remain unchanged]. And if the total power is positive, then the charging power command for all smoothed non-zero vehicles will be applied according to the same proportional coefficient. Scaling is performed to obtain the final power command. This includes vehicles that had already entered trickle flow mode. The trickle flow power... This indicates the trickle-down target instruction before the final feasible projection, and does not constitute a rigid lower limit for the final power output. If the residential base load... This has made the current available electric vehicle power budget If the value is 0, all electric vehicle charging commands are reset to zero; if the residential base load itself exceeds the safe active power limit... If the over-limit is recorded, it will be recorded as a basic load overload, rather than an electric vehicle dispatch over-limit.

[0044] The final charging power command is sent to the charging piles corresponding to each connected vehicle for execution, thereby completing the charging scheduling and control of electric vehicles.

[0045] After completing the charging scheduling and control steps for electric vehicles, the total active load trajectory of the transformer can be synthesized based on the basic residential load and the charging power instructions issued for each scheduling period. Based on the total active load trajectory of the transformer, offline thermal diagnosis of the distribution transformer and voltage screening of feeder nodes are performed to assess the safety feasibility after the charging power instructions for each scheduling period are issued and executed.

[0046] During the real-time online scheduling phase, the capacity guardrail layer is only responsible for calculating and outputting the final charging power command for electric vehicles, and does not directly participate in complex temperature and voltage calculations. Hot spot temperature diagnosis of distribution transformers and voltage margin screening of feeder nodes are performed in the offline phase. During offline analysis, the residential base load is superimposed with the previously generated charging power command to form the total load trajectory of the transformer. This trajectory is then used to calculate whether the transformer will overheat and age, and whether the voltage of the end node exceeds the limit.

[0047] like Figure 3 As shown, the process of offline calculation of the load factor, top oil temperature, hot spot temperature, aging acceleration factor, and equivalent aging factor based on the total active load trajectory of the transformer is as follows: Transformer Total Active Load Trajectory The calculation formula is: in, For the final power command, This is the basic load for residents.

[0048] Load factor in hot spot temperature playback of distribution transformer The calculation formula is: in For total active load, For power factor, This refers to the rated capacity of the transformer. The static derating factor is used in the example. We approximate it to 0.95.

[0049] Steady-state top oil temperature rise The calculation formula is: in For the rated top oil temperature rise, For load factor, This is the ratio of load loss to no-load loss under rated current. This refers to the oil index.

[0050] The transient update formula for top oil temperature is: in The current top oil temperature, For ambient temperature, The oil time constant, To schedule the time step, The steady-state top oil temperature rise is given by equation (19).

[0051] Hotspot temperature The calculation formula is: in The top oil temperature, For the rated hot spot temperature rise, For load factor, This represents the winding index.

[0052] Instant aging accelerator The calculation formula is: From hotspot temperature Sure, For every 6K increase in temperature relative to the 98℃ reference temperature, the insulation aging rate approximately doubles.

[0053] Equivalent aging factor during the evaluation period The calculation formula is: in To evaluate the number of steps in time, The instantaneous aging acceleration factor is obtained from equation (22).

[0054] The approximate formula for calculating the linear sensitivity of the three-phase forward-backward voltage screening method is as follows: in, This refers to the change in phase voltage. This is the voltage sensitivity matrix. This refers to the charging power vector for electric vehicles. In this embodiment... Instead of online learning through genetic programming priority rules or capacity guardrail layers, the system is constructed offline based on a three-phase forward-backward substitution feeder model of the target residential area. A 1kW incremental disturbance is applied to each node and each phase, and the three-phase forward-backward substitution calculation is performed independently. Corresponding sensitivity coefficients are generated based on the changes in node phase voltages before and after the disturbance, and these coefficients are then combined to form... For older residential areas lacking feeder data, the fixed sensitivity matrix in this embodiment is not directly used. Before actual deployment, it is necessary to obtain the feeder topology, line impedance, phase connection, and corresponding voltage monitoring information required to establish or calibrate the three-phase forward backward model of the substation based on existing on-site data or supplementary measurements. Then, the substation's data is regenerated according to the above-mentioned node-by-node, phase-by-phase 1kW disturbance method. If the necessary feeder parameters for establishing a three-phase forward-backward model of the power station cannot be obtained, supplementary field measurements should be taken first. Sensitivity matrices that have not been established or calibrated for the power station should not be used for deployment decisions. It is used only for offline voltage screening and deployment assessment, and does not participate in the power allocation of the online capacity guardrail layer.

[0055] Voltage margin Defined as: For a moment node Mutually voltage, The lower limit threshold of the phase voltage at the low-voltage node is 0.93 pu in this example; A value less than 0 indicates that there is a risk of voltage exceeding the limit at this time step.

[0056] Example 2 This embodiment provides a car charging scheduling system based on genetic programming and capacity barrier layers, such as... Figure 4 As shown, the system modules include: The status acquisition module is used to collect feature vectors of vehicles connected to charging piles in residential areas, monitor the basic load of residents, and read the parameters of the distribution transformer and preset safety control parameters. The genetic programming priority decision module is used to receive the feature vector output by the state acquisition module, input it into a pre-trained symbolic expression to calculate the priority score, and generate a priority sequence in descending order. The capacity guardrail layer execution module is used to receive the priority sequence output by the genetic programming priority decision module, and generate charging instructions for each scheduling period based on the distribution transformer parameters and preset safety control parameters through multi-stage safety allocation execution capacity constraint protection and instruction smoothing post-processing. The instruction issuing module is used to issue the charging instruction to the charging pile corresponding to each connected vehicle through the charging pile communication interface, and to record the execution log. The thermal diagnostic module is used to calculate the hot spot temperature and assess the equivalent aging based on the transformer thermal model according to the total active load trajectory of the transformer output by the instruction issuing module, and to feed back the generated thermal status report to the genetic programming priority decision module. The voltage screening module receives the charging command output by the thermal diagnostic module, performs voltage over-limit detection based on the sensitivity matrix of the power flow calculation method, and conducts bidirectional data interaction with the capacity guardrail layer execution module to complete the deployment feasibility assessment.

[0057] Example 3 This embodiment presents a specific application and effect verification process of a car charging scheduling method based on genetic programming and capacity barrier layers: In a specific implementation scenario, the rated capacity of the distribution transformer For 200kVA, static derating factor The safety load factor is 0.9. The initial margin coefficient is 0.85. The value is 0.98. The calculated safe active power limit for the distribution transformer is approximately 149.94 kW. If the secondary reserve margin factor is taken as 0.95, the calculated reserve buffer threshold is approximately 142.44 kW.

[0058] In this implementation scenario, the residential area serves 150 households, and scheduling evaluation was conducted using data from 200 test days. Under the condition of scarce baseline capacity, there is a daily power shortage of 26.6%. For example... Figure 5 As shown, the basic load of residents occupies a large amount of transformer capacity during the evening peak, which means that electric vehicle charging scheduling must complete service sequencing and power allocation within the limited remaining capacity.

[0059] The training of the priority rule tree in genetic programming can be performed using a non-dominated sorting genetic algorithm. In an optional implementation detail, the function set for genetic programming is as follows: The division method used is protected division; the terminal set is... ,in For the duration of vehicle connection, This is the current battery level. The current state of charge, For the average charging power obtained, Maximum AC charging power, To reach the state of charge. Genetic programming tree. Vehicle side feature vectors Mapped to priority score The priority score is used only for vehicle sorting. In an optional training configuration, the population size is set to 100, the number of generations is 100, a tournament selection mechanism of size 3 is used, the crossover probability is set to 0.95, the mutation probability is set to 0.20, the maximum tree depth is limited to 5, and the maximum number of nodes is limited to 20. The above training parameters are used to illustrate the specific implementation method and can be flexibly adjusted according to the historical data scale and computing resources during actual deployment.

[0060] like Figure 6 As shown, the charging scheduling method combining genetic programming sorting and capacity barrier layer achieves a better Pareto trade-off between average waiting time and maximum average user waiting time. Compared to traditional manual rules such as first-come-first-served, earliest deadline priority, least remaining time priority, least slack time priority, last arrival priority, equal budget allocation, and composite rules, genetic programming rules still maintain a significant service advantage within the capacity barrier layer execution.

[0061] like Figure 7 As shown, there are significant differences in the priority response surfaces of genetic programming rules and composite rules on vehicle state variables. This difference indicates that genetic programming does not simply replicate traditional manual rules, but rather forms a symbolic sorting logic that is highly compatible with capacity-scarce scenarios under the constraints of the training objective.

[0062] like Figure 8 As shown, the protection scheduling of the capacity guard layer not only limits the instantaneous peak active power, but also changes the thermal effect of the transformer load trajectory. In the baseline scenario, the full-stack capacity guard layer scheduling reduces the equivalent aging factor of the distribution transformer from 0.198 to 0.120, and the maximum hot spot temperature from 89.9℃ to 86.2℃.

[0063] like Figure 9 As shown, the non-dominated sorting genetic algorithm trained with 21 random seeds exhibits extremely high stability on the target value, with a final hypervolume of approximately 7071.4 ± 9.6 and an optimal average waiting time of only about 0.26 min for inter-seed variation. Although the size of the Pareto front varies with the random seeds, the convergence process of the service objective remains highly stable.

[0064] like Figure 10 As shown, the transformer capacity and operating pressure determine the effective operating boundary of the method in this embodiment. When the rated capacity of the transformer drops to 160kVA and the energy deficit reaches 38.9%, simply improving the sorting rules is insufficient to compensate for the severe shortage of physical capacity; when the transformer capacity increases to 315kVA, the improvement in service quality exhibits a diminishing marginal effect relative to the increase in capacity.

[0065] like Figure 11 and Figure 12 As shown, six-dimensional vehicle-side features are sufficient to represent the main ranking information in this priority learning task. In contrast, fifteen-dimensional extended features did not improve the supervolume or optimal average waiting time; instead, they increased the computation time for a single evaluation by approximately 33.1%. Therefore, this embodiment preferentially uses six-dimensional vehicle-side features that can be directly obtained from charging piles, vehicle session management systems, or community charging management platforms.

[0066] like Figure 13 As shown, when using the three-phase forward-backward substitution method for voltage screening, the voltage remains unbound under both reference phase mapping and end-bias mapping, with the worst lowest phase voltage being approximately 0.9593 pu and 0.9570 pu, respectively. However, under the extreme skew pressure scenario where all charging loads are concentrated on the first phase, the lowest phase voltage drops to approximately 0.9061 pu and a large number of bound time steps appear. This screening result indicates that the capacity guardrail layer can effectively constrain the total active power, but it cannot replace the three-phase forward-backward substitution method alone for screening feeder node voltages.

[0067] In summary, this embodiment addresses the adaptive vehicle sequencing problem through genetic programming priority rules, the online capacity safety control problem through a capacity guardrail layer, and the offline deployment evaluation problem through IEC thermal model playback and three-phase forward-backward substitution. The clear boundaries of responsibility for these three aspects enable independent auditing of scheduling performance, safety protection constraints, and engineering diagnostic results.

[0068] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A vehicle charging scheduling method based on genetic programming and capacity guardrail layer, characterized in that, Includes the following steps: Acquire vehicle-side state variables, residential basic load, power distribution transformer parameters, and preset safety control parameters for charging piles connected to residential areas; The vehicle-side state variables are input into a pre-trained genetic programming priority rule tree, and the priority score of each access vehicle is output. The access vehicles are then sorted according to the priority scores to obtain a priority sequence. The capacity barrier layer calculates the safe active power limit of the distribution transformer and the active power budget for the current scheduling period based on the parameters of the distribution transformer and the preset safety control parameters. Combined with the priority sequence, it converts the priority sequence of each access vehicle into a charging power instruction for each scheduling period. The capacity barrier layer is a deterministic safety execution unit independent of the genetic programming priority rule tree. The priority score is only used for sorting. The final charging power of each access vehicle is determined by the capacity barrier layer within the active power budget. The charging power instructions for each scheduling period are sent to the charging piles corresponding to each connected vehicle for execution, thereby completing the charging scheduling control of electric vehicles.

2. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The vehicle-side state variables include connection duration, current battery level, state of charge, average charging power received, maximum AC charging power, and state of charge when the vehicle arrives.

3. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The process by which the capacity guardrail layer calculates the safe active power limit and the active power budget of the distribution transformer based on the parameters of the distribution transformer is as follows: The rated capacity of the transformer in the distribution transformer parameters is multiplied by the static derating factor, safe load rate factor and primary reserve margin factor in the preset safety control parameters to obtain the safe active power limit of the distribution transformer. Calculate the difference between the safe active power limit of the distribution transformer and the current residential basic load. If the difference is greater than zero, then the difference is determined as the active power budget for electric vehicle charging during the current scheduling period. If the difference is less than or equal to zero, then the active power budget is zero.

4. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The step of converting the priority sequence of each access vehicle into charging power instructions for each scheduling period also includes a rank-sharing allocation step: The ranking weight of each vehicle is calculated based on the total number of currently connected vehicles and the ranking of each vehicle in the priority sequence, with the highest priority ranking corresponding to the largest weight. The ranking weights of each vehicle are normalized to obtain the normalized budget weights; Calculate the product of the normalized budget weight and the active power budget for each vehicle, and take the smaller of the product and the maximum AC charging power allowed by the vehicle or charging pile as the initial power allocation for the corresponding vehicle.

5. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The step of converting the priority sequence of each access vehicle into charging power instructions for each scheduling period also includes a trickle processing step: Determine whether the state of charge of the connected vehicle has reached the preset trickle threshold; When the state of charge of the connected vehicle reaches the preset trickle threshold, the charging power command of the vehicle is switched to the preset trickle power.

6. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The step of converting the priority sequence of each access vehicle into charging power instructions for each scheduling period also includes a reserved buffer trimming and backfilling step: The reserved buffer threshold is obtained by multiplying the secondary reserve margin coefficient in the preset safety control parameters with the safe active power limit of the distribution transformer. When the sum of the resident basic load and the current allocated power of each access vehicle exceeds the reserved buffer threshold, the allocated power of non-trickle-down vehicles is reduced in order of priority from low to high. The remaining available capacity is obtained by calculating the difference between the active power budget and the sum of the current allocated power of each connected vehicle. When the remaining available capacity is greater than zero, the remaining available capacity is allocated to vehicles that have not yet reached the maximum AC charging power in descending order of priority.

7. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The step of converting the priority sequence of each access vehicle into charging power commands for each scheduling period also includes hard truncation, ramp limiting, smoothing, and feasible projection steps: When the sum of the current allocated power of each connected vehicle exceeds the active power budget, the active power budget is divided by the sum of the current allocated power of each connected vehicle to obtain the hard cutoff ratio, and the charging power of vehicles with a current allocated power greater than zero is reduced proportionally according to the hard cutoff ratio. Limit the change in allocated power of each access vehicle in adjacent scheduling periods to no more than the set single-step change limit; The average power of the vehicle distribution is calculated using a three-step moving average to obtain the smoothed power. When the sum of the smoothed power of each access vehicle does not exceed the active power budget, the smoothed power of each access vehicle is used as the charging power instruction for each scheduling period; when the sum of the smoothed power of each access vehicle exceeds the active power budget and the total power is greater than zero, the smoothed power of each access vehicle is scaled according to the ratio of the active power budget to the sum of the smoothed power to obtain the final charging power instruction for each scheduling period.

8. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, After completing the charging scheduling and control steps for the electric vehicle, the method further includes: The total active load trajectory of the transformer is synthesized based on the basic load of residents and the charging power instructions issued for each scheduling period. Based on the total active load trajectory of the transformer, the distribution transformer thermal diagnosis and feeder node voltage screening are performed offline to assess the safety feasibility after the charging power instructions for each scheduling period are issued and executed.

9. The vehicle charging scheduling method based on genetic programming and capacity guardrail layer according to claim 1, characterized in that, The offline performance of distribution transformer thermal diagnosis and feeder node voltage screening based on the total active load trajectory of the transformer specifically includes: The load factor is calculated by dividing the total active load trajectory of the transformer by the product of the power factor, the rated capacity of the transformer, and the static derating factor. The steady-state top-level oil temperature rise is calculated based on the load factor, and the current top-level oil temperature is updated by combining the ambient temperature, oil time constant and scheduling time step. The hot spot temperature is calculated based on the top oil temperature, rated hot spot temperature rise, load factor, and winding index. The instantaneous aging acceleration factor is calculated based on the hot spot temperature, and the equivalent aging factor is obtained by averaging the instantaneous aging acceleration factor over a period of time within the evaluation period. Based on the three-phase forward-backward power flow calculation method, the phase voltage change is calculated by multiplying the voltage sensitivity matrix with the electric vehicle charging power vector. Calculate the voltage of each phase at each node in each time period, and subtract the lower limit threshold of the low-voltage node phase voltage to obtain the voltage margin; The period when the voltage margin is less than zero is statistically analyzed, and the statistical results of the lowest node phase voltage and the over-limit time step are obtained.

10. A vehicle charging scheduling system based on genetic programming and capacity guardrail layer, characterized in that, include: The status acquisition module is used to collect feature vectors of vehicles connected to charging piles in residential areas, monitor the basic load of residents, and read the parameters of the distribution transformer and preset safety control parameters. The genetic programming priority decision module is used to receive the feature vector output by the state acquisition module, input it into a pre-trained symbolic expression to calculate the priority score, and generate a priority sequence in descending order. The capacity guardrail layer execution module is used to receive the priority sequence output by the genetic programming priority decision module, and generate charging power instructions for each scheduling period by multi-stage safety allocation execution capacity constraint guarantee and instruction smoothing post-processing based on the distribution transformer parameters and preset safety control parameters. The instruction issuing module is used to issue the charging instruction to the charging pile corresponding to each connected vehicle through the charging pile communication interface, and to record the execution log. The thermal diagnostic module is used to calculate the hot spot temperature and assess the equivalent aging based on the transformer thermal model according to the total active load trajectory of the transformer output by the instruction issuing module, and to feed back the generated thermal status report to the genetic programming priority decision module. The voltage screening module receives the charging command output by the thermal diagnostic module, performs voltage over-limit detection based on the sensitivity matrix of the power flow calculation method, and conducts bidirectional data interaction with the capacity guardrail layer execution module to complete the deployment feasibility assessment.

Images