An orderly charging scheduling method for electric vehicles based on time-of-use electricity price response and priority ranking

Abstract

The application discloses a kind of based on demand prediction, electricity price correction and rolling feedback electric vehicle charging scheduling method and system, belong to electric vehicle orderly charging, charging station energy management and intelligent control technical field.This method collects charging pile operating state, transaction record, metering data and charging session information, and pretreats historical data;Future charging demand is predicted using EWMA and ARIMA fusion model, combined with peak flat valley time-of-use electricity price, demand response coefficient and SOC gap to establish charging priority scoring mechanism;Under the constraints of site capacity, pile rated power, vehicle acceptable power and power change rate, segmented charging power allocation scheme is generated.System dynamically adjusts pile end output through charging control interface, and realizes rolling optimization by combining metering feedback, SOC update, deviation monitoring and parameter calibration, thereby reducing peak load pressure, improving the economy, fairness and dynamic adaptability of charging scheduling.

Description

Technical Field

[0001] This invention belongs to the field of electric vehicle orderly charging and charging operation scheduling technology, and relates to an electric vehicle orderly charging scheduling method based on time-of-use electricity price response and priority ranking. In particular, it relates to an electric vehicle orderly charging scheduling method that combines charging demand time-series prediction, time-of-use electricity price, user demand response coefficient, multi-factor priority ranking, sorting greedy power allocation, and Open Charging Point Protocol (OCPP) power scheduling command issuance. Background Technology

[0002] With the continuous growth of electric vehicle ownership, the construction scale of public charging stations, community charging stations, and highway service area charging stations is constantly expanding. For charging operators, charging scheduling not only needs to meet the needs of users to reach the target state of charge before the target departure time, but also needs to take into account transformer capacity constraints, site basic load, time-of-use electricity price costs, and the engineering feasibility of charging pile power control.

[0003] In terms of charging pile access and centralized management, the central management system (CSMS) and charging piles typically communicate using the Open Charging Point Protocol (OCPP). OCPP enables charging pile access, status reporting, metering data uploading, transaction management, and power scheduling command issuance, providing a protocol foundation for orderly charging scheduling at charging stations. An engineering platform based on the OCPP server can undertake the basic functions of charging pile operation data acquisition and power control command issuance.

[0004] Existing electric vehicle charging scheduling methods mainly fall into the following categories: The first category uses historical averages or recent values ​​as the predicted charging demand for the next time period. While simple to calculate, these methods struggle to reflect intraday variations. The second category employs deep learning models such as LSTM and GRU for charging demand prediction. Although these models have strong fitting capabilities, they require numerous parameters, demanding training samples, and lack interpretability, hindering rapid deployment of small and medium-sized charging stations. The third category uses time-of-use pricing only for post-event electricity billing, failing to explicitly incorporate price signals into the charging session sequencing and power allocation process. The fourth category uses First-Come, First-Served (FCFS) or sorts charging sessions solely by State of Charge (SOC) from low to high, making it difficult to simultaneously consider user departure time, target power gap, and current electricity price factors. The fifth category uses rolling model predictive control, mixed-integer linear programming, or second-order cone programming for power optimization allocation. This results in high solution complexity, typically relying on commercial solvers and significant computational resources. The sixth category only outputs the optimal power curve at the simulation level, failing to directly map the power allocation results to the power scheduling object defined by the OCPP protocol, leading to a significant gap between algorithm simulation and engineering implementation.

[0005] Therefore, there is an urgent need to provide a lightweight, interpretable, easy-to-deploy method for orderly charging scheduling of electric vehicles that can directly interface with the OCPP protocol, so that time-of-use pricing can participate in charging demand correction and session sequencing, and complete real-time power allocation under the constraints of site capacity, charging pile rated power, vehicle acceptable power, and power change rate. Summary of the Invention

[0006] The technical problem this invention aims to solve is to address the issues of existing electric vehicle charging scheduling methods, such as overly simplistic or overly complex demand forecasting, lack of coupling between electricity price signals and sorting mechanisms, single sorting rules, high deployment threshold for optimization solutions, and disconnection between control commands and the OCPP protocol layer. This invention provides an orderly charging scheduling method for electric vehicles based on time-of-use electricity price response and priority sorting.

[0007] To achieve the above objectives, the present invention provides the following technical solution: An orderly charging scheduling method for electric vehicles based on time-of-use pricing response and priority ranking includes the following steps: S1. The central management system CSMS receives charging pile reports based on the Open Charging Point Protocol (OCPP), collects static information, real-time observation information, charging session information, allocable capacity of the site transformer, and basic load of the charging pile, and performs anomaly detection, anomaly replacement, time alignment and normalization on the real-time active power sequence to obtain the actual charging power demand of the site. S2. The actual charging demand power of the station is predicted by exponential weighted moving average (EWMA) for short-term forecasting and autoregressive integral moving average (ARIMA) for long-term forecasting. The EWMA model and the ARIMA model are then fused by reciprocal weighting based on the mean square error of the EWMA model and the ARIMA model on the historical validation set to obtain the fused predicted charging demand power of the station for future periods. S3. Based on the peak-hour electricity price, the normal-hour electricity price, the off-hour electricity price, and the user demand response coefficient, the fused predicted charging demand power is adjusted for price to obtain the effective charging demand power prediction value after price adjustment. S4. For each charging session, calculate a comprehensive priority score based on urgency, target power gap, and electricity price incentives. S5. Sort each charging session from high to low according to the comprehensive priority score, and determine the allocated charging power of each charging session in the current scheduling period by using a sorting greedy allocation method under the constraints of site capacity, charging pile rated power, vehicle acceptable power and power change rate. S6. The allocated charging power is encapsulated into a charging power scheduling object defined by OCPP according to the scheduling cycle, and pushed to the corresponding charging pile through the power scheduling distribution interface of CSMS. S7. Update the state of charge of the charging session based on the metering value returned by the charging pile, and execute S1 to S6 in a rolling manner according to a fixed scheduling cycle to form a closed-loop scheduling process of collection, prediction, sorting, allocation, distribution and feedback.

[0008] Furthermore, in S1, the charging pile reporting messages include access notifications, status notifications, metering values, start transactions, and end transactions; the charging pile static information includes the pile number. n connector k and rated maximum power The real-time observation information includes the pile's working status and real-time active power. and electricity The charging session information includes the target departure time. Initial state of charge, target state of charge Battery capacity and charging efficiency The anomaly detection uses a length of Sliding window and according to formula Identify outliers, among which and The mean and standard deviation of the real-time active power samples within the sliding window are respectively used to linearly interpolate and replace the identified outliers with adjacent normal sampling points.

[0009] Furthermore, in S2, the EWMA short-term forecast is based on the formula... In progress, among which For the site N During the period t EWMA charging demand power forecast, For the site N During the period t The actual charging power required. The smoothing coefficient is 0 < ≤1; ARIMA long-term forecast is based on the formula In progress, among which B For delay operators, It is an autoregressive polynomial. It is a moving average polynomial. p Let the order be the autoregressive order. d Let be the difference order. q The moving average order is... The sequence is a zero-mean white noise sequence; Reciprocal weighted fusion according to the formula , In progress, among which and These represent the mean squared errors of the EWMA and ARIMA models on the historical validation set, respectively. and These are the fusion weights for the EWMA and ARIMA predictions, respectively. h For predicting the number of steps and , H For prediction in the time domain.

[0010] Furthermore, in S3, the time period within a day is divided into peak time period sets. Normal Segment Set Hegu period collection Time period t Electricity price satisfy: when hour, ; when hour, ; when hour, ; and ; User demand response coefficient is calculated according to the formula Calculation, where The benchmark electricity price, For the response sensitivity coefficient and ≥0; the effective charging demand power forecast after electricity price adjustment is calculated according to the formula Calculation, where For the site N In the future t + h Forecast of effective charging demand power for a given period.

[0011] Furthermore, in S4, the degree of urgency is determined according to the formula... The target power shortage is calculated according to the formula. The electricity price incentive term is calculated according to the formula. The comprehensive priority score is calculated according to the formula. Calculation, where For charging sessions i During the period t The urgency For charging sessions i During the period t The current state of charge, For vehicle battery capacity, For charging efficiency, For charging sessions i The rated maximum charging power of the connected charging station, For charging sessionsi The target departure time For charging sessions i During the period t The target power shortage, For time period t Electricity price incentives, For charging sessions i During the period t The overall priority score, , and These are the weighting coefficients for urgency, target electricity gap, and electricity price incentives, respectively. ≥0、 ≥0、 ≥0、 + + =1.

[0012] Furthermore, in S5, the allocated charging power for a single charging session satisfy The total power allocation for all charging sessions during the same period meets the requirements. The allocated charging power of a single charging session in adjacent time periods meets the requirements. ,in This refers to the maximum acceptable charging power reported by the battery management system for the vehicle under its current state of charge. For time period t A collection of charging sessions that are currently charging or queuing. For the site N The transformer can allocate the total charging capacity. For time period t Non-charging base load, This represents the maximum power variation allowed in a single charging session between adjacent time periods.

[0013] Furthermore, in S5, Each charging session is arranged according to Power is allocated sequentially after sorting in descending order, and the power is allocated to the currently assigned charging session. i According to the formula Determine its allocated charging power, where For the site N During the period t The remaining allocatable capacity, j "Assigned" indicates that the session has higher priority than the charging session in the current order. i And for charging sessions that have already been assigned, the assignment will be completed according to the formula. Update charging session i In the next period of time, the state of charge, The time step of the scheduling cycle.

[0014] Furthermore, in S6, the charging power scheduling object includes a unique identifier, a scope of application, a time base, a power unit, and a segmented power sequence. The unique identifier is used to identify the power scheduling object generated within the current scheduling cycle. The scope of application is the power limit for a single charging session or the overall power limit for a single charging pile. The time base is absolute time, the power unit is watts, and the segmented power sequence includes the relative number of seconds at the start of the segment and the power limit for that segment. The power limit for that segment is based on the allocated charging power obtained in S5. It is confirmed that the charging power scheduling object is sent by the scheduling service to the target charging pile through the power scheduling distribution interface of CSMS, and enters the next scheduling cycle after the charging pile returns the confirmation status.

[0015] Furthermore, in S7, according to a fixed scheduling cycle The system continuously receives measurement feedback and updates the SOC, EWMA and ARIMA model parameters online, and monitors prediction deviations based on this feedback. The target completion rate is calculated according to the formula when the user leaves the site. Calculation, where For charging sessions i Target completion rate For users at the target departure time The actual state of charge at that time, For charging sessions i The set target state of charge.

[0016] An orderly charging scheduling system for electric vehicles includes a central management system (CSMS), a scheduling service, and multiple charging piles; The multiple charging piles are connected to the central management system CSMS via the Open Charging Point Protocol (OCPP). The scheduling service is also connected to the central management system CSMS. The scheduling service includes a data acquisition and preprocessing module, a demand forecasting module, an electricity price response correction module, a priority scoring module, a sorting greedy power allocation module, a protocol encapsulation and distribution module, and a closed-loop feedback module. The data acquisition and preprocessing module is used to receive charging pile reporting messages through the central management system CSMS, collect charging pile static information, real-time observation information, charging session information, site transformer allocable capacity and basic load, and perform anomaly detection, anomaly replacement, time alignment and normalization processing on the real-time active power sequence. The demand forecasting module is used to perform short-term forecasting using exponentially weighted moving average (EWMA) and long-term forecasting using autoregressive integral moving average (ARIMA) on the actual charging demand power of the site, and to perform inverse weighted fusion based on the mean square error of the EWMA model and the ARIMA model on the historical validation set to obtain the fused forecasted charging demand power of the site for future periods. The electricity price response correction module is used to correct the price of the fused predicted charging demand power based on the peak electricity price, the normal electricity price, the off-peak electricity price and the user demand response coefficient, so as to obtain the effective charging demand power prediction value after the electricity price correction. The priority scoring module is used to calculate a comprehensive priority score for each charging session based on urgency, target power shortage, and electricity price incentives. The sorting greedy power allocation module is used to sort each charging session from high to low according to the comprehensive priority score, and under the constraints of site capacity, charging pile rated power, vehicle acceptable power and power change rate, it uses a sorting greedy allocation method to determine the allocated charging power of each charging session in the current scheduling period. The protocol encapsulation and delivery module is used to encapsulate the allocated charging power into a charging power scheduling object defined by OCPP according to the scheduling cycle, and push it to the corresponding charging pile through the power scheduling delivery interface of the central management system CSMS. The closed-loop feedback module is used to update the state of charge of the charging session based on the metering value returned by the charging pile, and to update the charging scheduling process on a rolling basis according to a fixed scheduling cycle.

[0017] The beneficial effects of this invention are as follows: (1) The lightweight charging demand prediction method that combines EWMA and ARIMA improves the multi-step prediction accuracy while maintaining the interpretability of the model. Compared with the prediction method that only takes the historical average or the most recent value, it can better reflect the changes in charging demand. Compared with deep learning methods such as LSTM and GRU, it reduces the sample size requirement and training cost, making it easier to deploy in small and medium-sized charging operation scenarios.

[0018] (2) By using the user demand response coefficient, the time-of-use electricity price is directly introduced into the predicted demand correction and priority scoring, so that the electricity price signal is no longer only used for post-settlement, but plays a guiding role in the process of charging resource allocation.

[0019] (3) The comprehensive priority scoring method is weighted by three factors: urgency, target power gap and electricity price incentive. It takes into account the user's departure time, remaining charging demand and time-of-use electricity price. Compared with FCFS or a single rule that only sorts by SOC, it can reduce the probability that urgent charging users will be squeezed out by low-priority users.

[0020] (4) The sorting greedy power allocation method is adopted to complete the power allocation under the constraints of site capacity, charging pile rated power, vehicle acceptable power and power change rate. It does not require commercial solvers, has low computational complexity, and is suitable for real-time rolling scheduling.

[0021] (5) Encapsulate the power allocation results into a charging power scheduling object defined by OCPP and send it to the charging pile through CSMS. This can narrow the gap between algorithm simulation and engineering implementation and improve the engineering interoperability of the scheduling method.

[0022] (6) In this invention, the smoothing coefficient, ARIMA order, response sensitivity coefficient, weighting coefficient, power change rate constraint and scheduling cycle all have clear physical meanings and can be adjusted by the operator according to the site size, user behavior and electricity price rules, which facilitates engineering maintenance and fault location.

[0023] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a system architecture diagram upon which the method of this invention is based; Figure 2 This is a block diagram of the overall architecture of the method of the present invention; Figure 3 This is a flowchart illustrating the overall execution process of the method of the present invention. Detailed Implementation

[0025] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0026] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0027] Please see Figures 1-3 ,in: Figure 1 The system architecture diagram upon which the method of this invention is based consists of three parts: the charging pile (CP) side, the CSMS (which can be an open-source OCPP server such as SteVe during implementation), and an independent scheduling service. The CP side maintains a communication connection with the CSMS via the OCPP 1.6J protocol (SOAP / JSON-WS channel) to report access notifications, status notifications, metering values, start and end of transactions, and other messages. The CSMS internally includes modules such as the OCPP protocol stack, power scheduling interface, and database, used for protocol parsing, data persistence, and command relay. The independent scheduling service connects to the CSMS via a RESTful API and is responsible for executing core scheduling functions such as demand forecasting, electricity price adjustment, priority scoring, greedy power allocation, protocol encapsulation, and closed-loop feedback.

[0028] Figure 2 The overall architecture of the method of this invention comprises seven steps: S1 data acquisition and preprocessing, S2 charging demand prediction, S3 time-of-use pricing and demand response, S4 priority scoring, S5 sorted greedy power allocation, S6 instruction encapsulation and issuance, and S7 closed-loop feedback and rolling update. These steps are connected sequentially and return to S1 through the feedback link of S7 to form a closed loop, collectively achieving orderly charging scheduling of electric vehicles based on time-of-use pricing response and priority ranking.

[0029] Figure 3 The flowchart illustrates the overall execution process of the method of this invention, showing in detail the complete execution process from the start of the scheduling cycle, through S1 data acquisition and preprocessing, S2 charging demand prediction, S3 time-of-use pricing and demand response, S4 priority scoring, S5 sorted greedy power allocation, S6 instruction encapsulation and issuance, fault or anomaly judgment, and S7 closed-loop feedback and rolling update. This demonstrates the closed-loop rolling scheduling mechanism of the method of this invention in a dynamic environment.

[0030] The present invention will be further explained below with reference to various embodiments.

[0031] S1. Charging Data Acquisition and Preprocessing Based on OCPP Message Stream: Relying on the open-source OCPP 1.6J server (SteVe or similar can be used in implementation) as the central management system (CSMS), it subscribes to access notifications, status notifications, metering values, start and end transaction messages reported by charging piles through SOAP / JSON-WS channels, persisting transactions, pile status, and metering records to the database. The scheduling service connects to the database and REST interface as an independent service outside the CSMS, collecting the following four types of data: ① Static information of charging piles: including pile number Connector number Rated maximum charging power and its station number ; ② Real-time monitoring of charging piles: charging pile working status (available, ready, charging in progress, vehicle paused, stopped, faulty, etc.) from status reporting messages, and real-time active power from metering reporting messages. and cumulative active power ; ③ Charging session information: Session number from the start and end of the transaction messages. Associated connector number, user-reserved target departure time Initial state of charge Target state of charge Vehicle battery capacity and charging efficiency ; ④ Site power distribution information: the transformer's allocable capacity to the site and current base load (Including non-charging loads such as lighting and air conditioning).

[0032] During the data preprocessing stage, a sliding window anomaly detection method is used on the acquired real-time power sequences. A sampling point is identified as an anomaly if it meets the following conditions:

[0033] in, For charging piles connector During the period Real-time active power and They are respectively of length The mean and standard deviation of samples within a sliding window are calculated. For identified outliers, linear interpolation is used to replace adjacent normal sampling points. Subsequently, the multi-connector metering values ​​are time-aligned and normalized, and aggregated by site to obtain the actual charging power demand of each site, forming a unified basic dataset for subsequent modeling.

[0034] Furthermore, to enhance the adaptability of the scheduling service to the actual operating environment, this implementation caches the collected data according to the scheduling cycle and automatically updates the basic dataset when each scheduling cycle is triggered. This not only meets the needs of demand forecasting and electricity price adjustment in the future time domain, but also provides a unified input for subsequent priority scoring and power allocation, thus forming a continuous data support link of "collection-processing-update".

[0035] S2. Lightweight Charging Demand Forecasting Based on EWMA and ARIMA Fusion: Addressing the limited sample size and high interpretability requirements in charging operation scenarios, this implementation employs a lightweight forecasting method combining Exponentially Weighted Moving Average (EWMA) and Autoregressive Moving Average (ARIMA) to predict future charging demand at charging stations. The charging power demand of each step is predicted.

[0036] ①EWMA Short-Term Forecast: Used to capture trend changes over recent periods, its update relationship is as follows:

[0037] in, For the site During the period EWMA charging demand power forecast, For the site During the period The actual charging power required (the sum of the power of all charging sessions in progress during that period). For smoothing coefficients, The larger the value, the higher the weight of recent observations, and the faster the response to trend changes.

[0038] ②ARIMA Long-term forecasting: used to capture intraday patterns and cyclicalities, its basic form is:

[0039] in, For delay operator ( ), For autoregressive polynomials ( (where the autoregressive order is...) For the moving average polynomial ( (the order of the moving average) It is the difference order (used to stabilize non-stationary sequences). It is a zero-mean white noise sequence.

[0040] ③ Inverse MSE Weighted Fusion: The inverse weighted fusion is performed based on the mean squared error (MSE) of the two models on the historical validation set. The model with higher prediction accuracy (smaller MSE) has a larger inverse and is assigned a higher weight.

[0041]

[0042] in, and These represent the mean squared errors of the EWMA and ARIMA models on the historical validation set, respectively. and These are the fusion weights for the EWMA and ARIMA predictions, respectively. For the site In the future Time-segmented prediction of charging demand power. To predict the number of steps, For the prediction time domain, the resulting multi-step prediction sequence is input S3 for electricity price correction, and input S4 for priority scoring and power allocation in S5.

[0043] Compared to simple strategies that only use averages or recent values ​​as demand estimates, this implementation captures short-term trend changes through EWMA and intraday patterns and periodicities through ARIMA. It also achieves adaptive fusion of the two prediction results using a reciprocal MSE weighting strategy, providing more accurate multi-step demand forecasts while maintaining lightweight computation. Furthermore, compared to deep learning methods such as LSTM and GRU, this implementation does not require large-scale training samples and GPU resources, and its parameters have clear physical meanings, facilitating rapid deployment and operation in small to medium-sized charging operation scenarios.

[0044] S3. Time-of-use pricing and user demand response coefficient modeling: This implementation explicitly introduces time-of-use pricing into the sorting and allocation process, enabling price signals to directly influence charging scheduling rather than just participating in post-event settlement. The day is divided into three time periods: peak, flat, and valley.

[0045] in, For time period Electricity price (unit: CNY / kWh) , , Electricity prices are set for peak, off-peak, and valley periods, respectively, to meet [the needs of the public]. ; , , These are time sets representing peak, flat, and trough periods, respectively.

[0046] To characterize the degree of user responsiveness to electricity prices, a user demand response coefficient is introduced:

[0047] in, The user demand response coefficient (dimensionless) represents the user's response to electricity prices. The adjustment range of charging willingness relative to the benchmark electricity price; The benchmark electricity price (usually the normal electricity price) ); The response sensitivity coefficient is obtained by fitting historical user response data. When... When (peak electricity price is higher than the benchmark), This indicates that rising electricity prices are suppressing charging demand; when When the off-peak electricity price is lower than the benchmark, This indicates that lower electricity prices are incentivizing charging demand.

[0048] Price adjustments are made to the fusion forecast demand for S2 based on this coefficient:

[0049] in, For sites after electricity price adjustment In the future The effective charging demand power forecast for the time period. The revised demand is used to assess whether the future load of the site is close to the transformer capacity, and also to participate in the construction of the "electricity price incentive" item in the S4 priority scoring, so that the electricity price signal can directly affect the ranking and not just participate in the ex-post settlement.

[0050] This implementation method, by introducing a user demand response coefficient, elevates the time-of-use electricity price signal from a traditional "post-settlement tool" to a "pre-schedule guidance tool," effectively solving the problem that the electricity price signal does not explicitly play a guiding role in the allocation and sorting process in existing schemes.

[0051] S4. Three-factor weighted charging session priority score: This implementation abandons a single sorting rule (such as first-come, first-served or sorted only by SOC) and constructs a priority score weighted by three factors: "urgency, target power shortage, and electricity price incentive" as the basis for sorting each charging session in the current time period.

[0052] ① Urgency: Describes whether there is enough battery remaining to charge the target value, defined as:

[0053] in, For charging sessions During the period Urgency (dimensionless, range of values) ), For the target state of charge, For time period The current state of charge, Vehicle battery capacity (unit: kWh). The charging efficiency is dimensionless and is typically taken as 0.85 to 0.95. The rated maximum charging power (in kW) of the charging station connected to this session. Set the target departure time for the user. When the value approaches 1, it means that even if you keep charging at the rated power, you can only barely reach the target capacity before leaving the station, so it should be given high priority. When the value approaches 0, it indicates that there is still plenty of time remaining, and charging resources can be appropriately allocated.

[0054] ② Target power gap: Describes "how much power is still needed", in normalized form:

[0055] in, For charging sessions During the period The target power gap (dimensionless) is achieved through... The operation guarantees non-negativity: the gap is zero when the current SOC has exceeded the target value.

[0056] ③ Electricity price incentives: Appropriately lowering the priority of electricity during peak hours to guide load smoothing, defined as:

[0057] in, For time period Electricity price incentives (dimensionless). The benchmark electricity price, For time period The actual electricity price. Peak hour electricity prices are high ( )hour The priority of charging sessions during off-peak hours can be appropriately reduced to guide load smoothing; off-peak electricity prices are low ( )hour This can increase the priority of sorting to encourage off-peak electricity use. The impact of this item is diminished when user urgency is high. To take the lead and avoid inadvertently harming users who urgently need to top up.

[0058] ④ The overall priority score is defined as the weighted sum of the above three factors:

[0059] in, For charging sessions During the period Overall priority score (dimensionless). , , These are the weighting coefficients for urgency, target power shortage, and electricity price incentives, respectively, satisfying... The weights can be flexibly configured by the operator according to the scenario: when user experience is prioritized, the larger weight is selected. When prioritizing electricity price response, the larger price will be selected. When there is a power shortage, the larger one should be selected first. .

[0060] This implementation replaces the FCFS or a single rule based solely on SOC with a three-factor weighted scoring system, comprehensively considering user departure time constraints, remaining charging demand, and the incentive effect of electricity price signals. Compared to ranking methods that only consider a single dimension, this method achieves a balance between users' urgent charging needs, fairness, and economy, significantly reducing the probability of urgent charging users being squeezed out by lower-priority users.

[0061] S5. Greedy Power Allocation under Station Capacity Constraints: After obtaining the priority scores of all charging or queuing sessions, this implementation uses a greedy algorithm to allocate the power of each charging pile under station capacity constraints, avoiding the complexity of solving traditional model predictive control (MPC), mixed integer linear programming (MILP), or second-order cone relaxation (SOCP).

[0062] ① Power Constraint: The power of the connector corresponding to each charging session is constrained by the lower of the charging pile's rated power and the vehicle's dynamically acceptable power.

[0063] in, For charging sessions During the period Allocated charging power (unit: kW) The rated maximum charging power (in kW) of the charging station connected to this session. The maximum acceptable charging power (in kW) reported by the battery management system (BMS) at the current state of charge (SOC) of the vehicle.

[0064] ② Site capacity constraint: The sum of the power of all charging sessions during the same period shall not exceed the site's allocable capacity after deducting the base load.

[0065] in, For time period A collection of charging sessions that are currently charging or queuing. For transformer to station Total distributable charging capacity (unit: kW). For time period Non-charging base loads within the unit (including lighting, air conditioning, etc., unit: kW).

[0066] ③ Power change rate constraint: To prevent sudden changes in commands from impacting charging piles and the power grid, a power change rate constraint between adjacent time periods is introduced:

[0067] in, The maximum power variation allowed in a single charging session between adjacent time periods (unit: kW).

[0068] ④ Greedy allocation process: All charging sessions are sorted according to... Sort in descending order and allocate power to each session sequentially, taking the minimum value from four constraints (pile rated power, vehicle acceptable power, power from the previous period plus maximum change, and remaining allocable capacity at the site) each time:

[0069] in, For the site During the period The remaining allocatable capacity (unit: kW) is the total allocatable capacity of the site after deducting the base load and the power already allocated to higher priority sessions. Updated after allocation. And process the next session; when If this occurs, the remaining sessions will not be charged during this time period and will wait for the next scheduling cycle.

[0070] ⑤SOC Update: The state of charge is recursively deduced based on fundamental physical relationships:

[0071] in, For charging sessions In the next period The state of charge (dimensionless). The time step of the scheduling cycle (unit: h). The effective amount of electricity actually charged into the battery during this period (unit: kWh) divided by the battery capacity. This is the SOC increment.

[0072] The time complexity of the above sorting greedy algorithm is O(n). Quantity, of which For time period The number of charging sessions within the system. The solution time can be stabilized in the millisecond range, requiring no commercial solver, and the deployment threshold is significantly lower than traditional optimization methods such as MPC, MILP, and SOCP. In small and medium-sized operational scenarios, it can be deployed directly as a containerized scheduling service at low cost.

[0073] S6. Protocol Encapsulation and Issuance of Charging Power Scheduling Commands: In this implementation, the power allocation result obtained in S5 is encapsulated into a "Charging Power Scheduling Object" (i.e., a Charging Profile defined in OCPP) according to the OCPP 1.6J specification, and sent to the corresponding charging pile and connector through the CSMS power scheduling issuance interface. The core fields of the charging power scheduling object are mapped logically as follows: ① Unique identifier for scheduling object: Each power scheduling object generated in the current scheduling cycle for each charging session has a unique ID, which is used for reconciliation and retransmission determination between the charging pile and CSMS; ② Scope of application: This implementation adopts one of two operating modes: "power limit for a single session" or "overall power limit for a single charging pile" (corresponding to TxProfile and ChargePointMaxProfile respectively in OCPP 1.6J), and the scheduling service selects the mode based on whether there are concurrent sessions for the current charging pile; ③ Time base: Absolute time is used as the reference for the scheduling start time (i.e., Absolute mode in OCPP). ④ Power unit: The charging rate is measured in watts (W); ⑤ Segmented power sequence: according to scheduling cycle (e.g., 5 min or 15 min) is discretized into several segments, each segment is given in pairs with "segment start relative seconds" and "segment power limit", where "power limit" is obtained from S5. .

[0074] After encapsulation, the scheduling service pushes the power scheduling object to the target charging pile through the standard RESTful interface exposed by CSMS; the charging pile performs power limiting and then sends back a confirmation status. For charging sessions still in the queue, the scheduling service triggers them to start charging through the remote start interface exposed by CSMS; charging sessions not allocated in this cycle remain in a waiting state until the next scheduling cycle.

[0075] This implementation directly maps the "optimal power allocation result" at the simulation level to an OCPP standard protocol object and distributes it through the open-source CSMS, bridging the gap between simulation and engineering implementation. Compared with existing solutions that only output simulation curves, this method does not require an additional proprietary protocol adaptation layer and can be directly implemented using open-source OCPP servers such as SteVe, exhibiting good interoperability and engineering feasibility.

[0076] S7. Closed-loop feedback and rolling updates: The scheduling service operates on a fixed scheduling cycle. (e.g., 5 min or 15 min) Execute S1 to S6 in a loop to form a complete closed-loop scheduling mechanism. At the end of each scheduling cycle, perform the following operations: ① Metering feedback: CSMS receives metering messages reported by charging piles and obtains the actual active power for the current cycle. With cumulative active power ; ②SOC Update: The scheduling service calculates the current actual SOC of each charging session according to the SOC update formula in S5 based on the actual charging amount and the SOC of the previous period, and updates the remaining power demand. ③ Online update of the prediction model: Based on the latest collected actual charging demand power data, the EWMA model is retrained using a rolling window approach (updating the smoothing coefficients). ) and ARIMA model (update order) To adapt to changes in daily charging demand patterns; ④ Deviation Monitoring and Alarm: Compare the deviation between the actual power and the set power. If the deviation is continuous... The deviation in a scheduling cycle exceeds a preset threshold. If this occurs, the data recalibration process will be triggered and an alert will be sent to the operations personnel; ⑤ Enter the next scheduling cycle: After completing the above update, enter the next scheduling cycle and repeat S1 to S6 to form a complete closed loop of "collection → prediction → sorting → allocation → distribution → feedback".

[0077] The operational cost target for the scheduling service can be expressed as:

[0078] in, Operating costs (unit: CNY) For time period Time-of-use electricity price (unit: CNY / kWh) For charging sessions During the period Distributed charging power (unit: kW). The time step of the scheduling cycle (unit: h).

[0079] The target completion rate when users leave the site can be expressed as:

[0080] in, For charging sessions Target completion rate (dimensionless, range of values) ), For users at the target departure time The actual state of charge at that time, The target state of charge set for this session. This indicates that the target SOC has been fully achieved when the user leaves the site. This indicates that the target was not achieved. These indicators are continuously updated throughout the scheduling cycle and serve as a key basis for the operator's monthly performance review.

[0081] This implementation method uses a closed-loop feedback and rolling update mechanism to enable the scheduling service to continuously track changes in the operating status of charging stations and user demand. Based on online parameter updates and deviation monitoring, it can achieve adaptive adjustments, ensuring the stability and adaptability of the scheduling strategy in dynamic environments such as grid load fluctuations, changes in user orders, and changes in charging pile status.

[0082] Example 1: Single-site community slow charging scenario The following is a typical implementation case of a community slow charging station, with specific parameters.

[0083] Hardware deployment and platform configuration: One charging station is configured in a residential area. ), including 10 7 kW AC connectors ( Transformer can allocate capacity to charging stations kW (the remaining capacity is used for basic loads such as public lighting). The charging piles are permanently connected to the CSMS server via OCPP 1.6J / JSON-WS (this embodiment uses SteVe 3.x as the open-source OCPP server implementation), and the scheduling service is deployed as a container on the same server. Scheduling cycle. min, prediction time domain (i.e., 2 hours).

[0084] Key parameter settings are as follows: EWMA smoothing coefficient ARIMA order Response sensitivity coefficient benchmark electricity price CNY / kWh, peak / flat / valley electricity prices are respectively , , CNY / kWh, power change rate constraint kW, priority scoring weight (Urgency) (Target power shortage) (Electricity price incentives).

[0085] The operation process is as follows: 1) The CSMS receives and stores access notifications, status notifications, metering values, and other messages sent by the charging piles into the database; 2) The dispatch service samples the metering values ​​every 60 seconds and aggregates them by site to obtain the actual charging power demand of the site. ; 3) Execute S2 to output the fused predicted charging power demand for the next 8 steps (2 hours). ; 4) Execute S3, using the user demand response coefficient. The predicted demand is corrected to obtain the effective charging demand power prediction value. ; 5) Execute S4 to calculate the comprehensive priority score for each charging session within the current time period. ; 6) Execute S5, sort by priority in descending order, and then allocate charging power to each charging session in a greedy manner. ; 7) Execute S6 to encapsulate the allocation result into a charging power scheduling object (Charging Profile) and send it to the corresponding charging pile via CSMS; 8) Execute S7 to enter the next scheduling cycle, receive metering feedback and update model parameters online.

[0086] In terms of evaluation metrics, three core indicators can be used: Total Operating Cost (Cost), Sat Rate (Customer Satisfaction Rate), and Peak-to-Valuation Ratio (PAR).

[0087]

[0088]

[0089] in, User satisfaction rate (dimensionless, range of values) ), To assess the total number of charging sessions within a given time period, This is an indicator function (it takes the value 1 if the condition is true, and 0 otherwise); Peak-to-valley ratio (dimensionless) The numerator is the maximum total charging power of the stations across all time periods, and the denominator is the average total charging power of the stations across all time periods. The total number of time periods is used for evaluation. The closer PAR is to 1, the smoother the load curve, which is beneficial to power grid stability.

[0090] Three schemes were compared: A) using a fixed rate and first-come, first-served (FCFS) allocation; B) using a three-tiered time-of-use pricing system and allocation based solely on State of Charge (SOC); and C) the method of this invention. The expected results are as follows: Compared to scheme A, scheme C will reduce total operating costs by 10%–20%, decrease peak-to-valley parity (PAR) by 25%–35%, and increase user satisfaction to over 95%; compared to scheme B, scheme C will significantly improve peak-hour queuing congestion (urgency factors prioritize serving users with urgent charging needs) and further enhance user satisfaction.

[0091] Example 2: Public Fast Charging Station Scenario The following is a typical implementation case of a public fast charging station, with specific parameters.

[0092] Hardware deployment and platform configuration: One fast charging station is configured in a highway service area ( ), including 4 60 kW DC connectors ( Transformer allocable capacity kW. The charging pile is permanently connected to the CSMS server via OCPP 1.6J / JSON-WS, and the scheduling service is deployed in a containerized manner. Because fast-charging users are more sensitive to waiting time, the scheduling cycle has been shortened to [kW]. min, prediction time domain (i.e., 30 minutes).

[0093] Key parameters adjusted for fast charging scenarios: EWMA smoothing coefficient (Focusing more on recent changes to quickly respond to short-term fluctuations in fast charging traffic); Power change rate constraint kW (fast charging station power adjustment step size is larger); priority scoring weight , , (User experience is prioritized, and the urgency factor is much higher than that of electricity price incentives).

[0094] During holidays or peak traffic periods at service areas, a large number of vehicles may converge on charging stations within a short period. The dispatch service first obtains the status and metering values ​​of each fast charging pile through CSMS, and then performs EWMA and ARIMA fusion prediction on the actual charging power demand. Due to the rapid fluctuations in traffic at fast charging stations, a high smoothing coefficient is required. This allows the forecast results to quickly respond to recent observed changes. Subsequently, the scheduling service combines time-of-use pricing and user demand response coefficients to revise the forecasted demand and calculates a comprehensive priority score for each charging session.

[0095] For users with insufficient remaining range, short departure time, and significant battery deficit, their urgency and battery deficit are both high. Therefore, during the greedy power allocation process, they will be prioritized for higher charging power, ensuring that users who urgently need to recharge are not overtaken by those who only need short-distance charging. For users who only need short-term charging and are not in a rush to leave the station, the power may be appropriately reduced during peak hours (such as holiday peaks).

[0096] During off-peak electricity pricing periods, electricity price incentives... The increase (due to off-peak electricity prices being lower than the benchmark price) prioritizes some time-flexible charging sessions, thereby guiding users to replenish energy during off-peak hours, improving off-peak electricity utilization, and reducing overall operating costs.

[0097] Using the method in this embodiment, users who "urgently need to replenish energy to get somewhere" can be prioritized over users who "replenish energy for short distances" during peak traffic periods on holidays, resulting in a significantly higher average completion rate compared to Solution B. Guiding users to replenish energy during off-peak hours when electricity prices are in effect can reduce overall operating costs.

[0098] Example 3: Robust handling of concurrent multi-pile and single-pile faults This embodiment provides a robust handling method for multiple piles operating concurrently and single pile failures, used to verify the adaptability of the method of the present invention under abnormal operating conditions.

[0099] When a charging station reports a faulty status via a status notification message, the CSMS transmits this status to the scheduling service. During S1 data preprocessing, the scheduling service automatically removes the faulty connector from the set of allocable connectors for the current and next scheduling cycles. If the faulty connector was originally carrying a charging session, that charging session re-enters the scheduling set and re-participates in the S4 comprehensive priority scoring and S5 sorting greedy power allocation when an available connector becomes available or remote activation conditions are met.

[0100] When multiple charging piles are operating concurrently and the total charging power of the sites is close to the upper limit of the transformer's allocable capacity, the sorting greedy power allocation process naturally guarantees a high comprehensive priority score. Charging sessions with priority scores are given priority, and then the remaining capacity is allocated to the remaining charging sessions in descending order of priority score. If the site has remaining allocable capacity... If the power is insufficient to meet the full power requirements of a charging session, the session will be automatically reduced to the remaining capacity; if If the remaining charging sessions with lower priority are not charged during this period, the transformer will not be overloaded.

[0101] In the aforementioned fault scenarios and concurrent congestion scenarios, this invention still follows a fixed scheduling cycle. The system continuously executes S1 data acquisition, S2 demand forecasting, S3 electricity price adjustment, S4 priority scoring, S5 sorting and greedy allocation, S6 protocol issuance, and S7 feedback update, ensuring the charging scheduling process is robust and closed-loop. After a fault is recovered, the connector that has resumed normal operation is automatically re-added to the allocatable set in the next scheduling cycle.

[0102] This embodiment verifies that the method of the present invention can automatically complete fault isolation, session rescheduling and transformer overload protection without manual intervention under abnormal operating conditions, and has good engineering robustness and reliability.

[0103] In summary, this invention addresses the operational characteristics of electric vehicle charging scenarios, including simultaneous access by multiple vehicles, limited charging station locations, limited transformer capacity, time-of-use electricity price fluctuations, and varying user departure times. It achieves lightweight charging demand prediction through a weighted fusion of EWMA and the reciprocal MSE of ARIMA. The time-of-use electricity price is explicitly incorporated into the sorting and allocation process using a user demand response coefficient. A charging session priority score is constructed using a weighted approach of "urgency + target power gap + electricity price incentive." Power allocation is completed using a greedy sorting method under constraints of station transformer capacity, charging pile rated power, vehicle acceptable power, and power change rate. Finally, the allocation results are encapsulated into a charging power scheduling object according to the OCPP 1.6J specification and sent to the charging pile. The above method has the combined advantages of fast solution speed (millisecond level), low deployment threshold (no need for commercial solvers), strong interpretability (all model parameters have clear physical meaning), and good engineering interoperability (directly mapped to OCPP standard objects and distributed through open source CSMS). It can realize orderly charging scheduling of electric vehicles based on time-of-use pricing response and priority ranking in small and medium-sized charging operation scenarios at a low cost, and has good engineering application prospects.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for orderly charging scheduling of electric vehicles based on time-of-use pricing response and priority ranking, characterized in that, Includes the following steps: S1. The central management system CSMS receives charging pile reports based on the Open Charging Point Protocol (OCPP), collects static information, real-time observation information, charging session information, allocable capacity of the site transformer, and basic load of the charging pile, and performs anomaly detection, anomaly replacement, time alignment and normalization on the real-time active power sequence to obtain the actual charging power demand of the site. S2. The actual charging demand power of the station is predicted by exponential weighted moving average (EWMA) for short-term forecasting and autoregressive integral moving average (ARIMA) for long-term forecasting. The EWMA model and the ARIMA model are then fused by reciprocal weighting based on the mean square error of the EWMA model and the ARIMA model on the historical validation set to obtain the fused predicted charging demand power of the station for future periods. S3. Based on the peak-hour electricity price, the normal-hour electricity price, the off-hour electricity price, and the user demand response coefficient, the fused predicted charging demand power is adjusted for price to obtain the effective charging demand power prediction value after price adjustment. S4. For each charging session, calculate a comprehensive priority score based on urgency, target power gap, and electricity price incentives. S5. Sort each charging session from high to low according to the comprehensive priority score, and determine the allocated charging power of each charging session in the current scheduling period by using a sorting greedy allocation method under the constraints of site capacity, charging pile rated power, vehicle acceptable power and power change rate. S6. The allocated charging power is encapsulated into a charging power scheduling object defined by OCPP according to the scheduling cycle, and pushed to the corresponding charging pile through the power scheduling distribution interface of CSMS. S7. Update the state of charge of the charging session based on the metering value returned by the charging pile, and execute S1 to S6 in a rolling manner according to a fixed scheduling cycle to form a closed-loop scheduling process of collection, prediction, sorting, allocation, distribution and feedback.

2. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S1, the charging pile reporting messages include access notification, status notification, metering value, start transaction, and end transaction; the charging pile static information includes the pile number. n connector k and rated maximum power The real-time observation information includes the pile's working status and real-time active power. and electricity The charging session information includes the target departure time. Initial state of charge, target state of charge Battery capacity and charging efficiency The anomaly detection uses a length of Sliding window and according to formula Identify outliers, among which and The mean and standard deviation of the real-time active power samples within the sliding window are respectively used to linearly interpolate and replace the identified outliers with adjacent normal sampling points.

3. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S2, the EWMA short-term forecast follows the formula In progress, among which For the site N During the period t EWMA charging demand power forecast, For the site N During the period t The actual charging power required. The smoothing coefficient is 0 < ≤1; ARIMA long-term forecast is based on the formula In progress, among which B For delay operators, It is an autoregressive polynomial. It is a moving average polynomial. p Let the order be the autoregressive order. d It is the difference order. q The moving average order is... The sequence is a zero-mean white noise sequence; Reciprocal weighted fusion according to the formula , In progress, among which and These represent the mean squared errors of the EWMA and ARIMA models on the historical validation set, respectively. and These are the fusion weights for the EWMA and ARIMA predictions, respectively. h For predicting the number of steps and , H For prediction in the time domain.

4. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S3, the time period of a day is divided into peak time period sets. Normal Segment Set Hegu period collection Time period t Electricity price satisfy: when hour, ; when hour, ; when hour, ; and ; User demand response coefficient is calculated according to the formula Calculation, where The benchmark electricity price, For the response sensitivity coefficient and ≥0; the effective charging demand power forecast after electricity price adjustment is calculated according to the formula Calculation, where For the site N In the future t + h Forecast of effective charging demand power for a given time period.

5. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S4, the degree of urgency is calculated according to the formula. The target power shortage is calculated according to the formula. The electricity price incentive term is calculated according to the formula. The comprehensive priority score is calculated according to the formula. Calculation, where For charging sessions i During the period t The urgency For charging sessions i During the period t The current state of charge, For vehicle battery capacity, For charging efficiency, For charging sessions i The rated maximum charging power of the connected charging station, For charging sessions i The target departure time For charging sessions i During the period t The target power shortage, For time period t Electricity price incentives, For charging sessions i During the period t The overall priority score, , and These are the weighting coefficients for urgency, target electricity gap, and electricity price incentives, respectively. ≥0、 ≥0、 ≥0、 + + =1.

6. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S5, the allocated charging power for a single charging session satisfy The total power allocation for all charging sessions during the same period meets the requirements. The allocated charging power of a single charging session in adjacent time periods meets the requirements. ,in This refers to the maximum acceptable charging power reported by the battery management system for the vehicle under its current state of charge. For time period t A collection of charging sessions that are currently charging or queuing. For the site N The transformer can allocate the total charging capacity. For time period t Non-charging base load, This represents the maximum power variation allowed in a single charging session between adjacent time periods.

7. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 6, characterized in that, In S5, Each charging session is arranged according to Power is allocated sequentially after sorting in descending order, and the power is allocated to the currently assigned charging session. i According to the formula Determine its allocated charging power, where For the site N During the period t The remaining allocatable capacity, j "Assigned" indicates that the session has higher priority than the charging session in the current order. i And for charging sessions that have already been assigned, the assignment will be completed according to the formula. Update charging session i In the next period of time, the state of charge, The time step of the scheduling cycle.

8. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S6, the charging power scheduling object includes a unique identifier, a scope of application, a time base, a power unit, and a segmented power sequence. The unique identifier is used to identify the power scheduling object generated within the current scheduling cycle. The scope of application is the power limit for a single charging session or the overall power limit for a single charging pile. The time base is absolute time, and the power unit is watts. The segmented power sequence includes the relative number of seconds at the start of the segment and the power limit for that segment. The power limit for that segment is based on the allocated charging power obtained in S5. It is confirmed that the charging power scheduling object is sent by the scheduling service to the target charging pile through the power scheduling distribution interface of CSMS, and enters the next scheduling cycle after the charging pile returns the confirmation status.

9. The electric vehicle orderly charging scheduling method based on time-of-use pricing response and priority ranking according to claim 1, characterized in that, In S7, according to a fixed scheduling cycle The system continuously receives measurement feedback and updates the SOC, EWMA and ARIMA model parameters online, and monitors prediction deviations based on this feedback. The target completion rate is calculated according to the formula when the user leaves the site. Calculation, where For charging sessions i Target completion rate For users at the target departure time The actual state of charge at that time, For charging sessions i The set target state of charge.

10. An orderly charging scheduling system for electric vehicles, characterized in that, This includes the central management system CSMS, dispatch services, and multiple charging stations; The multiple charging piles are connected to the central management system CSMS via the Open Charging Point Protocol (OCPP). The scheduling service is also connected to the central management system CSMS. The scheduling service includes a data acquisition and preprocessing module, a demand forecasting module, an electricity price response correction module, a priority scoring module, a sorting greedy power allocation module, a protocol encapsulation and distribution module, and a closed-loop feedback module. The data acquisition and preprocessing module is used to receive charging pile reporting messages through the central management system CSMS, collect charging pile static information, real-time observation information, charging session information, site transformer allocable capacity and basic load, and perform anomaly detection, anomaly replacement, time alignment and normalization processing on the real-time active power sequence. The demand forecasting module is used to perform short-term forecasting using exponentially weighted moving average (EWMA) and long-term forecasting using autoregressive integral moving average (ARIMA) on the actual charging demand power of the site, and to perform inverse weighted fusion based on the mean square error of the EWMA model and the ARIMA model on the historical validation set to obtain the fused forecasted charging demand power of the site for future periods. The electricity price response correction module is used to correct the price of the fused predicted charging demand power based on the peak electricity price, the normal electricity price, the off-peak electricity price and the user demand response coefficient, so as to obtain the effective charging demand power prediction value after the electricity price correction. The priority scoring module is used to calculate a comprehensive priority score for each charging session based on urgency, target power shortage, and electricity price incentives. The sorting greedy power allocation module is used to sort each charging session from high to low according to the comprehensive priority score, and under the constraints of site capacity, charging pile rated power, vehicle acceptable power and power change rate, it uses a sorting greedy allocation method to determine the allocated charging power of each charging session in the current scheduling period. The protocol encapsulation and delivery module is used to encapsulate the allocated charging power into a charging power scheduling object defined by OCPP according to the scheduling cycle, and push it to the corresponding charging pile through the power scheduling delivery interface of the central management system CSMS. The closed-loop feedback module is used to update the state of charge of the charging session based on the metering value returned by the charging pile, and to update the charging scheduling process on a rolling basis according to a fixed scheduling cycle.

Images