CN-120634176-B - Method and system for configuring number of charging piles

CN120634176BCN 120634176 BCN120634176 BCN 120634176BCN-120634176-B

Abstract

The invention relates to the technical field of charging pile planning and discloses a method and a system for configuring the quantity of charging piles, wherein the method comprises the steps of collecting real-time data of charging stations, calculating queuing indexes under current configuration by using a queuing theoretical model based on the real-time data, packaging the queuing indexes into custom operators, and outputting queuing state vectors; the method comprises the steps of establishing a charging pile configuration model based on a value function, dynamically determining the number of parallel service modules required to be adjusted in each control period based on a queuing state vector, and establishing a double safety threshold constraint mechanism to correct the adjustment action output by the charging pile configuration model so as to realize the configuration of the number of charging piles. The method has the advantages of low sample learning capability, strong edge deployment performance and robustness and the like, and is suitable for intelligent operation scheduling scenes of a large-scale charging network.

Inventors

XU XIN
ZHANG YONG
LIN WEISHENG
MA SHUTONG
LIN WEIYAN

Assignees

山东精锐电器有限公司

Dates

Publication Date: 20260512
Application Date: 20250715

Claims (3)

1. A method for configuring the number of charging piles, comprising: collecting real-time data of a charging station, calculating queuing indexes under the current configuration by using a queuing theory model based on the real-time data, packaging the queuing indexes into custom operators, and outputting queuing state vectors; Constructing a charging pile configuration model based on a value function, and dynamically determining the number of parallel service modules required to be adjusted in each control period based on the queuing state vector; establishing a double safety threshold constraint mechanism, correcting the adjustment action output by the charging pile configuration model, and realizing the quantity configuration of the charging piles; the real-time data comprises charging station operation data, traffic flow data, queuing data and environment data; the charging station operation data comprises a starting/ending time stamp, duration, charging quantity and charging voltage current curve of each charging pile; The traffic flow data comprise the arrival rate of vehicles, the distribution of parking time length and the turnover rate of parking spaces; the queuing theory model comprises the following steps of calculating traffic intensity according to real-time data: ; Wherein, the Representing traffic intensity; Representing the vehicle arrival rate; representing the number of parallel service modules which are started; Representing the average service rate of a single parallel service module; When (when) When the vehicle is required to enter the queue, the probability of the vehicle entering the queue is calculated through an Erlang-C equation Further calculate queuing wait time: ; Wherein, the Representing queuing latency; Indicating the probability that the vehicle needs to enter the queue; When (when) When the charging station is saturated, the queuing probability is not calculated, the safety coverage branch is switched in, and the number of parallel service modules needing one-time opening is determined, and is expressed as: ; Wherein, the Representing the total number of parallel service modules to be started in the next period; indicating that the maximum value is taken; representing the least number of parallel service modules that are open; representing a preset upper limit of the safety utilization rate; representing the upward rounding operation, directly adjusting the number of the opened parallel service modules to be ; The queuing state vector includes, for example, a vehicle arrival rate Average service rate And the number of parallel service modules As a key input, calculate the probability that the vehicle needs to enter the queue based on the Erlang-C equation The partial derivative of the vehicle arrival rate for the key input is expressed as: ; Wherein, the Representing a differential sign; Compiling a forward formula and a partial derivative of key input together into a custom Autograd operator, wherein the forward formula is that 、 And Is calculated according to the formula; outputting queuing state vectors through tensor stitching: ; Wherein, the Representing a queuing state vector; Representing a splice operator; Representing an environmental data vector; outputting by a custom Autograd operator; The charging pile configuration model comprises defining output actions , Indicating the next period of content on or off A parallel service module for providing a service to the network according to Dueling-DQN architecture Input sharing feature extractor Obtaining a hidden vector; Sending the hidden vector into the first branch and the second branch simultaneously, calculating the state value of the first branch, calculating the motion advantage of the second branch, and synthesizing the first branch and the second branch Calculating the value, and calculating the loss by adopting a mean square TD-error; Outputting actions according to a greedy strategy According to Executing the dynamic adjustment of the number of the parallel service modules; The dual safety threshold constraint mechanism comprises the following steps of acquiring output actions Thereafter, an exponentially weighted moving average is used to estimate the arrival rate of the next cycle According to Calculating the lower limit pile number with average service rate ; Wherein, the Representing the number of lower limit piles; Number of open parallel service modules output by charging pile configuration model Theory is bound to If (if) Outputting the motion ; If it is Then cover as ; When the traffic intensity is smaller than the preset traffic intensity minimum value, outputting the action While still negative, then the coverage is 。
2. The method for configuring the number of charging piles according to claim 1, further comprising, according to the output action After the number of the parallel service modules is adjusted, instant rewards are obtained according to a rewarding function, wherein the rewarding function is expressed as: ; Wherein, the Representing an instant prize; a weighting coefficient representing the waiting cost; a weighting coefficient representing electricity price cost; a weighting coefficient representing a module switching cost; the time-sharing electricity price is represented; representing absolute value; Instead of a numerical result independently called from the queuing latency function, tensor output from the custom Autograd operator has a complete automatic differentiation attribute; in the counter-propagating phase, slave Initially, the first, second and shared feature extractors are passed sequentially along Dueling-DQN architecture, generating , Comprising three channels: ; Wherein, the Representing a loss function; according to the partial derivative formula in the custom Autograd operator, the error is further transformed into: ; influencing the weights of a shared feature extractor by means of the chain law The direction of latency degradation when key input changes are obtained in one back propagation.
3. A number configuration system of charging piles is applied to the number configuration method of the charging piles according to any one of claims 1 to 2, and is characterized by comprising, The collecting module is used for collecting real-time data of the charging station, calculating queuing indexes under the current configuration by using a queuing theory model based on the real-time data, packaging the queuing indexes into a custom operator and outputting a queuing state vector; The scheduling module is used for constructing a charging pile configuration model based on a value function and dynamically determining the number of parallel service modules required to be adjusted in each control period based on the queuing state vector; And the constraint module establishes a double safety threshold constraint mechanism, corrects the adjustment action output by the charging pile configuration model and realizes the quantity configuration of the charging piles.

Description

Method and system for configuring number of charging piles Technical Field The invention relates to the technical field of charging pile planning, in particular to a method and a system for configuring the number of charging piles. Background In recent years, the storage amount of electric vehicles is exponentially increased, and the peak-valley difference of charging requirements of cities and high-speed scenes is continuously increased. The traditional charging pile planning method is dependent on historical traffic flow and experience coefficients to configure the pile number at one time, so that the method is difficult to adapt to traffic surge caused by holidays, weather or sudden activities, and serious queuing in peak periods and idle equipment in valley periods are caused. Meanwhile, as the carbon trade and the electricity price of the required amount become severe, operators not only need to control the construction investment, but also have to grasp the power distribution in the operation period so as to reduce the peak load and the cost expenditure. In recent years, although a dynamic scheduling attempt based on reinforcement learning is available, a queuing model is generally regarded as a black box, a pile expansion and contraction strategy is learned in a trial-and-error mode, the sample efficiency is low and the real-time requirement of commercial stations is difficult to meet, and on the other hand, a safety strategy which completely depends on a threshold value lacks self-adaption capability and is easy to fail under extreme load, so that a comprehensive scheme combining queuing theory precision and deep learning decision speed is needed. Disclosure of Invention The present invention has been made in view of the above-described problems. Therefore, the method solves the technical problems that the sample efficiency of the existing charging pile quantity configuration method is low and the real-time requirement of commercial sites is difficult to meet. The charging pile quantity configuration method comprises the steps of collecting real-time data of a charging station, calculating queuing indexes under current configuration by using a queuing theory model based on the real-time data, packaging the queuing indexes into custom operators, and outputting queuing state vectors; Constructing a charging pile configuration model based on a value function, and dynamically determining the number of parallel service modules required to be adjusted in each control period based on the queuing state vector; And establishing a double safety threshold constraint mechanism, correcting the adjustment action output by the charging pile configuration model, and realizing the quantity configuration of the charging piles. The invention relates to a charging pile quantity configuration method, which is a preferable scheme, wherein the real-time data comprises charging station operation data, traffic flow data, queuing data and environment data; the charging station operation data comprises a starting/ending time stamp, duration, charging quantity and charging voltage current curve of each charging pile; the traffic flow data comprise the arrival rate of the vehicle, the parking time distribution and the turnover rate of the parking spaces. As a preferable scheme of the method for configuring the number of the charging piles, the queuing theory model comprises the following steps of calculating traffic intensity according to real-time data: ; Wherein, the Representing traffic intensity; Representing the vehicle arrival rate; representing the number of parallel service modules which are started; Representing the average service rate of a single parallel service module; When (when) When the vehicle is required to enter the queue, the probability of the vehicle entering the queue is calculated through an Erlang-C equationFurther calculate queuing wait time: ; Wherein, the Representing queuing latency; Indicating the probability that the vehicle needs to enter the queue; When (when) When the charging station is saturated, the queuing probability is not calculated, the safety coverage branch is switched in, and the number of parallel service modules needing one-time opening is determined, and is expressed as: ; Wherein, the Representing the total number of parallel service modules to be started in the next period; indicating that the maximum value is taken; representing the least number of parallel service modules that are open; representing a preset upper limit of the safety utilization rate; representing the upward rounding operation, directly adjusting the number of the opened parallel service modules to be 。 As a preferable mode of the method for configuring the number of the charging piles, the queuing state vector comprises the steps of adding the arrival rate of vehiclesAverage service rateAnd the number of parallel service modulesAs a key input, calculate the probability that the vehicle needs to enter