Method and system for electric vehicle virtual power plant to participate in power market scheduling

Abstract

The invention discloses a method and a system for an electric vehicle virtual power plant to participate in power market scheduling, and relates to the technical field of power system operation and scheduling. The method comprises the following steps: firstly, collecting multi-source information of electric vehicles, and dynamically dividing the multi-source information into three types of vehicles; predicting the scheduling capability of the two types of schedulable vehicles through a long-short-term memory network model; based on a conventional scheduling capability estimated value, a declared capacity is obtained by combining dynamic conservative coefficient reduction to participate in day-ahead bidding; and preferentially scheduling the common schedulable vehicle to fulfill the performance during real-time operation, and activating the emergency schedulable vehicle to complement energy through a request with a compensation price when a power gap meets an emergency triggering condition. According to the invention, collaborative optimization of scheduling reliability, economy and power grid security is realized, default risks are reduced, the resource utilization rate and operation benefits are improved, and reliable adjustment resources are provided for the power grid.

Description

Technical Field

[0001] This invention relates to the field of power system operation and dispatching technology, and in particular to a method and system for electric vehicle virtual power plants to participate in power market dispatching. Background Technology

[0002] With the rapid growth of electric vehicle (EV) ownership, their dispatch potential as mobile energy storage units has become a key direction for smart grid development. Aggregating dispersed EV resources through virtual power plant technology to participate in electricity market transactions and grid peak and frequency regulation is an important way to improve the absorption capacity of new energy sources and grid flexibility. However, existing technologies have significant drawbacks: Vehicle classification is crude, often using a binary division of "schedulable / unschedulable" without considering differences in user behavior, dynamic changes in vehicle status, and emergency scenario needs, resulting in low resource utilization. The lack of a risk management mechanism in market reporting means that reporting based on a single forecast value is prone to deviations in performance due to the randomness of user behavior, and may result in high penalties for default. The emergency response system is inadequate, lacking incentive mechanisms and triggering logic for rapid energy replenishment, making it difficult to cope with sudden power shortages in the power grid; The scheduling strategy was not fully integrated with multi-source data, resulting in insufficient prediction accuracy and scheduling adaptability.

[0003] Therefore, there is an urgent need for a virtual power plant scheduling method for electric vehicles that takes into account refined classification, controllable risk reporting, and emergency support, in order to solve the core pain points of existing technologies. Summary of the Invention

[0004] This invention aims to provide a hierarchical robust dispatching method and system for electric vehicle virtual power plants to participate in the electricity market. Through dynamic vehicle classification, intelligent predictive modeling, dynamic conservative reporting and emergency response closed loop, it achieves synergistic optimization of dispatching reliability, market economy and grid security, effectively reduces default risk and improves the utilization rate of electric vehicle resources and the operating efficiency of virtual power plants.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: This application provides a method for electric vehicle virtual power plants to participate in electricity market dispatch, including the following steps: Collect multi-source information on access electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify electric vehicles within the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles, and emergency power dispatchable vehicles. A pre-set prediction model is used to predict the conventional dispatching capacity of vehicles with ordinary power dispatchable and the emergency power dispatching capacity of vehicles with emergency power dispatchable within the target dispatching period, and the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity are obtained. Based on the prediction results, a conventional dispatch capacity estimate is selected and reduced by a dynamic conservative coefficient to obtain the final declared capacity, which is then used to participate in the day-ahead bidding in the electricity market; wherein, the dynamic conservative coefficient is related to the ratio of the conventional dispatch capacity estimate to the emergency power dispatch capacity estimate; In real-time operation, priority is given to dispatching vehicles with regular power availability to fulfill contracts; when a power shortage is detected that meets the preset emergency triggering conditions, the emergency response mechanism is activated, which sends a dispatch request with a compensation price to emergency power available vehicles and activates the vehicles that accept the request to make up for the power shortage.

[0006] Optionally, the classification rule is to determine the vehicle category by comprehensively considering at least one of the following factors: the user's active mode selection on the terminal, the vehicle's real-time state of charge and the user's set expected time to leave the vehicle, and the user profile generated based on the user's historical charging behavior.

[0007] Optionally, the prediction model is a long short-term memory network model, whose inputs include vehicle historical data, real-time status data, weather forecasts, and date type data after feature engineering.

[0008] Optionally, the value of the dynamic conservative coefficient is also dynamically adjusted based on at least one of the following factors: historical forecast error, weather forecast data, and electricity market price volatility.

[0009] Optionally, the emergency triggering condition is a first preset percentage threshold where the real-time schedulable power is lower than the contracted power for a first preset duration, or a second preset percentage threshold where the ultra-short-term predicted power is lower than the contracted power.

[0010] Optionally, the dynamic partitioning of parked electric vehicles within the target scheduling period includes: Based on a preset estimation module, the number of parked electric vehicles and their corresponding categories within the target scheduling period are estimated. The estimation module is trained based on historical data.

[0011] Optionally, the estimation module is a deep learning model.

[0012] This application provides a system for electric vehicle virtual power plants to participate in electricity market dispatch for implementing the above method, including: The collection module is used to collect multi-source information of the connected electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify the electric vehicles within the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles and emergency power dispatchable vehicles. The prediction module is used to predict the conventional dispatching capacity of vehicles with ordinary power dispatchable vehicles and the emergency power dispatching capacity of vehicles with emergency power dispatchable vehicles within the target dispatching period using a preset prediction model, and to obtain the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity. The application module is used to select a conventional dispatch capacity estimate based on the prediction results, and then reduce it using a dynamic conservative coefficient to obtain the final application capacity, thereby participating in the day-ahead bidding in the electricity market; wherein, the dynamic conservative coefficient is related to the ratio of the conventional dispatch capacity estimate to the emergency power dispatch capacity estimate; The dispatch module is used to prioritize dispatching vehicles with regular power availability to fulfill contracts during real-time operation; when a power shortage is detected that meets the preset emergency triggering conditions, the emergency response mechanism is activated by sending dispatch requests with compensation prices to emergency power available vehicles and activating the vehicles that accept the requests to make up for the power shortage.

[0013] This application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0014] This application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0015] In summary, the method for electric vehicle virtual power plants to participate in the power market dispatch provided in this application, through the construction of a three-tiered collaborative control system of "classification forecasting, conservative bidding, and emergency response," has produced the following significant beneficial effects: By refining the classification of electric vehicle resources and using a dynamic conservative coefficient to reduce the declared capacity, sufficient buffer capacity is reserved for system operation, fundamentally reducing the default risk caused by forecast uncertainty and greatly enhancing the ability of virtual power plants to fulfill market contracts. Under the premise of ensuring basic returns, emergency resources are mobilized through on-demand activation and compensated access, effectively balancing market returns and risk costs, overcoming the dilemma of "high returns accompanied by high risks" or "low risks leading to low returns" in traditional solutions. A complete management closed loop is formed from day-ahead market risk prevention to real-time operational risk response, significantly enhancing the ability of virtual power plants to cope with uncertainty and improving the robustness and stability of system operation. Through differentiated classification and incentive strategies, heterogeneous electric vehicle loads are transformed into hierarchically controllable dispatch resources, greatly improving resource utilization efficiency and aggregate value, and realizing the effective transformation of "load" into "resource." Making electric vehicle virtual power plants a reliable regulatory resource for the power grid provides a new technical approach to mitigate the fluctuations of new energy sources and enhance the grid's regulation capabilities, thus having significant social benefits. Attached Figure Description

[0016] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 This is a flowchart illustrating a method for a virtual power plant of an electric vehicle to participate in power market dispatch, provided in one embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the structure of an electric vehicle virtual power plant participating in the power market dispatch system provided in one embodiment of this application.

[0019] Figure 3 This is a schematic diagram of an electronic device structure provided in one embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Figure 1 This is a flowchart illustrating a method for a virtual power plant of an electric vehicle to participate in electricity market dispatch, provided in one embodiment of this application. Figure 1 As shown, the method includes the following:

[0022] Step S110: Collect multi-source information of the access electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify the electric vehicles in the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles and emergency power dispatchable vehicles. The virtual power plant control center collects multi-source information from all connected electric vehicles in real time through a data communication network. This information includes, but is not limited to: real-time vehicle status data (such as current state of charge (SOC), vehicle location, and connection status), user-defined data (such as desired departure time and charging mode selected on the mobile app or in-vehicle terminal), historical behavior data (such as charging start time and duration over a past period), and external environmental data (such as weather forecasts and date types).

[0023] Based on the above multi-source information, the system dynamically classifies electric vehicles within the target scheduling period (e.g., the next 24 hours) into three categories according to preset classification rules: Ordinary dispatchable electric vehicles (V1): These are vehicles whose charging behavior can accept optimized dispatch instructions from virtual power plants and participate in grid interaction within a specific time period. Typically, users of these vehicles have actively selected the "economic dispatch" mode, and their real-time SOC is within a suitable dispatch range (e.g., between 20% and 80%). At the same time, the user-set departure time has a sufficient time difference with the current time (e.g., greater than 4 hours) to ensure the feasibility of dispatch operations.

[0024] Regular charging vehicles (V2): These are vehicles for which users have set a specific charging completion time, or vehicles that must be charged immediately due to low battery levels (e.g., SOC below 20%) and are therefore unable to participate in the charging schedule. These vehicles are considered fixed loads that must be served, and their charging needs have the highest priority.

[0025] Emergency Power Dispatchable Vehicles (V3): These are vehicles that users are willing to provide backup capacity in the event of a grid emergency, but typically require substantial financial compensation. These users usually have proactively registered for the "emergency backup" service, and their vehicles currently have a high State of Charge (SOC) (e.g., above 50%), enabling them to discharge power to the grid (V2G). In non-emergency situations, these vehicles are treated as stationary loads and behave similarly to V2 vehicles.

[0026] This refined dynamic classification lays a solid foundation for subsequent differentiated scheduling and risk management.

[0027] Step S120: Use a preset prediction model to predict the conventional dispatching capacity of vehicles with ordinary power dispatchable and the emergency power dispatching capacity of vehicles with emergency power dispatchable within the target dispatching period, and obtain the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity. After classifying the vehicles, the system uses a pre-set prediction model to quantitatively predict the dispatching capacity of different types of vehicles in the next dispatching cycle. Specifically, the prediction model (e.g., using a machine learning model such as a Long Short-Term Memory network LSTM) will predict the regular dispatching capacity of ordinary power dispatchable vehicles (V1) and the emergency power dispatching capacity of emergency power dispatchable vehicles (V3).

[0028] The predictive model is input to multi-source information processed by feature engineering, including historical vehicle charging sequences, real-time State of Charge (SOC), weather forecast features (such as temperature and weather type coding), and date types (such as weekday and weekend coding). The model outputs two key estimates for a future period (e.g., a 24-hour period at 15-minute intervals): a regular dispatch capacity estimate (i.e., the total dispatchable power range available to the V1 vehicle cluster) and an emergency power dispatch capacity estimate (i.e., the total reserve power that the V3 vehicle cluster can activate in an emergency). This step transforms uncertain behavior into quantifiable dispatch resources, providing data support for market decision-making.

[0029] Step S130: Based on the prediction results, select the estimated value of conventional dispatch capacity and reduce it using a dynamic conservative coefficient to obtain the final declared capacity, and participate in the day-ahead bidding in the electricity market accordingly; wherein, the dynamic conservative coefficient is related to the ratio of the estimated value of conventional dispatch capacity to the estimated value of emergency power dispatch capacity; The core of this step is to adopt a proactive risk-averse strategy when participating in day-ahead bidding in the electricity market. Based on the prediction results of step S120, the system selects only the estimated conventional dispatch capacity as the base capacity for participating in market bidding. To offset prediction errors and uncertainties in real-time operation, the system introduces a dynamic conservative coefficient α (0 < α < 1) to reduce this base capacity, thereby obtaining a lower-risk final bid capacity (P_bid = α × estimated conventional dispatch capacity).

[0030] The key innovation lies in the determination logic of the dynamic conservative coefficient α. It is positively correlated with the ratio of the estimated conventional dispatch capacity to the estimated emergency power dispatch capacity. This design embodies profound systems thinking: when the V1 conventional resources are more abundant than the V3 emergency resources (larger ratio), it indicates a solid system foundation, allowing for relative optimism (larger α value); conversely, when the V3 emergency resources account for a higher proportion (smaller ratio), it suggests that the system's ability to cope with emergencies is relatively dependent on external incentives, requiring a more cautious bidding strategy (smaller α value). Based on this final declared capacity, virtual power plants generate bidding curves and participate in day-ahead bidding in the electricity market, thereby reducing the risk of default due to over-declaration at the source.

[0031] In step S140, during real-time operation, priority is given to dispatching ordinary power dispatchable vehicles to fulfill contracts; when a power shortage is detected to meet the preset emergency triggering conditions, the emergency response mechanism is activated by sending a dispatch request with a compensation price to the emergency power dispatchable vehicles and activating the vehicles that accept the request to make up for the power shortage.

[0032] During the real-time operation phase of the electricity market, virtual power plants prioritize dispatching ordinary dispatchable vehicles (V1) and strictly fulfill their contractual obligations in the day-ahead market by issuing charging and discharging control commands to these vehicles.

[0033] The system simultaneously monitors the deviation between the actual dispatchable power and the contracted power in real time. Once the system detects that the power gap meets the preset emergency trigger conditions (for example, the real-time dispatchable power is continuously lower than 95% of the contracted power for more than 5 minutes, or the future power gap based on ultra-short-term forecasts will exceed 3% of the contracted power), it will immediately determine it as an emergency and automatically activate the emergency response mechanism.

[0034] Once activated, the system sends a dispatch request to the user terminals (such as mobile apps) of all emergency power dispatchable vehicles (V3). This request explicitly includes the required power output, the dispatch duration, and an attractive compensation price. Users can choose whether to accept the dispatch based on the compensation price. The system then activates all V3 vehicles that accept the request, enabling them to quickly fill the power gap by stopping charging or discharging into the grid, thus reliably ensuring the fulfillment of contractual obligations and avoiding hefty penalties for breach of contract.

[0035] Through the coordinated operation of the above four steps, this invention constructs a complete robust scheduling closed loop from prediction and pricing to execution, effectively solving the core challenges faced by electric vehicle virtual power plants when participating in the electricity market.

[0036] In some embodiments, the classification rule is to determine the vehicle category by comprehensively considering at least one of the following factors: the user's active mode selection on the terminal, the vehicle's real-time state of charge and the user's set expected time to leave the vehicle, and the user profile generated based on the user's historical charging behavior.

[0037] Specifically, the vehicle classification rules adopt a multi-factor fusion decision-making mechanism, comprehensively considering information from the following dimensions to achieve accurate assessment and classification of the dispatch potential of electric vehicles: 1. Classification based on user-initiated mode selection; Users can actively choose their preferred mode of participation in virtual power plant dispatching via a mobile application or in-vehicle terminal. When a user selects "Economic Dispatch Mode," their vehicle is prioritized as a regular power dispatchable vehicle (V1); when a user selects "Immediate Charging Mode," it is classified as a regular charging vehicle (V2); and when a user selects "Emergency Standby Mode" and completes the relevant agreement signing, their vehicle will be classified as an emergency power dispatchable vehicle (V3). This classification method, based on the user's explicit wishes, respects the user's right to choose and provides a legal basis for subsequent differentiated dispatching.

[0038] 2. System verification based on vehicle status and user plans; Even if a user selects a certain mode, the system will still verify whether the vehicle's real-time status meets the basic requirements of that mode. Specifically, the system will comprehensively consider the vehicle's real-time state of charge (SOC) and the user's set expected departure time. For example, even if a vehicle is preset to Class V1, if its current SOC is below 20% (urgently needs charging) or the departure time is less than 1 hour (insufficient dispatch time), the system will temporarily adjust it to Class V2 to ensure that the user's basic travel needs are guaranteed. Conversely, for vehicles with a longer expected departure time (e.g., more than 4 hours) and a moderate SOC, the system can suggest that they participate in Class V1 dispatch even if no mode is actively selected.

[0039] 3. Intelligent supplementation based on user behavior profiles; For users who haven't explicitly selected a scheduling mode, the system builds user profiles by analyzing their historical charging behavior, enabling intelligent classification. The system continuously learns users' charging habits, including but not limited to: the regularity of charging times, average charging duration, and responsiveness to historical electricity price signals. Based on these characteristics, the system uses clustering algorithms to divide users into different groups. For example, highly regular commuter users (charging at fixed times daily for stable durations) are automatically classified into class V1; while users with more random charging behavior are tended to be classified into class V2 to control scheduling risks.

[0040] This multi-factor fusion classification mechanism, through the triple judgment of "user active selection - system status verification - behavior profile analysis", not only ensures the accuracy and reliability of the classification results, but also fully considers individual differences and real-time status changes, laying a solid foundation for building a high-quality schedulable resource pool.

[0041] In some embodiments, the prediction model is a long short-term memory network model, whose inputs include feature-engineered vehicle historical data, real-time status data, weather forecasts, and date-type data.

[0042] The system employs a Long Short-Term Memory (LSTM) network model as its core prediction tool, which was specifically chosen for its excellent time-series data processing capabilities. The detailed implementation of this prediction model is described below: 1. Input feature engineering processing; The input to the prediction model is carefully designed and processed multi-source feature data, mainly including: Vehicle historical data includes charging power sequences, SOC change trajectories, and charging start times for the same period over the past 7-30 days. This sequence data is regularized into fixed-length time window samples and normalized.

[0043] Real-time vehicle status data includes real-time parameters such as current SOC value, vehicle access status, and current charging power. This data has been standardized to eliminate the influence of dimensions.

[0044] Weather forecast data includes meteorological parameters such as temperature, humidity, and precipitation probability. These data are encoded as numerical features, and special identifiers are set for extreme weather conditions.

[0045] Date type data: including classification codes for weekdays, weekends, and statutory holidays, as well as special seasonal characteristics (such as summer cooling season and winter heating season indicators).

[0046] 2. LSTM model architecture design; The prediction model uses a multi-layer LSTM network structure, with the following specific configuration: Input layer: Receives 96-dimensional feature vectors (feature combinations corresponding to 24 hours × 4 15-minute time periods) after feature engineering.

[0047] Hidden layers: Contains a 3-layer LSTM network, each layer is equipped with 128 neurons, and Dropout regularization (dropout=0.2) is used between layers to prevent overfitting.

[0048] Output layer: Employs a fully connected layer to output two key forecast values ​​every 15 minutes for the next 24 hours: the total dispatchable power of ordinary power dispatchable vehicles and the total reserve capacity of emergency power dispatchable vehicles.

[0049] 3. Model training and optimization; The model training process employs the following strategies: Loss function: Use the Huber loss function, which is insensitive to outliers and can improve the robustness of the model.

[0050] Optimization algorithm: The Adam optimizer is used, the initial learning rate is set to 0.001, and a learning rate decay strategy is adopted.

[0051] Training data: Historical operational data from the past 12 months was used and divided into training, validation, and test sets in a 7:2:1 ratio; Early stopping mechanism: When the validation set loss no longer decreases for 10 consecutive epochs, training is automatically stopped to preserve the best model parameters; 4. Predictive output and uncertainty quantification; The model not only outputs point predictions, but also quantifies the uncertainty of the predictions using Monte Carlo Dropout technology: Dropout is enabled multiple times during the inference phase to perform random forward propagation; Calculate the prediction interval for each time point based on the distribution of multiple prediction results; The lower limit of the 95% confidence interval is used as a conservative scheduling reference, and the upper limit is used as an optimistic scheduling reference. This LSTM-based prediction framework effectively captures the long-term dependencies and periodic patterns of electric vehicle charging behavior. Furthermore, through robust feature engineering and uncertainty quantification, it provides reliable data support for subsequent market decisions. The model's rolling update mechanism (retraining every 24 hours) ensures its ability to adapt to the slow changes in charging behavior patterns.

[0052] In some embodiments, the value of the dynamic conservative coefficient is also dynamically adjusted based on at least one of the following factors: historical forecast error, weather forecast data, and electricity market price volatility.

[0053] The determination of the dynamic conservatism coefficient is a multi-factor driven adaptive process. Its value comprehensively considers the real-time changes in the internal and external environment of the system to achieve the optimal balance between risk and return. Specifically, the dynamic adjustment of this coefficient is based on one or more of the following factors: 1. A self-correction mechanism based on historical prediction errors; The system continuously tracks the deviation between historical forecasts and actual scheduling values, establishing a dynamic evaluation system for forecast accuracy. Specifically, it quantifies forecast reliability by calculating the mean absolute percentage error (MAPE) or root mean square error (RMSE) over the most recent scheduling periods (e.g., the past 7 days). When historical forecast errors remain consistently high, the system automatically lowers the conservatism coefficient to more cautiously address uncertainty; conversely, when forecast accuracy steadily improves, the coefficient is appropriately increased to maximize market gains. This self-correcting mechanism based on empirical data enables the system to continuously learn from operational experience and optimize decision-making quality.

[0054] 2. Adjustment of risk warnings based on weather forecast data; The system accesses authoritative weather forecast data and sets risk adjustment factors for specific weather conditions. For example, when extreme high temperatures (which may lead to a surge in air conditioning load) or heavy rain (which may affect travel plans and thus change charging behavior) are forecast, the system will activate a preventative adjustment strategy, significantly reducing the conservative coefficient value. This adjustment is based on statistical analysis of the volatility of charging behavior under similar weather conditions in historical data, enabling the system to take defensive measures in advance before weather-driven uncertainties increase.

[0055] 3. Market risk management based on electricity market price volatility; The system monitors daily and real-time market price fluctuations in real time and calculates price volatility for specific periods. When drastic changes in the market environment are detected and price volatility exceeds a preset threshold, it indicates increased market risk. At this time, the system automatically reduces the conservative coefficient and decreases the order volume to avoid potential risks caused by drastic price fluctuations. Conversely, during periods of stable market environment and steady price trends, the system tends to adopt a more aggressive coefficient strategy to capture more market opportunities.

[0056] In actual operation, the above factors collectively influence the final coefficient value through weighted fusion. The system assigns dynamic weights to different factors, with prediction error as the basic weight, weather factors as event-based weights, and market fluctuations as immediate weights. Through this multi-dimensional, adaptive decision-making framework, the dynamic conservative coefficient is no longer a fixed parameter but becomes an intelligent indicator reflecting the overall risk status of the system, ensuring that the virtual power plant maintains the optimal level of risk management in complex and ever-changing environments.

[0057] In some embodiments, the emergency triggering condition is a first preset percentage threshold where the real-time schedulable power is lower than the contracted power for a first preset duration, or a second preset percentage threshold where the ultra-short-term predicted power is lower than the contracted power.

[0058] The emergency triggering conditions are designed to fully consider both real-time operational anomaly detection and short-term future risk warnings, ensuring the timeliness and accuracy of emergency response through a dual-judgment mechanism (in some embodiments, emergency response may be ensured based on only one judgment mechanism). Specific implementation methods are as follows: 1. Emergency triggering based on real-time deviation; When the system detects that the real-time dispatchable power is lower than a first preset percentage threshold (e.g., 95%) of the contracted power, it does not immediately trigger an emergency response. Instead, it initiates a continuous monitoring timer. Only when this power deficit persists for a first preset duration (e.g., 5 minutes) will the emergency mechanism be formally triggered. This dual judgment method of "threshold + duration" effectively filters out false triggers caused by measurement noise or instantaneous fluctuations, ensuring the necessity of the emergency response. At the same time, the 5-minute duration provides dispatchers with a window for manual intervention, enhancing the system's controllability.

[0059] 2. Early Warning Triggering Based on Predicted Power Shortfall: To further enhance the system's foresight, this invention also establishes early warning triggering conditions based on ultra-short-term load forecasting. The system utilizes a rolling ultra-short-term forecasting algorithm (such as time series forecasting based on ARIMA or neural networks) to predict the schedulable power for the next 15-30 minutes. When the forecast indicates that the power for a future period will be lower than a second preset percentage threshold (e.g., 97%) of the contracted power, the system will trigger an emergency response in advance, even if the current real-time power still meets the requirements. This early warning mechanism buys valuable time for the V3 resource activation response process, including the time required for sending requests to users, user confirmation of responses, and equipment startup, thereby preventing the actual occurrence of power shortages.

[0060] 3. Adaptive Optimization of Triggering Conditions In practical applications, the aforementioned percentage thresholds and duration parameters are not fixed but are adaptively optimized based on system operating experience. The system records the effectiveness data of each emergency response, including indicators such as response latency and the adequacy of power replenishment, and periodically adjusts the parameter settings of the triggering conditions based on this data. For example, during periods with higher system reliability requirements, the percentage threshold will be appropriately increased (i.e., more sensitive); while during periods with slower user response times, the warning lead time will be appropriately extended.

[0061] This carefully designed emergency triggering condition not only ensures the system's rapid response capability when a real power shortage occurs, but also provides proactive protection for the system through a predictive early warning mechanism. At the same time, it avoids malfunctions caused by system fluctuations, comprehensively improving the reliability and economy of virtual power plant operation.

[0062] In some embodiments, the dynamic classification of parked electric vehicles within a target scheduling period includes: estimating the parked electric vehicles and their corresponding categories within the target scheduling period based on a preset estimation module; wherein the estimation module is trained based on historical data. The estimation module is a deep learning model.

[0063] The system uses an estimation module trained on historical data to achieve refined classification of electric vehicles that are out of service during the target scheduling period. This module, as the core of the classification decision, is implemented as follows: The estimation module employs a deep learning model architecture, specifically a multi-task learning framework. This framework includes a shared feature extraction layer and multiple dedicated output layers, enabling it to simultaneously perform two related tasks: vehicle presence prediction and category determination. The main structure of the model includes: Input layer: Receives standardized multidimensional feature vectors with a dimension of 256; Feature extraction layer: A 3-layer fully connected network is used, with 512 neurons in each layer, and the ReLU activation function is used; Multi-task output layer: There is a prediction branch: outputs the probability that a vehicle will be in a stopped state at each time period within the target scheduling cycle; Category determination branch: Outputs the probability distribution of vehicles belonging to each scheduling category; 2. Training Data Construction and Feature Engineering: The training data for the model comes from massive amounts of data accumulated from historical operations, and has undergone a rigorous cleaning and labeling process. The feature dimensions include vehicle static attributes (vehicle model, battery capacity), user historical behavior features (distribution of charging time periods, average charging time, and dispatch command response rate over the past 30 days), real-time status features (current SOC, access location), environmental features (date type, season marker), and user preference features (extracted through a latent semantic model). The sample labels are based on historical actual scheduling records and have been reviewed and confirmed by operations experts to ensure the accuracy of the labels; Data augmentation employs time-series sliding window technology to increase the diversity of training samples; 3. The training process of the model training and optimization policy estimation module adopts a staged optimization strategy: The first stage of pre-training: using massive amounts of historical data, with the goal of minimizing cross-entropy loss, to train basic classification capabilities; The second phase of fine-tuning: Based on recent data, a course learning strategy is adopted to adjust the difficulty of training samples from easy to difficult. The third phase of online learning involves continuously receiving feedback data after deployment and fine-tuning the model parameters weekly. 4. Dynamic Reasoning and Uncertainty Management In the actual reasoning process, the system employs integrated prediction and uncertainty quantification techniques: The probability distribution of the prediction results is obtained by sampling multiple times using Monte Carlo Dropout. Set a confidence threshold; when the prediction uncertainty is high, classify the vehicle into the default category with lower risk. For boundary cases, initiate a rule-based post-processing workflow to ensure the reasonableness of the classification results; This classification scheme, based on a deep learning estimation module, can automatically learn effective discrimination patterns from complex high-dimensional features, exhibiting stronger adaptability and accuracy compared to classification methods based on fixed rules. Through continuous learning and online optimization, this module can track the slow changes in user behavior patterns, maintaining long-term effective classification performance and providing a reliable resource evaluation basis for the entire scheduling system.

[0064] The collection module is used to collect multi-source information of the connected electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify the electric vehicles within the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles and emergency power dispatchable vehicles. The prediction module is used to predict the conventional dispatching capacity of vehicles with ordinary power dispatchable vehicles and the emergency power dispatching capacity of vehicles with emergency power dispatchable vehicles within the target dispatching period using a preset prediction model, and to obtain the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity. The application module is used to select a conventional dispatch capacity estimate based on the prediction results, and then reduce it using a dynamic conservative coefficient to obtain the final application capacity, thereby participating in the day-ahead bidding in the electricity market; wherein, the dynamic conservative coefficient is related to the ratio of the conventional dispatch capacity estimate to the emergency power dispatch capacity estimate; The dispatch module is used to prioritize dispatching vehicles with regular power availability to fulfill contracts during real-time operation; when a power shortage is detected that meets the preset emergency triggering conditions, the emergency response mechanism is activated by sending dispatch requests with compensation prices to emergency power available vehicles and activating the vehicles that accept the requests to make up for the power shortage.

[0065] The system embodiments of this application can be used to execute the method embodiments of this application. For details not disclosed in the system embodiments of this application, please refer to the method embodiments of this application.

[0066] Figure 2 The diagram shown is a block diagram of an electric vehicle virtual power plant participating in the power market dispatch system according to an embodiment of this application. Figure 2 As shown, the device includes: The collection module 21 is used to collect multi-source information of the access electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify the electric vehicles within the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles and emergency power dispatchable vehicles. The prediction module 22 is used to predict the conventional dispatching capacity of ordinary power dispatchable vehicles and the emergency power dispatching capacity of emergency power dispatchable vehicles within the target dispatching period using a preset prediction model, and to obtain the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity. The application module 23 is used to select the estimated value of conventional dispatch capacity based on the prediction results, and reduce it by a dynamic conservative coefficient to obtain the final application capacity, and participate in the day-ahead bidding in the power market accordingly; wherein, the dynamic conservative coefficient is related to the ratio of the estimated value of conventional dispatch capacity to the estimated value of emergency power dispatch capacity; The scheduling module 24 is used to prioritize scheduling ordinary power dispatchable vehicles to fulfill contracts during real-time operation; when a power gap is detected that meets the preset emergency triggering conditions, the emergency response mechanism is activated by sending a scheduling request with a compensation price to the emergency power dispatchable vehicles and activating the vehicles that accept the request to make up for the power gap.

[0067] Below, for reference Figure 3 This describes an electronic device according to embodiments of the present application. Figure 3 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0068] like Figure 3 As shown, the electronic device 300 includes one or more processors 310 and memory 320.

[0069] The processor 310 may be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device 300 to perform desired functions.

[0070] The memory 320 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 310 may execute the program instructions to implement the electric vehicle virtual power plant participation in electricity market dispatching methods and / or other desired functions described in the various embodiments of this application above. Various contents, such as category correspondences, may also be stored in the computer-readable storage medium.

[0071] In one example, the electronic device 300 may also include an input device 330 and an output device 340, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0072] In addition, the input device 330 may also include, for example, a keyboard, mouse, interface, etc. The output device 340 can output various information to the outside, including analysis results, etc. The output device 340 may include, for example, a display, speaker, printer, and communication network and its connected remote output devices, etc.

[0073] Of course, for the sake of simplicity, Figure 3Only some of the components of the electronic device relevant to this application are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device may include any other suitable components depending on the specific application.

[0074] In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the electric vehicle virtual power plant participation in electricity market dispatch methods according to various embodiments of this application as described in the "Exemplary Methods" section above.

[0075] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0076] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the electric vehicle virtual power plant participation in electricity market dispatch method according to various embodiments of this application as described in the "Exemplary Methods" section above.

[0077] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0078] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for electric vehicle virtual power plants to participate in electricity market dispatch, characterized in that, Includes the following steps: Collect multi-source information on access electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify electric vehicles within the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles, and emergency power dispatchable vehicles. A pre-set prediction model is used to predict the conventional dispatching capacity of vehicles with ordinary power dispatchable and the emergency power dispatching capacity of vehicles with emergency power dispatchable within the target dispatching period, and the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity are obtained. Based on the prediction results, a conventional dispatch capacity estimate is selected and reduced by a dynamic conservative coefficient to obtain the final declared capacity, which is then used to participate in the day-ahead bidding in the electricity market; wherein, the dynamic conservative coefficient is related to the ratio of the conventional dispatch capacity estimate to the emergency power dispatch capacity estimate; In real-time operation, priority is given to dispatching vehicles with regular power availability to fulfill contracts; when a power shortage is detected that meets the preset emergency triggering conditions, the emergency response mechanism is activated, which sends a dispatch request with a compensation price to emergency power available vehicles and activates the vehicles that accept the request to make up for the power shortage.

2. The method according to claim 1, characterized in that, The classification rule is based on at least one of the following factors to determine the vehicle category: the user's active mode selection on the terminal, the vehicle's real-time state of charge and the user's set expected time to leave the vehicle, and the user profile generated based on the user's historical charging behavior.

3. The method according to claim 1, characterized in that, The prediction model is a long short-term memory network model, and its inputs include vehicle historical data, real-time status data, weather forecasts, and date-type data after feature engineering.

4. The method according to claim 1, characterized in that, The value of the dynamic conservative coefficient is also dynamically adjusted based on at least one of the following factors: historical forecast error, weather forecast data, and electricity market price volatility.

5. The method according to claim 1, characterized in that, The emergency triggering condition is either a first preset percentage threshold where the real-time schedulable power is lower than the contracted power for a first preset duration, or a second preset percentage threshold where the ultra-short-term predicted power is lower than the contracted power.

6. The method according to claim 1, characterized in that, The dynamic partitioning of parked electric vehicles within the target scheduling period includes: Based on a preset estimation module, the number of parked electric vehicles and their corresponding categories within the target scheduling period are estimated. The estimation module is trained based on historical data.

7. The method according to claim 6, characterized in that, The estimation module is a deep learning model.

8. A system for electric vehicle virtual power plants participating in electricity market dispatch to implement the method of any one of claims 1 to 7, characterized in that, include: The collection module is used to collect multi-source information of the connected electric vehicles, and based on the multi-source information and preset classification rules, dynamically classify the electric vehicles within the target scheduling period into ordinary power dispatchable vehicles, ordinary charging vehicles and emergency power dispatchable vehicles. The prediction module is used to predict the conventional dispatching capacity of vehicles with ordinary power dispatchable vehicles and the emergency power dispatching capacity of vehicles with emergency power dispatchable vehicles within the target dispatching period using a preset prediction model, and to obtain the estimated values ​​of conventional dispatching capacity and emergency power dispatching capacity. The application module is used to select a conventional dispatch capacity estimate based on the prediction results, and then reduce it using a dynamic conservative coefficient to obtain the final application capacity, thereby participating in the day-ahead bidding in the electricity market; wherein, the dynamic conservative coefficient is related to the ratio of the conventional dispatch capacity estimate to the emergency power dispatch capacity estimate; The scheduling module is used to prioritize scheduling vehicles with regular power availability to fulfill contracts during real-time operation; when a power shortage is detected that meets the preset emergency triggering conditions, the emergency response mechanism is activated by sending a scheduling request with a compensation price to emergency power available vehicles and activating the vehicles that accept the request to make up for the power shortage.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images