Electric vehicle orderly charging method, system and equipment integrating load prediction and dynamic price excitation and storage medium

Abstract

The invention discloses an electric vehicle orderly charging method, system and device integrating load prediction and dynamic price excitation and a storage medium, and the method comprises the following steps: collecting power grid historical data and electric vehicle charging behavior data through a load prediction module; a long short-term memory network is adopted to carry out day-ahead load prediction and intra-day load prediction so as to evaluate the load state of the power distribution network in real time; the scheduling optimization module carries out multi-target charging scheduling solution by utilizing a whale optimization algorithm and combining a Pareto frontier based on a load prediction result, the electricity price scheduling module implements a dynamic electricity price mechanism, and the electricity price is composed of a basic electricity price and a dynamic electric vehicle bearing capacity scheduling service fee; the scheduling service charge fluctuates in real time according to the predicted power grid load; the user incentive module gives incentive points to users actively participating in peak shifting charging, and the points can be used for directly deducting scheduling services generated during subsequent charging; safe and stable operation of a power system is ensured.

Description

Technical Field

[0001] This invention relates to the field of orderly charging technology for electric vehicles, and specifically to an orderly charging method, system, device, and storage medium for electric vehicles that integrates load forecasting and dynamic price incentives. Background Technology

[0002] In recent years, electric vehicles (EVs) have become an important alternative to traditional gasoline-powered vehicles, and their market penetration rate is growing at an unprecedented rate. However, the large-scale grid-connected charging of EVs, especially the disorderly charging behavior driven by user habits, has brought severe challenges to the existing power distribution network. Most users tend to charge their vehicles in the evening after get off work, which overlaps with the peak hours for residential electricity consumption. This can easily lead to a surge in regional load, increase the peak-valley difference in grid load, and in severe cases, may cause transformer overload, line voltage exceeding limits, and other problems, affecting the safe and stable operation of the power grid. Summary of the Invention

[0003] Purpose of the invention: This invention provides a method, system, device, and storage medium for orderly charging of electric vehicles that integrates load forecasting and dynamic price incentives, in order to solve the problems existing in the background art.

[0004] Technical Solution: The present invention provides an orderly charging method for electric vehicles that integrates load forecasting and dynamic price incentives, comprising the following steps: collecting historical power grid data and electric vehicle charging behavior data, and using a long short-term memory network to perform day-ahead and intraday load forecasting to assess the load status of the distribution network in real time; based on the load forecasting results, using the whale optimization algorithm combined with the Pareto front to solve multi-objective charging scheduling, and implementing a dynamic electricity price mechanism according to the solution results, wherein the electricity price consists of a base electricity price and a dynamic electric vehicle carrying capacity scheduling service fee, the scheduling service fee fluctuating in real time according to the predicted power grid load; and awarding incentive points to users who actively participate in off-peak charging according to the dynamic electricity price mechanism, the points of which can be used to directly offset the scheduling service fees incurred during subsequent charging.

[0005] Furthermore, historical power grid data and electric vehicle charging behavior data are collected, and long short-term memory (LSTM) networks are used for day-ahead and intraday load forecasting to assess the load status of the distribution network in real time. Specifically, the following steps are taken: power grid operation data and electric vehicle charging data are collected, and day-ahead load forecasting is performed using an LSM neural network model to output the load curve for the next day; intraday load forecasting is performed, and the day-ahead forecasting results are rolled over using real-time monitoring data to cope with random load fluctuations; the load forecasting results are linked with the distribution network real-time monitoring system, and an early warning is triggered when the forecasted load exceeds a safety threshold; among these, day-ahead forecasting provides forward-looking planning for dispatch optimization, and intraday forecasting ensures the adaptability of dispatching strategies through real-time correction. The linkage between load forecasting and monitoring forms a forecast-response closed loop.

[0006] Furthermore, based on the load forecasting results, a multi-objective charging scheduling solution is obtained using the whale optimization algorithm combined with the Pareto front, as follows: Electric vehicle charging scheduling is modeled as a multi-objective optimization problem, with objectives including minimizing grid load fluctuations and minimizing user charging costs; a multi-population automatic learning whale optimization algorithm is used for solving the problem, improving search efficiency through integer encoding, Euclidean operator initialization, multi-population evolution, and automatic learning mechanisms; the Pareto optimal solution set is output as the charging scheduling scheme; the scheduling optimization module relies on the load status data provided by the load forecasting module and considers the dynamic electricity price signal generated by the electricity price scheduling module to balance grid security and user economy.

[0007] Furthermore, the dynamic electricity pricing mechanism implemented based on the solution results is as follows: The dynamic electricity price consists of a base price and a dispatch service fee. The dispatch service fee includes a day-ahead dispatch service fee and a real-time price correction factor. The day-ahead dispatch service fee is formulated based on the day-ahead load forecast curve and is released in advance. The real-time price correction factor is dynamically calculated based on the deviation between the real-time load and the forecast load to ensure that the price signal is flexibly adjusted according to the grid status. The end-user electricity price is the sum of the base price, the day-ahead dispatch service fee, and the real-time price correction factor. Among them, the electricity price dispatch module receives the forecast data from the load forecast module and generates a price signal according to the dispatch optimization module's dispatch requirements to directly guide users' charging behavior.

[0008] Furthermore, according to the dynamic electricity pricing mechanism, incentive points are awarded to users who actively participate in off-peak charging, specifically as follows: A grid-friendliness point system is introduced, where users earn points by charging during off-peak hours or discharging during peak hours; the point acquisition rules are linked to the grid load status, with more points earned per unit of charging when the load is lower; points are directly used to offset dispatch service fees incurred by users during charging; among these, the user incentive module is closely coupled with the electricity price dispatch module, and the point mechanism strengthens the guiding role of dynamic electricity pricing on user behavior, forming a "contribution-reward" closed loop and enhancing user participation.

[0009] The present invention discloses an orderly charging system for electric vehicles that integrates load forecasting and dynamic price incentives, comprising:

[0010] The load forecasting module is used to collect grid and electric vehicle charging data and perform day-ahead and intraday load forecasting through a long short-term memory network.

[0011] The scheduling optimization module is used to solve multi-objective charging scheduling problems based on load forecasting results, using the whale optimization algorithm and Pareto front.

[0012] The electricity price dispatch module is used to implement a dynamic electricity price mechanism and generate user electricity prices consisting of a base electricity price and a dispatch service fee.

[0013] The user incentive module is used to reward users for participating in off-peak charging through points, which can be used to offset dispatch service fees.

[0014] Furthermore, the load forecasting module collects grid operation data and electric vehicle charging data, performs day-ahead load forecasting using a long short-term memory neural network model, and outputs the next day's load curve; it also performs intraday load forecasting, using real-time monitoring data to continuously correct the day-ahead forecasting results to cope with random load fluctuations; and it links the load forecasting results with the distribution network real-time monitoring system, triggering an early warning when the predicted load exceeds a safety threshold. Among these features, day-ahead forecasting provides forward-looking planning for dispatch optimization, while intraday forecasting ensures the adaptability of dispatching strategies through real-time correction. The linkage between load forecasting and monitoring forms a forecast-response closed loop.

[0015] Furthermore, in the scheduling optimization module, electric vehicle charging scheduling is modeled as a multi-objective optimization problem, with objectives including minimizing grid load fluctuations and minimizing user charging costs. A multi-population automatic learning whale optimization algorithm is used for solving the problem, improving search efficiency through integer encoding, Euclidean operator initialization, multi-population evolution, and automatic learning mechanisms. The Pareto optimal solution set is output as the charging scheduling scheme. The scheduling optimization module relies on load status data provided by the load forecasting module and considers dynamic electricity price signals generated by the electricity price scheduling module to balance grid security and user economics.

[0016] Furthermore, in the electricity price dispatch module, the dynamic electricity price consists of a base price and a dispatch service fee. The dispatch service fee includes a day-ahead dispatch service fee and a real-time price correction factor. The day-ahead dispatch service fee is formulated based on the day-ahead load forecast curve and is released in advance. The real-time price correction factor is dynamically calculated based on the deviation between the real-time load and the forecast load to ensure that the price signal is flexibly adjusted according to the grid status. The end-user electricity price is the sum of the base price, the day-ahead dispatch service fee, and the real-time price correction factor. The electricity price dispatch module receives forecast data from the load forecast module and generates price signals based on the dispatch requirements of the dispatch optimization module to directly guide users' charging behavior.

[0017] Furthermore, the user incentive module introduces a grid-friendliness points system, where users earn points by charging during off-peak hours or discharging during peak hours. The points earning rules are linked to the grid load status, with more points earned per unit of charging when the load is lower. These points are directly used to offset the dispatch service fees incurred by users during charging. The user incentive module is closely coupled with the electricity price dispatch module, and the points mechanism strengthens the guiding role of dynamic electricity prices on user behavior, forming a "contribution-reward" closed loop and enhancing user participation.

[0018] An electronic device according to the present invention includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described herein.

[0019] The present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method described herein.

[0020] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention uses the carrying capacity of the distribution network as the objective function, combines day-ahead and intraday forecast information, and optimizes the charging strategy of electric vehicles through dynamic electricity price adjustment, orderly charging user incentives and intelligent scheduling, thereby improving the carrying capacity of the distribution network, avoiding power system overload, and ensuring the safe and stable operation of the power system. Attached Figure Description

[0021] Figure 1 This is a flowchart of the present invention;

[0022] Figure 2 This is a comparison chart of dynamic time-of-use electricity prices and static time-of-use electricity prices under different numbers of electric vehicles connected to the grid according to the present invention;

[0023] Figure 3 This is a comparison chart of ordered and disordered charging loads under different numbers of electric vehicles connected, as presented in this invention. Detailed Implementation

[0024] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0025] like Figure 1 As shown, this invention provides a method for orderly charging of electric vehicles that integrates load forecasting and dynamic price incentives, including the following steps: A load forecasting module collects grid and charging data, uses LSTM for load forecasting, and evaluates the distribution network load status in real time. A scheduling optimization module, based on the forecast results, uses the whale optimization algorithm and Pareto front to perform multi-objective scheduling, reducing load fluctuations and peak-shifting charging. A price scheduling module implements a dynamic price mechanism, with the price including a base price and an EV carrying capacity scheduling service fee based on load fluctuations. The service fee adjusts with load changes. A user incentive module awards incentive points to users participating in peak-shifting charging; these points can be used to offset subsequent charging costs, encouraging long-term user participation.

[0026] The load forecasting module includes:

[0027] By collecting and recording information such as voltage and current in the power grid, power grid load data is obtained. Power grid load can be calculated using the following formula:

[0028]

[0029] in, Let be the grid power at time t; This refers to the instantaneous power of the power grid. Let be the voltage at time t; Let be the current at time t; This refers to the phase angle of voltage and current.

[0030] In an AC system, power Instantaneous power fluctuates periodically over time, and is typically represented by the effective value (RMS value) to indicate the magnitude of grid voltage, current, and power. Information such as electric vehicle charging power, battery level, and charging period is collected from charging stations.

[0031] (1) Day-ahead load forecast

[0032] In the optimized scheduling system for orderly charging of electric vehicles, load pre- The layer collects historical load data from the distribution network and electric vehicle charging / discharging behavior data, and uses machine learning algorithms to perform day-ahead and intraday load forecasting. This enables real-time monitoring and assessment of the distribution network load status. Day-ahead load forecasting aims to predict the load change trend of the distribution network within the next 24 hours, including regular load and electric vehicle charging load, one day in advance. Day-ahead load forecasting can be represented as a function approximation problem:

[0033]

[0034] in Indicates the next day Forecast values ​​of load at any given time. Represents the historical load sequence. This represents other feature vectors that affect the load (such as season, weekday, user behavior patterns, etc.).

[0035] 1) Long Short-Term Memory Neural Network

[0036] To capture the time-series characteristics and nonlinear patterns of load, this invention employs machine learning models such as Long Short-Term Memory (LSTM) neural networks to train historical data. LSTM models can memorize long-sequence dependencies, making them suitable for modeling the daily cycles and seasonal fluctuations of distribution network loads. The prediction model takes load data from the past several days and related influencing factors as input, and outputs load forecasts for each time period of the following day. The input vector at time t contains historical load and other auxiliary features. For the hidden state of the LSTM, it can be described as follows:

[0037]

[0038] in, These are the forget gate, input gate, and output gate vectors, respectively. For the state of the memory unit, It is the Sigmoid activation function. This indicates element-wise multiplication.

[0039] 2) The objective function for current prediction

[0040] During training, assume the observed values ​​are The predicted value is The total number of samples is The commonly used mean squared error (MSE) can be selected as the loss function:

[0041]

[0042] The parameters of the prediction model are obtained by iteratively solving using gradient descent. The optimal solution is found by adjusting the model parameters to improve prediction accuracy by minimizing the mean squared error loss between the predicted and actual values. After training, the latest historical data is input into the model at each daily scheduling cycle, i.e., the feature input for the next day. It can output various time periods of the day. Load forecast ,in This represents the total number of time periods within a day, from which the load forecast curve for the entire next day is obtained.

[0043] To enhance the accuracy of day-ahead forecasts, this invention incorporates factors such as grid seasonality trends and electric vehicle user charging behavior patterns into the model. For example, summer and winter load curves exhibit regular peak-hour differences, and user charging demand patterns differ between weekdays and weekends. By including these factors in the training data, the LSTM model can automatically learn the corresponding pattern characteristics, thereby predicting load changes that more closely reflect reality.

[0044] The day-ahead forecasting model of this invention provides the scheduling optimization module with detailed load forecasts for the next day, including hourly baseline load levels and projected electric vehicle charging demand. The overall goal is to lay the foundation for subsequent scheduling, minimize the peak-to-valley difference in the power grid, and avoid local overloads.

[0045] (2) Intraday load forecast

[0046] Because actual load is affected by real-time factors, relying solely on day-ahead forecasts may be insufficient to cope with random fluctuations and unforeseen events in power grid operation. Therefore, this invention designs an intraday load forecasting mechanism that continuously updates the load forecast for the remaining time of the day. Intraday forecasts utilize load data obtained from real-time monitoring of the day to continuously correct previous forecast results, thereby improving the real-time performance and accuracy of the forecasts.

[0047] Specifically, at each time step (e.g., every hour or every 15 minutes) within a day, the latest load monitoring data is compared with previous forecasts, the forecast error is calculated, and the model input is updated accordingly, enabling the model to output new load forecasts for the next time period and even the remainder of the day. This rolling forecast strategy allows the model to "adapt" to load deviations that occur within the day. The following correction formula can be used:

[0048]

[0049] in This is the predicted value from the previous period. for The measured load at any given time The gain coefficient is determined empirically (or adaptively determined by algorithms such as Kalman filtering).

[0050] Through the above corrections, the load for future periods can be appropriately adjusted based on the latest deviations, making intraday forecasts closer to real-time conditions. Simultaneously, a preliminary assessment is conducted using forecast data before the day, and intraday adjustments are made to the daytime results through a comprehensive strategy, thereby dynamically updating the system's response capability. This invention tracks and adjusts daytime load plans to ensure that load forecasts promptly reflect actual operating conditions.

[0051] While completing the intraday rolling forecast, it is also necessary to determine whether the load status of the distribution network exceeds the preset warning line:

[0052]

[0053] in Based on the basic load forecast, This represents the predicted charging and discharging load of the electric vehicle fleet at time t. If we determine... If the safety threshold of a certain line or transformer in the distribution network is exceeded, an emergency dispatch plan needs to be triggered.

[0054] When sudden large-scale charging events or load surges occur, the intraday forecasting model quickly identifies anomalies and updates the forecast curve, ensuring that the dispatching strategy can adapt to the new load level. This rolling optimization forecasting method improves the robustness of load forecasting, enabling distribution network dispatching to effectively track target load levels and dynamically adjust charging control strategies.

[0055] Furthermore, the model monitoring and prediction includes:

[0056] (1) Prediction model optimization and accuracy improvement

[0057] To ensure the accuracy and stability of load forecasting, this invention makes several improvements in model optimization and error correction. First, cross-validation and hyperparameter optimization are introduced during the model training phase to ensure the machine learning model has good generalization ability, thus maintaining high-accuracy predictions even on unknown data. Second, considering the uncertainties in distribution network load and electric vehicle charging behavior, interval forecasting and probabilistic forecasting methods are used to improve the model's ability to describe uncertainty. The load forecast value at a certain moment is denoted as... Define its upper and lower limits of fluctuation range. as follows:

[0058]

[0059] in and The fluctuation coefficients are estimated based on historical errors, representing the relative fluctuation range of the load that may be lower or higher than the predicted value.

[0060] In this way, the prediction result is no longer a single value, but a confidence interval, representing the possible range of load variation. This interval can be referenced during scheduling decisions to optimize performance while ensuring a safety margin, thereby mitigating the risks posed by prediction errors. Literature also points out that large load prediction errors can adversely affect scheduling results, necessitating consideration of prediction uncertainty to improve scheduling robustness. Therefore, this invention uses interval prediction and error correction to minimize the impact of load prediction errors on the system.

[0061] In addition, the accuracy can be improved by incorporating multi-source data during the model optimization process. This includes integrating meteorological data (temperature, rainfall, etc.) to improve the prediction of climate-related loads such as air conditioning, and integrating usage data of electric vehicle charging stations to understand the randomness of vehicle access.

[0062] (2) Real-time monitoring and forecasting linkage of distribution network load status

[0063] The prediction layer of this invention is closely integrated with the real-time monitoring of the distribution network load status, forming a closed-loop prediction-monitoring system. Voltage, current, and power load data are collected at nodes such as the low-voltage side of the distribution network transformer and feeder outlets. This real-time load data is transmitted to the energy management platform via a communication network to update load status information. The operation platform, on the one hand, publishes this real-time load information to charging facilities, enabling charging stations to understand the current load margin of the distribution network; on the other hand, it inputs the real-time data into the prediction model to correct the model's internal state.

[0064] The prediction layer makes intraday prediction corrections based on the latest measured data to ensure that the prediction results are synchronized with the actual operating status. Real-time monitoring allows the system to detect abnormal load conditions in a timely manner. When the distribution network load is detected to be approaching the safety limit, the system can issue an alert through the prediction layer and take measures in conjunction with the scheduling module to temporarily reduce the charging power of some charging piles or delay the charging start of some electric vehicles. This linkage between monitoring and prediction improves the system's proactive defense capability.

[0065] This invention integrates power distribution network load monitoring data and electric vehicle charging demand forecasts through an operation platform to achieve comprehensive data analysis, collect charging demand information and make forecasts, and report to the power grid dispatch center to assist decision-making: on the one hand, it accurately grasps the current load level, and on the other hand, it predicts future load trends, and combines the two to optimize decision-making.

[0066] To fully utilize real-time data, the prediction layer of this invention employs adaptive data assimilation technology. Whenever a new load monitoring value arrives, it is fused with the model's prior predictions. The model state is then updated by resetting the hidden layer state of the recurrent neural network, thereby improving the next prediction. Supported by real-time data, load forecasting can be refreshed at a high frequency, ensuring that scheduling decisions are always based on the latest information. Through real-time monitoring of the distribution network load status, an organic combination of day-ahead and intraday forecasts is achieved: day-ahead forecasts provide a global trend, while intraday forecasts are refined and adjusted using real-time data. This approach maximizes the efficiency and flexibility of electric vehicle charging scheduling while ensuring grid security.

[0067] Furthermore, the scheduling optimization module includes:

[0068] This invention proposes a multi-swarm automatic learning whale optimization algorithm (MAWOA) for electric vehicle scheduling. This algorithm is derived from the simulation of humpback whale hunting behavior and has been deeply improved for discrete scheduling problems. Compared with the standard whale optimization algorithm (WOA), MAWOA has stronger global search capabilities and faster convergence speed.

[0069] It mainly consists of the following innovative parts:

[0070] Integer encoding: The charging scheduling scheme for each electric vehicle is encoded with integers to adapt to discrete scheduling periods and make calculations simpler.

[0071] Euclidean operator: The Euclidean distance operator is introduced in the initialization phase to increase the diversity of initial solutions and avoid the algorithm getting trapped in local optima too early.

[0072] Multi-population evolution: Dividing the whale population into multiple subpopulations for collaborative search improves the algorithm's global search efficiency and solution quality.

[0073] Automatic learning mechanism: After each iteration, an automatic learning mechanism of neighborhood search is introduced to speed up the search for Pareto optimal solution set.

[0074] Model building:

[0075] (1) Objective function

[0076] This invention models the Ordered Electric Vehicle Charging Schedule (OWSP) as a multi-objective optimization problem, and mainly considers the following two conflicting objectives:

[0077] Objective 1: Minimize grid load fluctuations ( )

[0078] The goal is to smooth the total load curve of the power grid and reduce peak-to-valley differences by scheduling the charging time of electric vehicles. The total load of the power grid consists of conventional load and electric vehicle charging load, and the load is calculated based on the power grid's load throughout the day (divided into...). Its volatility is measured by the standard deviation of the total load over a given period.

[0079]

[0080] in, yes Total grid load during the period ; It is the predicted base load; Are all electric vehicles in Total charging load during the period; It is the average value of the total load.

[0081] Objective 2: Minimize the total charging cost for users ( )

[0082] The goal is to encourage users to charge their devices during periods of lower electricity prices, thereby reducing their overall costs. This requires a dynamic pricing mechanism. Assuming electricity prices... It varies at different times.

[0083]

[0084] in, It represents the total number of electric vehicles participating in the dispatch; It is a vehicle During the period The charging power; It is a time period Electricity price; and They are vehicles The start and end times of charging; It is the length of a single time period (e.g., 0.5 hours).

[0085] Final optimized model:

[0086] Combining the above objectives, the final multi-objective optimization problem can be defined as:

[0087]

[0088] in This represents a charging scheduling scheme.

[0089] (2) Constraints

[0090] Power grid security constraints: any time period The total load must not exceed the maximum carrying capacity of the distribution network. .

[0091]

[0092] User charging demand constraint: Each electric vehicle Must leave at the expected time Before that, ensure that the battery's state of charge (SOC) reaches the target value set by the user. .

[0093]

[0094] in It's the battery capacity. It refers to charging efficiency.

[0095] Charging power constraint: The charging power of the vehicle cannot exceed the maximum power allowed by the charging station or the vehicle itself. .

[0096]

[0097] (3) Solution method based on MAWOA

[0098] 1) Encoding method

[0099] This invention employs an integer encoding method. Assume a day is divided into... Each scheduling period (e.g., if 24 hours are divided into 30-minute intervals, there are 48 intervals). For A vehicle, a scheduling scheme (i.e., the location of a whale) can be represented by a length of... vector express:

[0100]

[0101] in, It is an integer representing the first... The starting time for charging electric vehicles.

[0102] 2) Euclidean operators

[0103] To ensure diversity of initial solutions, Euclidean distance is used for selection after randomly generating the initial population. (Two whales are considered, representing two scheduling schemes.) and The distance between them is defined as:

[0104]

[0105] If the distance between the two whales is less than the set threshold If so, one of them will be regenerated to ensure that there are sufficient differences between the initial schemes.

[0106] 3) Multi-population evolution

[0107] The total number is whale populations are divided into A subpopulation. The division steps are as follows:

[0108] Based on one of the objective functions (such as...) Sort all whales.

[0109] The whale ranked first is placed in subgroup 1, the second one in subgroup 2, and so on. to the first Subgroup. Then the first Then they return to the first subgroup, and so on, until all whales have been allocated.

[0110] Each subpopulation performs subsequent evolutionary operations independently, enhancing global search capabilities.

[0111] 4) Adaptive modification of whale hunting behavior

[0112] Whale hunting behavior is divided into three stages, and the formula for position update (i.e., iterative optimization of the scheduling scheme) is as follows:

[0113] Encircling Prey: The whale moves closer to the current optimal solution (best scheduling scheme).

[0114]

[0115] Bubble-net Attacking: Simulates two synchronous behaviors: spiraling approach and shrinking encirclement.

[0116]

[0117] Search for Prey: When At this time, a random individual is selected for exploration in order to escape the local optimum.

[0118]

[0119] In these formulas, This is the current scheduling scheme. This is the current optimal solution. It is a randomly selected scheme. It is a coefficient that controls the behavior of the algorithm.

[0120] 5) Automatic learning mechanism

[0121] After each subpopulation completes a hunting iteration, an automatic learning mechanism is introduced to perform neighborhood search. For each whale... Generate near its current location A new candidate location (new scheduling scheme):

[0122]

[0123] in, It is the first Among the candidate solutions, the first one is... The charging start time of the vehicle; rand is a random number between (0, 1); step represents the search step size (e.g., 1 or 2 time intervals) used to fine-tune the charging time.

[0124] The algorithm will evaluate this A new solution is proposed, and its non-dominated solutions are recorded. If the new solution is better than the current solution, the whale will move to the new location; otherwise, it will remain stationary, thus speeding up the algorithm's convergence to the Pareto optimal front.

[0125] Furthermore, the real-time electricity price dispatch module includes:

[0126] The electricity price dispatch module in this invention is a core hub connecting the power grid, charging agents, and electric vehicle (EV) users. Its main function is to transmit real-time carrying capacity information of the distribution network to users through a flexible and dynamic electricity price mechanism, thereby guiding users to participate in charging and discharging autonomously and orderly, achieving peak shaving and valley filling, ensuring grid stability, and maximizing the overall interests of the alliance (agents and users).

[0127] (1) Composition of dynamic electricity price

[0128] To differentiate itself from traditional fixed time-of-use pricing, this invention proposes a dynamic pricing system. It consists of two parts: the base electricity price and carrying capacity scheduling service fee .

[0129]

[0130] Basic electricity price Reflecting the basic commodity attributes of electricity in the electricity market, the time-of-use price (i.e., peak, flat, and valley prices) published by grid aggregators is mainly determined by generation costs, transmission and distribution costs, etc., and remains unchanged within a certain period (such as a day or a quarter), providing users with a predictable basic charging cost.

[0131] Dispatch service fees are composed of day-ahead dispatch service fees. and real-time price correction factor Joint decision:

[0132]

[0133] 1) Formulation and publication of the dispatch service fee

[0134] The day before each scheduling day, this module first uses the day-ahead total load forecast curve output by the data analysis and forecasting layer. To formulate and publish the dispatch service fee for the following day. The purpose of this fee is to provide a benchmark price expectation for optimized scheduling the following day.

[0135] The calculation model remains consistent with previous models, but its input is explicitly defined as the day-ahead forecast load:

[0136] .

[0137] Among them, the total power grid load predicted recently; These are the power grid's safe load, off-peak load, and ultimate capacity, respectively. These are the maximum service fee, the maximum subsidy, and the congestion response coefficient, respectively.

[0138] The day-ahead service fee will be released to users and the dispatch system 24 hours in advance, serving as the basis for users to participate in the day-ahead charging plan and for the system to conduct preliminary resource optimization.

[0139] 2) Dynamic adjustment of intraday real-time prices

[0140] Once the actual scheduling day begins, the actual load may deviate from the previous day's forecast due to unforeseen events or forecast errors. To address this, this module introduces a real-time price correction factor. This factor is dynamically calculated based on the deviation between the real-time monitored load and the day-ahead forecasted load.

[0141] First, define the real-time load deviation. :

[0142]

[0143] in, This is the current total load of the power grid. When... This indicates that the actual load exceeded expectations, increasing pressure on the power grid and necessitating further suppression of charging. This indicates that the actual load is lower than expected, the grid margin is increased, and charging can be further incentivized.

[0144] Real-time price correction factor It is designed as a function sensitive to deviations and with saturation boundaries to ensure the effectiveness and stability of price signals. This invention models it using the hyperbolic tangent function (tanh):

[0145]

[0146] in, The maximum price correction factor (in yuan / kWh) determines the maximum price fluctuation range for real-time correction. The price sensitivity coefficient controls the rate at which the correction factor responds to load deviations. This is the hyperbolic tangent function. The advantage of choosing this function is that it is effective when the deviation... When the deviation is small, the price correction is approximately linear; when the deviation is very large, the correction factor will smoothly approach its saturation value. This avoids the instability of prices increasing indefinitely due to extreme disturbances, thus enhancing the robustness of the system.

[0147] 3) Formation and application of end-user electricity prices

[0148] Combining the two stages above, at any given time The dynamic electricity price that users ultimately face The complete model is as follows:

[0149]

[0150]

[0151] in This represents a piecewise function based on day-ahead forecasted load.

[0152] (2) Model fusion based on user selection behavior

[0153] The effectiveness of dynamic electricity pricing ultimately depends on whether users are willing to change their charging behavior based on price signals. This invention uses the idea of ​​incentivizing users to change their travel routes through compensation, treating dynamic electricity pricing as an economic incentive signal.

[0154] A user's charging decision can be modeled as a choice problem. When connecting to a charging station, a user can choose the "plug and charge" uncoordinated charging mode. ), or an ordered charging / discharging mode that "accepts scheduling" ( ).

[0155] User selection mode The probability of success depends not only on physical factors such as the vehicle's SOC and parking duration, but more importantly on the cost difference between the two modes. Users choosing the orderly charging mode (…) The decision function of can be expressed as:

[0156]

[0157] in, User During the period Select mode The willingness or probability; Incompatible charging modes The price (usually high and fixed); This invention proposes a dynamic electricity price. and These represent the user's parking duration and the vehicle's current battery level, respectively.

[0158] The calculated service fee directly affects The value. During peak electricity consumption periods, A significant increase, making If the difference becomes smaller or even negative, users will be more inclined to accept the scheduling (choice) due to economic considerations. (Mode) and postpone charging. Conversely, during off-peak hours, The price reduction due to subsidies will incentivize users to choose to charge during this period.

[0159] Ultimately, the goal of scheduling the MAWOA optimization module is to find a set of optimal solutions. The value allows the user to choose the objective function of the entire system under the selected model (such as grid friendliness). Total revenue To achieve optimal results.

[0160] Furthermore, the user incentive module includes:

[0161] The user incentive module of this invention aims to establish a long-term, win-win interactive mechanism, which motivates users to participate in orderly charging by quantifying their contributions to the power grid and providing corresponding rewards.

[0162] (1) Points-based reward mechanism: Power grid friendliness points

[0163] This module introduces a reward system called "Grid Friendliness Points". Users accumulate points by performing charging and discharging behaviors that benefit the grid. These points can be directly used for economic benefits, thus forming a complete incentive loop.

[0164] (2) Rules for obtaining points

[0165] The contribution of a user's charging behavior to the power grid is the sole criterion for earning points. Points are primarily earned based on the user's responsiveness to the dynamic price signals emitted by the "electricity price dispatch module."

[0166] Rule 1: Earn points by charging during off-peak hours

[0167] When users choose to charge during off-peak hours when the grid load is low, the system will award points based on the amount of charging done during those off-peak hours.

[0168] The formula for calculating integrals is as follows:

[0169]

[0170] in, User During the period Points earned; User During the period The amount of electricity charged (kWh); This is the predicted total load on the power grid; It is the preset target value for off-peak load; It is the integral conversion factor (for example, 100 base points are awarded for each kilowatt-hour of electricity).

[0171] The lower the grid load, the more credits a user can earn per unit of charging.

[0172] Rule 2: Peak discharge (V2G) contribution integral

[0173] For vehicles that support vehicle-to-grid (V2G) functionality, when users discharge electricity back into the grid during peak electricity consumption periods, this represents the highest value contribution to the grid. Therefore, the system will award significantly higher reward points than for off-peak charging to incentivize users to provide valuable peak-shaving capacity to the grid during critical times.

[0174] (3) Rules for using points

[0175] The sole purpose of the points is to directly offset the "capacity scheduling service fee" that users need to pay when charging, thus coupling the incentive mechanism with the pricing mechanism.

[0176] Deduction rules: Set the exchange rate, for example, every 100 points can deduct 1 yuan of dispatch service fee (the specific amount will be determined according to the actual situation).

[0177] Deduction limit: The amount of points deducted for each charge cannot exceed the total dispatch service fee incurred that time.

[0178] The user's final charging cost It can be represented as:

[0179]

[0180] in, It is the number of points the user chooses to use; It's the points redemption rate.

[0181] (4) Forming a closed loop of business and technology

[0182] During peak electricity consumption periods, in order to avoid high costs and earn points, users proactively shift their charging time to off-peak hours when electricity prices and service fees are both zero or negative. The user's charging behavior during off-peak hours is quantified by the system as "grid friendliness points" and stored in their account. When the user inevitably needs to charge during non-off-peak hours in the future and faces certain service fees, their previously accumulated points can be used to offset these fees, thus achieving a closed loop of "contribution" and "reward".

[0183] In this embodiment, an orderly charging system for electric vehicles that integrates load forecasting and dynamic price incentives includes:

[0184] The load forecasting module is used to collect grid and electric vehicle charging data and perform day-ahead and intraday load forecasting through a long short-term memory network.

[0185] The scheduling optimization module is used to solve multi-objective charging scheduling problems based on load forecasting results, using the whale optimization algorithm and Pareto front.

[0186] The electricity price dispatch module is used to implement a dynamic electricity price mechanism and generate user electricity prices consisting of a base electricity price and a dispatch service fee.

[0187] The user incentive module is used to reward users for participating in off-peak charging through points, which can be used to offset dispatch service fees.

[0188] The system specifically includes the following four core functional modules: data analysis and prediction module, real-time electricity price dispatch module, dispatch optimization module, and user incentive module.

[0189] (1) Data Analysis and Prediction Module

[0190] The data analysis and prediction module provides reliable data support for subsequent optimization decisions. This module consists of the following units:

[0191] Data Acquisition and Processing Unit: This unit is responsible for acquiring information from multiple data sources. Its inputs include: 1) Historical and real-time load data collected from the distribution network SCADA system or smart meters, such as voltage, current, and power at key nodes (low-voltage side of transformers, feeders, etc.); 2) Electric vehicle charging behavior data obtained from the charging pile management platform, such as vehicle access time, charging duration, and charging amount; 3) External multi-source auxiliary data, such as meteorological data (temperature, humidity, affecting air conditioning load) and calendar data (weekdays / weekends / holidays, affecting user travel patterns). The collected raw data undergoes cleaning, noise reduction, and format normalization to form a standard dataset usable by the model.

[0192] Day-ahead Forecasting Unit: This unit forms the foundation for day-ahead planning. It incorporates a deep learning model based on a Long Short-Term Memory (LSTM) network. This model takes load data, weather, and date type from the past 30 days as input sequences. Leveraging the powerful time-series feature capture capabilities of the LSTM network, it learns the daily, weekly, and seasonal trends of the grid load. Its output is a day-ahead basic load forecast curve covering the next 24 hours (e.g., 96 points at 15-minute granularity). .

[0193] Intraday Rolling Forecasting Unit: To address uncertainties and unforeseen events in actual operation, this unit is responsible for real-time correction of forecast results. It operates on a short cycle (e.g., every 15 minutes), updating the forecast results with the latest real-time load data acquired by the acquisition unit. The forecast is compared with the most recent forecast to calculate the forecast error. Then, a rolling forecast strategy or Kalman filtering algorithm is used to dynamically adjust and update the load forecast curve for the remaining future periods. This unit ensures that the system's assessment of the power grid condition always closely matches the actual situation.

[0194] (2) Real-time electricity price dispatch module

[0195] The real-time electricity price dispatch module is the system's "command stick," transforming the invisible grid carrying capacity into quantifiable economic signals that are visible to users. Its core is the generation and publication of dynamic electricity prices through a two-stage pricing model.

[0196] Day-ahead service fee calculation unit: During the day-ahead planning phase, this unit obtains the day-ahead load forecast curve output by the forecasting module. It is based on a preset power grid safe load threshold. and off-peak load threshold The daytime scheduling service fee for each time period of the following day is calculated using a piecewise function. This cost is an estimate by the power grid of future load pressure and will be released in advance.

[0197] Real-time Correction Factor Calculation Unit: During the intraday execution phase, the core task of this unit is to reflect the immediacy of prices. It calculates load deviations in real time. When the actual load exceeds the forecast, it uses a bounded, nonlinear hyperbolic tangent (tanh) model to calculate a positive real-time price correction factor. This is used to increase charging costs; conversely, when the actual load is lower than the forecast, the calculated correction factor is negative, which is equivalent to providing additional real-time charging subsidies.

[0198] Price Synthesis and Publication Unit: This unit will display the fixed base electricity price. The service fee set recently and real-time correction factor The three factors combined form the final, real-time dynamic electricity price that users need to pay. The price is pushed to charging stations and user apps in real time via API.

[0199] (3) Scheduling optimization module

[0200] The scheduling optimization module is the "brain" of the system, responsible for finding the globally optimal charging scheduling scheme while satisfying the constraints of all parties.

[0201] Model Building Unit: Upon receiving a user's charging request, this unit immediately constructs a multi-objective optimization problem. It transforms the user's requirements (such as "must be fully charged before 7 AM") into constraints for the model. Simultaneously, it defines two core objective functions: 1) Minimize the fluctuation of the power grid load. Its calculation relies on load data provided by the prediction module; 2) minimizing user charging costs Its calculation depends on the dynamic electricity price provided by the electricity price module.

[0202] MAWOA Solver: To solve the aforementioned complex, discrete, multi-objective optimization problems, this module employs an improved multi-swarm automatic learning whale optimization algorithm (MAWOA). This algorithm uses integer encoding to represent charging periods, leverages Euclidean operators and a multi-swarm strategy to ensure the breadth of the search and avoid getting trapped in local optima; simultaneously, it utilizes an automatic learning mechanism for neighborhood search, accelerating convergence and enabling the finding of a set of high-quality Pareto optimal solutions in a short time.

[0203] Strategy Output Unit: After the solution is completed, the module selects the most balanced solution from the Pareto optimal solution set (or provides several options for the user to choose from) and outputs it as the recommended "smart charging strategy".

[0204] (4) User Incentive Module

[0205] The user incentive module maintains user engagement by building a complete business loop.

[0206] Integral Calculation Unit: This unit monitors users' charging behavior in real time. When a user charges according to the scheduling strategy, during the system-determined off-peak hours (e.g., ... After charging is completed, this unit will calculate and reward a corresponding amount of "grid friendliness points" based on the amount of "valley filling" electricity it contributed, using an integral formula.

[0207] Points Account Management Unit: This unit maintains an independent points account for each registered user. It is responsible for adding (rewarding) and deducting (using) points, and provides users with a function to query points details.

[0208] Points Deduction Unit: This unit is linked to the electricity pricing module and payment system. When a user's charging activity generates a positive "dispatch service fee," this unit will automatically check the user's points balance during the payment settlement process and provide the option to deduct the service fee using points, thus realizing the value of the user's contribution.

[0209] Figure 2 This invention relates to the dynamic electricity price generated by the system for end users in application scenarios with varying numbers of electric vehicles (EVs). The comparison curves are shown below. This embodiment aims to verify the adaptability, responsiveness, and effectiveness of the price signal of the two-stage dynamic pricing mechanism proposed in this invention under different scales of charging load pressure.

[0210] To conduct simulation verification of this embodiment, we set the following scenario:

[0211] Simulation object: A typical urban residential area power distribution network, with its transformer limit capacity. The safe load threshold is 2.5MW. Set at 2.0MW (80% capacity), with a target value for off-peak load. It is set to 1.0MW.

[0212] Base load and electricity price: The base load curve for a typical workday in this region is used as the background load. Base electricity price The standard peak-flat-valley model was adopted (peak period: 1.2 yuan / kWh, average period: 0.7 yuan / kWh, valley period: 0.4 yuan / kWh). These two basic conditions remained unchanged in all comparison scenarios.

[0213] Variable parameters: To simulate different levels of charging load pressure, this embodiment sets up five independent simulation scenarios. The only variable is the number of electric vehicles connected to this scheduling system, which are 500, 1000, 1500, 2000, and 2500 vehicles, respectively. It is assumed that the travel and charging demands of these vehicles follow a uniform statistical distribution pattern.

[0214] Figure 2 The horizontal axis represents 24 hours in a day, and the vertical axis represents the electricity price (unit: yuan / kWh). The graph contains six curves: one is the baseline electricity price curve, and the other five correspond to the dynamic electricity price curves for end users calculated by the system of this invention under the five scenarios mentioned above.

[0215] Analysis 1: Price Signal Classification and Pressure Transmission during Peak Hours (18:00-22:00)

[0216] like Figure 2 As shown, the baseline load during the evening peak period from 18:00 to 22:00 is already at a high level. As the number of connected electric vehicles increases from 500 to 2500, the system's forecasting module predicts the total daily load... . also rose significantly as a result.

[0217] In a scenario with 500 vehicles: the predicted peak total load may just reach or slightly exceed the safety threshold. Therefore, the day-ahead dispatch service fee calculated by the electricity pricing module... It is a small positive value. For example... Figure 2 As shown, the corresponding dynamic electricity price curve (marked as "500 vehicles" in the figure) is only slightly higher than the base electricity price.

[0218] In a scenario with 1500 vehicles: the predicted peak total load has significantly exceeded [the target]. According to the exponential growth model of day-to-day service fees (Formula (2)), This value was calculated to be relatively high. Therefore, the price curve for "1500 vehicles" in the graph was significantly inflated, creating a substantial gap with the base electricity price.

[0219] In a scenario with 2500 vehicles: the predicted total load is now approaching its maximum capacity. The exponential effect of service fees is fully activated, generating punitive prices. As shown in the figure, the price curve for "2500 vehicles" reaches its highest value at the peak, sending a strong signal to users that the power grid is about to be overloaded, thus most effectively suppressing charging behavior during this period and ensuring grid safety.

[0220] In this way, the pricing mechanism of the present invention achieves precise linkage between price signals and load pressure. The more vehicles there are and the greater the pressure, the stronger the price signal becomes, demonstrating a high degree of adaptability.

[0221] Analysis 2: Price incentive depth adjustment during off-peak hours (02:00-05:00)

[0222] During the off-peak electricity hours in the early morning, the scheduling optimization module will schedule the vast majority of scheduling charging tasks during this period.

[0223] In scenarios with 500 and 1000 vehicles: even if all vehicles are charging during this period, the predicted total load... It may still be lower than the target value for off-peak load. At this point, the electricity pricing module will calculate a negative day-ahead dispatch service fee, thus providing a charging subsidy. For example... Figure 2 As shown, both curves are lower than the base electricity price during off-peak hours, forming a "price trough".

[0224] In a scenario with 2,500 vehicles: Due to the large number of vehicles requiring dispatch, even during off-peak hours, the concentrated charging load can increase the predicted total load. Exceeding the trough threshold However, it remains within a safe range. In this situation, the day-ahead dispatch service fee... It is calculated as zero. As shown in the figure, its corresponding price curve is the same as the basic off-peak electricity price. Although there is no additional subsidy, it is still the lowest price of the day, which is enough to encourage users to charge here.

[0225] The pricing mechanism of this invention can intelligently adjust the "depth" of incentives according to the scale of load scheduling, providing subsidies when the load is sufficient and avoiding unnecessary fiscal expenditures when large-scale "valley filling" is needed, thus achieving effective utilization of incentive resources.

[0226] Analysis 3: Dynamic Reflection of Intraday Real-Time Corrections

[0227] Observe carefully Figure 2 The five dynamic electricity price curves in the model are not smooth curves, but rather have numerous subtle, high-frequency spikes or fluctuations superimposed on the overall trend. This is not noise, but rather the real-time price adjustment factor in the two-stage pricing mechanism of this invention. The direct manifestation of its effectiveness.

[0228] The system's electricity pricing module immediately detected this deviation within a 15-minute inspection cycle. It used the tanh function model to calculate a negative value. This negative price adjustment will be added to the user's electricity price in real time. This caused all five curves in the graph to experience a synchronized, slight downward "dip" around 14:00. This instantaneous price reduction, though small, was enough to incentivize price-sensitive vehicles that happened to need charging during that time to charge immediately, thus taking advantage of this valuable, unforeseen grid margin.

[0229] pass Figure 2 Simulation results demonstrate that the proposed two-stage dynamic pricing method is feasible and efficient. It not only formulates forward-looking and differentiated pricing plans based on daily load forecasts of varying scales, but also fine-tunes price signals at the second or minute level based on real-time daily load monitoring, ensuring that prices remain a true "barometer" of the grid's current carrying capacity. This pricing strategy, combining planning and flexibility, is far superior to traditional, fixed time-of-use pricing, maximizing user participation in orderly charging and discharging and achieving optimal allocation of grid resources.

[0230] Figure 3 This is a simulation comparison chart demonstrating the significant effect of the scheduling strategy described in this invention on smoothing grid load fluctuations and improving the carrying capacity of electric vehicles. This embodiment selects the distribution network area described in Embodiment 3 as the object, and compares and analyzes the changes in the total grid load curve under two modes: disordered charging and ordered charging using this invention.

[0231] Analysis 1: Load superposition effect in disordered charging scenarios (control group)

[0232] Please see Figure 3 The figure illustrates a typical disordered charging scenario without the intervention of the system of this invention. In this scenario, we assume that user behavior follows normal habits, namely, plugging the electric vehicle into the charging station and starting charging immediately after returning home from get off work (usually 18:00-20:00).

[0233] The bottom area in the figure represents the basic load curve of the transformer area when there is no electric vehicle load. It can be seen that there is a peak in residential electricity consumption between 19:00 and 21:00 in the evening.

[0234] The upper area in the diagram represents the disordered electric vehicle charging load. Due to the convergence of user charging times, this load increases sharply after 19:00, reaching its peak around 20:30, forming a charging peak.

[0235] The solid line at the top represents the total grid load, which is the direct superposition of the base load and the electric vehicle charging load. As shown in the figure, the superposition of the two peaks results in an extremely sharp load spike in the total load of the distribution area around 20:30.

[0236] In the simulation of this embodiment, we analyzed the impact under different vehicle sizes:

[0237] When 500 vehicles are connected, the combined peak traffic is already close to the safe threshold. .

[0238] When 1500 vehicles are connected (e.g.) Figure 3 As shown in the representative case, the total load peak has far exceeded the safety threshold, causing serious overload pressure on the transformer and line.

[0239] When 2,500 vehicles are connected, the simulated peak total load may even exceed the transformer's maximum capacity. In actual operation, this will trigger the protection device to trip, resulting in a power outage in the area, indicating that at this scale, disorderly charging is completely unbearable for the power grid.

[0240] Analysis 2: Peak shaving and valley filling effects after applying the orderly scheduling of this invention

[0241] Please see Figure 3 The figure shows the total grid load curve after the orderly charging scheduling of the system of this invention is fully applied under the same vehicle access demand and base load background.

[0242] Load reduction during peak hours (18:00-22:00):

[0243] During this period, the system's real-time electricity price dispatch module generated a high dynamic dispatch service fee based on predicted and real-time congestion conditions, resulting in higher electricity prices for end users. The nighttime charging rate is significantly higher than other times of day. Simultaneously, the scheduling optimization module plans the most economically efficient charging strategy for all users. Guided by both price and strategy, the vast majority of users choose to accept the scheduling. Therefore, as... Figure 3 As shown, during the evening peak hours, there is almost no new electric vehicle charging load, and the total load curve is basically the same as the baseline load curve.

[0244] Load filling during off-peak hours (02:00-07:00):

[0245] In stark contrast, a broad and stable charging load platform emerged during the early morning off-peak period. This was the result of the scheduling optimization module intelligently time-shifting and stacking all schedulable charging demands throughout the day (e.g., charging tasks for 1500 vehicles) using the MAWOA algorithm. The system ensured that the total height of this charging platform remained within a safe threshold at all times. The following is a summary of the points provided. Meanwhile, the electricity pricing module offered a negative service fee (i.e., charging subsidy) during this period and awarded points through the user incentive module, further solidifying user participation.

[0246] Analysis 3: The effect of improving load-bearing capacity under different vehicle sizes

[0247] A key advantage of this invention lies in the scalability of its scheduling capabilities. Although Figure 2 Only a comparison of 1500 vehicles was shown, but full-scenario simulations of 500 to 2500 vehicles show that:

[0248] As the number of connected vehicles increases from 500 to 2,500, the system is able to consistently offload the vast majority of the charging load from peak hours.

[0249] During off-peak hours, the charging load platform will "rise" accordingly. For example, with 500 vehicles, the charging load platform is lower; with 2,500 vehicles, the charging load platform will be higher, making fuller use of the grid capacity during off-peak hours, but it will still be strictly controlled within the safety threshold.

[0250] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.

[0251] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described herein.

Claims

1. A method for orderly charging of electric vehicles that integrates load forecasting and dynamic price incentives, characterized in that, Includes the following steps: Historical power grid data and electric vehicle charging behavior data are collected, and long short-term memory networks are used for day-ahead and intraday load forecasting to assess the load status of the distribution network in real time. Based on the load forecast results, a multi-objective charging scheduling solution is obtained using the whale optimization algorithm combined with the Pareto front. A dynamic electricity pricing mechanism is implemented based on the solution results, where the electricity price consists of a base price and a dynamic electric vehicle carrying capacity scheduling service fee, which fluctuates in real time according to the predicted power grid load. According to the dynamic electricity pricing mechanism, users who actively participate in off-peak charging are given incentive points, which can be used to directly offset the scheduling service fee incurred during subsequent charging.

2. The method for orderly charging of electric vehicles integrating load forecasting and dynamic price incentives according to claim 1, characterized in that, Historical power grid data and electric vehicle charging behavior data are collected, and long short-term memory (LSTM) networks are used for day-ahead and intraday load forecasting to assess the load status of the distribution network in real time. Specifically: Power grid operation data and electric vehicle charging data are collected; day-ahead load forecasting is performed using an LSTM neural network model, outputting the next day's load curve; intraday load forecasting is performed, and the day-ahead forecast results are rolled over using real-time monitoring data to address random load fluctuations; the load forecast results are linked with the distribution network real-time monitoring system, triggering an early warning when the predicted load exceeds a safety threshold; day-ahead forecasting provides forward-looking planning for dispatch optimization, while intraday forecasting ensures the adaptability of dispatch strategies through real-time correction, forming a forecast-response closed loop through the linkage of load forecasting and monitoring.

3. The method for orderly charging of electric vehicles integrating load forecasting and dynamic price incentives according to claim 1, characterized in that, Based on load forecasting results, a multi-objective charging scheduling solution is developed using the whale optimization algorithm combined with the Pareto front, as follows: Electric vehicle charging scheduling is modeled as a multi-objective optimization problem, with objectives including minimizing grid load fluctuations and minimizing user charging costs. The whale optimization algorithm, employing multi-population automatic learning, is used to solve the problem, improving search efficiency through integer encoding, Euclidean operator initialization, multi-population evolution, and automatic learning mechanisms. The Pareto optimal solution set is output as the charging scheduling scheme. The scheduling optimization module relies on load status data provided by the load forecasting module and considers dynamic electricity price signals generated by the electricity price scheduling module to balance grid security and user economics.

4. The method for orderly charging of electric vehicles integrating load forecasting and dynamic price incentives according to claim 1, characterized in that, The dynamic electricity pricing mechanism implemented based on the solution results is as follows: The dynamic electricity price consists of a base price and a dispatch service fee. The dispatch service fee includes a day-ahead dispatch service fee and a real-time price correction factor. The day-ahead dispatch service fee is determined based on the day-ahead load forecast curve and is released in advance. The real-time price correction factor is dynamically calculated based on the deviation between the real-time load and the forecast load to ensure that the price signal is flexibly adjusted according to the grid status. The end-user electricity price is the sum of the base price, the day-ahead dispatch service fee, and the real-time price correction factor. The electricity price dispatch module receives the forecast data from the load forecast module and generates a price signal based on the dispatch optimization module's dispatch requirements to directly guide users' charging behavior.

5. The method for orderly charging of electric vehicles integrating load forecasting and dynamic price incentives according to claim 1, characterized in that, According to the dynamic electricity pricing mechanism, users who actively participate in off-peak charging will be given incentive points as follows: A grid friendliness point system will be introduced, and users will obtain points by charging during off-peak hours or discharging during peak hours; the point acquisition rules are linked to the grid load status, and the lower the load, the more points will be obtained per unit of charging. Points are directly used to offset the dispatch service fees incurred when users charge their devices. The user incentive module is closely coupled with the electricity price dispatch module. The points mechanism strengthens the guiding role of dynamic electricity prices on user behavior, forming a "contribution-reward" closed loop and increasing user participation.

6. An orderly charging system for electric vehicles integrating load forecasting and dynamic price incentives, characterized in that, include: The load forecasting module is used to collect grid and electric vehicle charging data and perform day-ahead and intraday load forecasting through a long short-term memory network. The scheduling optimization module is used to solve multi-objective charging scheduling problems based on load forecasting results, using the whale optimization algorithm and Pareto front. The electricity price dispatch module is used to implement a dynamic electricity price mechanism and generate user electricity prices consisting of a base electricity price and a dispatch service fee. The user incentive module is used to reward users for participating in off-peak charging through points, which can be used to offset dispatch service fees.

7. The electric vehicle orderly charging system integrating load forecasting and dynamic price incentives according to claim 6, characterized in that, The load forecasting module collects grid operation data and electric vehicle charging data, performs day-ahead load forecasting using a long short-term memory neural network model, and outputs the next day's load curve; it also performs intraday load forecasting, using real-time monitoring data to continuously correct the day-ahead forecasting results to cope with random load fluctuations; and it links the load forecasting results with the distribution network real-time monitoring system, triggering an early warning when the predicted load exceeds a safety threshold. Among these features, day-ahead forecasting provides forward-looking planning for dispatch optimization, while intraday forecasting ensures the adaptability of dispatching strategies through real-time correction. The linkage between load forecasting and monitoring forms a forecasting-response closed loop.

8. The electric vehicle orderly charging system integrating load forecasting and dynamic price incentives according to claim 6, characterized in that, In the scheduling optimization module, electric vehicle charging scheduling is modeled as a multi-objective optimization problem, with objectives including minimizing grid load fluctuations and minimizing user charging costs. A whale optimization algorithm with multi-population automatic learning is used for solving the problem, improving search efficiency through integer encoding, Euclidean operator initialization, multi-population evolution, and automatic learning mechanisms. The Pareto optimal solution set is output as the charging scheduling scheme. The scheduling optimization module relies on load status data provided by the load forecasting module and considers dynamic electricity price signals generated by the electricity price scheduling module to balance grid security and user economics.

9. The electric vehicle orderly charging system integrating load forecasting and dynamic price incentives according to claim 6, characterized in that, In the electricity price dispatch module, the dynamic electricity price consists of a base price and a dispatch service fee, which includes a day-ahead dispatch service fee and a real-time price adjustment factor. The day-ahead dispatch service fee is based on the day-ahead load forecast curve and is released in advance; the real-time price correction factor is dynamically calculated based on the deviation between the real-time load and the forecast load to ensure that the price signal is flexibly adjusted according to the grid status; the end-user electricity price is the sum of the base price, the day-ahead dispatch service fee and the real-time price correction factor; among them, the electricity price dispatch module receives the forecast data from the load forecast module and generates price signals according to the dispatch optimization module's dispatch requirements to directly guide users' charging behavior.

10. The electric vehicle orderly charging system integrating load forecasting and dynamic price incentives according to claim 6, characterized in that, In the user incentive module, a grid-friendliness points system is introduced, where users earn points by charging during off-peak hours or discharging during peak hours. The points acquisition rules are linked to the grid load status, with more points earned per unit of charging when the load is lower. Points are directly used to offset the dispatch service fees incurred when users charge their devices. The user incentive module is closely coupled with the electricity price dispatch module. The points mechanism strengthens the guiding role of dynamic electricity prices on user behavior, forming a "contribution-reward" closed loop and increasing user participation.

11. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in claims 1-5.

12. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the method as described in claims 1-5.

Images