CN-122026336-A - Electric vehicle charging pile demand prediction and layout optimization method

CN122026336ACN 122026336 ACN122026336 ACN 122026336ACN-122026336-A

Abstract

The invention discloses a method for predicting the requirements and optimizing the layout of electric vehicle charging piles. The method comprises the steps of carrying out double-dimensional uncertainty quantification, respectively predicting probability confidence intervals of charging demand fluctuation and equipment faults through a demand prediction model and a fault prediction model, coupling to generate a combined uncertainty scene set, carrying out robust redundancy layout solution, constructing a multi-objective optimization function by taking the combined scene set as a constraint, adopting an improved genetic algorithm solution capable of dynamically distributing weights to obtain a configuration scheme of a hot backup and a cold backup redundancy pile, carrying out dynamic iterative optimization, updating the model and the scene set based on real-time data, and dynamically adjusting the layout scheme according to the model and the scene set to form a closed loop. The invention realizes the optimal balance of charging service continuity, resource utilization rate and economy under the double uncertainties of demand fluctuation and equipment failure.

Inventors

ZHAO TIANYU
SHANG JIAYI
GONG YINGFEI
ZHENG YAMIN
YU HAO
CHEN CHEN
HE WEIBIN

Assignees

国网浙江省电力有限公司杭州供电公司

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. The electric automobile charging pile demand prediction and layout optimization method is characterized by comprising the following steps of: Step S1, two-dimensional uncertainty quantization: Collecting charging demand data and charging pile operation data of a target area, and preprocessing and extracting features to obtain a demand feature set and an equipment feature set; based on the demand feature set, predicting probability distribution and demand confidence interval of charging demand fluctuation in a future period through a demand prediction model based on a Bayesian neural network; Based on the equipment characteristic set, predicting a fault confidence interval of the fault probability and the fault duration of the charging pile in the future period through a fault prediction model; Coupling the demand confidence interval and the fault confidence interval to generate a combined uncertainty scene set containing various combined scenes and occurrence probabilities thereof; Step S2, solving a robust redundant layout: constructing a multi-objective optimization function with the charging service satisfaction rate, the resource utilization rate and the full life cycle cost as targets by taking the combined uncertainty scene set as a constraint; Solving the function by adopting an improved non-dominant ranking genetic algorithm, wherein the algorithm dynamically distributes uncertainty weights according to the requirement fluctuation intensity and the equipment aging degree; And determining the configuration scheme of the redundant charging piles in the target area through a redundant configuration model according to the solving result, wherein the configuration scheme comprises the types, the number and the layout positions of the hot backup redundant piles and the cold backup redundant piles.
2. The electric vehicle charging pile demand prediction and layout optimization method according to claim 1, wherein in the step S1, the collected charging demand data at least includes: Real-time operation parameters of the regional power grid, user travel track and discrete charging behavior record data and regional development dynamic data, wherein the regional power grid real-time operation parameters are acquired through a load sensor, the user travel track and discrete charging behavior record data are acquired through a vehicle terminal or a traffic management system, and the regional development dynamic data are acquired through a planning department interface; The charging pile operation data at least comprises: The method comprises the steps of obtaining operation parameters of a core component of the charging pile through a test sensor and obtaining a historical fault record through an operation and maintenance system.
3. The electric vehicle charging pile demand prediction and layout optimization method according to claim 1, wherein the collecting the charging demand data and the charging pile operation data in step S1 specifically includes: acquiring demand data, namely acquiring real-time operation parameters of a regional power grid through a load sensor, acquiring user behavior data comprising travel tracks and charging time length through an application program interface of a vehicle terminal or a traffic management platform, and acquiring regional development dynamic data through a planning data interface; Acquiring equipment data, namely acquiring operation parameters of core components of the charging pile through a test sensor arranged on the charging pile; and acquiring auxiliary data, namely acquiring a power grid load redundancy threshold value through a power grid dispatching system interface and acquiring extreme weather early warning information through a meteorological data interface.
4. The electric vehicle charging pile demand prediction and layout optimization method according to claim 1, wherein the obtaining the demand confidence interval based on the demand prediction model in the step S1 specifically includes: model training, namely training the Bayesian neural network by taking the preprocessed historical charging demand data as a supervision tag; the interval prediction is to take feature data of a prospective period as input based on a trained model, output a charging demand prediction interval under different confidence levels and represent uncertainty of demand fluctuation in a probability distribution form; And dividing the scenes, namely sampling from the probability distribution by adopting a Monte Carlo simulation method to generate a plurality of requirement scenes covering different fluctuation intensities, and forming the instantiation expression of the requirement confidence interval.
5. The electric vehicle charging pile demand prediction and layout optimization method according to claim 1, wherein the generating the joint uncertainty scene set in step S1 specifically includes: coupling operation, namely coupling operation is carried out on the demand fluctuation probability distribution represented by the demand confidence interval and the equipment fault probability distribution represented by the fault confidence interval to construct a joint probability model representing the two-dimensional uncertainty; Generating a plurality of joint scenes with determined probabilities through sampling calculation based on the two-dimensional joint probability distribution model, wherein each joint scene is jointly defined by a requirement fluctuation interval under a specific confidence level and equipment fault combinations under the specific probabilities; And weight distribution and constraint extraction, namely distributing a weight for each generated joint scene, wherein the weight is equal to the occurrence probability of the scene in the two-dimensional joint probability distribution model, and determining the extreme joint scene with the occurrence probability lower than a preset threshold value as a hard constraint condition which must be met in subsequent optimization solution.
6. The electric vehicle charging pile demand prediction and layout optimization method according to claim 1, wherein the improved non-dominant ranking genetic algorithm in step S2 dynamically assigns weights according to demand fluctuation intensity and equipment aging degree, and specifically comprises: calculating scene intensity indexes representing the overall demand fluctuation range according to the combined uncertainty scene set, and calculating regional aging indexes representing the overall health state of equipment according to the fault confidence interval; Dynamically generating weight coefficient combinations for balancing the relation between a service continuity target and a cost reliability target according to real-time values of the scene intensity index and the regional aging index in each iteration of an algorithm through a preset mapping rule; Weight application-applying the dynamically generated weight coefficient combinations to the solving process of the multi-objective optimization function, so that the searching direction of the algorithm can adaptively tend to alleviate the more prominent risk dimension in the current uncertainty characteristic.
7. The electric vehicle charging pile demand prediction and layout optimization method according to claim 6, wherein the constructing and solving the multi-objective optimization function in the step S2 specifically includes: function construction, namely the constructed multi-objective optimization function is expressed as: , wherein, In order for the service to meet the rate, For the full life-cycle cost of the device, In order for the utilization of the resources to be achieved, 、、 Respectively obtaining preset service satisfaction rate target weight, cost target weight and resource utilization rate target weight; Coefficient dynamic coupling, namely coupling the service satisfaction rate target weight and the value of the cost target weight with the dynamically generated weight coefficient combination instead of a fixed constant, so that the core weighing relation of the multi-target optimization function can be adjusted in real time along with the system risk characteristics represented by the combined uncertainty scene set generated in the step S1; And (3) a solution set mapping output, namely optimizing the multi-objective optimization function through the improved non-dominant ordered genetic algorithm, and selecting a scheme from a final non-dominant solution set, wherein the scheme directly outputs redundant pile configuration parameters corresponding to the current weight tendency.
8. The method for predicting the demand and optimizing the layout of the charging piles for the electric vehicle according to claim 1, wherein the determining the number of the redundant charging piles in the step S2 is specifically implemented by the following steps: the parameter dynamic acquisition, namely, key input parameters in the redundant configuration model, namely, the requirement fluctuation floating rate and the regional average fault probability, are obtained from the specific statistical values in the requirement confidence interval and the fault confidence interval which are newly output in the step S1 in real time; Substituting the dynamically acquired parameters into the redundancy configuration model, wherein the logic of the redundancy configuration model is configured to enable a resolving result to be more prone to increase redundancy resources for preventing the demand spike when the demand fluctuation floating rate is obviously increased; And executing the strategy, namely combining the calculated total redundancy requirement quantity with the layout positions obtained by the multi-objective optimization function to jointly form the configuration scheme.
9. The electric vehicle charging pile demand prediction and layout optimization method according to claim 1, further comprising step S3 of dynamic iterative optimization: after the scheme is implemented, the dynamic requirements and equipment operation data of the target area are collected in real time; rolling and updating the demand prediction model and the fault prediction model based on the dynamic data so as to update the demand confidence interval, the fault confidence interval and the combined uncertainty scene set; dynamically adjusting the configuration scheme of the redundant charging pile according to the updated scene set and a preset trigger condition related to the requirement fluctuation or equipment failure; In the step S3, data is collected in real time to drive closed loop iteration of the model and the scheme, which specifically includes: After the configuration scheme is implemented, continuously acquiring real-time operation feedback data of the target area through a deployed data acquisition node network, wherein the feedback data comprises an actual charging demand sequence and a charging pile operation state sequence; The feedback characteristic processing is used for immediately processing the real-time operation feedback data and extracting dynamic demand characteristics and dynamic equipment health characteristics for model updating; And (3) model iteration driving, namely inputting the extracted dynamic demand characteristics and dynamic equipment health characteristics into a demand prediction model and a fault prediction model in the step (S1) as incremental training data so as to drive updating of the demand confidence interval, the fault confidence interval and the combined uncertainty scene set.
10. The electric vehicle charging pile demand prediction and layout optimization method according to claim 9, wherein the dynamic iterative optimization is implemented in step S3, specifically by the following closed-loop procedure: Based on the real-time operation feedback data, rolling and retraining the demand prediction model and the fault prediction model in a period not exceeding a preset updating time, and updating the demand confidence interval, the fault confidence interval and the combined uncertainty scene set according to the rolling and retraining; When the updated combined uncertainty scene set indicates that the system risk feature change reaches a set threshold, automatically triggering to take the updated scene set as input, and re-executing the robust redundant layout solving process of the step S2; And (3) carrying out difference analysis on the newly solved new configuration scheme and the current operation scheme, and generating an executable adjustment instruction set to complete seamless switching and dynamic deployment from the old scheme to the new scheme.

Description

Electric vehicle charging pile demand prediction and layout optimization method Technical Field The invention relates to the technical field of management of electric automobile charging infrastructure, in particular to a method for forecasting requirements and optimizing layout of electric automobile charging piles. Background With the continuous increase of the electric vehicle conservation amount, the planning and layout of the charging infrastructure has become an important foundation for supporting the large-scale application of the electric vehicle. The configuration quantity, the spatial distribution and the operation reliability of the charging piles directly influence the charging convenience of users, the continuity of charging service and the overall utilization efficiency of related resources. Meanwhile, investment cost, operation and maintenance cost and long-term operation benefit are also required to be considered in the construction of the charging facilities, so that higher requirements are put forward for the prediction of the requirements of the charging piles and the optimization of the layout. In an actual running environment, the charging requirements of the electric automobile show obvious space-time fluctuation characteristics, and the charging requirements of different areas and different time periods are comprehensively influenced by various factors, so that the electric automobile has instability and uncertainty. In addition, as an electrical device that operates for a long period of time, the performance state of the charging pile may vary with the use time, the load level, and the environmental conditions, and malfunction of the device or degradation of the performance is unavoidable, which further increases the uncertainty of the charging service supply side. However, the existing charging pile demand analysis and layout planning methods generally have difficulty in comprehensively reflecting the complex characteristics. On the one hand, the analysis of the charging requirement is mostly based on a deterministic result, and it is difficult to effectively describe the fluctuation range and risk level of the requirement in the future operation process, so that the situation that the service capability is not matched with the actual requirement easily occurs in the actual application. On the other hand, in the layout planning process, it is often assumed that the charging facility can continuously and stably run, and a systematic consideration is lacking in equipment running state change and potential faults, so that the reliability of charging service is insufficient in part of running scenes, or an excessive configuration strategy is adopted to avoid risks, so that resource waste is caused. Further, the existing scheme generally lacks comprehensive analysis capability for superposition influence of various uncertain factors, and when demand fluctuation and equipment operation abnormality occur simultaneously, the adaptability and robustness of the existing layout scheme are obviously insufficient, and the continuity and the overall operation efficiency of charging service are difficult to guarantee. In addition, the related planning results are mainly configured in a one-time static mode, and a mechanism for continuous correction and dynamic adjustment based on operation data is lacked, so that the charging facility layout is difficult to adapt to regional development change and operation condition evolution for a long time. Therefore, a method for comprehensively analyzing the charging demand change and the equipment operation uncertainty under a complex operation environment and realizing continuous optimization of the charging facility layout on the basis of the comprehensive analysis is needed, so that the service reliability, the resource utilization efficiency and the overall operation stability of the charging system are improved. Disclosure of Invention Aiming at the defects existing in the prior art, the invention aims to provide a method for predicting the requirements and optimizing the layout of an electric vehicle charging pile, which is used for realizing the robust balance among the continuity of charging service, the utilization rate of resources and the full life cycle cost under the complex scene of multiple uncertainty superposition. In order to achieve the purpose, the invention provides the following technical scheme that the electric vehicle charging pile demand prediction and layout optimization method comprises the following steps: Step S1, two-dimensional uncertainty quantization: Collecting charging demand data and charging pile operation data of a target area, and preprocessing and extracting features to obtain a demand feature set and an equipment feature set; based on the demand feature set, predicting probability distribution and demand confidence interval of charging demand fluctuation in a future period through a demand prediction model based on a