A multi-source heterogeneous adjustable resource unified modeling method and system

Abstract

This application provides a unified modeling method and system for multi-source heterogeneous adjustable resources, including: integrating the number of vehicles on the network and the dwell time of vehicles according to the boundary time-varying description, and determining the improvement index of aggregation credibility; determining whether it is necessary to iteratively update the boundary time-varying description according to the aggregation credibility improvement index, until a committable capacity space that meets the scheduling reliability requirements is output; using the committable capacity space to simulate the fulfillment rate, inputting historical scheduling plans and actual vehicle off-network records, replaying and verifying the completion of scheduling instructions under the capacity constraint, and outputting the capacity boundary; encapsulating the adjustable capacity of charging piles reconstructed based on dwell behavior with the adjustment capabilities of other types of resources, realizing the aggregation and invocation of heterogeneous resources under a unified semantics.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a unified modeling method and system for multi-source heterogeneous adjustable resources. Background Technology

[0002] In the field of energy management and intelligent transportation integration, the resource scheduling and optimization of electric vehicle charging stations is particularly crucial. This area directly relates to improving energy efficiency and user experience, and is an important pillar for promoting green travel and the development of smart grids. With the rapid growth in the number of electric vehicles, how to rationally allocate charging resources to meet dynamic demand has become a key issue that the industry urgently needs to address. However, current research and practice, when dealing with the adjustable capacity of charging station resources, often neglect the impact of dynamic changes in vehicle dwell time on the overall resource boundary. Many methods only focus on the number of vehicles, assuming that a larger number means more adjustable resources, but fail to deeply consider the actual limitations caused by differences in vehicle dwell time. This neglect leads to a significant discrepancy between resource assessment and actual availability, especially when facing complex usage scenarios, where scheduling effects are often unsatisfactory. Furthermore, because dwell time changes constantly with vehicle plugging and unplugging events, vehicles with short dwell times may frequently enter and leave the charging network. This frequent change not only lowers the committable range of resources but also causes sudden jumps in the resource boundary curve. For example, during a peak period, a large number of vehicles briefly stop to charge and then quickly leave. Although the total number of vehicles seems considerable, the actual available charging time is very limited. This causes the scheduling system to be unable to accurately predict and commit to resource capacity, thus affecting users' charging plans and the stability of the power grid. Therefore, accurately identifying and responding to jumps in resource boundaries caused by short-staying vehicles when vehicles are frequently plugged in and out and the duration of their stops is uncertain has become a key issue in improving the reliability of resource scheduling. Summary of the Invention

[0003] This invention provides a unified modeling method for multi-source heterogeneous adjustable resources, including: By obtaining the timing of plug-in / plug-out events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion information of capacity bag boundary nodes from the charging pile network, a real-time status data set of vehicles on the network is obtained. The distribution characteristics of the remaining dwell time of vehicles are analyzed, and the actual reduction of the upper limit of adjustable capacity is quantified. When the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the stability threshold set by the system, the system identifies the period of sudden increase in event density per unit time and marks the end of the period as the potential jump point where the upper edge of the adjustable capacity may jump upward. Extract the capacity bag boundary node addition and deletion records associated with each potential transition point location, analyze whether the newly added boundary node is mainly triggered by the concentrated access of short-staying vehicles, and if it is confirmed that the transition is caused by unsustainable short-term access, then reset the capacity accumulation starting point at the transition point and generate a boundary time-varying description. Based on the time-varying description of the boundary, the number of vehicles on the network and the duration of vehicle stay are integrated to determine the index for improving aggregation credibility; Determine whether it is necessary to iteratively update the boundary time-varying description based on the aggregation reliability improvement index until the output of the promised capacity space that meets the scheduling reliability requirements is reached; The fulfillment rate is simulated using the committable capacity space. Historical scheduling plans and actual vehicle off-grid records are input, and the completion of scheduling instructions under the capacity constraint is verified by replaying the data. The capacity boundary is then output. The adjustable capacity of charging piles reconstructed based on dwell behavior is encapsulated with the adjustment capabilities of other types of resources, enabling the aggregation and invocation of heterogeneous resources under a unified semantic framework.

[0004] Furthermore, by acquiring the timing of plug-in / plug-out events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion information of capacity bag boundary nodes from the charging pile network, a real-time status data set of vehicles on the network is obtained. The distribution characteristics of the remaining dwell time of vehicles are analyzed, and the actual reduction in the upper limit of adjustable capacity is quantified, including: The charging pile network's communication interface collects plug-in / plug-out event timestamps, synchronously reads the remaining dwell time of vehicles on the network, and records the dynamic addition and deletion logs of capacity bag boundary nodes. The timestamps of the plugging / unplugging events, the remaining dwell time, and the changes in boundary nodes are matched and associated with vehicle identifiers, and abnormal data is removed to obtain a set of real-time status data. The real-time status data set is divided into time slices, the remaining dwell time field is extracted, the distribution pattern is described by kernel density estimation, the peak position and quantile are extracted, the short-term fluctuation amplitude is measured, and the distribution characteristics are obtained. Based on the distribution characteristics, short-stay vehicles are identified, and the capacity occupancy share of short-stay vehicles is statistically calculated. The unavailable capacity ratio is then weighted and calculated as a reduction in the adjustable capacity.

[0005] Furthermore, based on the distribution characteristics, short-stay vehicles are identified, and the capacity share occupied by short-stay vehicles is statistically calculated. A weighted average is then used to calculate the capacity unavailability ratio, which serves as a reduction in the adjustable capacity. This includes: based on the distribution characteristics, vehicles whose remaining stay time falls within the left tail quantile interval of the distribution are identified as short-stay vehicles; the capacity share corresponding to the capacity bag boundary node occupied by the short-stay vehicles is statistically calculated; and the capacity share is weighted and calculated based on the proportion of short-stay vehicles in the total number of vehicles in the network, resulting in the capacity unavailability ratio caused by early departure, which serves as a reduction in the adjustable capacity.

[0006] Furthermore, when the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the system's set stability threshold, the system identifies periods of sudden increase in event density per unit time and marks the end of these periods as potential jump points where the adjustable capacity upper edge may jump upwards, including: The sum of the proportion of vehicles with short stops and the aforementioned reduction amount is obtained by normalizing and summing. If the sum exceeds a preset stability threshold, a sliding time window scan is triggered; The density of insertion and removal events is statistically analyzed using the sliding time window to form a density sequence. The density increment is obtained by using the difference method to determine the period of sudden increase in event density. For the period of sudden increase, review the vehicle dwell time, eliminate false sudden increase segments, retain the real sudden increase segments dominated by vehicles with short dwell times, and mark the end time of the real sudden increase segment as the potential jump point position.

[0007] Furthermore, the process involves extracting the capacity bag boundary node addition and deletion records associated with each potential transition point, analyzing whether the newly added boundary nodes are mainly triggered by concentrated access from short-staying vehicles, and if it is confirmed that the transition is caused by unsustainable short-term access, then the capacity accumulation starting point is reset at the transition point, and a time-varying boundary description is generated, including: Extract the addition and deletion records of boundary nodes centered on the potential transition point to form a change spectrum; For the newly added boundary node subset in the aforementioned variation spectrum, the proportion of short-stay vehicles triggering the event is statistically analyzed. If the trigger ratio exceeds the preset attribution threshold and the time interval between the generation of new nodes is lower than the preset threshold, then the jump is determined to be caused by short-term access, and a true jump point is obtained. Using the true transition point as the reset point, the capacity contribution of short-staying vehicles is eliminated, and the remaining vehicle capacity is used as the new starting point to smoothly connect and generate the boundary time-varying description.

[0008] Furthermore, the step of integrating the number of vehicles on the network and the vehicle dwell time based on the boundary time-varying description to determine the improvement index of aggregation reliability includes: Extract the number of vehicles on the network and the remaining dwell time corresponding to the time-varying description of the boundary, and discretize them to obtain the duration distribution; Using the number of vehicles on the network and the distribution of time tiers as input, an aggregation model is used to calculate the promised capacity share of each tier, and the sum of the promised capacity shares of each tier is used to obtain the integrated reliable capacity value. After arranging the curves according to time sections to form an integrated credibility capacity curve, the curves are compared with the original curves, and the average credibility value is calculated to obtain the aggregate credibility improvement index.

[0009] Furthermore, the step of determining whether to iteratively update the boundary time-varying description based on the aggregation reliability improvement index until a committable capacity space that meets the scheduling reliability requirements is output includes: Compare the improvement index with the preset scheduling reliability threshold; If the improvement index is lower than the preset scheduling reliability threshold, an iterative update instruction is triggered to tighten the short stay determination fraction, re-execute the starting point reset process, and update the improvement index. If the improvement index is not lower than the preset scheduling reliability threshold, a convergence termination instruction is triggered, and the current integrated reliable capacity curve is output as the promised capacity space.

[0010] Furthermore, the simulation of fulfillment rate using the committable capacity space involves inputting historical scheduling plans and actual vehicle off-grid records, replaying and verifying the completion of scheduling instructions under the capacity constraint, and outputting the capacity boundary, including: Using the promised capacity space as a constraint, historical scheduling plans and vehicle off-network records are retrieved, aligned, and merged to form a replay dataset; The power adjustment command is replayed according to the time section. Vehicles whose off-grid time is later than the section are filtered out. The power is accumulated to obtain the capacity contribution. The lower edge value is taken with the constraint to calculate the command completion degree and form the fulfillment rate sequence. For sections in the fulfillment rate sequence where the completion rate is lower than a preset lower limit, the constraint values ​​are adjusted downwards, and the capacity boundary is generated by splicing them together.

[0011] Furthermore, the adjustable capacity of the charging pile reconstructed based on dwell behavior is encapsulated with the adjustment capabilities of other types of resources to achieve aggregation and invocation of heterogeneous resources under a unified semantic framework, including: Using the aforementioned capacity boundary as the adjustable capacity input for the charging pile, parameters of energy storage and interruptible load are retrieved; The capacity boundary and the parameters are mapped according to a preset time-series availability field set to generate each resource entry; The entries are encapsulated according to standard specifications and the scale is normalized to obtain unified semantic resource standard entries; The resource standard entry data is processed according to time segments to generate aggregated adjustable capability time series, which is exposed to the upper-level power grid dispatch through calling interface to realize the aggregation and calling of the heterogeneous resources.

[0012] This invention provides a unified modeling system for multi-source heterogeneous adjustable resources, comprising: The data acquisition and analysis module is used to acquire information on the timing of plugging and unplugging events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion of capacity bag boundary nodes from the charging pile network. It obtains a set of real-time status data of vehicles on the network, analyzes the distribution characteristics of the remaining dwell time of vehicles, and quantifies the actual reduction of the upper edge of adjustable capacity. The jump point identification module is used to identify the period of sudden increase in event density per unit time when the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the stability threshold set by the system, and to mark the end of the period as the potential jump point position where the upper edge of the adjustable capacity may jump upward. The boundary description generation module is used to extract the addition and deletion records of capacity bag boundary nodes associated with each potential jump point, analyze whether the newly added boundary nodes are mainly triggered by the concentrated access of short-staying vehicles, and if it is confirmed that the jump is caused by unsustainable short-term access, then the capacity accumulation starting point is reset at the jump point to generate a time-varying boundary description. The credibility determination module is used to integrate the number of vehicles on the network and the vehicle dwell time based on the boundary time-varying description to determine the improvement index of aggregate credibility; The iterative update module is used to determine whether the boundary time-varying description needs to be iteratively updated based on the aggregation reliability improvement index, until the output of the promised capacity space that meets the scheduling reliability requirements is reached. The fulfillment simulation module is used to simulate the fulfillment rate using the committable capacity space. It takes historical scheduling plans and actual vehicle off-grid records as input, replays and verifies the completion of scheduling instructions under the capacity constraint, and outputs the capacity boundary. The resource encapsulation module is used to encapsulate the adjustable capacity of charging piles reconstructed based on dwell behavior, along with the adjustment capabilities of other types of resources, to achieve the aggregation and invocation of heterogeneous resources under a unified semantic framework.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a unified modeling method for multi-source heterogeneous adjustable resources. It proposes an integrated solution for complex business scenarios in charging pile networks, including reduced adjustable capacity due to short-stayed vehicles, transition point identification, and heterogeneous resource aggregation and scheduling. This invention acquires real-time data on plug-in / plug-out events and vehicle dwell time to analyze the capacity reduction impact of short-stayed vehicles, identify potential transition points, reset the capacity accumulation starting point, and generate a realistic and usable time-varying boundary description, thereby optimizing aggregation reliability. Simultaneously, this invention utilizes a committed capacity space to drive fulfillment rate simulation, verifying the completion of scheduling instructions. Finally, it standardizes and encapsulates the adjustable capacity of charging piles with other resources such as energy storage and interruptible loads, achieving aggregation and invocation under a unified semantic framework. The core innovation of this invention lies in solving the consistency problem in heterogeneous resource allocation through dwell behavior reconstruction and temporal availability expression, significantly improving the reliability and efficiency of resource scheduling. Attached Figure Description

[0014] Figure 1 This is a flowchart of a unified modeling method for multi-source heterogeneous adjustable resources according to the present invention.

[0015] Figure 2This is a schematic diagram of the structure of a unified modeling system for multi-source heterogeneous adjustable resources according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0017] like Figure 1 This embodiment of a unified modeling method and system for multi-source heterogeneous adjustable resources may specifically include: S101. Obtain the time of plugging / unplugging events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion information of capacity bag boundary nodes from the charging pile network to obtain a set of real-time status data of vehicles on the network. Analyze the distribution characteristics of the remaining dwell time of vehicles and quantify the actual reduction of the upper edge of adjustable capacity.

[0018] The timestamp of each plug-in / plug-out event is collected from the charging pile network's terminal communication interface. The remaining dwell time declared by the owner for each vehicle in the network is simultaneously read, and dynamic addition / deletion logs of capacity bag boundary nodes generated by vehicle access and disconnection are recorded. The timestamp of plug-in / plug-out events, remaining dwell time, and boundary node changes are matched and associated using the vehicle's unique identifier. Anomalies such as timestamp misalignment and missing dwell time fields are removed to obtain a real-time status data set of in-network vehicles. For this real-time status data set, time slices are divided with a fixed step size. The remaining dwell time field is extracted from each time slice. Kernel density estimation is used to depict the distribution of remaining dwell time across different time slices. The peak position, tail decay slope, and left tail quantile are extracted from this distribution. The short-term variation amplitude is measured by combining the morphological differences between adjacent time slices to obtain the distribution characteristics of the remaining dwell time. Based on the distribution characteristics, vehicles whose remaining dwell time falls within the left tail quantile interval of the distribution are identified as short-staying vehicles. The capacity share corresponding to the capacity bag boundary node occupied by the short-staying vehicles is calculated. The capacity share is weighted and converted in combination with the proportion of short-staying vehicles in the total number of vehicles in the network to obtain the proportion of capacity unavailability caused by early departure, which is used as the actual reduction of the upper edge of adjustable capacity.

[0019] In one embodiment, the charging pile network is deployed in a commercial park or public parking lot. The communication interface at the charging pile adopts a message reporting mechanism in coordination with the host computer. Whenever an electric vehicle inserts or removes the charging gun, the charging pile immediately pushes an event message to the dispatching backend. The message carries the charging pile number, the vehicle's unique identifier, and the timestamp of the insertion / removal event. The timestamp accuracy is usually controlled at the second level.

[0020] Specifically, the remaining dwell time for each vehicle on the network is calculated by converting the departure time scheduled by the owner on the vehicle's infotainment system or mobile application. The remaining dwell time is obtained by subtracting the current time from the scheduled departure time, in minutes. The capacity bag refers to the power adjustment range that the charging pile network can promise to the upper-level power grid at a certain time segment. The upper and lower edges of this range change incrementally or decrementally with vehicle access and disconnection. Each change corresponds to a boundary node on the time-varying curve of the capacity bag. The addition and deletion logs of boundary nodes are written to disk in real time by the capacity bag maintenance process in the scheduling backend.

[0021] It should be noted that the scheduling backend performs external join matching on the three types of records: plug-in / plug-out event timestamp, remaining stay duration, and boundary node change, based on the vehicle's unique identifier. For timestamp misalignment caused by delayed message arrival and missing stay duration fields caused by the vehicle owner's failure to make an appointment, these are removed from the matching results to obtain a set of real-time status data of vehicles on the network. This set is organized with vehicles as rows and the three types of fields as columns.

[0022] Specifically, the real-time status data set of vehicles on the network is divided into time slices with a fixed step size. In one implementation, the step size is 15 minutes. Each time slice corresponds to a sample group of the remaining dwell time of all vehicles on the network within that slice. Kernel density estimation is used to perform nonparametric density fitting on the sample group. Kernel density estimation is a method to reconstruct the probability density by replacing each sample point with a kernel function centered on it and then averaging the results. The commonly used kernel function is the Gaussian kernel, and the bandwidth parameter is determined empirically based on the sample group size.

[0023] Specifically, the following features are extracted from the density curve output by kernel density estimation: the peak position is the remaining dwell time value corresponding to the point of maximum density, reflecting the mode dwell time level of vehicles on the network; the tail decay slope is obtained by taking the reciprocal of the horizontal span of the density curve from the peak height to one-tenth of its value on the right side of the peak, characterizing the rate of decay of the proportion of long-staying vehicles in the sample; the left tail quantile is the remaining dwell time value corresponding to the cumulative distribution function equal to 0.1 or 0.2, serving as the threshold for short-stay determination. The sum of the absolute values ​​of the point-by-point differences in the kernel density curves between adjacent time slices is used to measure the short-term fluctuation amplitude, thus obtaining the distribution characteristics of the remaining dwell time.

[0024] Preferably, vehicles with a remaining dwell time lower than the value corresponding to the left tail quantile are classified as short-stay vehicles. In the capacity bag boundary nodes, the boundary node triggered by vehicle access corresponds to a capacity contribution interval extending from the access time to the scheduled departure time, the length of which is determined by the vehicle's remaining dwell time. Further, the area of ​​the capacity contribution interval corresponding to each boundary node triggered by a short-stay vehicle is summed to obtain the cumulative capacity share of short-stay vehicles. Then, the cumulative capacity share is weighted and calculated using the ratio of the number of short-stay vehicles to the total number of vehicles in the network as a weight, yielding the proportion of capacity unavailability caused by early departure. This proportion is multiplied by the value of the upper edge of the capacity bag in the current time slice to obtain the actual reduction in the adjustable capacity upper edge.

[0025] Understandably, for scenarios with a high proportion of missing vehicle owner reservation information, the sliding average of the vehicle's historical dwell time can be used as a substitute estimate of the remaining dwell time and input into the matching and association process. The resulting reduction amount is continuously updated as the time slice moves forward.

[0026] S102. When the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the stability threshold set by the system, identify the period of sudden increase in event density per unit time and mark the end of the period as the potential jump point position where the upper edge of the adjustable capacity may jump upward.

[0027] For the proportion of short-staying vehicles in the total number of vehicles on the network and the actual reduction in the upper limit of adjustable capacity, the proportion and the actual reduction are normalized according to their respective value ranges and then summed to obtain a total value. If the total value exceeds a pre-calibrated stability threshold, a sliding time window scanning process is triggered, and a start signal for the sliding time window scanning is obtained. When the start signal for the sliding time window scanning is received, a sliding time window with a preset fixed window width and a preset step length is pushed forward on the insertion and removal event timestamp sequence. For each sliding time window, the number of insertion and removal events falling into the window is counted and divided by the window width to obtain the insertion and removal event density at the corresponding time of the sliding time window. The insertion and removal event density sequence is formed by sequentially splicing the sequence. The density increment of adjacent time moments is obtained by first-order difference on the insertion and removal event density sequence. For continuous segments where the density increment exceeds a preset surge threshold, it is determined as a period of sudden increase in event density. For the period of sudden increase in event density, the start and end times of the period are extracted. The unique vehicle identifier and remaining dwell time of the plug-in / plug-out events in the period are back-checked. False surge segments dominated by vehicles with dwell time higher than the left tail quantile are removed from the back-check results. True surge segments dominated by vehicles with short dwell times are retained. The corresponding position of the end time of the true surge segment on the time-varying curve of the capacity bag is marked as the potential jump point position of the adjustable capacity upper edge upward jump.

[0028] Specifically, regarding the proportion of short-staying vehicles and the actual reduction in the upper limit of adjustable capacity, the former is a dimensionless percentage, while the latter is a power reduction value in the kilowatt range. Since their dimensions are inconsistent, the proportion of short-staying vehicles and the actual reduction need to be mapped to the 0-1 interval using maximum-minimum normalization according to their respective value ranges. The implementation method of maximum-minimum normalization is to subtract the minimum value within the current sliding window from the value, and then divide by the difference between the maximum and minimum values. After normalization, the values ​​are added together to obtain the sum, denoted as s.

[0029] In one embodiment, the pre-calibrated stability threshold is obtained based on offline statistics.

[0030] Specifically, several stable periods where the capacity bag boundary nodes did not change are selected from historical operation records. A sample of the sum value corresponding to each stable period is statistically analyzed, and the 0.95 quantile of the sample is taken as the stability threshold. When s exceeds the stability threshold, a sliding time window scanning process is triggered, and a start signal for the sliding time window scanning is obtained; when s does not exceed the stability threshold, the scanning process remains dormant.

[0031] It should be noted that the window width and step length of the sliding time window are selected based on the average arrival rhythm of events in the charging pile network. In one embodiment, the window width is 5 minutes and the step length is 1 minute. A window width that is too short will result in sparse count samples, causing fluctuations in density estimation, while a window width that is too long will smooth out true bursts of activity.

[0032] Specifically, as the sliding time window moves forward on the insertion / removal event timestamp sequence, a counting operation is performed on each sliding time window. Insertion / removal event entries falling within the start and end range of that time window are scanned in the timestamp sequence, and the event count is accumulated to obtain c. Then, c is divided by the window width w (in minutes) to obtain the insertion / removal event density d corresponding to the center time of that sliding time window, where d = c / w, and the unit is times per minute. All d values ​​corresponding to the sliding time windows are concatenated in chronological order to form the insertion / removal event density sequence. This sequence is arranged with equal steps, reflecting the density of insertion / removal events on the time axis.

[0033] In one embodiment, a first-order hysteresis is added to the insertion / removal event density sequence, and the arithmetic mean of three adjacent sampling points is taken before being written into the sequence to smooth out sporadic random fluctuations. Further, a first-order difference is applied to the insertion / removal event density sequence. The first-order difference refers to the operation of subtracting the adjacent preceding value from the subsequent value in the sequence, and the difference result is denoted as Δd. A positive value of Δd indicates that the insertion / removal events tend to be dense, while a negative value indicates that they tend to be sparse. All Δd values ​​are arranged in chronological order to form a density increment sequence.

[0034] Preferably, the surge threshold is pre-calibrated based on three times the standard deviation of the Δd samples during historical stable periods. When several consecutive sampling points in the density increment sequence have Δd values ​​exceeding the surge threshold, this consecutive segment is identified as a period of sudden event density increase, and its start and end times are recorded.

[0035] Specifically, for the period of sudden increase in event density, the unique vehicle identifier and remaining dwell time corresponding to each plug-in / plug event within that period are queried from the real-time status data set of vehicles on the network. Plug-in / plug events triggered by vehicles with remaining dwell time higher than the left tail percentile are included in the background segment, while plug-in / plug events triggered by vehicles with remaining dwell time lower than the left tail percentile are included in the short-stop dominant segment. If the background segment dominates during the period of sudden increase in event density, that period is determined to be a false surge segment and is removed; if the short-stop dominant segment dominates, it is retained as a true surge segment.

[0036] In one embodiment, the dominance relationship is determined by a simple majority of events.

[0037] Understandably, the end time corresponding to each of the aforementioned real burst segments is projected onto the time-varying capacity bag curve. This time-varying capacity bag curve is obtained by plotting the upper edge values ​​of the capacity output by the scheduling backend at each time step, and the corresponding positions are recorded as potential transition point positions. Multiple potential transition point positions are arranged in chronological order to form a transition point position sequence, which serves as the position index for subsequent capacity boundary smoothing processing.

[0038] S103. Extract the capacity bag boundary node addition and deletion records associated with each potential transition point location, analyze whether the newly added boundary node is mainly triggered by the concentrated access of short-staying vehicles, and if it is confirmed that the transition is caused by unsustainable short-term access, then reset the capacity accumulation starting point at the transition point and generate a boundary time-varying description.

[0039] For each potential transition point in the potential transition point location sequence, a preset observation span is extended forward and backward from the potential transition point location as the center. The addition and deletion records of the boundary nodes of the content bag within the observation span are extracted. The addition and deletion records include the generation time of each newly added boundary node, the corresponding vehicle unique identifier and contribution capacity value, and the disappearance time and corresponding contribution capacity value of each deleted boundary node. The addition and deletion records within the observation span are arranged in chronological order to obtain the boundary node variation spectrum around the potential transition point location. For the subset of newly added boundary nodes in the boundary node variation spectrum located before the potential jump point, the remaining dwell time of the vehicle corresponding to each newly added boundary node is extracted. The ratio of the number of vehicle entries with remaining dwell time lower than the left tail quantile in the subset of newly added boundary nodes to the total number of entries in the subset is calculated to obtain the short dwell trigger ratio of the newly added boundary nodes. If the short dwell trigger ratio exceeds a pre-calibrated attribution threshold, and the generation times of the newly added boundary nodes are all less than a preset dense interval threshold, then the jump at the potential jump point is determined to be caused by unsustainable short-term access, and a true jump point is obtained after trigger determination. Using the time corresponding to the true jump point as the reset point for capacity accumulation, the capacity bag time-varying curve is reset at the reset point. The capacity component contributed by short-dwelling vehicles before the reset point is removed, and the capacity share continuously committed by the remaining on-network vehicles after the reset point is used as the new accumulation starting point. The two segments of the curve before and after the reset point are aligned and smoothed to remove step discontinuities on the curve, resulting in a continuous and realistic time-varying description of the boundary.

[0040] Specifically, each potential jump point corresponds to a moment on the time-varying capacity bag curve where an upward step jump is likely to occur. Fine-grained attribution analysis around each potential jump point is crucial to preventing short-term false spikes in capacity, which could lead to unfulfilled promises to the upstream power grid. The observation span is selected based on the typical charging station network's entry and exit rhythm. In one implementation, the observation span is set to 30 minutes forward and 30 minutes backward, covering the complete link of an electric vehicle from plugging in, protocol handshake, power ramp-up, to plugging out. An observation span that is too narrow will result in insufficient samples, while one that is too wide will introduce remote nodes unrelated to the potential jump point locations.

[0041] Specifically, each entry in the added / deleted records is written to disk in real time by the capacity bag maintenance process in the scheduling background, using structured fields. A newly added boundary node entry includes the generation time *ta*, the vehicle's unique identifier *v*, and the contributed capacity value *pa*; a deleted boundary node entry includes the disappearance time *td*, the vehicle's unique identifier *v*, and the corresponding contributed capacity value *pd*. *ta* and *td* use second-level timestamps, while *pa* and *pd* are in kilowatts. Further, all added / deleted records within the observation span are arranged chronologically to form a boundary node change spectrum around the potential transition point. This boundary node change spectrum uses time as the horizontal axis and the algebraic sum of contributed capacity as the vertical axis. Upward steps represent the addition of new boundary nodes, and downward steps represent the departure of deleted boundary nodes; the step height reflects the capacity increment.

[0042] In one embodiment, for a subset of newly added boundary nodes in the boundary node change spectrum that are located before the potential transition point, the remaining dwell time field of the corresponding vehicle is accessed one by one for each of the newly added boundary nodes.

[0043] Specifically, the number of vehicle entries in the newly added boundary node subset whose remaining dwell time is lower than the left tail quantile is denoted as ns, and the total number of entries in this subset is denoted as nt. The short dwell time trigger ratio r = ns / nt. The left tail quantile is the value obtained in the previous step, and the remaining dwell time threshold corresponding to the cumulative distribution function is usually equal to 0.1 or 0.2. The attribution threshold is pre-calibrated based on the 0.9 quantile of r samples under the same statistical caliber during historical stable periods. In one embodiment, the attribution threshold of r is set to 0.6. When r exceeds the attribution threshold, it indicates that the capacity increase before the potential jump point is mainly contributed by short-dwelling vehicles.

[0044] Specifically, the dense interval threshold is used to constrain the generation time interval between newly added boundary nodes. In one embodiment, the dense interval threshold is set to 2 minutes. When the difference between adjacent generation times of the newly added boundary node is lower than the dense interval threshold, and r also exceeds the attribution threshold, the potential jump point is determined to be a true jump point; if any condition is not met, it is determined to be a false jump and discarded. Further, the time corresponding to each true jump point is taken as the reset point tr. The capacity accumulation starting point refers to the baseline ordinate of the capacity bag time-varying curve near tr used to accumulate new contributions.

[0045] Specifically, the starting point of the time-varying curve of the capacity bag is reset at tr, and the sum of the capacity components contributed by short-staying vehicles in the newly added boundary node subset is subtracted from the ordinate at the left limit of tr.

[0046] Specifically, the capacity contribution component is calculated as a proportion of the actual promised charging time of each short-stay vehicle to its original contribution range.

[0047] For example, if a vehicle originally contributes 7 kilowatts of capacity and has an original contribution interval of 60 minutes, but its actual remaining commitment time after the tr period is only 10 minutes, then the vehicle's capacity contribution is retained after being reduced to a ratio of 10 / 60.

[0048] It is understood that the endpoint alignment process makes the right endpoint of the curve before the reset point and the left endpoint of the curve after the reset point strictly equal on the ordinate; the smooth connection process uses a three-point moving average for several sampling points near tr to remove the sharp step discontinuity caused by the start point reset, and obtains a continuous and time-varying description of the boundary that reflects the true availability.

[0049] Preferably, for three typical scenarios in the charging pile network—afternoon rush hour, midday charging, and nighttime long parking—the observation span can be set to 35 minutes, 20 minutes, and 90 minutes respectively, thereby adapting to the event rhythm of different time periods and ensuring that the boundary time-varying description can reflect the true trend of the promised capacity in all scenarios.

[0050] S104. Based on the time-varying boundary description, integrate the number of vehicles on the network and the vehicle dwell time to determine the index for improving aggregation credibility.

[0051] The number of vehicles on the network corresponding to the time-varying description of the boundary and the current remaining dwell time of each vehicle on the network are extracted for each time segment. The remaining dwell time is discretized according to a preset unit time to obtain a duration tier distribution. The number of vehicles on the network and the duration tier distribution together constitute the input vector of the aggregation model. The input vector is arranged according to time segments to obtain an aggregated input sequence with time segments aligned. For each time segment in the aggregated input sequence, the aggregation model is used to integrate the number of vehicles on the network and the duration tier distribution. For each tier in the duration tier distribution, the number of vehicle entries corresponding to that tier is multiplied by the median dwell time of that tier to obtain the promised capacity share of that tier. The promised capacity shares of all tiers are summed one by one to obtain the integrated reliable capacity value of that time segment. The integrated reliable capacity values ​​are arranged according to time segments to form an integrated reliable capacity curve. The integrated reliable capacity curve is compared point by point with the original capacity bag curve under the same time segment without starting point reset. For each time segment, the ratio of the integrated reliable capacity value to the corresponding capacity value of the original curve is taken as the reliability value of that time segment. The average value of the reliability value is calculated along all time segments to obtain the improvement index of the aggregate reliability.

[0052] Specifically, the continuous, time-varying boundary description reflecting true availability is output from the front-end process and presented as a continuous curve with time on the horizontal axis and the upper edge of the capacity bag's committable value on the vertical axis. Based on this time-varying boundary description, the number of vehicles on the network and the vehicle dwell time are synchronously aggregated, and a reliability improvement index is output.

[0053] It should be noted that at each time segment, the dispatch backend polls the charging piles to obtain online snapshots. These online snapshots contain three types of fields: the unique vehicle identifier of all currently connected charging piles, the remaining dwell time, and the current rated power of the charging pile. The number of entries in the online snapshots is counted to obtain the number of vehicles N in service at that time segment; the remaining dwell time field is read for each entry to obtain a sample group of remaining dwell time for that time segment.

[0054] Specifically, the remaining dwell time sample group is discretized according to a preset unit time to obtain the duration tier distribution. In one embodiment, the preset unit time is 15 minutes. The remaining dwell time between 0 and 15 minutes is assigned to the first tier, between 15 and 30 minutes to the second tier, and so on, until the upper limit of the 6-hour tier is covered. The value of each tier is represented by the midpoint of the upper and lower bounds of the tier, called the median dwell time of that tier, corresponding to 7.5 minutes and 22.5 minutes respectively. The number of vehicle entries falling into each tier is the number of vehicle entries for that tier. Further, the number of vehicles on the network and the duration tier distribution are concatenated into a vector. The first component of the vector is N, and the remaining components are arranged according to the tier order corresponding to the number of vehicle entries, to obtain the input vector of this time segment. The input vector is arranged in the order of the time segment to form an aggregated input sequence.

[0055] In one embodiment, the aggregation model is a weighted accumulation structure based on gear level.

[0056] Specifically, the input vector for each time segment is integrated as follows: the product of the number of vehicle entries ni in the i-th gear and the median dwell time τi in that gear is taken to obtain the promised capacity share qi = ni × τi × Pavg for that gear, where Pavg is the average rated power at the pile end, obtained by taking the average value from the rated power field of the online snapshot. The integrated reliable capacity value Q for that time segment is obtained by summing qi for all gears. The weighted accumulation method upgrades the simple vehicle count to a joint measure of vehicle count and dwell time, avoiding a large number of short-stop vehicles falsely inflating the promised capacity. In one embodiment, long-stop vehicles with gear numbers exceeding a preset upper limit are also included to avoid the aggregation result being insensitive to the tail end of long-stop vehicles.

[0057] It is understood that the integrated reliable capacity value Q is arranged according to time sections to obtain the integrated reliable capacity curve. The integrated reliable capacity curve is aligned with the boundary time-varying description on the time axis, and the value of the vertical axis does not exceed the value of the corresponding position of the boundary time-varying description.

[0058] Specifically, the original capacity bag curve without starting point reset is obtained from the upper edge trend of the capacity bag retained by the scheduling backend before executing the starting point reset process. For the integrated reliable capacity curve and the original capacity bag curve, at each time segment, the ratio of the integrated reliable capacity value Q to the corresponding capacity value Q0 of the original curve is taken, denoted as the reliability value ρ=Q / Q0; ρ is located in the range of 0 to 1, and the closer it is to 1, the closer the aggregated output is to the upper edge of the original statement.

[0059] Preferably, the arithmetic mean of ρ is calculated along all time segments, and the mean is further multiplied by the proportion of non-short-term parking periods as a weighting correction term to obtain the improved index of aggregate credibility. In one implementation, for weekday midday recharging scenarios, the improved index value is relatively small, reflecting strong disturbance from short-term parking vehicles; for nighttime long-term parking scenarios, the improved index value is relatively large, reflecting a high degree of consistency between the credible capacity and the original claim.

[0060] S105. Determine whether it is necessary to iteratively update the time-varying boundary description based on the aggregation reliability improvement index until the output of the promised capacity space that meets the scheduling reliability requirements is reached.

[0061] The improvement index of the aggregation reliability is obtained, and compared with a preset scheduling reliability threshold. If the improvement index is lower than the scheduling reliability threshold, an iterative update instruction for the boundary time-varying description is triggered; if the improvement index is not lower than the scheduling reliability threshold, a convergence termination instruction is triggered, and the iterative triggering determination result is obtained. When the iterative update instruction is received, the left tail quantile of the short dwell determination is tightened one level by a preset step size in the direction of smaller values, and the starting point reset process of the boundary time-varying description is re-executed to update the improvement index; when the convergence termination instruction is received, the current integrated reliability capacity curve is output as the committable capacity space that meets the scheduling reliability requirements.

[0062] Specifically, the scheduling reliability threshold is obtained through offline statistics from the scheduling backend. Several stable periods are selected from historical operation records where capacity requests were submitted to and actually fulfilled by the superior power grid. Samples of improvement indicators corresponding to each stable period are statistically analyzed, and the 0.05 quantile of these samples is taken as the scheduling reliability threshold. In one embodiment, the scheduling reliability threshold is set to 0.85. When the improvement indicator is lower than this value, an iterative update instruction for the boundary time-varying description is triggered; when the improvement indicator is not lower than this value, a convergence termination instruction is triggered.

[0063] It should be noted that the preset step size of the left tail quantile is set according to the cumulative distribution function scale.

[0064] In one embodiment, the left tail quantile values ​​are 0.20, 0.18, 0.15, 0.12, and 0.10, respectively. When the iterative update instruction is received, the left tail quantile value is tightened by one level from the current value in the order of the scale towards smaller values. The threshold for determining short-stay vehicles thus becomes more stringent; lightly short-stay vehicles that were originally included in the short-stay set are removed from the set, and the capacity contribution component removed by the starting point reset process decreases accordingly, resulting in an overall increase in the vertical axis of the integrated reliable capacity curve. Based on the updated integrated reliable capacity curve, the reliability value is recalculated and the average is calculated to obtain the updated improvement index.

[0065] Understandably, the updated improvement index re-enters the threshold comparison process, repeating this cycle until the convergence termination instruction is received.

[0066] Preferably, an upper limit is set for the number of tightening cycles at the left tail quantile. If the convergence termination command is not triggered after the number of cycles reaches this upper limit, the current integrated reliable capacity curve is forcibly used as the output. The integrated reliable capacity curve is the committable capacity space that meets the scheduling reliability requirements. After being reported to the superior power grid dispatching system, it is used as the power regulation boundary committed by the charging pile network to the outside world.

[0067] S106. Simulate the fulfillment rate using the committable capacity space. Input historical scheduling plans and actual vehicle off-grid records, replay and verify the completion of scheduling instructions under the capacity constraint, and output the capacity boundary.

[0068] The promised capacity space is obtained and used as the capacity constraint for the fulfillment rate simulation. Historical scheduling plans are retrieved from the scheduling backend, containing power adjustment command values ​​issued to the charging pile network at each time segment. Actual vehicle off-grid records are retrieved from the vehicle pile logs, containing the actual off-grid time of each vehicle. The historical scheduling plans and actual vehicle off-grid records are aligned and merged according to time segments to obtain a replay dataset. The replay dataset is loaded into the fulfillment rate simulation, and the power adjustment command values ​​in the historical scheduling plans are replayed sequentially according to time segments. For each power adjustment command value, vehicle entries whose actual off-grid time is later than that time segment are selected from the actual vehicle off-grid records. The rated power of the corresponding pile for each vehicle entry is accumulated to obtain the capacity contribution. The lower edge value of this contribution is taken at the same time segment as the capacity constraint to obtain the upper limit of power that can be actually fulfilled at that time segment. The ratio of the upper limit of power to the power adjustment command value is taken as the command completion degree. The command completion degrees are arranged according to time segments to form a fulfillment rate sequence. For the fulfillment rate sequence, for each time segment where the completion rate of the instruction is lower than the preset fulfillment lower limit, the value of the capacity constraint at the same time segment is located, and the value is corrected downward. The correction amount is the capacity concession amount calculated proportionally from the difference between the completion rate of the instruction and the preset fulfillment lower limit. The corrected capacity constraint sequence is spliced ​​together by time segments to serve as a capacity boundary that can be used for unified scheduling of multi-source heterogeneous resources.

[0069] Specifically, the fulfillment rate simulation module is deployed in the offline calculation partition of the scheduling backend and mainly consists of three parts connected in series: a data loading subunit, a call playback subunit, and a completion calculation subunit. The operation of the fulfillment rate simulation module does not affect the real-time control of the charging pile network by the scheduling backend. It only completes historical playback verification before the capacity boundary is released and outputs a relatively stable capacity boundary for subsequent unified scheduling of multi-source heterogeneous resources.

[0070] It should be noted that the historical scheduling plan is provided by the instruction archive in the scheduling backend. In one embodiment, the instruction archive retains the issuance records for nearly 30 calendar days at 5-minute time intervals. The issuance record for each time interval includes the power adjustment instruction value for the charging pile network, the start and end times of the target fulfillment window, and the instruction number. The power adjustment instruction value is usually in kilowatts, with positive values ​​representing upward adjustment (increasing charging power) and negative values ​​representing downward adjustment (reducing charging power).

[0071] Specifically, the actual vehicle off-grid records are compiled from the events of charging station disconnection. Each vehicle off-grid entry includes a unique vehicle identifier, the corresponding charging station number, and the actual off-grid time. The actual off-grid time is recorded on a disc according to the charging station's clock, with an accuracy typically controlled at the second level.

[0072] In one embodiment, entries where the deviation between the pile-end clock and the scheduling backend clock exceeds a preset tolerance are drift-corrected before being written. Further, the historical scheduling plan and the actual vehicle departure records are aligned and merged by time segments. Using the time segments of the scheduling plan as a skeleton, the index of the time segment to which each actual departure time belongs is appended to the corresponding position in the skeleton to form a playback dataset.

[0073] In one embodiment, the playback subunit proceeds sequentially from the earliest time segment of the playback dataset, reading the power adjustment command value of that time segment for each shot and executing the following fulfillment calculation process.

[0074] Specifically, the actual vehicle off-grid records are scanned, and vehicle entries whose actual off-grid time is later than the specified time segment are selected. These vehicle entries represent those still connected to the network at that time segment. The rated power of the charging pile corresponding to each vehicle entry is read from the charging pile network's pile-end archive, denoted as Pi, where i is the vehicle entry number. The Pi values ​​for all vehicle entries are summed to obtain the capacity contribution C for that time segment. Simultaneously, the capacity constraint value U is taken from the corresponding position of the capacity constraint at that time segment. The upper limit of power is the smaller value between C and U, denoted as L=min(C,U). The power adjustment command value R for that time segment is compared with L to obtain the command completion rate η=L / R, where η is between 0 and 1. η close to 1 indicates that the command can be fully fulfilled, while a smaller η indicates that the command cannot be fully fulfilled due to capacity constraints or insufficient connected vehicles.

[0075] It is understood that the fulfillment rate sequence is obtained by arranging the completion rates η of each instruction in chronological order. In one embodiment, the arithmetic mean of η for three adjacent time segments is written into the sequence to smooth out random jitter in individual samples.

[0076] Preferably, the preset performance limit is determined based on the minimum fulfillment rate agreed upon by the upper-level power grid for frequency regulation or backup auxiliary services, and a commonly used value is 0.9.

[0077] In one embodiment, for time segments where the instruction completion rate η is less than 0.9, the deviation d = 0.9 - η is multiplied by a preset conversion factor α to obtain the capacity concession amount for that time segment. In one implementation, α is taken as 1. The corrected value is obtained by subtracting Δ from the corresponding value U of the capacity constraint at that time segment. .

[0078] Specifically, after applying the downward correction to all time segments, the corrected values ​​U' are spliced ​​together in the order of the time segments to form a capacity boundary curve, which serves as the capacity boundary that can be used for unified scheduling of multi-source heterogeneous resources. The capacity boundary, along with the adjustable range of energy storage and the adjustable range of interruptible loads, are aggregated under a unified time-series coordinate system and reported to the superior power grid for unified use in intraday or real-time scheduling.

[0079] S107. The adjustable capacity of the charging pile reconstructed based on dwell behavior is encapsulated with the adjustment capabilities of other types of resources to realize the aggregation and invocation of heterogeneous resources under a unified semantics.

[0080] The capacity boundary is used as the input for the adjustable capacity of the charging pile in the time-series availability dimension. For energy storage resources, their state of charge curve, maximum charging and discharging power, and sustainable discharge duration are retrieved. For interruptible loads, their load reporting value, interruptible power limit, and maximum allowable interruption duration are retrieved. The retrieved parameters corresponding to the energy storage resources and the retrieved parameters corresponding to the interruptible loads constitute a set of original parameters for heterogeneous resources. The capacity boundary and the original parameter set of heterogeneous resources are mapped according to a pre-agreed set of time-series availability fields, which includes time section index, adjustable power, adjustable power, sustainable duration, response delay, and ramp rate. After mapping, adjustable capacity entries for charging piles, adjustable capacity entries for energy storage, and adjustable capacity entries for interruptible loads are obtained. The three types of entries share the same time-series coordinate reference. For the adjustable capacity entries of the charging pile, the adjustable capacity entries of the energy storage, and the adjustable capacity entries of the interruptible load, each entry is encapsulated according to a preset standard encapsulation protocol. A resource type label, a unique resource identifier, and a unit of measurement are added to the beginning of the entry. The value sequence corresponding to the time-series availability field set is filled into the body of the entry. A version number and a timestamp are appended to the end of the entry. The adjustable power and adjustable power in the body of each entry are scaled to a unified kilowatt dimension, and the response delay is scaled to a unified second dimension. After encapsulation, a resource standard entry with unified semantics is obtained. The resource standard entries are aligned along the time cross-section and arranged in the same field order. The adjustable power and adjustable power in each of the resource standard entries are summed by substituting their values ​​according to time segments. The maximum value is taken for response delay, and the minimum value is taken for ramp rate, to obtain the aggregated adjustable capacity vector under the time segment. The aggregated adjustable capacity vector is spliced ​​according to the time segment to form the aggregated adjustable capacity time sequence. The aggregated adjustable capacity time sequence is exposed to the upper-level power grid dispatch through a pre-agreed calling interface. The calling interface receives the power adjustment instructions issued by the upper-level power grid dispatch and distributes them according to the resource type tags in the resource standard entries, to obtain the aggregation and calling results of heterogeneous resources under a unified semantics.

[0081] The unified modeling framework for multi-source heterogeneous adjustable resources uses time-series availability as the common semantic layer, and uniformly characterizes adjustable resources of different forms as a set of adjustable capacity fields arranged according to time sections. This makes the adjustable capacity of charging piles, adjustable capacity of energy storage, and adjustable capacity of interruptible loads appear as entries with the same structure in the dispatching background, which facilitates the issuance of unified instructions by the upper-level power grid dispatching.

[0082] Specifically, the time-series availability field set is a set of public field definitions exposed by heterogeneous resources, including six items: time segment index, adjustable power, adjustable power, sustainability duration, response latency, and ramp rate. The time segment index is numbered at 5-minute intervals, covering a rolling scheduling cycle of the next 24 hours. Adjustable power refers to the power margin that a resource can increase at a given time segment, measured in kilowatts; adjustable power refers to the power margin that can be reduced, also measured in kilowatts. Sustainability duration refers to the longest number of minutes a resource can continuously operate at either the adjustable power or the adjustable power without leaving the adjustable state. Response latency refers to the number of seconds from receiving the instruction until the power change reaches the target value. Ramp rate refers to the maximum kilowatt value that the power can change per second.

[0083] Specifically, for the capacity boundary obtained from the pre-processing stage, the value of the capacity boundary curve at that cross-section is read according to the time cross-section and written into the adjustable power field; the value of the rated reduceable capacity of the charging pile network at that cross-section is written into the adjustable power field; the committable duration span corresponding to the capacity boundary is written into the sustainable duration field; and the fixed value from the charging pile controller's handshake at the protocol layer to the power response is written into the response delay field, thus obtaining the adjustable capacity entry for the charging pile. Further, for the energy storage resource, its state of charge curve is retrieved. The state of charge refers to the ratio of the energy storage device's current remaining dischargeable capacity to its rated capacity, with a value between 0 and 1. The maximum charging and discharging power is divided into maximum charging power and maximum discharging power, which are written into the adjustable power field and the adjustable power field, respectively. The sustainable discharge duration is obtained by multiplying the state of charge by the rated capacity and then dividing by the maximum discharge power, and is written into the sustainable duration field, thus obtaining the adjustable capacity entry for the energy storage.

[0084] It is understandable that after retrieving the load reporting value for interruptible loads, the upper limit of interruptible power is written into the adjustable power field, the adjustable power field is set to zero in the interruptible load entry, and the longest allowed interruption duration is written into the sustainable duration field to obtain the adjustable capacity entry for interruptible loads.

[0085] In one embodiment, the standard packaging specification stipulates that each adjustable capacity entry consists of three segments: a header, a body, and a tail.

[0086] Specifically, the header includes a resource type label, a unique resource identifier, and a unit of measurement. The resource type label can be one of three types: charging pile, energy storage, or interruptible load. The unique resource identifier is assigned by the scheduling backend to avoid duplicate names. The unit of measurement specifies that adjustable power is expressed in kilowatts, continuous duration in minutes, and response delay in seconds. The body is filled with the value sequence corresponding to the time-series availability field set, arranged in time-section index order. The tail appends a version number and a timestamp. The version number distinguishes the differences of the same resource in different release cycles, and the timestamp records the time the entry was written to disk.

[0087] Specifically, the dimension normalization process is performed after the main data is entered. If the original values ​​of adjustable power and adjustable power are in megawatts, they are multiplied by 1000 and uniformly converted to kilowatts; if the original value of response delay is in milliseconds, it is divided by 1000 and uniformly converted to seconds. The normalized resource standard entries are aligned along the time segment and arranged in the same field order for easy column-by-column reading. Further, the aggregation operation is performed on all resource standard entries along the same time segment: the adjustable power of each entry is replaced by the sum of the adjustable power values ​​to obtain the total adjustable power of that time segment; the adjustable power of each entry is replaced by the sum of the adjustable power values ​​to obtain the total adjustable power; the maximum value of the response delay reflects the response constraint of the slowest resource; the minimum value of the ramp rate reflects the ramp constraint of the slowest resource; and the six fields are combined to form the aggregated adjustable capacity vector for that time segment.

[0088] It is understood that the calling interface exposes multi-channel interfaces such as adjustable power sequences, adjustable power sequences, response delay sequences, and ramp rate sequences. When the upper-level grid dispatcher initiates a request, the calling interface returns the aggregated adjustable capacity timing sequence and receives the power adjustment command issued by the upper-level grid dispatcher. The power adjustment command carries the target power value and the target time segment. The calling interface internally allocates the target power value according to the proportion of adjustable power or adjustable power in each of the resource standard entries, and then distributes the allocated sub-commands to the charging pile dispatching channel, energy storage dispatching channel, and interruptible load dispatching channel according to the resource type tag.

[0089] Preferably, when the state of charge of the energy storage device is close to its safe lower limit, the calling interface is used to force the adjustable power field of the standard entry of the energy storage resource to be set to zero, so as to avoid the distributed component exceeding the actual discharge capacity; for the adjustable capacity entry of the charging pile, the capacity boundary value is reset after the position marked by the true jump point in the preceding link, so that the aggregated output is consistent with the actual achievable level.

[0090] like Figure 2 This invention provides a unified modeling system for multi-source heterogeneous adjustable resources, mainly comprising: The data acquisition and analysis module is used to acquire information on the timing of plugging and unplugging events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion of capacity bag boundary nodes from the charging pile network. It obtains a set of real-time status data of vehicles on the network, analyzes the distribution characteristics of the remaining dwell time of vehicles, and quantifies the actual reduction of the upper edge of adjustable capacity. The jump point identification module is used to identify the period of sudden increase in event density per unit time when the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the stability threshold set by the system, and to mark the end of the period as the potential jump point position where the upper edge of the adjustable capacity may jump upward. The boundary description generation module is used to extract the addition and deletion records of capacity bag boundary nodes associated with each potential jump point, analyze whether the newly added boundary nodes are mainly triggered by the concentrated access of short-staying vehicles, and if it is confirmed that the jump is caused by unsustainable short-term access, then the capacity accumulation starting point is reset at the jump point to generate a time-varying boundary description. The credibility determination module is used to integrate the number of vehicles on the network and the vehicle dwell time based on the boundary time-varying description to determine the improvement index of aggregate credibility; The iterative update module is used to determine whether the boundary time-varying description needs to be iteratively updated based on the aggregation reliability improvement index, until the output of the promised capacity space that meets the scheduling reliability requirements is reached. The fulfillment simulation module is used to simulate the fulfillment rate using the committable capacity space. It takes historical scheduling plans and actual vehicle off-grid records as input, replays and verifies the completion of scheduling instructions under the capacity constraint, and outputs the capacity boundary. The resource encapsulation module is used to encapsulate the adjustable capacity of charging piles reconstructed based on dwell behavior, along with the adjustment capabilities of other types of resources, to achieve the aggregation and invocation of heterogeneous resources under a unified semantic framework.

[0091] It should be noted that the above examples are merely some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments and many variations are possible. All variations that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should be considered within the scope of protection of this invention.

Claims

1. A unified modeling method for multi-source heterogeneous adjustable resources, characterized in that, The method includes: By obtaining the timing of plug-in / plug-out events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion information of capacity bag boundary nodes from the charging pile network, a real-time status data set of vehicles on the network is obtained. The distribution characteristics of the remaining dwell time of vehicles are analyzed, and the actual reduction of the upper limit of adjustable capacity is quantified. When the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the stability threshold set by the system, the system identifies the period of sudden increase in event density per unit time and marks the end of the period as the potential jump point where the upper edge of the adjustable capacity may jump upward. Extract the capacity bag boundary node addition and deletion records associated with each potential transition point location, analyze whether the newly added boundary node is mainly triggered by the concentrated access of short-staying vehicles, and if it is confirmed that the transition is caused by unsustainable short-term access, then reset the capacity accumulation starting point at the transition point and generate a boundary time-varying description. Based on the time-varying description of the boundary, the number of vehicles on the network and the duration of vehicle stay are integrated to determine the index for improving aggregation credibility; Determine whether it is necessary to iteratively update the boundary time-varying description based on the aggregation reliability improvement index until the output of the promised capacity space that meets the scheduling reliability requirements is reached; The fulfillment rate is simulated using the committable capacity space. Historical scheduling plans and actual vehicle off-grid records are input, and the completion of scheduling instructions under the capacity constraint is verified by replaying the data. The capacity boundary is then output. The adjustable capacity of charging piles reconstructed based on dwell behavior is encapsulated with the adjustment capabilities of other types of resources, enabling the aggregation and invocation of heterogeneous resources under a unified semantic framework.

2. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, The process involves acquiring information from the charging pile network, including the timing of plug-in / plug-out events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion of capacity bag boundary nodes, to obtain a real-time status data set of vehicles on the network. This data is then used to analyze the distribution characteristics of the remaining dwell time of vehicles and quantify the actual reduction in the adjustable capacity upper limit. The charging pile network's communication interface collects plug-in / plug-out event timestamps, synchronously reads the remaining dwell time of vehicles on the network, and records the dynamic addition and deletion logs of capacity bag boundary nodes. The timestamps of the plugging / unplugging events, the remaining dwell time, and the changes in boundary nodes are matched and associated with vehicle identifiers, and abnormal data is removed to obtain a set of real-time status data. The real-time status data set is divided into time slices, the remaining dwell time field is extracted, the distribution pattern is described by kernel density estimation, the peak position and quantile are extracted, the short-term fluctuation amplitude is measured, and the distribution characteristics are obtained. Based on the distribution characteristics, short-stay vehicles are identified, and the capacity occupancy share of short-stay vehicles is statistically calculated. The unavailable capacity ratio is then weighted and calculated as a reduction in the adjustable capacity.

3. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 2, characterized in that, Based on the distribution characteristics, short-stay vehicles are identified, their capacity occupancy share is statistically analyzed, and a weighted average of these vehicles is used to calculate the unavailable capacity ratio, which serves as a reduction in the adjustable capacity. This includes: Based on the distribution characteristics, vehicles whose remaining dwell time falls within the left tail quantile interval of the distribution are identified as short-stay vehicles. The capacity share corresponding to the capacity bag boundary node occupied by the short-stay vehicles is calculated. The capacity share is weighted and converted in combination with the proportion of short-stay vehicles in the total number of vehicles on the network, so as to obtain the proportion of capacity unavailability caused by early departure, which is used as the reduction amount of the adjustable capacity.

4. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, When the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the system's set stability threshold, the system identifies periods of sudden increase in event density per unit time and marks the end of these periods as potential jump points where the adjustable capacity upper edge may jump upwards, including: The sum of the proportion of vehicles with short stops and the aforementioned reduction amount is obtained by normalizing and summing. If the sum exceeds a preset stability threshold, a sliding time window scan is triggered; The density of insertion and removal events is statistically analyzed using the sliding time window to form a density sequence. The density increment is obtained by using the difference method to determine the period of sudden increase in event density. For the period of sudden increase, review the vehicle dwell time, eliminate false sudden increase segments, retain the real sudden increase segments dominated by vehicles with short dwell times, and mark the end time of the real sudden increase segment as the potential jump point position.

5. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, The process involves extracting the capacity bag boundary node addition and deletion records associated with each potential transition point, analyzing whether the newly added boundary nodes are mainly triggered by concentrated access from short-staying vehicles, and if it is confirmed that the transition is caused by unsustainable short-term access, then the capacity accumulation starting point is reset at the transition point, and a time-varying boundary description is generated, including: Extract the addition and deletion records of boundary nodes centered on the potential transition point to form a change spectrum; For the newly added boundary node subset in the aforementioned variation spectrum, the proportion of short-stay vehicles triggering the event is statistically analyzed. If the trigger ratio exceeds the preset attribution threshold and the time interval between the generation of new nodes is lower than the preset threshold, then the jump is determined to be caused by short-term access, and a true jump point is obtained. Using the true transition point as the reset point, the capacity contribution of short-staying vehicles is eliminated, and the remaining vehicle capacity is used as the new starting point to smoothly connect and generate the boundary time-varying description.

6. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, The step of integrating the number of vehicles on the network and the vehicle dwell time based on the boundary time-varying description to determine the improvement index of aggregation reliability includes: Extract the number of vehicles on the network and the remaining dwell time corresponding to the time-varying description of the boundary, and discretize them to obtain the duration distribution; Using the number of vehicles on the network and the distribution of time tiers as input, an aggregation model is used to calculate the promised capacity share of each tier, and the sum of the promised capacity shares of each tier is used to obtain the integrated reliable capacity value. After arranging the curves according to time sections to form an integrated credibility capacity curve, the curves are compared with the original curves, and the average credibility value is calculated to obtain the aggregate credibility improvement index.

7. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, The step of determining whether to iteratively update the boundary time-varying description based on the aggregation reliability improvement index until a committable capacity space that meets the scheduling reliability requirements is output includes: Compare the improvement index with the preset scheduling reliability threshold; If the improvement index is lower than the preset scheduling reliability threshold, an iterative update instruction is triggered to tighten the short stay determination fraction, re-execute the starting point reset process, and update the improvement index. If the improvement index is not lower than the preset scheduling reliability threshold, a convergence termination instruction is triggered, and the current integrated reliable capacity curve is output as the promised capacity space.

8. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, The simulation of fulfillment rate using the committable capacity space involves inputting historical scheduling plans and actual vehicle off-grid records, replaying and verifying the completion of scheduling instructions under the capacity constraint, and outputting the capacity boundary, including: Using the promised capacity space as a constraint, historical scheduling plans and vehicle off-network records are retrieved, aligned, and merged to form a replay dataset; The power adjustment command is replayed according to the time section. Vehicles whose off-grid time is later than the section are filtered out. The power is accumulated to obtain the capacity contribution. The lower edge value is taken with the constraint to calculate the command completion degree and form the fulfillment rate sequence. For sections in the fulfillment rate sequence where the completion rate is lower than a preset lower limit, the constraint values ​​are adjusted downwards, and the capacity boundary is generated by splicing them together.

9. The unified modeling method for multi-source heterogeneous adjustable resources according to claim 1, characterized in that, The adjustable capacity of charging piles reconstructed based on dwell behavior is encapsulated with the adjustment capabilities of other types of resources to achieve aggregation and invocation of heterogeneous resources under a unified semantic framework, including: Using the aforementioned capacity boundary as the adjustable capacity input for the charging pile, parameters of energy storage and interruptible load are retrieved; The capacity boundary and the parameters are mapped according to a preset time-series availability field set to generate each resource entry; The entries are encapsulated according to standard specifications and the scale is normalized to obtain unified semantic resource standard entries; The resource standard entry data is processed according to time segments to generate aggregated adjustable capability time series, which is exposed to the upper-level power grid dispatch through calling interface to realize the aggregation and calling of the heterogeneous resources.

10. A unified modeling system for multi-source heterogeneous adjustable resources, characterized in that, The system includes: The data acquisition and analysis module is used to acquire information on the timing of plugging and unplugging events, the remaining dwell time of a single vehicle, and the dynamic addition and deletion of capacity bag boundary nodes from the charging pile network. It obtains a set of real-time status data of vehicles on the network, analyzes the distribution characteristics of the remaining dwell time of vehicles, and quantifies the actual reduction of the upper edge of adjustable capacity. The jump point identification module is used to identify the period of sudden increase in event density per unit time when the sum of the proportion of short-staying vehicles and their capacity reduction exceeds the stability threshold set by the system, and to mark the end of the period as the potential jump point position where the upper edge of the adjustable capacity may jump upward. The boundary description generation module is used to extract the addition and deletion records of capacity bag boundary nodes associated with each potential jump point, analyze whether the newly added boundary nodes are mainly triggered by the concentrated access of short-staying vehicles, and if it is confirmed that the jump is caused by unsustainable short-term access, then the capacity accumulation starting point is reset at the jump point to generate a time-varying boundary description. The credibility determination module is used to integrate the number of vehicles on the network and the vehicle dwell time based on the boundary time-varying description to determine the improvement index of aggregate credibility; The iterative update module is used to determine whether the boundary time-varying description needs to be iteratively updated based on the aggregation reliability improvement index, until the output of the promised capacity space that meets the scheduling reliability requirements is reached. The fulfillment simulation module is used to simulate the fulfillment rate using the committable capacity space. It takes historical scheduling plans and actual vehicle off-grid records as input, replays and verifies the completion of scheduling instructions under the capacity constraint, and outputs the capacity boundary. The resource encapsulation module is used to encapsulate the adjustable capacity of charging piles reconstructed based on dwell behavior, along with the adjustment capabilities of other types of resources, to achieve the aggregation and invocation of heterogeneous resources under a unified semantic framework.

Images