A method for a charging station operator to participate in market bidding based on artificial intelligence

Abstract

The present application relates to the technical field of power system load management and market declaration control, and particularly relates to a method for market bidding of charging station operators based on artificial intelligence. The method comprises the following steps: receiving site archives, historical charging curves, user charging habit data, user constraint data and real-time state data of multiple charging stations, generating charging station label records and interruption tolerance constraint records; combining external prediction data and power grid load information, generating adjustable load prediction curves according to transaction time periods and interruption tolerance levels; reading market rules, income parameters and deviation assessment parameters of multiple market channels, determining main market channel records; generating quantity and price declaration combinations and site allocation records, retracting target charging station bidding quantities when queuing exceeds limits and making up the retraction difference, and generating final declaration forms. The present application keeps the association of site allocation, retraction sources and supplement sources, and effectively improves the coherence of aggregated market declaration and site real-time state.

Description

Technical Field

[0001] This invention relates to the field of power system load management and power market declaration control technology, and in particular to a method for charging station operators to participate in market bidding based on artificial intelligence. Background Technology

[0002] In the field of power system load management and electricity market bidding control technology, existing solutions for charging station operators to participate in market bidding typically receive historical charging data, real-time electricity price information, weather forecasts, grid load information, charging station capacity, and user charging habit data. These data are then combined with charging demand forecasting, a two-tier bidding model, market clearing rules, or in-station load management to generate bidding records or scheduling results. While such solutions can handle basic processing such as charging demand forecasting, market bid generation, and charging pile power adjustment in conventional scenarios, they are prone to limitations in scenarios where multiple charging stations aggregate to participate in market bidding, including insufficient connection between user constraints and adjustable load records, asynchronous station allocation results and station-end real-time status, and discontinuity between bidding records and performance status records.

[0003] Existing solutions often rely on unified load forecasting results, fixed charging strategy classifications, or single market channel rules for processing. When changes occur in charging station queuing status, user waiting constraints, new energy output forecasting data, or market deviation assessment parameters, there is a lack of stable connection between charging station load forecasting, quantity and price declaration combinations, and site allocation records. This can easily lead to difficulties such as difficulty in locating the source of bidding volume, difficulty in timely mapping queuing over-limit status to market declaration records, and lack of consistent constraints in supplementing site selection.

[0004] For the joint processing of site profiles, historical charging curves, user constraint data, adjustable load forecast curves, market channel parameter records, and real-time queuing status of multiple charging stations, existing technologies still have common shortcomings in constraint recording, application allocation, queuing status determination, and final application generation. Therefore, it is necessary to address the issue of transforming real-time queuing status into bid volume reduction and supplementation to form the final application, based on interruption tolerance constraint records and site allocation records, within the context of the charging station operator's aggregated market application scenario. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for charging station operators to participate in market bidding based on artificial intelligence, comprising:

[0006] S100: Receives site files, historical charging curves, user charging habit data, user constraint data, and real-time status data from multiple charging stations; binds these data according to charging station type, charging pile time period information, historical load characteristics, and user constraint data; and generates charging station tag records and interruption tolerance constraint records. The interruption tolerance constraint records include interruption tolerance level, maximum allowable waiting time, and user compensation cost parameters.

[0007] S200. Based on the charging station tag records and the interruption tolerance constraint records, generate an adjustable load forecast curve according to the transaction period and interruption tolerance level;

[0008] S300. Based on the adjustable load forecast curve, read the market rules, revenue parameters and deviation assessment parameters of multiple market channels, generate market channel parameter records and determine the main investment market channel record.

[0009] S400. Based on the main investment market channel records, optimize the solution according to the adjustable load, user compensation cost parameters and deviation assessment expectations of each interruption tolerance level, and generate volume and price declaration combinations and site allocation records.

[0010] S500: Based on the site allocation record, receive the real-time queuing status of each charging station, compare the real-time queuing status with the maximum allowed waiting time, reduce the bid amount of the target charging station when a queuing limit exceedance event is generated, and generate a final application form that binds the site allocation record and the reduction and supplementation record.

[0011] First, select charging stations with the same tags as the target charging station. These same tags include the same charging station type, the same or matching public charging pile attributes, private charging pile attributes, battery swapping attributes, operational vehicle-specific attributes, and the same interruption tolerance level. Eliminate charging stations that have already generated queue overrun events. Then, read the adjustable load margin of the candidate charging stations, the power adjustment range in the market channel parameter record, and the response speed requirements to determine the replenishment amount. If the adjustable load margin of the replenishment charging stations with the same tags is less than the reduction difference, then determine the replenishment amount based on the charging station tag record, selecting charging stations corresponding to adjacent interruption tolerance levels. When the replenishment amount equals the reduction difference, the replenishment result is written to the completion status. When the replenishment amount is less than the reduction difference, a reduction application record or a performance risk record is generated.

[0012] Furthermore, the charging station tag record includes the station identifier, fast charging ratio, slow charging ratio, battery swapping attribute, public charging pile attribute, private charging pile attribute, operating vehicle exclusive attribute, rated power, number of available charging piles, charging pile time period information, and historical load characteristics;

[0013] The binding includes establishing an association between the charging pile time period information, historical load characteristics, and user constraint data under the site identifier.

[0014] Furthermore, the interruption tolerance constraint record also includes the maximum interruptible duration, the maximum power reduction ratio, the minimum reserved charging power, the historical cancellation rate, and the historical response success rate;

[0015] The interruption tolerance level is generated based on the historical cancellation rate, the historical response success rate, and the historical queuing status, and is updated based on the execution feedback records.

[0016] Furthermore, the process of generating adjustable load forecast curves includes data access mapping, load splitting, and confidence labeling.

[0017] Furthermore, the data access mapping process includes: receiving external inputs by the adjustable load forecasting module, and writing data source, forecast release time, forecast period, and version identifier for each type of input; when the forecast time granularity of different sources is inconsistent, the external inputs are mapped to a specific time period based on the trading period specified in the market rules; the weather forecast includes temperature, precipitation, wind speed, and weather conditions; the surrounding POI event data includes activities in the surrounding business districts of the charging station, passenger flow events at transportation hubs, shuttle bus times in the park, and activity records in public parking areas; the new energy output forecast data includes one or more of wind power output forecasts, photovoltaic power output forecasts, and distributed new energy output forecasts.

[0018] Furthermore, the load splitting process includes: first, dividing multiple charging stations into transaction periods according to the station identifier and charging pile time period information in the charging station tag record; then, splitting the charging load in each transaction period into load segments corresponding to different levels according to the interruption tolerance level; subsequently, mapping the weather forecast, holiday calendar, and surrounding POI event data to each transaction period to generate the baseline load for the corresponding transaction period; the adjustable load prediction curve includes the transaction period, level identifier, baseline load, adjustable load, delayable load, interruptible load, and confidence interval; when generating the adjustable load, reading the maximum power reduction ratio and the minimum reserved charging power, first calculating the difference between the baseline load corresponding to the level and the minimum reserved charging power, and then limiting the difference according to the maximum power reduction ratio; when generating the delayable load, reading the reservation period, queuing period, and the maximum allowed waiting time, when the reservation period overlaps with the transaction period, only the charging demand that can be postponed within the maximum allowed waiting time is written into the delayable load; when generating the interruptible load, reading the maximum interruptible duration and the historical response success rate.

[0019] Furthermore, the confidence level marking process includes: the confidence interval is generated jointly by the prediction deviation records of the same historical transaction period and the real-time status data; for stations with a historical response success rate lower than the rule configuration value, a low confidence mark is written into the interruptible load; if the real-time status data in the tag record of a certain charging station is a communication anomaly, then the adjustable load margin of that station in the current transaction period is written into the low confidence state; if the external prediction data is missing a certain transaction period, then the data of the same transaction period in the previous version is read, and a version continuation mark is written into the adjustable load prediction curve.

[0020] Furthermore, generating the final application form that binds the site allocation record and the backsliding and replenishment record also includes: queuing status monitoring processing, backsliding amount calculation processing, replenishment site screening processing, and application form output processing;

[0021] The queuing status monitoring process includes: the queuing status monitoring module matching the real-time queuing status with the site identifier and transaction period in the site allocation record to form a site queuing status record; the real-time queuing status includes the current number of vehicles in the queue, average waiting time, longest waiting time, number of available charging piles, and real-time power; comparing the average waiting time or longest waiting time with the maximum allowed waiting time in the interruption tolerance constraint record, and generating a queuing limit violation event when the average waiting time is greater than the maximum allowed waiting time or the longest waiting time is greater than the maximum allowed waiting time; if the waiting time field in the real-time queuing status is missing, the current number of vehicles in the queue, the number of available charging piles, and the average service time of the same historical period are read to generate an estimated waiting time, and an estimated flag is written into the queuing limit violation event.

[0022] Furthermore, the shrinkage amount calculation process includes: the shrinkage compensation module reads the allocated bid amount of the target charging station in the corresponding transaction period, and then reads the adjustable load margin of the target charging station; when the adjustable load margin is less than the allocated bid amount, or when a queue over-limit event has been generated, the difference between the allocated bid amount and the adjustable load margin is written into the shrinkage amount; when the maximum allowable waiting time corresponding to the queue over-limit event exceeds the limit by a large margin, the shrinkage amount is set to the total allocated bid amount of the target charging station in that transaction period.

[0023] The supplementary charging station screening process includes: first, selecting charging stations with the same tags as the target charging station, where the same tags include the same charging station type, the same or matching public charging pile attributes, private charging pile attributes, battery swapping attributes, operational vehicle-specific attributes, and the same interruption tolerance level; eliminating charging stations that have already generated queuing over-limit events, and then reading the adjustable load margin of the candidate charging stations, the power adjustment range and response speed requirements in the market channel parameter record to determine the supplementary amount; if the adjustable load margin of the supplementary charging stations with the same tags is less than the reduction difference, then based on the charging station tag record, selecting charging stations corresponding to adjacent interruption tolerance levels to determine the supplementary amount; when the supplementary amount equals the reduction difference, the supplementary result is written to the completion status; when the supplementary amount is less than the reduction difference, a reduction application record or a performance risk record is generated.

[0024] Furthermore, the declaration form output processing includes: if the market declaration window is not yet closed, updating the final declaration form according to the reduction declaration record; if the market declaration window is already closed, writing the performance risk record into the risk status field of the final declaration form; the final declaration form includes market channel, trading period, declared electricity or capacity, declared price, site allocation record, pullback and replenishment record, market rule version, tag version, and model version, and the pullback and replenishment record includes target site, queue over-limit event, pullback amount, replenishment site, replenishment amount, replenishment result, and risk status.

[0025] The key innovations of this invention include:

[0026] (1) Bind the site files, historical charging curves, user charging habit data, user constraint data and real-time status data of multiple charging stations according to the charging station type, charging pile time period information and historical load characteristics to form charging station tag records and interruption tolerance constraint records containing interruption tolerance level, maximum allowable waiting time and user compensation cost parameters.

[0027] (2) Using the charging station tag record and the interruption tolerance constraint record as a common index, weather forecasts, holiday calendars, surrounding POI event data, new energy output forecast data and grid load information are mapped to trading periods and interruption tolerance levels to form an adjustable load forecast curve. Market channel parameter records and main investment market channel records are generated in combination with market rules, revenue parameters and deviation assessment parameters.

[0028] (3) The main investment market channel record, the adjustable load forecast curve and the interruption tolerance constraint record are introduced into the quantity and price declaration combination generation process to form a site allocation record; when the real-time queuing status exceeds the maximum allowable waiting time, the bidding quantity of the target charging station is reduced based on the site allocation record, and the supplementary charging station with the same label is selected based on the charging station label record to supplement the reduction difference, and a final declaration form binding the site allocation record and the reduction supplementary record is generated.

[0029] The following are its main beneficial effects:

[0030] (1) To address the problem of insufficient connection between user constraints and adjustable load records in the existing scheme, the binding of charging station tag records and interruption tolerance constraint records is used to make the site attributes, historical load characteristics, maximum allowable waiting time and user compensation cost parameters form constraint records under the same site identifier, which facilitates subsequent adjustable load prediction and application allocation.

[0031] (2) To address the issue of discontinuous connection between the unified load forecast results and the multi-market reporting rules in the existing scheme, an adjustable load forecast curve is generated according to the trading period and the interruption tolerance level. The main investment market channel record is determined by combining the market channel parameter record, so that the adjustable load, market rules, revenue parameters and deviation assessment parameters form a corresponding relationship before the quantity and price are reported.

[0032] (3) To address the issue of the inconsistency between the site allocation results and the real-time queuing status at the site in the existing scheme, the real-time queuing status is mapped to the reduction and supplementation record in the final application form through queuing limit exceeding event, bid volume reduction and same label supplementation processing, so that the declared electricity or capacity, site allocation source, reduction source and supplementation source are kept related in the same application record. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating a method for a charging station operator to participate in market bidding based on artificial intelligence, as provided in an embodiment of this application. Detailed Implementation

[0034] Example 1: Refer to Figure 1 This is a flowchart illustrating a method for charging station operators to participate in market bidding based on artificial intelligence, according to an embodiment of the present invention. The process may include at least steps S100-S500:

[0035] S100: Receives site files, historical charging curves, user charging habit data, user constraint data, and real-time status data of multiple charging stations, binds them according to charging station type, charging pile time period information, historical load characteristics, and user constraint data, and generates charging station tag records and interruption tolerance constraint records. The interruption tolerance constraint records include interruption tolerance level, maximum allowable waiting time, and user compensation cost parameters.

[0036] S200: Based on the charging station tag record and the interruption tolerance constraint record, receive weather forecasts, holiday calendars, surrounding POI event data, new energy output forecast data and grid load information, and generate an adjustable load forecast curve according to the transaction period and interruption tolerance level;

[0037] S300. Based on the adjustable load forecast curve, read the market rules, revenue parameters and deviation assessment parameters of multiple market channels, generate market channel parameter records and determine the main investment market channel record.

[0038] S400. Based on the main investment market channel record, the adjustable load forecast curve and the interruption tolerance constraint record, optimize the solution according to the adjustable load amount, user compensation cost parameters and deviation assessment expectations of each interruption tolerance level, and generate the quantity and price declaration combination and site allocation record.

[0039] S500: Receive the real-time queuing status of each charging station based on the site allocation record, compare the real-time queuing status with the maximum allowed waiting time, reduce the bid amount of the target charging station when a queuing limit exceedance event is generated, and filter the supplementary charging stations with the same tags based on the charging station tag record to supplement the reduction difference, and generate a final application form that binds the site allocation record and the reduction supplement record.

[0040] S100 receives site files, historical charging curves, user charging habit data, user constraint data, and real-time status data from multiple charging stations. It binds these data according to charging station type, charging pile time period information, historical load characteristics, and user constraint data, generating charging station tag records and interruption tolerance constraint records.

[0041] S100 is executed by the station-side data access module, tag generation module, and interruption tolerance binding module in the aggregated bidding server. The station-side data access module is connected to the charging pile management system, charging station operation management system, charging service platform, and station-side monitoring equipment; the tag generation module receives standardized fields generated by the station-side data access module; and the interruption tolerance binding module generates interruption tolerance constraint records based on user constraint data and historical status records. The multiple charging stations constitute a set of charging stations participating in the same application cycle within the management scope of the charging station operator; each charging station uses its site identifier as the merge key, and the site identifier remains unique within the same application cycle.

[0042] The site profile includes site identifier, access area, charging station type, rated power, number of available charging piles, battery swapping attribute, public charging pile attribute, private charging pile attribute, and vehicle-specific attribute. The historical charging curve includes charging power, charging volume, number of charging sessions, and power change sequence at each time point within a historical trading day. The user charging habit data includes user arrival time, charging duration, reservation time, queuing time, charging request power, and historical cancellation records. The user constraint data includes maximum allowed waiting time, maximum interruptible time, maximum power reduction ratio, minimum reserved charging power, and user compensation cost parameters. The real-time status data includes the number of currently occupied charging piles, the number of available charging piles, real-time power, station communication status, and the number of vehicles currently in the queue. The above inputs can be generated by different systems. When the aggregated bidding server receives the inputs, it generates access records according to the site identifier, sampling time, and data source, retaining the original fields and access time.

[0043] In S100, the charging station type is used to distinguish between fast charging stations, slow charging stations, battery swapping stations, public charging stations, private charging stations, and stations dedicated to commercial vehicles; the charging pile time period information indicates the availability, reservation status, and historical operating status of the charging pile in different transaction periods; the historical load characteristics are extracted from the historical charging curve by transaction period, including average charging power, peak power, off-peak power, power variation amplitude, and historical adjustable load margin. The tag generation module first matches the station identifier in the station file with the station identifier in the historical charging curve; then it establishes a correlation between the matched historical load characteristics and the charging pile time period information; subsequently, it writes the user constraint data under the same station identifier to form a charging station tag record.

[0044] The charging station tag record includes station identifier, fast charging ratio, slow charging ratio, battery swapping attribute, public charging pile attribute, private charging pile attribute, vehicle-specific attribute, rated power, number of available charging piles, charging pile time period information, and historical load characteristics. In one implementation, the fast charging ratio is the ratio of the total rated power of fast charging piles to the total rated power of all charging piles in the station; the slow charging ratio is the ratio of the total rated power of slow charging piles to the total rated power of all charging piles in the station; the battery swapping attribute is recorded based on whether the station provides battery swapping services; the public charging pile attribute, private charging pile attribute, and vehicle-specific attribute are recorded using the station's service object field. The tag generation module does not change the original content of each charging station's station file, but rather creates tag records under the station identifier that can be called by subsequent transaction periods.

[0045] The interruption tolerance constraint record is indexed by the interruption tolerance level and records the maximum allowable waiting time, maximum interruptible duration, maximum power reduction ratio, minimum reserved charging power, user compensation cost parameters, historical cancellation rate, and historical response success rate. The interruption tolerance binding module reads user constraint data and historical queuing status, mapping the maximum allowable waiting time to a waiting constraint field; the maximum interruptible duration to a continuous interruption constraint field; the maximum power reduction ratio and minimum reserved charging power to power adjustment constraint fields; and the user compensation cost parameters to a reporting cost side field. The historical cancellation rate is calculated from the number of queue exits and the number of queue sessions in the historical queuing status; the historical response success rate is calculated from the number of completions and the number of triggers in the demand response execution record.

[0046] In one implementation, the interruption tolerance levels include Level 1, Level 2, and Level 3. Level 1 corresponds to a shorter maximum allowable waiting time and a lower maximum power reduction ratio, while Level 3 corresponds to a longer maximum allowable waiting time and a higher maximum power reduction ratio. The level names serve only as field identifiers, and the correspondence between levels and specific thresholds is maintained by a rule configuration table. In another implementation, the interruption tolerance levels are generated based on the station's service objects and user constraint data. For example, the maximum allowable waiting time for dedicated stations for operational vehicles is given by the operational shift time and reservation period, while the maximum allowable waiting time for public charging stations is given by historical queuing status and historical cancellation rate.

[0047] S100 is activated when the market declaration window opens, site profiles are updated, user constraint data is updated, or real-time status data arrives. If a charging station lacks historical charging curves, the tag generation module retains the site profile and real-time status data, and writes a historical load feature missing marker into the charging station tag record. If the user constraint data lacks the maximum allowable waiting time, the interruption tolerance binding module reads the most recent valid interruption tolerance constraint record for that site. If there are conflicts between site identifiers from different sources, the site-side data access module writes the conflict record into the abnormal access record and temporarily does not include that site in the site set for this round of declaration. The above abnormal access records are saved together with the original input records for subsequent manual review or re-access in the next declaration cycle.

[0048] In one implementation scenario, the aggregated bidding server receives site profiles for stations A, B, and C; station A is a public fast-charging station, station B is a private slow-charging station, and station C is a dedicated station for commercial vehicles. The station-side data access module receives historical charging curves from the past 30 days and real-time status data for the current day from the three stations. The tag generation module generates three charging station tag records based on the charging time period information and historical load characteristics of each station. The interruption tolerance binding module generates corresponding interruption tolerance constraint records based on the historical cancellation rate, historical response success rate, and user constraint data of each station. The maximum allowed waiting time is recorded as 15 minutes for station A, 45 minutes for station B, and 10 minutes for station C. These fields are used as inputs for generating adjustable load forecast curves in S200, as a reference source for user compensation cost parameters in S400, and as a comparison basis for queuing limit exceeding events in S500.

[0049] S100 ultimately outputs the charging station tag record and the interruption tolerance constraint record. The site identifier, charging station type, charging pile time period information, and historical load characteristics in the charging station tag record are entered into S200, serving as the grouping basis for generating adjustable load forecast curves by transaction time period. The interruption tolerance level, maximum allowable waiting time, and user compensation cost parameters in the interruption tolerance constraint record are simultaneously entered into S200, S400, and S500, participating in the generation of adjustable load forecast curves, the generation of quantity-price declaration combinations, and the comparison of queue over-limit events in subsequent steps. The execution feedback record, after being generated in subsequent steps, can be returned to S100 to update the historical response success rate and user compensation cost parameters of the corresponding site.

[0050] S200: Based on the charging station tag records and the interruption tolerance constraint records, receive weather forecasts, holiday calendars, surrounding POI event data, new energy output forecast data and grid load information, and generate adjustable load forecast curves according to the transaction period and interruption tolerance level.

[0051] S200 is executed by the adjustable load forecasting module in the aggregated bidding server. The adjustable load forecasting module receives charging station tag records and interruption tolerance constraint records output from S100, and receives weather forecasts, holiday calendars, surrounding POI event data, renewable energy output forecast data, and grid load information from external data interfaces. In the surrounding POI event data, POI stands for Point of Interest; in this step, the surrounding POI event data includes activity records in the surrounding commercial area, passenger flow events at transportation hubs, shuttle bus schedules within the park, and activity records in public parking areas.

[0052] The weather forecast includes temperature, precipitation, wind speed, and weather conditions corresponding to the trading period; the holiday calendar includes weekdays, weekends, statutory holidays, and adjusted workdays; the renewable energy output forecast data includes one or more of wind power output forecasts, photovoltaic power output forecasts, and distributed renewable energy output forecasts; the grid load information includes regional base load, distribution area load, or distribution network node load. When receiving external input, the adjustable load forecasting module writes the data source, forecast release time, forecast period, and version identifier for each type of input; when the forecast time granularity of different sources is inconsistent, the adjustable load forecasting module uses the trading period specified in the market rules as a benchmark to perform time period mapping on the external input.

[0053] The adjustable load forecast curve is a load record formed according to the transaction period and interruption tolerance level; its fields include transaction period, level identifier, baseline load, adjustable load, deferred load, interruptible load, and confidence interval. The baseline load is generated by mapping historical charging curves, weather forecasts, holiday calendars, and surrounding POI event data for the corresponding transaction period; the adjustable load is generated based on the maximum power reduction ratio and the minimum reserved charging power; the deferred load is generated based on the reservation period, queuing period, and maximum allowable waiting time; the interruptible load is generated based on the maximum interruptible duration, user constraint data, and historical response success rate. The confidence interval is generated jointly by the forecast deviation records of historical similar transaction periods and the real-time status data.

[0054] In S200, the adjustable load forecasting module first divides multiple charging stations into trading periods according to the station identifiers and charging pile time information in the charging station tag records. Then, based on the interruption tolerance levels in the interruption tolerance constraint records, it breaks down the charging load within each trading period into load segments corresponding to different levels. Next, it maps weather forecasts, holiday calendars, and surrounding POI event data to each trading period to generate the baseline load for that trading period. Subsequently, it generates a market-side demand status field based on renewable energy output forecast data and grid load information. This field records the renewable energy output level, regional load status, and time period markers eligible for demand response within the trading period. All of these intermediate fields are then incorporated into the adjustable load forecasting curve generation process.

[0055] In one implementation, the adjustable load forecasting module uses a time-series forecasting model to predict baseline load. The inputs to the forecasting model include historical charging curves, weather forecasts, holiday calendars, and surrounding POI event data; the output of the forecasting model is the baseline load corresponding to the trading period. In another implementation, the adjustable load forecasting module generates baseline load using a rule-based mapping method. For example, during statutory holidays and when there are large-scale events in the surrounding POI event data, charging curves from similar historical dates are used as baseline load reference records. All of the above implementations do not change the processing order of generating adjustable load forecast curves according to trading periods and interruption tolerance levels in this step.

[0056] When generating adjustable load, the adjustable load prediction module reads the maximum power reduction ratio and minimum reserved charging power from the interruption tolerance constraint record. For a certain interruption tolerance level within a certain transaction period, it first calculates the difference between the baseline load corresponding to that level and the minimum reserved charging power; then, it limits the difference according to the maximum power reduction ratio to generate the adjustable load. When generating deferred load, the adjustable load prediction module reads the reservation period, queuing period, and maximum allowed waiting time. When the reservation period overlaps with the transaction period, only charging demands that can be postponed within the maximum allowed waiting time are written into the deferred load. When generating interruptible load, the adjustable load prediction module reads the maximum interruptible duration and historical response success rate. For sites with a historical response success rate lower than the rule configuration value, the interruptible load is marked with a low reliability flag.

[0057] The triggering conditions for S200 include the generation of S100 output records, the update of external forecast data versions, the opening of the market declaration window, and the arrival of the rolling trading period. If the external forecast data lacks a certain trading period, the adjustable load forecasting module reads the data of the same trading period from the previous version and writes a version continuation mark into the adjustable load forecasting curve; if the real-time status data in a charging station tag record shows a communication anomaly, the adjustable load forecasting module writes the adjustable load margin of that station in the current trading period into a low-confidence state; if there is a time inconsistency between the new energy output forecast data and the grid load information, the adjustable load forecasting module uses the trading period in the market rules as the benchmark for truncation and retains the original time period information of the truncated data.

[0058] In one implementation scenario, the transaction periods are 9:00-9:15 and 9:15-9:30. Station A is labeled as a fast-charging public charging station, with a maximum allowable waiting time of 15 minutes and a maximum power reduction ratio of 20% for its Level 1 designation. Station B is labeled as a slow-charging private charging station, with a maximum allowable waiting time of 45 minutes and a maximum power reduction ratio of 40% for its Level 3 designation. The adjustable load forecasting module receives rainfall markers from weather forecasts, weekday markers from holiday calendars, and commercial area activity markers from surrounding POI event data. It then generates baseline loads for the two transaction periods. Based on the interruption tolerance constraints recorded for stations A and B, it generates adjustable load, delayed load, and interruptible load for Level 1 and Level 3, respectively. The resulting adjustable load forecast curve includes transaction period, level identifier, and station source fields for the S300 to read.

[0059] The adjustable load forecast curve output by S200 is fed into S300 as the capacity boundary for generating market channel parameter records and main investment market channel records. The trading period and level identifier in the adjustable load forecast curve are used to match the reporting granularity of the market channel; the adjustable load, delayable load, and interruptible load are used to calculate the reporting electricity or capacity corresponding to the market channel; the confidence interval and low confidence marker are used by S300 to record status when reading deviation assessment parameters. S200 simultaneously retains both the input data version and the forecast version. When S500 subsequently generates a rollback and supplementary record, the corresponding forecast version can be written into the final reporting form.

[0060] S300. Based on the adjustable load forecast curve, read the market rules, revenue parameters and deviation assessment parameters of multiple market channels, generate market channel parameter records and determine the main investment market channel record.

[0061] S300 is executed by the market channel parameter parsing module and the upper-level market selection module in the aggregated bidding server. The market channel parameter parsing module receives the adjustable load forecast curve output by S200 and reads the market rules, revenue parameters and deviation assessment parameters of multiple market channels from the market trading platform, rule configuration table or local market rule base; the upper-level market selection module generates the main investment market channel record based on the market channel parameter record, the adjustable load forecast curve and the deviation assessment parameters.

[0062] The multiple market channels include the day-ahead electricity market, the frequency regulation ancillary services market, and the valley-filling demand response market. Revenue parameters for the day-ahead electricity market include time-of-use pricing and bid price limits; revenue parameters for the frequency regulation ancillary services market include capacity rates and mileage rates; and revenue parameters for the valley-filling demand response market include unit subsidy standards and response time requirements. The market rules include bid granularity, response speed requirements, regulation accuracy requirements, and power regulation range; the deviation assessment parameters include allowable deviation percentages, deviation assessment unit prices, assessment time windows, and deviation calculation methods.

[0063] The market channel parameter parsing module first reads the rule version of the corresponding market channel according to the trading period in the adjustable load forecast curve; then it maps the revenue parameters, deviation assessment parameters, and constraint fields of each market channel; subsequently, it generates a market channel parameter record. The market channel parameter record includes the market channel, trading period, time-of-use tariff, capacity rate, mileage rate, unit subsidy standard, application granularity, response speed requirements, adjustment accuracy requirements, power adjustment range, and deviation assessment parameters. If a market channel is not open for application during a certain trading period, the market channel parameter parsing module writes the market channel status corresponding to that trading period as "unavailable for application."

[0064] When the upper-level market selection module determines the main investment market channel record, it first reads the adjustable load, deferred load, interruptible load, and confidence interval from the adjustable load forecast curve; then it reads the declaration granularity, response speed requirements, adjustment accuracy requirements, and power adjustment range from the market channel parameter record; subsequently, it performs rule verification on the market channel for each trading period. For the frequency regulation ancillary service market, the verification includes whether the adjustable load in the adjustable load forecast curve covers the power adjustment range and whether the real-time status data meets the response speed requirements; for the valley filling demand response market, the verification includes whether the deferred load and interruptible load cover the response amount corresponding to the unit subsidy standard; for the day-ahead electricity market, the verification includes whether the declared electricity or capacity conforms to the declaration granularity and market rules.

[0065] After the market channel passes the rule verification, the upper-level market selection module calculates the risk-adjusted return for each trading period based on the market channel parameter records, adjustable load forecast curve, and deviation assessment parameters. The risk-adjusted return is the value obtained by subtracting user compensation costs and deviation assessment expectations from the return parameters; whereby user compensation costs are referenced from the user compensation cost parameter in the interruption tolerance constraint record, and deviation assessment expectations are generated from the deviation assessment parameters, the confidence interval in the adjustable load forecast curve, and historical response deviation records. The upper-level market selection module sorts the risk-adjusted returns according to the trading period and writes the market channel that has passed verification and is ranked first into the main investment market channel record.

[0066] In one implementation, the primary investment market channel record includes the trading session, market channel, return parameters, deviation assessment parameters, reporting granularity, response speed requirements, adjustment accuracy requirements, power adjustment range, and rule version. In another implementation, when the top-ranked market channel is in a low-confidence state, the upper-level market selection module simultaneously generates a backup market channel record. The backup market channel record uses the same field structure as the primary investment market channel record and includes a backup status flag. The backup market channel record does not change the S400's reference order of the primary investment market channel record; it is only used as a replacement input when the primary investment market channel record is updated by the rule version or overwritten by a market closure status.

[0067] S300 is activated when the adjustable load forecast curve is generated, the market rule version is updated, the market trading platform releases the declaration window, or the deviation assessment parameters are updated. If the market rule version and the revenue parameter version are inconsistent, the market channel parameter parsing module writes the version inconsistency status into the market channel parameter record and calls the currently valid version in the rule configuration table; if the deviation assessment parameters of a certain market channel are missing, the upper-level market selection module writes the market channel into the pending review status and does not participate in this round of sorting; if all market channels fail the rule verification, the aggregated bidding server generates a record of no market declaration status and temporarily does not write the corresponding trading period into the main investment market channel record.

[0068] In one implementation scenario, during the trading session from 9:00 to 9:15, the adjustable load forecast curve shows that the second-level adjustable load is 800kW, and the deferable load is 300kW. The market channel parameter parsing module reads the market rules of the day-ahead energy market, the frequency regulation ancillary service market, and the valley filling demand response market. The frequency regulation ancillary service market requires the response speed to not exceed the time requirement in the rule configuration table, and the power adjustment range is 500kW to 1000kW; the valley filling demand response market requires the response amount to be no less than 300kW. The upper-level market selection module calculates the risk-adjusted returns of the three market channels respectively, verifies the response speed requirements, adjustment accuracy requirements, and power adjustment range; and writes the market channel that passes the verification and ranks first in risk-adjusted returns into the main investment market channel record for the trading session from 9:00 to 9:15.

[0069] The market channel parameter records and main investment market channel records output by S300 are input into S400. The order granularity, response speed requirements, adjustment accuracy requirements, power adjustment range, and deviation assessment parameters in the market channel parameter records serve as constraints for S400's optimization solution; the trading session, market channel, and return parameters in the main investment market channel records serve as market-side inputs for the volume-price order combination. S300 also retains a version of the market rules; when S500 generates the final order form, this version of the market rules is bound together with the pullback and compensation records.

[0070] S400: Based on the main investment market channel record, the adjustable load forecast curve, and the interruption tolerance constraint record, optimize the solution according to the adjustable load amount, user compensation cost parameters, and deviation assessment expectations for each interruption tolerance level, and generate quantity-price declaration combination and site allocation record.

[0071] S400 is executed by the lower-level volume and price reporting module in the aggregated bidding server. This module receives the main investment market channel record and market channel parameter record output from S300, receives the adjustable load forecast curve output from S200, and reads the interruption tolerance constraint record output from S100. The main investment market channel record provides the trading period, market channel, and market rules; the adjustable load forecast curve provides the adjustable load amount for each interruption tolerance level; and the interruption tolerance constraint record provides user compensation cost parameters, minimum reserved charging power, maximum interruptible duration, and maximum power reduction ratio.

[0072] The quantity-price declaration combination includes the transaction period, declared electricity or capacity, declared price, and market channel; the site allocation record includes the site identifier, transaction period, allocated bid quantity, corresponding interruption tolerance level, and adjustable load margin. The quantity-price declaration combination generates a declaration record for the market trading platform; the site allocation record generates a bid quantity source record for each charging station. Before generating the quantity-price declaration combination, the lower-level quantity-price declaration module first segments the adjustable load forecast curve according to the declaration granularity in the main investment market channel record; then maps the segmented adjustable load quantity to each interruption tolerance level; and finally establishes a correlation between the user compensation cost parameter of each level and the corresponding adjustable load quantity.

[0073] In S400, the deviation assessment expectation is generated from the deviation assessment parameters, the confidence interval of the adjustable load forecast curve, and historical response deviation records. For a specific trading period and a specific interruption tolerance level, the lower-level volume-price declaration module reads the adjustable load volume and confidence interval for that level; then it reads the allowable deviation ratio and deviation assessment unit price from the deviation assessment parameters; subsequently, it generates the corresponding deviation assessment expectation. If the historical response deviation record shows that a certain site has experienced response deviations multiple times within the same trading period, the lower-level volume-price declaration module writes a response deviation flag in the site allocation record corresponding to that site, and reads this flag when generating the allocation bid volume.

[0074] The optimization solution is constrained by the reporting granularity corresponding to the main investment market channel record, the adjustable load volume of each interruption tolerance level, the minimum reserved charging power, user compensation cost parameters, and deviation assessment expectations. Specifically, the lower-level volume-price reporting module first calculates the adjustable load volume that can participate in the reporting for each interruption tolerance level according to the trading period; then, it rounds or splits the adjustable load volume according to the reporting granularity of the market channel; subsequently, it writes the user compensation cost parameters and deviation assessment expectations into the cost-side record; finally, under the conditions of not exceeding the adjustable load volume of the corresponding level, not being lower than the minimum reserved charging power, and not violating the reporting granularity of the market channel, it generates a volume-price reporting combination. After the optimization solution is completed, the reported electricity or capacity in the volume-price reporting combination is allocated to multiple charging stations, forming a site allocation record.

[0075] In one implementation, site allocation records are prioritized based on the same interruption tolerance level. The lower-level volume and price declaration module first filters charging station tag records that match the corresponding transaction period in the main investment market channel records; then it reads the adjustable load margin and interruption tolerance level of the corresponding charging station; subsequently, it sorts them from largest to smallest adjustable load margin and allocates the allocation bid volume item by item. In another implementation, site allocation records are sorted according to historical response success rate, with sites with higher historical response success rates being written into the allocation record first; when two sites have the same historical response success rate, the number of available charging piles in the real-time status data is read as the sorting field. All of the above allocation rules generate site allocation records with the same structure.

[0076] S400 starts after the main investment market channel record is generated, and can also be restarted when the market rule version changes, the adjustable load forecast curve is updated, or the deviation assessment parameters are updated. If the adjustable load is less than the market channel declaration granularity during a certain trading period, the lower-level volume and price declaration module writes the trading period into an undeclarable state and does not generate a volume and price declaration combination; if the user compensation cost parameters of a certain level are missing, the lower-level volume and price declaration module reads the valid parameters of the previous declaration period of that level and writes the parameter reuse mark in the volume and price declaration combination; if the minimum reserved charging power of a certain site conflicts with the allocated bid quantity during the optimization solution process, the allocated bid quantity of that site is set to 0, and the site is written into the unallocated record.

[0077] In one implementation scenario, the S300-generated main investment market channel record shows participation in the valley filling demand response market during the 9:00-9:15 trading period, with a declaration granularity of 100kW. The adjustable load forecast curve output by the S200 shows that the second-level delayed load is 520kW, and the third-level delayed load is 360kW. The lower-level quantity and price declaration module reads the user compensation cost parameters corresponding to the second and third levels, and reads the deviation assessment parameters; then, it matches the declaration granularity of the declared capacity, generating a quantity and price declaration combination with a declared capacity of 800kW. The lower-level quantity and price declaration module allocates this 800kW to stations A, B, and D, forming the station identifier, trading period, allocated bid quantity, corresponding interruption tolerance level, and adjustable load margin fields, respectively; these fields are entered into the S500 for calculating the reduction amount after a queuing over-limit event and for selecting supplementary charging stations.

[0078] The quantity-price declaration combination and site allocation records output by S400 are input into S500. The transaction period, declared electricity or capacity, declared price, and market channel in the quantity-price declaration combination are used to generate the final declaration form; the site identifier, allocated bid quantity, corresponding interruption tolerance level, and adjustable load margin in the site allocation records are used for target charging station bid quantity reduction and supplementary charging station screening. S400 also retains the source of deviation assessment expectations and user compensation cost parameters. When subsequent execution feedback records are returned, the actual execution quantity and measured compensation cost values ​​can be correlated with the aforementioned source fields.

[0079] S500: Receive the real-time queuing status of each charging station based on the site allocation record, compare the real-time queuing status with the maximum allowed waiting time, reduce the bid amount of the target charging station when a queuing limit exceedance event is generated, and filter the supplementary charging stations with the same tags based on the charging station tag record to supplement the reduction difference, and generate a final application form that binds the site allocation record and the reduction supplement record.

[0080] S500 is executed by the queuing status monitoring module, the rollback and replenishment module, and the application form output module in the aggregated bidding server. The queuing status monitoring module receives the site allocation records output by S400 and receives the real-time queuing status of each charging station from the charging pile management system, the charging station on-site queuing system, or the station-side real-time monitoring equipment; the rollback and replenishment module reads the charging station tag records and interruption tolerance constraint records generated by S100; the application form output module reads the quantity and price application combination generated by S400, and generates the final application form by binding the rollback and replenishment records with the site allocation records.

[0081] The real-time queuing status includes the current number of vehicles in the queue, average waiting time, longest waiting time, number of available charging piles, and real-time power. The current number of vehicles in the queue is generated from vehicle entry records or the queuing system; the average waiting time is obtained by averaging the waiting times of the vehicles in the queue; the longest waiting time is the maximum value among the waiting times of the vehicles in the queue; the number of available charging piles is generated from the occupancy and fault status in the charging pile management system; and the real-time power is generated from the charging station metering system or charging pile power records. The queuing status monitoring module matches the real-time queuing status with the station identifier and transaction period in the station allocation record to form a station queuing status record.

[0082] The queuing status monitoring module compares the average waiting time or the longest waiting time with the maximum allowable waiting time recorded in the interruption tolerance constraint record. When the average waiting time exceeds the maximum allowable waiting time, or the longest waiting time exceeds the maximum allowable waiting time, the queuing status monitoring module generates a queuing limit violation event. The queuing limit violation event includes the target charging station's site identifier, transaction period, corresponding interruption tolerance level, current number of vehicles in the queue, average waiting time, longest waiting time, and event generation time. If the waiting time field in the real-time queuing status is missing, the queuing status monitoring module reads the current number of vehicles in the queue, the number of available charging piles, and the average service time of the same historical period to generate an estimated waiting time, and writes the estimated flag into the queuing limit violation event.

[0083] Upon receiving a queuing limit exceedance event, the retraction and replenishment module determines the retraction amount based on the allocated bid amount in the site allocation record and the adjustable load margin of the target charging station. Specifically, the retraction and replenishment module first reads the allocated bid amount of the target charging station for the corresponding transaction period; then it reads the adjustable load margin and real-time queuing status of the target charging station; when the adjustable load margin is less than the allocated bid amount, or a queuing limit exceedance event has been generated, the difference between the allocated bid amount and the adjustable load margin is written into the retraction amount; when the maximum allowable waiting time corresponding to the queuing limit exceedance event exceeds the limit by a large margin, the retraction amount can be set to the total allocated bid amount of the target charging station for that transaction period. The above retraction amount is written into the target site retraction field in the retraction and replenishment record.

[0084] When selecting supplementary charging stations, the retraction and replenishment module reads the charging station tag records and first selects charging stations with the same tags as the target charging station. The same tags include the same charging station type, the same or matching public charging pile attributes, private charging pile attributes, battery swapping attributes, operational vehicle-specific attributes, and the same interruption tolerance level. The retraction and replenishment module compares the charging stations with the same tags with the real-time queuing status, eliminating charging stations that have already generated queuing overrun events; then it reads the adjustable load margin of the candidate charging stations, the power adjustment range in the market channel parameter records, and the response speed requirements; subsequently, it determines the replenishment amount corresponding to each supplementary charging station. If the adjustable load margin of supplementary charging stations with the same tags is less than the retraction difference, the retraction and replenishment module filters charging stations corresponding to adjacent interruption tolerance levels based on the charging station tag records, and determines the replenishment amount according to the adjustable load margin, real-time queuing status, and market channel parameter records.

[0085] The retraction and replenishment record includes the target station, queuing over-limit event, retraction amount, replenishment station, replenishment amount, replenishment result, and risk status. The retraction and replenishment module summarizes the retraction amount of the target charging station and the replenishment amount of the replenishment station; when the replenishment amount equals the retraction difference, the replenishment result is written to the completion status; when the replenishment amount is less than the retraction difference, the retraction and replenishment module generates a reduction declaration record or a performance risk record. The reduction declaration record includes the transaction period, the original declared electricity or capacity, the declared electricity or capacity after the reduction, and the source of the reduction; the performance risk record includes the transaction period, the unreplenished amount, the target station identifier, and the risk status. If the market declaration window is not yet closed, the declaration output module updates the final declaration based on the reduction declaration record; if the market declaration window has been closed, the declaration output module writes the performance risk record into the risk status field of the final declaration.

[0086] In one implementation, the final application form includes a market channel, trading period, declared electricity or capacity, declared price, site allocation record, pullback and replenishment record, market rule version, tag version, and model version. The market rule version originates from the market channel parameter record in S300; the tag version originates from the charging station tag record in S100; and the model version originates from the prediction version written when generating the adjustable load prediction curve in S200. The site allocation record in the final application form retains the original allocation field generated in S400, while adding a pullback and replenishment record reference field; this reference field records the target site, pullback amount, replenishment site, and replenishment amount. In another implementation, the final application form also includes a performance control curve, which records the bid amount, pullback amount, replenishment amount, and adjustable load margin of each charging station according to the site identifier and trading period.

[0087] S500 starts after the site allocation record is generated, and can also restart when the real-time queue status is updated, a queue over-limit event is generated, the replenishment capacity is insufficient, or the market declaration window status changes. If the real-time queue status of a replenishment charging station changes after the replenishment amount is generated, the queue status monitoring module regenerates the site queue status record for that station; the shrink replenishment module recalculates the replenishment amount and updates the shrink replenishment record. If multiple target charging stations generate queue over-limit events simultaneously within the same trading period, the shrink replenishment module sorts them according to the generation time and shrinkage amount of the queue over-limit event; it processes the queue over-limit events with earlier generation times first, and then processes the queue over-limit events with larger shrinkage amounts at the same time. If the shrink replenishment record is inconsistent with the market rule version in the final declaration form, the declaration form output module suspends the output of the final declaration form for that trading period and rereads the current valid market channel parameter record generated by S300.

[0088] In one implementation scenario, during the 9:00-9:15 trading session, Station A's site allocation record shows an allocated bid volume of 300kW and an adjustable load margin of 280kW. The queuing status monitoring module receives Station A's real-time queuing status, with an average waiting time of 18 minutes and a maximum waiting time of 23 minutes. Station A's interruption tolerance constraint record shows a maximum allowed waiting time of 15 minutes. The queuing status monitoring module generates a queuing limit violation event for Station A. The rollback and replenishment module reads Station A's allocated bid volume and adjustable load margin, recording the rollback amount as 20kW. Simultaneously, it reads the charging station tag record and filters for Stations B and C, which have the same tags as Station A. Station B has no queuing limit violation event and an adjustable load margin of 30kW, while Station C has an adjustable load margin of 10kW. The rollback and replenishment module writes the 20kW replenishment amount into Station B's replenishment record, generating a rollback and replenishment record. The declaration form output module writes Station A's rollback amount, Station B's replenishment amount, site allocation record, and market rule version into the final declaration form.

[0089] In another implementation scenario, the reduction shortfall at station A is 120kW, and the total adjustable load margin of supplementary charging stations with the same label is 80kW. The reduction and supplementation module continues to filter charging stations corresponding to adjacent interruption tolerance levels, obtaining an adjustable load margin of 20kW for station D. The total supplementation amount is 100kW, still less than the reduction shortfall; the reduction and supplementation module generates a performance risk record for the 20kW shortfall. If the market application window is still open, the application output module generates a reduction application record based on the shortfall and updates the declared electricity or capacity in the final application form; if the market application window is closed, the application output module writes the performance risk record into the final application form and sends the corresponding performance control curve to the charging pile management system.

[0090] The S500 also receives execution feedback records. These records include actual execution volume, response deviations, queuing exceedance events, and measured compensation costs. After the trading session ends, the aggregated bidding server matches the execution feedback records with the rollback and replenishment records. The matching criteria include site identifier, trading session, market channel, and final application form version. After matching, the execution feedback records and rollback and replenishment records are returned to the corresponding charging station tag records and interruption tolerance constraint records in the S100. Updates to historical response success rates, historical cancellation rates, user compensation cost parameters, and maximum allowable waiting time are written into the input of the next application cycle. The final application form serves as the application object for the market trading platform, the performance control curve serves as the execution object for the charging pile management system, and the rollback and replenishment records serve as the historical state source when generating site allocation records again in subsequent trading sessions.

Claims

1. A method for charging station operators to participate in market bidding based on artificial intelligence, characterized in that, include: S100: Receives site files, historical charging curves, user charging habit data, user constraint data, and real-time status data from multiple charging stations; binds these data according to charging station type, charging pile time period information, historical load characteristics, and user constraint data; and generates charging station tag records and interruption tolerance constraint records. The interruption tolerance constraint records include interruption tolerance level, maximum allowable waiting time, and user compensation cost parameters. S200. Based on the charging station tag records and the interruption tolerance constraint records, generate an adjustable load forecast curve according to the transaction period and interruption tolerance level; S300. Based on the adjustable load forecast curve, read the market rules, revenue parameters and deviation assessment parameters of multiple market channels, generate market channel parameter records and determine the main investment market channel record. S400. Based on the main investment market channel records, optimize the solution according to the adjustable load, user compensation cost parameters and deviation assessment expectations of each interruption tolerance level, and generate volume and price declaration combinations and site allocation records. S500: Based on the site allocation record, receive the real-time queuing status of each charging station, compare the real-time queuing status with the maximum allowed waiting time, reduce the bid amount of the target charging station when a queuing limit exceedance event is generated, and generate a final application form that binds the site allocation record and the reduction and supplementation record.

2. The method according to claim 1, characterized in that, The charging station tag record includes the station identifier, fast charging ratio, slow charging ratio, battery swapping attribute, public charging pile attribute, private charging pile attribute, operating vehicle exclusive attribute, rated power, number of available charging piles, charging pile time period information, and historical load characteristics; The binding includes establishing an association between the charging pile time period information, historical load characteristics, and user constraint data under the site identifier.

3. The method according to claim 1, characterized in that, The interruption tolerance constraint record also includes the maximum interruptible duration, the maximum power reduction ratio, the minimum reserved charging power, the historical cancellation rate, and the historical response success rate. The interruption tolerance level is generated based on the historical cancellation rate, the historical response success rate, and the historical queuing status, and is updated based on the execution feedback records.

4. The method according to claim 1, characterized in that, The process of generating adjustable load forecast curves includes data access mapping, load splitting, and confidence labeling.

5. The method according to claim 4, characterized in that, The data access mapping process includes: receiving external inputs by the adjustable load forecasting module, and writing data source, forecast release time, forecast period, and version identifier for each type of input; when the forecast time granularity of different sources is inconsistent, the external inputs are mapped to the trading period specified in the market rules; the weather forecast includes temperature, precipitation, wind speed, and weather conditions; the surrounding POI event data includes activities in the surrounding business districts of the charging station, passenger flow events at transportation hubs, shuttle bus times in the park, and activity records in public parking areas; the renewable energy output forecast data includes one or more of wind power output forecasts, photovoltaic power output forecasts, and distributed renewable energy output forecasts.

6. The method according to claim 4, characterized in that, The load splitting process includes: first, dividing multiple charging stations into transaction periods according to the station identifier and charging pile time period information in the charging station tag record; then, splitting the charging load in each transaction period into load segments corresponding to different levels according to the interruption tolerance level; and then mapping the weather forecast, holiday calendar and surrounding POI event data to each transaction period to generate the baseline load for the corresponding transaction period. The adjustable load forecast curve includes the transaction period, level identifier, baseline load, adjustable load, deferred load, interruptible load, and confidence interval; When generating the adjustable load, the maximum power reduction ratio and the minimum reserved charging power are read. First, the difference between the baseline load corresponding to the level and the minimum reserved charging power is calculated, and then the difference is limited according to the maximum power reduction ratio. When generating the deferred load, the reservation period, the queuing period, and the maximum allowed waiting time are read. When the reservation period overlaps with the transaction period, only the charging demand that can be postponed within the maximum allowed waiting time is written into the deferred load. When generating the interruptible load, read the maximum interruptible duration and the historical response success rate.

7. The method according to claim 4, characterized in that, The confidence level marking process includes: the confidence interval is generated by combining the prediction deviation records of historical similar transaction periods and the real-time status data; for sites with historical response success rates lower than the rule configuration value, a low confidence level is written into the interruptible load. If the real-time status data in a charging station tag record indicates a communication anomaly, the adjustable load margin of that station in the current trading period will be written to a low-confidence status. If the external forecast data lacks a certain trading period, the data for the same trading period in the previous version will be read, and a version continuation mark will be written into the adjustable load forecast curve.

8. The method according to claim 1, characterized in that, The final declaration form that binds the site allocation record and the shrinkage and replenishment record also includes: queuing status monitoring processing, shrinkage amount calculation processing, replenishment site screening processing, and declaration form output processing; The queuing status monitoring process includes: the queuing status monitoring module matching the real-time queuing status with the station identifier and transaction period in the station allocation record to form a station queuing status record; the real-time queuing status includes the current number of vehicles in the queue, average waiting time, longest waiting time, number of available charging piles, and real-time power; comparing the average waiting time or the longest waiting time with the maximum allowed waiting time in the interruption tolerance constraint record, and generating the queuing limit violation event when the average waiting time is greater than the maximum allowed waiting time or the longest waiting time is greater than the maximum allowed waiting time; If the waiting time field in the real-time queuing status is missing, the estimated waiting time is generated by reading the current number of vehicles in the queue, the number of available charging piles, and the average service time of the same period in history, and an estimated flag is written in the queue over-limit event.

9. The method according to claim 8, characterized in that, The shrinkage calculation process includes: the shrinkage compensation module reads the allocated bid amount of the target charging station in the corresponding transaction period, and then reads the adjustable load margin of the target charging station; when the adjustable load margin is less than the allocated bid amount, or when a queue over-limit event has been generated, the difference between the allocated bid amount and the adjustable load margin is written into the shrinkage amount; when the maximum allowable waiting time corresponding to the queue over-limit event exceeds the limit by a large margin, the shrinkage amount is set to the total allocated bid amount of the target charging station in that transaction period. The supplementary site screening process includes: first selecting charging stations with the same tags as the target charging station. The same tags include the same charging station type, the same or matching public charging pile attributes, private charging pile attributes, battery swapping attributes, operating vehicle exclusive attributes, and the same interruption tolerance level. After removing charging stations that have already generated queue over-limit events, the adjustable load margin of candidate charging stations, the power adjustment range and response speed requirements in the market channel parameter record are read to determine the replenishment amount. If the adjustable load margin of replenishment charging stations with the same label is less than the reduction difference, the replenishment amount is determined by filtering charging stations corresponding to adjacent interruption tolerance levels based on the charging station label record. When the replenishment amount is equal to the reduction difference, the replenishment result is written to the completion status. When the replenishment amount is less than the reduction difference, a reduction declaration record or performance risk record is generated.

10. The method according to claim 8, characterized in that, The declaration form output processing includes: if the market declaration window is not yet closed, updating the final declaration form according to the reduction declaration record; if the market declaration window is already closed, writing the performance risk record into the risk status field of the final declaration form; the final declaration form includes market channel, trading period, declared electricity or capacity, declared price, site allocation record, pullback and replenishment record, market rule version, tag version, and model version, and the pullback and replenishment record includes target site, queue over-limit event, pullback amount, replenishment site, replenishment amount, replenishment result, and risk status.

Images