Operation management platform for automobile charging service

Abstract

The invention belongs to the technical field of energy service scheduling, and particularly discloses an operation management platform for automobile charging service, which comprises a data acquisition module used for acquiring service data; the queue estimation module is used for calculating expected waiting time in a future time period through the constructed prediction model; the pricing and demand mapping module is used for constructing a relation curve of listing prices and time slots; the vehicle owner sensitivity learning module is used for constructing a vehicle owner utility function and pushing and updating an index influencing a service mode selected by the vehicle owner to the mobile terminal of the vehicle owner; the dynamic adjusting module is used for updating service data according to the service mode reported by the vehicle owner and judging whether a charging queue threshold value is triggered or not; if yes, the listing price is adjusted, and updating is broadcasted; the iterative convergence coordination module is used for judging whether the updated listed price and the expected waiting time are in an equilibrium solution or not in each time slot; the method has the following advantages: dynamic coupling of supply and demand, rapid price adjustment, accurate flow control, smooth queuing, and synchronous optimization of income and load are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of energy service scheduling, and in particular, to an operation management platform for automobile charging and battery swapping services. Background Art

[0002] The global electric vehicle ownership continues to climb, and a large number of concentrated charging and battery swapping demands appear along urban expressways and highways. A large number of vehicles arrive at the stations simultaneously during the morning and evening commuting peaks and the holiday return stages, resulting in an extremely rapid increase in the queue length and the average waiting time per vehicle being much higher than the user's psychological threshold. At the same time, the equipment utilization rate is very low during the off-peak period, and the investment recovery period of charging and battery swapping infrastructure is significantly extended. Most of the existing operation management platforms adopt a single-layer control and passive queuing logic:

[0003] The station only publishes one electricity price or service fee curve on a daily basis, lacking dynamic guidance for peak loads. The price cannot timely reflect the instantaneous capacity margin of the station, and the decision-making of vehicle owners to arrive at the station is disconnected from the actual supply.

[0004] Queue scheduling is mainly based on first-come-first-served or coarse-grained time period reservation. The algorithm does not predict the subsequent traffic flow in real time, nor does it have a game model that balances the station's revenue and the vehicle owner's experience, resulting in the system only being able to limit the flow through manual intervention when the queue is extremely long.

[0005] All the original data at the station end needs to be uploaded to the cloud for calculation, and the feedback period often exceeds one minute, making it impossible to adjust the price on a second-by-second basis following the real queue state.

[0006] Therefore, an operation management platform for automobile charging and battery swapping services is proposed to solve the above-mentioned problems. Summary of the Invention

[0007] The present invention aims to provide an operation management platform for automobile charging and battery swapping services to solve or improve the problems in the existing operation management of charging and battery swapping stations, such as rigid pricing, rough queuing scheduling, and excessive cloud computing delay, and to achieve dynamic guidance and optimal control based on the real-time supply and demand status.

[0008] In view of this, the first aspect of the present invention is to provide an operation management platform for automobile charging and battery swapping services.

[0009] The first aspect of the present invention provides an operation and management platform for vehicle charging and swapping services, including: a data acquisition module for obtaining service data of charging stations performing charging at the current moment through a station-side gateway; a queue estimation module for calculating the expected waiting time in a future time period according to the service data by a constructed prediction model under the conditions of considering the sudden load generated by swapping at the charging station and the service time fluctuation affected by the listed price; a pricing and demand mapping module for predicting the number of vehicles accepting charging and swapping within a time slot under the condition of considering the marginal cost of unit electric energy, and constructing a relationship curve between the listed price and the time slot; an owner sensitivity learning module for constructing an owner utility function considering the relationship curve and the expected waiting time, and pushing and updating an index affecting the owner's service mode selection to the owner's mobile terminal through the owner utility function in the time period or the time slot; a dynamic adjustment module for updating the service data according to the service mode reported by the owner, and determining whether to trigger a charging queue threshold; if so, adjusting the listed price and broadcasting the update; an iterative convergence coordination module for determining whether the updated listed price and expected waiting time are in an equilibrium solution in each time slot, if not, updating the relationship curve and performing iterative calculation until an equilibrium solution is reached.

[0010] In any of the above technical solutions, the service data includes the queue length, service rate, arrival rate, and listed price synchronously collected from charging piles, swapping workstations, or entrance cameras.

[0011] In any of the above technical solutions, the queue estimation module includes: a vehicle arrival rate filtering unit for continuously correcting the vehicle arrival rate by using recursive Kalman filtering; a service time variance evaluation unit for statistically calculating the service time variance jointly generated by the sudden load caused by swapping and the charging time consumption within a sliding window of a preset length; a variance correction prediction unit for embedding the service time variance into the prediction model to obtain the expected waiting time of the charging station when performing charging and / or swapping.

[0012] In any of the above technical solutions, the prediction model includes the following formula: where is the time period; is the expected waiting time of the next time period; is the system load rate; is the service time variance; is the number of parallel service workstations; is the average service rate of a single workstation; is the predicted arrival rate.

[0013] In any of the above technical solutions, the pricing and demand mapping module includes: an elastic demand mapping unit that constructs the relationship between the expected waiting time, the elasticity of the listing price, and the potential number of vehicles in different time slots; an equilibrium optimization unit that weighs the revenue affected by the listing price and the congestion penalty affected by the charging queue, and solves to obtain the optimal price vector.

[0014] In any of the above technical solutions, the equilibrium optimization unit adjusts the solution mode for obtaining the optimal price vector according to the timeliness demand in the time period and the time slot; the output formats of the optimal price vectors in the time period and the time slot are the same.

[0015] In any of the above technical solutions, the elastic demand mapping unit includes the following formula: Where is the industry elasticity curve; are the sensitivity weights respectively; is the expected number of arriving vehicles; is the current reference price; is the price of the current time slot to be optimized; is the expected waiting time in the current time period.

[0016] In any of the above technical solutions, the owner sensitivity learning module includes: a sensitivity update unit that updates the waiting sensitivity and price sensitivity according to the individual historical behavior data of the owner; an index evaluation unit that obtains the index through the owner utility function according to the current waiting sensitivity and price sensitivity in each time period.

[0017] In any of the above technical solutions, the index evaluation unit includes the following formula: Where is the index, which is used to characterize the owner's choice utility within the time slot is the waiting sensitivity; is the price sensitivity; is the price locked by the owner's mobile terminal in the current time slot or time period.

[0018] In any of the above technical solutions, the time length of the time period is related to the number of owners received by the platform, and / or the time length of the time slot is related to the charging time of the vehicle.

[0019] The beneficial effects of the present invention compared with the prior art:

[0020] Through the linkage between the data acquisition module and the queue estimation module, the platform can, based on the synchronous data of charging piles, battery swapping workstations, and entrance cameras, perceive the instantaneous queue length, service rate, and arrival rate changes at the station in real time, and then dynamically predict the future expected waiting time, so as to closely couple the pricing and guidance strategies with the actual supply and demand status, and avoid queue out - of - control or equipment idleness caused by information lag.

[0021] Through the pricing - demand mapping module, according to the relationship between the marginal cost of unit electric energy and the elastic demand of vehicles, a curve of the corresponding relationship between the listed price and time slots is dynamically generated, and combined with the second - level feedback adjustment of the dynamic adjustment module, precise control of the arrival behavior of vehicle owners by the listed price is achieved. It can effectively inhibit congestion during peak periods, promote equipment utilization during off - peak periods, and significantly improve the overall operation revenue and the stability of the station load.

[0022] Through the iterative convergence coordination module, within each time slot, continuously detect whether the equilibrium is reached between the arrival volume and waiting time prediction after the adjustment of the listed price. If not, continue to update the relationship curve and perform iterative calculations until dynamic equilibrium is achieved, ensuring that the price adjustment and service load converge synchronously, significantly improving the system stability and prediction accuracy, and avoiding decision - making lag and frequent manual intervention caused by static pricing in traditional systems.

[0023] The additional aspects and advantages of the embodiments according to the present invention will become apparent in the following description part, or be learned through the practice of the embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above - mentioned and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, where:

[0025] Figure 1 is a block diagram of the platform logic structure of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] In order to more clearly understand the above - mentioned objects, features, and advantages of the present invention, the present invention will be further described in detail below in conjunction with the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0027] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0028] Please refer to Figure 1 , and the following describes an operation management platform for automotive charging and battery swapping services according to some embodiments of the present invention.

[0029] An embodiment of the first aspect of the present invention provides an operation management platform for vehicle charging and swapping services. In some embodiments of the present invention, as Figure 1 shown, the platform includes:

[0030] A data acquisition module, configured to obtain service data of the charging station performing charging at the current moment through a station-side gateway.

[0031] A queue estimation module, which calculates the expected waiting time in a future time period according to the service data by means of a constructed prediction model under the conditions of considering the sudden load generated by the charging station performing battery swapping and the service time fluctuation affected by the listed price.

[0032] A pricing and demand mapping module, configured to predict the number of vehicles accepting charging and battery swapping within a time slot under the condition of considering the marginal cost of unit electric energy, and construct a relationship curve between the listed price and the time slot.

[0033] An owner sensitivity learning module, configured to construct an owner utility function by considering the relationship curve and the expected waiting time, and push and update an index affecting the owner's choice of service mode to the owner's mobile terminal through the owner utility function in a time period or a time slot.

[0034] A dynamic adjustment module, which updates the service data according to the service mode reported by the owner, and determines whether to trigger a charging queue threshold; if so, adjusts the listed price and broadcasts the update.

[0035] An iterative convergence coordination module, configured to determine whether the updated listed price and the expected waiting time are in an equilibrium solution in each time slot; if not, update the relationship curve and perform iterative calculation until an equilibrium solution is reached.

[0036] The present invention provides an operation and management platform for automobile charging and battery swapping services. The data acquisition module completes the high-fidelity collection of multi-dimensional real-time information through the three-layer collaborative mechanism of station-side gateway-edge bus-security pipeline. In principle, the platform embeds a Gigabit Ethernet interface inside each national standard DC charging gun, each liquid-cooled battery swapping mechanical arm, and each box transformer measurement and control device, forming a full-duplex star topology with the station-side gateway; the platform deploys the IEEE1588 precision clock protocol on the physical link to unify all sampling frames to the nanosecond time base, thereby ensuring that high-speed quantities such as instantaneous current peak, contact resistance, and battery core temperature can be aligned one by one in space. The gateway side loads a lightweight eBPF filter program in real time: first, the quartile distance group is eliminated in the ring buffer to remove the spike stray pulses generated by the connector jitter at the microsecond scale; then the Savitzky–Golay convolution kernel is applied to perform local polynomial smoothing on the high-frequency noise, which not only retains key dynamic features such as slope and acceleration, but also reduces the variance in the first-order and second-order differentials, thereby improving the separability of subsequent feature learning from the root. All packets that have passed the time calibration and filtering are packaged according to the five-tuple of "service index-millisecond timestamp-physical channel-engineering unit-original byte", enter the ZeroMQ-based publish-subscribe bus, and are synchronously transmitted to the edge inference engine and the cloud feature lake. In order to protect the collection link from malicious implants, the gateway firmware completes an SM2 two-way handshake with the platform root certificate during the startup phase, and appends a 256-bit HMAC-SM3 fingerprint to the end of each frame of the message. Any tampering will be immediately discarded on the edge side.

[0037] Unified time base and noise suppression allow the electrochemical response curves of the same vehicle in multiple scenarios such as battery replacement, fast charging, and balanced charging to be strictly aligned with sub-millisecond accuracy, significantly improving the convergence speed and accuracy of SoH estimation, abnormal arc light identification, and instantaneous power prediction models. Secondly, the eBPF pipeline is implemented in the gateway kernel state, and the data interception-processing-forwarding delay is less than 150 microseconds. Compared with the traditional "whole packet direct push" solution of caching first and then reporting, the average edge delay is shortened by 92%, providing sufficient iteration budget for the upper layer's 30-second rolling Stackelberg game. Thirdly, the five-tuple packet header allows any subsequent functional module to subscribe-parse-backtrack by field, without the maintenance cost of multiple versions of parsers. Historical data and real-time streams are absolutely consistent in physical quantity dimensions, which is convenient for digital twins and playback simulation. Finally, the full-link national secret signature and fingerprint verification ensure the isolation of the data plane and the control plane, prevent forged pile-side messages from squeezing out service capabilities, and the engine offline rate and false shutdown rate have decreased by 0.47 percentage points, directly improving the annual availability of the site and shortening the investment recovery period.

[0038] When performing battery swapping services at a charging station, since the battery swapping process usually triggers a higher user arrival density per unit time compared to the traditional charging process, the queue estimation module specifically constructs a dynamic perception mechanism for sudden loads in the prediction model. By real-time monitoring the number of battery swapping trigger events and the changing trend of the queue at the battery swapping positions, the module can form a weighted arrival rate adjustment coefficient based on the sudden load, accurately mapping the abnormal increase in the number of arriving users in a short time to the future queue model, so as to estimate the possible service congestion risk brought by it before the arrival of the battery swapping peak.

[0039] In terms of service time modeling, the queue estimation module not only considers the basic service capabilities of charging or battery swapping equipment, but also introduces the listed price as an influencing factor. Since the change in the listed price will directly affect the user's choice behavior, such as when the price is high, users may tend to charge for a short time or perform quick battery swapping, and when the price is low, they may choose slow charging or deep energy replenishment. Therefore, the module dynamically adjusts the mean and variance parameters of the service time when predicting the service time distribution. Specifically, through the price-service time elasticity function trained by historical service data, the real-time change in the listed price is mapped within the fluctuation range of the time required for each single service, enabling the model to accurately capture the subtle change trend of the service time under different price strategies and further improving the overall prediction accuracy.

[0040] In the model solving process, the queue estimation module uses the multi-server queuing system as the theoretical basis and introduces a service time fluctuation correction term. According to the current arrival rate, the adjusted service time distribution characteristics, and the number of service positions at the station, the module calculates the expected waiting time for multiple future small time periods, such as every 5 minutes, in a rolling manner. The model output not only includes the predicted value of the average waiting time for the next time period, but also can give the queuing trend change curve for a longer period according to the set window accumulation, such as the next 30 minutes or 1 hour, providing real-time reference for subsequent price guidance and vehicle owner decision-making.

[0041] Since the model can sensitively capture the characteristics of sudden loads caused by battery swapping at the charging station, it can effectively prevent the risk of queue out-of-control caused by the battery swapping peak and greatly improve the platform's response ability to extreme load fluctuations. Compared with the traditional static prediction method, the queuing error of the system during the battery swapping tide is significantly reduced, effectively ensuring service continuity during the peak period.

[0042] Dynamically considering the impact of the listed price on the service time enables the prediction results to be adjusted in real time with the change of the price strategy, avoiding the problem of the disconnection between price changes and the actual energy replenishment rhythm, fundamentally improving the accuracy and effectiveness of price guidance, and ensuring that pricing measures and queuing management form a synchronous linkage.

[0043] Since it is able to continuously calculate the future expected waiting time based on real-time service data, the platform can trigger corresponding price adjustment or guidance strategies in advance before the queuing volume exceeds the standard, and actively guide vehicle owners to refuel during appropriate time periods, thus achieving preventive peak control and reducing the passive emergency measures that rely on manual flow limiting or forced station closure after the fact.

[0044] As the basic input parameter for subsequent modules such as price optimization, vehicle owner decision-making, and queue feedback, the expected waiting time significantly enhances the link consistency and closed-loop stability of the entire operation management platform through highly accurate prediction, providing solid data support for achieving second-level dynamic scheduling and dual optimization of revenue and experience.

[0045] In the process of predicting the number of vehicles receiving services, the pricing and demand mapping module uses the marginal cost of unit electric energy as the benchmark for the lower limit of pricing, and incorporates the basic economic constraint conditions of the supply side into the prediction model. Through the real-time obtained service data, the current electricity procurement price of the station, and the operation cost data, the module accurately calculates the lowest acceptable electricity price level corresponding to providing charging or swapping services for each vehicle per unit time. This marginal cost (in yuan per kilowatt-hour) serves as the basis for pricing decisions, ensuring that there is no risk of price inversion in all subsequent listed price formulations and guaranteeing the platform to maintain positive revenue in each time slot.

[0046] In the demand forecasting section, the pricing and demand mapping module introduces the changing pattern of vehicle price sensitivity into the modeling. By collecting historical service data, an empirical elasticity function between the listed price and the arrival probability of vehicle owners is established, which can deduce the changing trend of the number of vehicles receiving charging or swapping services in each time slot according to different listed price levels. Specifically, the module calculates the probability of a vehicle choosing this time slot by means of exponential mapping, with the listed price and the predicted value of the waiting time in the current time slot as independent variables, and multiplies it by the basic arrival volume.

[0047] Since the pricing is based on the marginal cost of unit electric energy, the platform can ensure positive revenue for each transaction under any load condition, avoid economic losses caused by random price cuts, and enhance the overall operation stability of the station. By incorporating both waiting time and price changes into the demand forecasting function, the module can accurately reflect the actual refueling selection logic of users, enabling the listed price to not only regulate the number of vehicle arrivals but also synchronously affect the arrival time and service type selection of vehicle owners, truly realizing the deep linkage between price and user behavior. The real-time optimization of the curve relationship between the listed price and the time slot enables the platform to automatically adapt to supply and demand changes at different times, forming an intelligent dynamic scheduling mechanism that actively cuts peaks during peak hours and promotes sales actively during off-peak hours, significantly alleviating the queuing congestion or equipment vacancy problems caused by traditional fixed prices.

[0048] The vehicle owner sensitivity learning module is used to dynamically construct the utility function of vehicle owners on the basis of the relationship curve between the listed price and time slots and the expected waiting time already constructed by the platform, further combining the individual characteristics of each vehicle owner. By comprehensively considering the listed price, expected waiting time, price sensitivity, and time sensitivity of vehicle owners, this module evaluates the preference degree of vehicle owners for choosing charging or swapping services in each time slot, and in each time period or time slot, it pushes the impact index calculated based on the utility function to the mobile device of the vehicle owner to assist the vehicle owner in making refueling decisions. At the same time, it continuously updates and optimizes the individual sensitivity parameters according to the actual selection behavior of the vehicle owner to continuously improve the accuracy and guiding effect of subsequent pushes.

[0049] The vehicle owner sensitivity learning module first takes the relationship curve between the listed price and time slots output by the platform in the current rolling cycle, and the predicted value of the expected waiting time corresponding to the time slot, as the basic input. For each vehicle owner in the online state or about to arrive at the station, the module initializes or updates their price sensitivity parameter and waiting time sensitivity parameter according to information such as their historical charging and swapping records, refueling preferences, current battery power, and remaining range requirements. These sensitivity parameters reflect the tolerance and reaction speed of vehicle owners to price changes and queuing time changes, and are the core basis for constructing personalized utility functions.

[0050] In specific applications, the vehicle owner sensitivity learning module calculates the comprehensive utility value of each time slot for the vehicle owner by combining the current listed price and expected waiting time for several future optional time slots. The lower the utility value, the more attractive the time slot is and the easier it is to be selected by the vehicle owner. For the convenience of users' understanding and interface presentation, the utility value is normalized and converted into an impact index with intuitive numerical values and easy to compare. The higher the impact index, the more favorable the time period is for the vehicle owner in terms of comprehensive factors such as price, queuing, and refueling convenience. Through the mobile application or in-vehicle intelligent terminal, the impact index of several future time slots is pushed to the vehicle owner and displayed in the form of lists, graphs, recommended prompts, etc., to guide the vehicle owner to independently select the optimal time period for charging or swapping services. The pushed content includes both price information and estimated waiting time prompts, enabling the vehicle owner to make a more rational decision that meets their own needs while understanding the refueling cost and time cost.

[0051] More importantly, the vehicle owner sensitivity learning module does not fixedly use the initial sensitivity parameters, but dynamically optimizes the sensitivity modeling by continuously tracking the actual response results of the vehicle owner. When the vehicle owner selects a certain time slot for charging after receiving the push, the module adjusts the price sensitivity and waiting time sensitivity of the vehicle owner according to the deviation between the actual selection and the predicted optimal time slot. For example, if the vehicle owner frequently prefers a lower price and tolerates a longer waiting time, the system will automatically lower the weight of their price sensitivity and increase the weight of the waiting time; if the vehicle owner tends to choose a time slot with a short waiting time, even if the price is slightly higher, the time sensitivity will be appropriately adjusted. This dynamic update process adopts an incremental learning mechanism, which can gradually converge to a sensitivity model that more truly and accurately reflects the vehicle owner's behavior habits without increasing the operation burden on the vehicle owner.

[0052] At the beginning of each time slot, the platform first initializes the listing price setting and waiting time estimation for this time slot based on the vehicle owner response behavior output by the vehicle owner sensitivity learning module in the previous control cycle, combined with the listing price and demand prediction curve provided by the pricing and demand mapping module, and the expected waiting time prediction result provided by the queue estimation module.

[0053] Based on this, the iterative convergence coordination module first checks whether the current listing price level and the expected waiting time prediction can match each other. Specifically, it judges whether the listing price has achieved the set control objectives in terms of guiding the vehicle owner to select time slots, balancing the queuing load, and controlling the number of arriving vehicles, such as parameters like the target vehicle quantity, tolerable queuing length, and maximum allowable waiting time. If the system detects a deviation between the actual prediction result and the control objective, it indicates that there is no ideal equilibrium relationship between the current listing price and the vehicle owner's response.

[0054] At this time, the iterative convergence coordination module will dynamically adjust the listing price according to the deviation direction and magnitude, and synchronously correct the demand mapping curve. For example, if the system predicts that the number of arriving vehicles is higher than the set target, or the expected waiting time is higher than the tolerance threshold, the module will actively increase the listing price to increase the price and suppress some potential demand; conversely, if the predicted number of vehicles is insufficient or the waiting time is too short, the module will appropriately lower the listing price to attract more vehicle owners to charge during this time slot, thereby optimizing the equipment utilization rate and revenue level of the station.

[0055] After each adjustment of the listing price, it will synchronously trigger the pricing and demand mapping module to regenerate the relationship curve between the new listing price and the arrival volume, and call the queue estimation module to update the expected waiting time according to the latest prediction result. The new listing price and waiting time are input into the iterative convergence coordination module again for a new round of equilibrium detection.

[0056] The above process is carried out in a rolling loop at a fixed iteration frequency or within the maximum iteration count limit. Each iteration aims to reduce the deviation between the posted price guiding effect and the actual demand response. When it is detected that the changes in the posted price and the expected waiting time are both below the preset convergence threshold in two consecutive iterations, the module determines that the system has reached a stable equilibrium. At this time, the posted price and the waiting time prediction are locked as the service strategy finally released in this time slot for reference by the subsequent vehicle owner decision-making module and are officially pushed to the vehicle owner side application.

[0057] If the system fails to converge within the limited maximum iteration count, the module will automatically trigger an emergency convergence mechanism and push the posted price and waiting time that are currently the closest to the equilibrium as a sub-optimal strategy to ensure that the time slot update cycle will not be overly extended and to guarantee the coherence and real-time nature of the overall platform operation rhythm.

[0058] To sum up, the data acquisition module collects service data at the millisecond level, the queue estimation module predicts the expected waiting time at the second level, and the dynamic adjustment module instantaneously adjusts the posted price. The three form a high-frequency closed loop. The platform always makes the number of arriving vehicles match the station service capacity, and the phenomenon of out-of-control queuing is significantly reduced. The queue estimation module identifies sudden load changes during battery swapping, and the pricing and demand mapping module uses the price lever to suppress peaks or boost valleys in advance, making the load curve tend to be smoother. The waiting time during the charging peak decreases, the equipment utilization rate during the low valley period increases, and the demand on the power grid side is more stable. The pricing and demand mapping module takes the marginal cost of unit electric energy as the bottom line to ensure a positive profit for each transaction; the vehicle owner sensitivity learning module pushes a personalized influence index to guide vehicle owners to find a balance between waiting costs and monetary costs, and the vehicle owner experience and the station profit increase simultaneously. The dynamic adjustment module immediately broadcasts a new posted price when it detects that the queue reaches the threshold, and the iterative convergence coordination module locks it after confirming that the price-waiting combination enters the equilibrium. The traditional "minute-level or even hour-level" price adjustment lag is compressed to the "second level", and the peak flow limiting demand is significantly reduced. The iterative convergence coordination module records the convergence trajectory of each round, and the vehicle owner sensitivity learning module records each behavior feedback. The two data streams enter the model training pipeline. The longer the platform runs, the smaller the prediction error and the price adjustment oscillation amplitude, and the long-term operation cost continues to decrease.

[0059] In any of the above embodiments, the service data includes the queue length, service rate, arrival rate, and posted price synchronously collected from charging piles, battery swapping workstations, or entrance cameras.

[0060] In this embodiment, high-precision time synchronization modules are respectively embedded in three types of on-site devices: charging piles, battery swapping stations, and entrance cameras. The control boards of the charging piles and battery swapping stations output the service rate and listed price in real time. The entrance camera uses a vehicle recognition algorithm to capture the arrival rate and aligns the frame numbers through the same time pulse. All data items - queue length, service rate, arrival rate, and listed price - are packaged under the same millisecond-level time tag, and after being aggregated by the station-side gateway, they are immediately sent to the edge bus to realize the service data stream of "spatially distributed sampling, absolutely synchronized time, and unified field encapsulation". In this way, any subsequent module can directly retrieve the complete data according to the unified time axis without the need for secondary alignment or splicing.

[0061] In any of the above embodiments, the queue estimation module includes:

[0062] A vehicle arrival rate filtering unit for continuously correcting the vehicle arrival rate using recursive Kalman filtering.

[0063] A service time variance evaluation unit for statistically calculating the variance of the service time caused by both the sudden load generated by battery swapping and the charging time in a sliding window of a preset length.

[0064] A variance correction and prediction unit for embedding the service time variance into a prediction model to obtain the expected waiting time of the charging station during charging and / or battery swapping.

[0065] In this embodiment, the vehicle arrival rate filtering unit receives the vehicle arrival rate curve from the data acquisition module in real time, uses recursive Kalman filtering to smooth the spike noise, recognition errors, and camera occlusion jitters in the original curve in real time, and automatically adjusts the gain coefficient according to the predicted value of the previous moment, taking into account both tracking agility and noise suppression stability. After filtering, a continuous, stable arrival rate estimate that is sensitive to sudden traffic flows is obtained, providing a reliable input for subsequent queuing predictions.

[0066] The service time variance evaluation unit statistically calculates the actual service duration of charging and battery swapping tasks in a sliding window of a set length, incorporating single battery swaps that are quickly completed, deep DC fast charging with a long duration, and possible abnormal retries into the same time series, thus completely characterizing the discrete degree of service duration. Phenomena such as the variance increase during peak battery swapping moments and the variance reduction caused by slow charging at night are retained on the same time axis, providing an accurate fluctuation quantification index for subsequent models.

[0067] The variance correction prediction unit directly embeds the service time variance index obtained in the previous step into the waiting time prediction model, allowing the fluctuation level to have an explicit impact on the result when calculating the queuing evolution. When the variance increases, the prediction model automatically raises the waiting time estimate; when the variance decreases, the model lowers the waiting time to avoid wasting service capacity due to excessive conservatism. The corrected waiting time prediction, together with the arrival rate filtering result, is synchronously pushed to the pricing and demand mapping module and the vehicle owner sensitivity learning module.

[0068] Specifically, the vehicle arrival rate filtering unit is calculated by the following formula:

[0069]

[0070] In the formula, are the estimated covariance and the observation noise respectively; is the Kalman gain at time t, which controls the fusion ratio of the actual observation value and the predicted value; is the prediction for time t + 1 under the condition of the observed information at time t; is the prior prediction at time t, that is, the estimate before the filtering update; is the actual observed vehicle arrival rate at time t, that is, the original sensor count.

[0071] Specifically, the service time variance evaluation unit is calculated by the following formula:

[0072]

[0073] In the formula, is the time taken for a single charging / recharging; is the sliding window width; is the mean of the service durations in the most recent w time instants, that is, the average service time.

[0074] In any of the above embodiments, the prediction model includes the following formula:

[0075]

[0076] Among them, is the time period; is the expected waiting time for the next time period; is the system load rate; is the service time variance; is the number of parallel service stations; is the factorial of c; is the average service rate of a single station; is the predicted arrival rate.

[0077] In any of the above embodiments, the pricing and demand mapping module includes:

[0078] The elastic demand mapping unit constructs the relationships between the expected waiting time, the elasticity of the listed price, and the potential number of vehicles in different time slots.

[0079] The equilibrium optimization unit is used to balance the revenue affected by the listed price and the congestion penalty affected by the charging queue, and solve to obtain the optimal price vector.

[0080] In this embodiment, based on the real-time collected service data and combined with the feedback of the owners' arrival behavior in the previous cycle, the elastic demand mapping unit dynamically constructs the elastic mapping relationship between the expected waiting time and the listed price on the number of owners arriving. According to the change trend of the number of vehicles arriving at the station in different time periods, the unit divides the whole day into several time slots, and respectively fits the reaction intensity of the owners to the extension of the waiting time and the sensitivity to the change of the listed price within each time slot. Through such time-segmented modeling, the platform can accurately depict the specific impact of different combinations of listed prices and waiting times on the number of potential owners under different operating conditions, so as to form a detailed, multi-time-slot, multi-dimensional potential traffic flow prediction relationship diagram.

[0081] After the elastic relationship is established, the equilibrium optimization unit takes over the elastic prediction result and begins to comprehensively consider the revenue growth brought by the increase in the listed price and the congestion penalty brought by the elongation of the charging queue, and conducts multi-objective trade-off optimization. First, calculate the theoretical revenue corresponding to the current listed price level in each time slot, and simultaneously estimate the decline in customer experience and the loss of queuing resource occupancy caused by the increase in vehicle waiting time, and combine the two into a unified evaluation index. Subsequently, within the set upper and lower price limits, the equilibrium optimization unit iteratively adjusts the listed price of each time slot according to the principle of maximizing revenue and acceptable queuing costs until a set of optimal price vectors is obtained. At the same time, to avoid the overly drastic change in the listed price in different time periods from affecting the owners' expectations, a price change smoothing constraint is introduced during the optimization process to ensure that the finally generated listed price curve is continuous, stable, and easy to understand in the time dimension.

[0082] In any of the above embodiments, the equilibrium optimization unit adjusts the solution mode for obtaining the optimal price vector according to the timeliness requirements in the time period and time slot; the output formats of the optimal price vectors in the time period and time slot are the same.

[0083] In this embodiment, the equilibrium optimization unit first determines, according to the current platform operation rhythm, whether it is in a wide time period price adjustment mode mainly based on hours or enters a slot-level fine price adjustment mode mainly based on minutes or even finer granularity. In the time period regulation mode, the platform focuses on the supply-demand balance and the overall revenue trend in the medium and long term. Therefore, the module adopts an optimization strategy with a lower update frequency and a higher comprehensive revenue weight. By solving the listing price trajectory covering the entire time period, it preferentially smooths the price fluctuations in the large cycle, ensures the stability and consistency of price adjustment, and prevents disturbing the decision-making rhythm of car owners due to frequent price adjustments.

[0084] In the slot regulation mode, especially in dynamic scenarios such as sudden peak of battery swapping and concentrated temporary traffic flow, the platform needs to respond quickly at a finer time resolution. At this time, the equilibrium optimization unit adopts an optimization mode with a higher update frequency and a stronger queuing penalty weight according to the rapid fluctuations of the real-time arrival rate and the expected waiting time, and finely adjusts the listing price for each independent slot, striving to suppress sudden queuing backlogs within a time window from seconds to minutes, dynamically disperse the arrival times of car owners at the station, and achieve rapid load peak shaving.

[0085] Whether using time period optimization or slot optimization, the format of the optimal price vector finally output by the equilibrium optimization unit remains the same, which is a price list arranged according to the time index. Each record corresponds to a specific future service window and includes the corresponding listing price, valid time range, and version number. This is convenient for the subsequent car owner sensitivity learning module, dynamic adjustment module, and iterative convergence coordination module. When receiving and processing the listing price push, they do not need to distinguish the source, and only need to parse according to the time index order to complete the update of the guidance strategy and the collection of behavior feedback.

[0086] Specifically, taking the slot as an example, the discrete MIP mode is adopted for medium concurrency scenarios, and the model construction includes:

[0087] Leader layer objective function:

[0088]

[0089] In the formula, is the marginal cost of unit electric energy, that is, the lowest internal energy supply cost for the platform to provide one degree of electricity; is the queuing penalty weight coefficient, which measures the negative impact of excessive queuing volume on the overall benefit; is the price fluctuation penalty weight coefficient, which controls the penalty intensity for the change of the listing price relative to the benchmark price, avoids violent fluctuations in the price curve, and improves the car owner experience; is the square of the queuing length.

[0090] Follower layer KKT constraint:

[0091]

[0092] In the formula, is the partial derivative of the owner's utility function with respect to the charging time, representing the change rate of the comprehensive utility felt by the owner when the charging time changes slightly; is the Lagrange multiplier related to the lower bound of the charging time selection, used to detect whether the boundary condition of the earliest charging time is reached; is the charging time actually selected by the owner; is the earliest charging time that the owner is allowed to select, that is, the lower bound; is the Lagrange multiplier related to the upper bound of the charging time selection, used to detect whether the boundary condition of the latest charging time is reached; is the latest charging time that owner i is allowed to select, that is, the upper bound.

[0093] Discretization and solution are as follows: Represent the owner's decision with mutually exclusive 0-1 variables and add SOS1 constraints. The master-slave layer is merged into a single mixed complementary integer programming (MPEC-MIP). First, perform sparse rearrangement on the CPU and then upload it to the GPU; GPU-HiGHS uses the two-phase simplex method, and the goal is to obtain a feasible optimal solution within 25 seconds. If a feasible solution is not obtained within 25 seconds, the controller immediately degrades to the continuous Mean-Field mode and retains 5 seconds to complete the analytical solution and broadcast.

[0094] Specifically, taking time slots as an example, for high-concurrency scenarios, the continuous Mean-Field mode is adopted, and the model construction includes:

[0095] The owner's strategy adopts the probability density :

[0096]

[0097] In the formula, is the average waiting and price sensitivity of the owner group; is the temperature factor.

[0098] Leader layer general function:

[0099]

[0100] In the formula, is the comprehensive benefit of the platform over the entire planning period and is a function of the listed price strategy p. The goal is to maximize it; is the total number of potential owners or the total arrival demand in the current period, used to normalize the demand density.

[0101] Analyze the first-order conditions:

[0102]

[0103] Container constraint correction:

[0104] where is the predicted expected waiting time; is the service capacity per unit time of time slot τ, which is the product of the number of parallel workstations and the single-workstation rate.

[0105] If it is necessary to satisfy , starting from the analytical value and entering L-BFGS-B, it only takes 5-8 iterations to converge, and the time consumption is about 3 seconds.

[0106] Furthermore, the rolling iterative solution is:

[0107]

[0108] where is the set representing all time slots, that is, all time periods or time intervals defined in the optimization process; is the predicted queue length at time slot under the k-th iteration; is the posted price at time slot under the k-th iteration; is the convergence threshold of the queue length, indicating how many vehicles within the queue prediction change can be considered convergent; is the convergence threshold of the posted price, indicating how many yuan within the posted price change can be considered convergent; Both conditions must be satisfied simultaneously to determine the overall convergence.

[0109] In any of the above embodiments, the elastic demand mapping unit includes the following formula:

[0110]

[0111] where is the industry elasticity curve; are the sensitivity weights respectively; is the expected number of arriving vehicles; is the current reference price; is the price of the current time slot to be optimized; is the expected waiting time of the current time period.

[0112] In any of the above embodiments, the owner sensitivity learning module includes:

[0113] A sensitivity update unit that updates the waiting sensitivity and price sensitivity according to the individual historical behavior data of the owner.

[0114] An exponential evaluation unit that obtains an index through the owner utility function according to the current waiting sensitivity and price sensitivity in each time period.

[0115] In this embodiment, the sensitivity update unit uses the historical behavior data of each car owner as the basic input to dynamically update the car owner's sensitivity to waiting time and price changes. The platform records in real time the selection behavior of each car owner when facing different listing prices and expected waiting times, and extracts key features including arrival time slots, actual selection time slots, price levels, waiting times, and actual completed charging and swapping behaviors. After collecting a certain amount of sample data, the sensitivity update unit continuously calculates the behavioral elasticity of each car owner to price changes and waiting time changes through an incremental learning method. If the car owner shows a strong willingness to transfer when facing price increases, the system automatically increases its price sensitivity; if the car owner tends to endure a longer wait but insists on replenishing energy, the system automatically adjusts its waiting sensitivity. The update mechanism is not only adjusted based on a single behavior, but also weighted averaged through long-term behavior trajectories to prevent short-term abnormal behaviors from causing drastic fluctuations in sensitivity parameters, ensuring that the model converges stably and gradually fits the real preferences of the car owner.

[0116] The index evaluation unit calculates the comprehensive index for each time period based on the current waiting sensitivity and price sensitivity of each car owner, combined with the listing price and expected waiting time information generated in real time by the platform, using the car owner's utility function. The index evaluation unit uses the utility function as the calculation core, uniformly converts the price cost and waiting cost into a negative utility value for the car owner, and further converts the comprehensive utility value of different time periods into a standardized impact index through index mapping. The impact index is presented in the form of intuitive charts or scores on the car owner's mobile terminal to guide the car owner to choose the optimal charging time or service method. The difference in index values for different time periods reflects the comprehensive attractiveness of each time period, helping car owners quickly understand and compare the charging costs under different options.

[0117] Specifically, the sensitivity update unit is calculated by the following formula:

[0118]

[0119] In the formula, is the waiting sensitivity for updates, is the price sensitivity to be updated; To wait for the memory factor; is the cost memory factor; is the response indication of the owner in the rth record.

[0120] In any of the above embodiments, the index evaluation unit includes the following formula:

[0121]

[0122] in, is an index used to characterize the owner In time slot the selection utility within; is the waiting sensitivity; is the price sensitivity; is the price locked by the vehicle owner's mobile device in the current time slot or time period.

[0123] In any of the above embodiments, the time length of the time period is related to the number of vehicle owners received by the platform, and / or the time length of the time slot is related to the charging time of the vehicle.

[0124] In this embodiment, for the setting of the time period, when the number of vehicle owners accessing the platform is small, the platform uses a longer time period for operation prediction and listing price update to improve the overall stability and calculation efficiency of system scheduling. When the number of vehicle owners is relatively scarce, such as at night or during off-peak periods, the arrival frequency of vehicles is low and changes smoothly. The platform can formulate listing prices and load balancing strategies in units of hours or even longer time periods. The principle of this design is that under the condition of little change in vehicle flow, adjusting prices too frequently is likely to cause unnecessary fluctuations and interfere with the experience of vehicle owners. Smoother regulation over a longer time period better meets the needs of low-load scenarios. When the number of vehicle owners received by the platform rises rapidly, especially during peak hours or sudden battery swapping tides, the arrival frequency of vehicles increases sharply, and the queuing state changes quickly and fluctuates greatly. The platform then automatically shortens the time period length and enters a higher-frequency price and queuing regulation rhythm, dynamically adapting to congestion changes with a refresh cycle from seconds to minutes. This linkage mechanism between the time period length and the number of vehicle owners enables the platform to flexibly adjust the scheduling granularity under different vehicle flow scales, maintaining both operational continuity and avoiding unnecessary resource consumption caused by overly fine-grained scheduling.

[0125] In the setting of the time slot, the platform strictly determines the time slot length dynamically according to the actual service time required for vehicle charging or battery swapping. When vehicles mainly use fast charging or battery swapping and the single-vehicle service time is generally short, the platform sets the time slot length relatively compactly, usually between five and ten minutes. In this way, the dynamic changes of vehicle queuing, charging, and leaving the station within each time slot can be captured in a timely manner and fed back into the listing price and queuing estimation model, ensuring that the service state and price guidance strategy are adjusted synchronously. On the contrary, when the charging mode at the station is mainly slow charging and deep energy replenishment, and the single-vehicle charging duration is long, such as thirty minutes to one hour, the platform automatically extends the time slot length, making the energy replenishment behavior covered within each time slot have stronger continuity and predictability, and avoiding excessive adjustment of prices or service strategies due to short-term fluctuations.

[0126] Furthermore, for each vehicle, the time lengths of the time slot and the time period can be different.

[0127] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above-mentioned functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working process of the units and modules in the above system can refer to the corresponding process in the foregoing method embodiment and will not be elaborated here.

[0128] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this disclosure.

[0129] In the embodiments provided in this disclosure, it should be understood that the disclosed device / electronic device and method can be implemented in other ways. For example, the device / electronic device embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. Multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical, mechanical or other form.

[0130] When the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present disclosure, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. The computer program can include computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0131] The above embodiments are only used to illustrate the technical solutions of the present disclosure, and are not intended to limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present disclosure, and should all be included in the protection scope of the present disclosure.

Claims

1. An operation and management platform for vehicle charging and battery swapping services, characterized in that, Including: A data acquisition module, configured to obtain service data of the charging station performing charging at the current moment through a substation gateway; A queue estimation module, which calculates the expected waiting time in a future time period according to the service data by means of a constructed prediction model under the conditions of considering the sudden load generated by the charging station performing battery swapping and the service time fluctuation affected by the listed price; A pricing and demand mapping module, configured to predict the number of vehicles accepting charging and battery swapping within a time slot under the condition of considering the marginal cost of unit electric energy, and construct a relationship curve between the listed price and the time slot; A vehicle owner sensitivity learning module, configured to construct a vehicle owner utility function considering the relationship curve and the expected waiting time, and push and update an index affecting the vehicle owner's service mode selection to the vehicle owner's mobile terminal through the vehicle owner utility function in the time period or the time slot; A dynamic adjustment module, which updates the service data according to the service mode reported by the vehicle owner and determines whether to trigger a charging queue threshold; If so, adjust the listed price and broadcast the update; An iterative convergence coordination module, configured to determine whether the updated listed price and the expected waiting time are at an equilibrium solution in each time slot, and if not, update the relationship curve and perform iterative calculation until reaching the equilibrium solution.

2. The operation management platform for automobile charging and battery swapping services according to claim 1, wherein The service data includes the queue length, service rate, arrival rate, and listed price synchronously collected from charging piles, battery swapping workstations, or entrance cameras.

3. The operation management platform for vehicle charging and battery swapping services according to claim 2, wherein The queue estimation module includes: A vehicle arrival rate filtering unit, configured to continuously correct the vehicle arrival rate by using recursive Kalman filtering; A service time variance evaluation unit, which statistically calculates the service time variance jointly generated by the sudden load generated by battery swapping and the charging time consumption within a sliding window of a preset length; A variance correction prediction unit, which embeds the service time variance into the prediction model to obtain the expected waiting time of the charging station when performing charging and / or battery swapping.

4. The operation and management platform for vehicle charging and swapping services according to claim 3, wherein, The prediction model includes the following formula: ； Among them, is the time period; is the expected waiting time for the next time period; is the system load rate; is the variance of the service time; is the number of parallel service stations; is the average service rate of a single station; is the predicted arrival rate.

5. The operation management platform for automobile charging and battery swapping services according to claim 1, characterized in that, The pricing and demand mapping module includes: An elastic demand mapping unit, which constructs a relationship between the expected waiting time, the elasticity of the listed price, and the potential number of vehicles in different time slots; An equilibrium optimization unit, configured to balance the revenue affected by the listed price and the congestion penalty affected by the charging queue, and solve to obtain an optimal price vector.

6. The operation management platform for vehicle charging and battery swapping services according to claim 5, characterized in that, The equilibrium optimization unit adjusts the solution mode for obtaining the optimal price vector according to the time efficiency requirements in the time period and the time slot; the output formats of the optimal price vectors in the time period and the time slot are the same.

7. The operation management platform for vehicle charging and battery swapping services according to claim 5, wherein The elastic demand mapping unit includes the following formula: ； Among them, is the industry elasticity curve; are the sensitivity weights respectively; is the expected number of arriving vehicles; is the current reference price; is the price of the current time slot to be optimized; is the expected waiting time of the current time period.

8. The operation management platform for automobile charging and battery swapping services according to claim 7, wherein The vehicle owner sensitivity learning module includes: A sensitivity update unit, which updates the waiting sensitivity and price sensitivity according to the individual historical behavior data of the vehicle owner; An index evaluation unit, which obtains the index through the vehicle owner utility function according to the current waiting sensitivity and price sensitivity in each time period.

9. The operation management platform for vehicle charging and battery swapping services according to claim 8, characterized in that, The index evaluation unit includes the following formula: ； Among them, is the said exponent, used to characterize the car owner in the time slot the selection utility within; is the waiting sensitivity; is the price sensitivity; is the price locked by the mobile terminal of the car owner in the current time slot or time period.

10. The operation and management platform for vehicle charging and battery swapping services according to claim 1, characterized in that, The time length of the time period is related to the number of vehicle owners received by the platform, and / or the time length of the time slot is related to the charging time of the vehicle.

Images