An intelligent control method and system for operation of a vehicle charging pile

Abstract

This invention provides an intelligent control method and system for the operation of electric vehicle charging piles, relating to the field of charging pile control technology. The method includes: acquiring historical operation monitoring data of a target site; determining multiple power operation events for each charging pile and constructing a power event sequence; extracting multiple event state features of the target site and constructing multiple window event state vectors; calculating multiple state difference degrees of the target site and constructing state difference curves; segmenting multiple operation trajectory segments of the target site; extracting the site operation state vector for each operation trajectory segment and identifying multiple operation mechanism modes; generating an inter-pile response dataset for each operation mechanism mode and constructing a directed coupling graph; determining multiple behavioral coupling partitions of the operation mechanism mode; extracting the adjustment upper limit features of the operation mechanism mode and constructing a collaborative control set; and optimizing the online operation control of multiple charging piles based on the collaborative control set. This invention improves the stability of multi-pile collaborative operation at charging stations.

Description

Technical Field

[0001] The present invention relates to the technical field of charging pile control, and particularly to an intelligent control method and system for the operation of vehicle charging piles. Background Art

[0002] With the popularization of electric vehicles, charging stations with multiple charging piles operating concurrently are becoming increasingly common. Due to the sharing of constraints such as power distribution capacity and loop capacity among multiple charging piles within the station, actions such as the start, power increase and decrease of charging sessions are likely to have a linkage impact among multiple devices. For example, after the power of a certain charging pile is lowered, other charging piles compensate and increase, which in turn triggers new power adjustments, resulting in reciprocating power fluctuations and frequent adjustments within the station, thereby causing a decrease in charging efficiency, a reduction in throughput capacity, and even phenomena such as session instability.

[0003] Some stations adopt methods such as total power limit, or threshold derating and peak shifting control based on instantaneous monitoring to achieve the operation control of multiple charging piles. In this process, if a single pile is used as the adjustment object, it is easy to have the problem of "pressing the gourd and lifting the ladle", and it is difficult to continuously and effectively suppress the fluctuations and frequent adjustments caused by device linkage under different stations and different operation stages. Therefore, a running control method for the multi-pile concurrent scenario is needed to more effectively reduce the linkage fluctuations and frequent adjustments while meeting the station-level constraints, and improve the operation stability and service capacity of the station. Summary of the Invention

[0004] The present invention provides an intelligent control method and system for the operation of vehicle charging piles, which can reduce the reciprocating power fluctuations and frequent adjustment phenomena under the condition that the power supply capacity within the station is limited and there is a linkage impact among devices, and improve the operation stability and charging service capacity of the station while ensuring safety constraints.

[0005] In the first aspect of the present invention, an intelligent control method for the operation of vehicle charging piles is provided, including: Obtain the historical operation monitoring data of the target station, where the historical operation monitoring data includes the historical station-level operation data of the target station and the historical pile-level operation data corresponding to multiple charging piles within the target station respectively; Determine multiple power operation events of each charging pile according to the historical pile-level operation data, and construct a power event sequence for each charging pile; Traverse multiple power event sequences through a preset trajectory window, extract multiple event state features of the target station within each preset trajectory window, and construct multiple window event state vectors of the target station; Calculate multiple state difference degrees of the target station according to the window event state vectors and construct a state difference curve, and identify multiple segmentation nodes to divide multiple operation trajectory segments of the target station; Extract the station operation status vector for each operation trajectory segment and identify multiple operation mechanism modes, generate the pile-to-pile response dataset for each operation mechanism mode, and construct a directed coupling graph; Based on the directed coupling graph, multiple behavioral coupling partitions of the operation mechanism mode are determined, the adjustment upper limit features of the operation mechanism mode are extracted, and a collaborative control set is constructed. Based on the collaborative control set, online operation control optimization is performed on multiple charging piles.

[0006] Preferably, multiple power operation events for each charging pile are determined based on historical charging pile operation data, and the power event sequence for each charging pile is constructed as follows: Based on historical pile-level operation data, the pile-level output power sequence and operation status of multiple charging piles are determined, and the output power sequence of each pile-level is smoothed to obtain the pile-level smooth power sequence. Calculate multiple power differential features of each pile-level smooth power sequence on a preset differential window, construct a pile-level power differential sequence for each charging pile, determine the change threshold of each pile-level smooth power sequence based on the pile-level power differential sequence, and determine the power increase and power decrease events of the charging pile based on the change threshold. Within a preset jitter detection window, the number of alternations between power increase and power decrease events is counted. Based on the number of alternations, multiple power jitter events of the charging pile are determined. A power event sequence of the charging pile is constructed based on the timestamps of multiple power operation events.

[0007] Preferably, calculating multiple state difference degrees of the target station based on the window event state vector and constructing a state difference curve, and identifying multiple segmentation nodes to segment multiple running trajectory segments of the target station includes: For the window event state vector, multiple target charging piles in the charging state within each preset trajectory window are determined according to the operating status of the charging piles. Multiple event state features are extracted based on the power event sequence of the multiple target charging piles. The event state features include adjustment behavior density, adjustment behavior coverage, jitter behavior density, and behavior synchronization rate. The window event state vector of the target station in different preset trajectory windows is constructed based on the multiple event state features. Calculate the state difference degree of the target station between any two adjacent preset trajectory windows based on the window event state vector, and construct a state difference curve based on multiple state difference degrees. Multiple local peaks in the state difference curve are identified, the target state difference threshold is determined based on the state difference curve, local peaks with state difference greater than the target state difference threshold are marked as segmentation nodes, and the state difference curve is segmented according to multiple segmentation nodes to obtain multiple running trajectory segments of the target station.

[0008] Preferably, the process of extracting the station operation state vector for each operation trajectory segment, identifying multiple operation mechanism modes, generating a pile-to-pile response dataset for each operation mechanism mode, and constructing a directed coupling graph includes: The state vectors of multiple window events contained in each running trajectory segment are fused to construct the station running state vector of each running trajectory segment. The state similarity between any two running trajectory segments is calculated based on the station running state vector. Based on the state similarity, pattern recognition is performed on multiple running trajectory segments to generate multiple running mechanism modes of the target station. Based on the operating mechanism mode to which each operating trajectory segment belongs, candidate response datasets for each operating mechanism mode are extracted from multiple sets of historical pile-level operating data, including pile-level power differential sequences and operating status of multiple charging piles; Based on historical station-level operation data, the station-level smoothed power sequence of the target station is extracted. Based on the station-level smoothed power sequence, the station-driven stripping process is performed on each candidate response dataset. This includes constructing the station-level power difference sequence corresponding to the station-level smoothed power sequence, calculating the driving ratio coefficient between the station-level power difference sequence and each pile-level power difference sequence in the candidate response dataset, performing the station-driven stripping process on the pile-level power difference sequence based on the driving ratio coefficient, calculating the residual feature of each power difference feature in the pile-level power difference sequence, and constructing the pile-level power residual sequence for each charging pile. Based on the pile-level power residual sequence, a directed response analysis is performed on any two charging piles in the candidate response dataset to calculate the directed response weight between any two charging piles. Using multiple charging piles as nodes of the directed coupling graph, a directed coupling graph of the operation mechanism mode is constructed by combining multiple directed response weights.

[0009] Preferably, the adjustment upper limit features of the extracted operating mechanism mode include: For behavioral coupling partitions, if the directed response weights between any two charging piles are greater than the preset response strength threshold, the directed edges between the two charging piles are merged into undirected coupling edges, and an undirected coupling graph about multiple charging piles is constructed. Based on the multiple undirected coupling edges, multiple connected components in the undirected coupling graph are extracted to obtain multiple behavioral coupling partitions of the operation mechanism mode. The number of behavioral coupling partitions in the adjustment state within each preset trajectory window of the candidate response dataset is counted. The number of behavioral coupling partitions in the preset trajectory window and the jitter behavior density are combined to construct a partition linkage risk curve. The critical inflection point of the partition linkage risk curve is identified, and the number of behavioral coupling partitions corresponding to the critical inflection point is used as the adjustment upper limit feature of the operation mechanism mode.

[0010] Preferably, optimizing the online operation control of multiple charging piles based on a collaborative control set includes: Collect online operation monitoring data of the target site, identify the target operation mechanism mode of the target site based on the online operation monitoring data, identify the real-time operation status of each behavior coupling partition with respect to the online operation monitoring data according to the multiple behavior coupling partitions of the target operation mechanism mode, monitor the number of behavior coupling partitions of the target site in the adjustment state in real time, and impose adjustment constraints based on the adjustment upper limit characteristics of the target operation mechanism mode.

[0011] A second aspect of the present invention provides an intelligent control system for the operation of a car charging station, used to implement the above-mentioned intelligent control method for the operation of a car charging station, comprising: The operation data acquisition module is used to acquire historical operation monitoring data of the target site. The historical operation monitoring data includes historical station-level operation data of the target site, as well as historical pile-level operation data corresponding to multiple charging piles within the target site. The event extraction module is used to determine multiple power operation events for each charging pile based on historical charging pile operation data, and to construct a power event sequence for each charging pile. The event state analysis module traverses multiple power event sequences through preset trajectory windows, extracts multiple event state features of the target station within each preset trajectory window, and constructs multiple window event state vectors for the target station. The trajectory segmentation module is used to calculate the degree of difference between multiple states of the target station based on the window event state vector and construct the state difference curve, and identify multiple segmentation nodes to segment multiple trajectory segments of the target station. The operation mechanism identification module is used to extract the station operation status vector of each operation trajectory segment and identify multiple operation mechanism modes, generate the pile response dataset of each operation mechanism mode and construct a directed coupling map; The operation control optimization module is used to determine multiple behavioral coupling partitions of the operation mechanism mode based on the directed coupling graph, extract the adjustment upper limit features of the operation mechanism mode and construct a collaborative control set, and perform online operation control optimization for multiple charging piles based on the collaborative control set.

[0012] The present invention has the following beneficial effects: Based on the historical operation monitoring data of the target site, this invention eventizes the continuous power change behavior during the concurrent operation of multiple charging piles and performs windowed aggregation characterization. Through state difference analysis, it segments the operation trajectories and identifies different operation mechanism modes. On this basis, it further conducts analysis of the response relationship between the piles, strips the common driving component of the overall site power to highlight the true linkage response structure between the charging piles, and then constructs a directed coupling map and extracts the behavior coupling partitions. At the same time, by combining the correlation between the concurrent regulation scale and the jitter behavior in each partition under different operation mechanism modes, it extracts the regulation upper limit characteristics matching the mechanism mode and forms a cooperative control set, enabling online control to implement concurrent regulation constraints with the coupling partition as the regulation unit at different operation stages, thereby reducing the compensatory rebound and linkage diffusion caused by single-pile granularity control, reducing the power reciprocating fluctuation and frequent regulation phenomena, and improving the stability and controllability of the multi-pile cooperative operation of the charging station. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic flow chart of an intelligent control method for the operation of an electric vehicle charging pile provided by an embodiment of this invention.

[0014] Figure 2 It is a schematic structural diagram of an intelligent control system for the operation of an electric vehicle charging pile provided by an embodiment of this invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] In order to enable those skilled in the art to better understand the technical solutions in this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this invention. It should be understood that the specific embodiments described herein are only used to explain this invention and are not used to limit this invention.

[0016] Figure 1 It is a schematic flow chart of an intelligent control method for the operation of an electric vehicle charging pile provided by an embodiment of this invention. In the charging station scenario where multiple charging piles operate concurrently, multiple charging piles share the site power supply capacity and operation constraints, and the power adjustment behaviors of each charging pile during the charging process may affect each other, resulting in problems such as power reciprocating fluctuation, frequent regulation, and a decrease in the overall charging efficiency of the site in some operation stages. To improve the operation stability and service ability of the site, an embodiment of this invention provides an intelligent control method for the operation of an electric vehicle charging pile. By structuring the historical operation monitoring data of the site and identifying the operation mechanism, a set of cooperative control strategies matching the operation mechanism is formed, and based on this, the online operation control of multiple charging piles in the station is optimized to reduce unnecessary linkage regulation and improve the operation effect of the site.

[0017] Please refer to Figure 1 , the intelligent control method for the operation of an electric vehicle charging pile includes the following steps: Step S1: Obtain historical operation monitoring data of the target site. The historical operation monitoring data includes historical station-level operation data of the target site, as well as historical pile-level operation data corresponding to multiple charging piles within the target site.

[0018] In this embodiment, the historical operation monitoring data of the target site includes, but is not limited to, historical station-level operation data characterizing the overall operation status and changes in station-level constraints of the site, such as time series data corresponding to station-level operation parameters such as total station power, station incoming line voltage / circuit voltage, and the number of concurrent charging at the site; and historical pile-level operation data characterizing the operation process and charging session behavior of each charging pile, such as time series data corresponding to pile-level operation parameters such as output power, temperature, and voltage of each charging pile, as well as information such as charging session status. By simultaneously collecting station-level and pile-level data, a joint perspective of the overall and individual components of the site is provided for subsequent analysis.

[0019] It should be noted that the historical operational monitoring data of the target site may include various station-level and pile-level monitoring quantities to comprehensively characterize the operation of the site and equipment. In this embodiment, to facilitate explanation and reduce dependence on the types of monitoring quantities, the total power of the site and the output power and charging session status of each charging pile are preferably used as the main data sources for analysis, and power operation events and subsequent trajectory status representations are constructed based on the above power-related data. Other monitoring quantities such as voltage, current, and temperature can be used as auxiliary features to enhance the stability of event recognition or enrich the operational status representation, but they do not constitute a necessary condition for achieving the technical effects of this invention in this embodiment.

[0020] Step S2: Determine multiple power operation events for each charging pile based on historical charging pile operation data, and construct a power event sequence for each charging pile.

[0021] In this embodiment, continuous power operation data is converted into discrete power operation events to provide a unified description of key power change behaviors during operation. Power operation events can characterize typical behaviors such as power adjustment, power reduction, and alternating triggering of charging piles. Multiple identified power operation events are formed into a power event sequence ordered by time, which facilitates subsequent windowed aggregation and mechanism identification of the concurrent operation of multiple charging piles within the site based on the event sequence.

[0022] As an optional implementation, the process of constructing the power event sequence for each charging station specifically includes: Based on historical charging pile operation data, the output power sequence and operating status of multiple charging piles are determined. The output power sequence reflects the change of power of the charging pile over time during historical operation, while the operating status information characterizes the different operating states of the charging pile at each time point, such as idle, connected, charging, or terminated.

[0023] The output power sequence of each pile level is smoothed to obtain a smoothed power sequence of the pile level. The smoothing process can be moving average, moving median or other noise reduction smoothing methods. This embodiment does not specifically limit it. The purpose is to reduce the influence of sampling noise, short-term spikes or transient jitter on the power change judgment, so as to make the subsequent difference and event judgment more stable and reliable.

[0024] Multiple power differential features are calculated for each charging pile-level smoothed power sequence within a preset differential window. For example, the power mean corresponding to several sampling points within each differential window is calculated, and the difference between the power mean corresponding to two adjacent differential windows is calculated to obtain multiple power differential features and construct the charging pile-level power differential sequence. In this way, the continuous power sequence is converted into a differential sequence that is easier to identify changing trends, thereby capturing typical changing behaviors such as power increase and power decrease.

[0025] Based on the pile-level power differential sequence, a change threshold for the smoothed power sequence at the pile level is determined. This change threshold distinguishes between normal minor fluctuations and effective power changes to avoid noise-triggered events. Those skilled in the art can adaptively set the threshold based on historical differential statistical characteristics, such as determining the change threshold using statistical measures like quantiles, mean, and standard deviation of the pile-level power differential sequence. Preferably, the change threshold can be further constrained by operational status information, for example, only participating in threshold statistics or event determination when the charging pile is charging, thereby improving the effectiveness of event identification.

[0026] After determining the change threshold, power increase and decrease events are identified based on the comparison between the pile-level power difference sequence and the change threshold: when the power difference characteristic is greater than or equal to the change threshold, it is determined as a power increase event; when the power difference characteristic is less than or equal to a negative change threshold, it is determined as a power decrease event. For differential changes within the change threshold range, no event is generated to reduce the number of invalid events and highlight key power adjustment behaviors.

[0027] Within a preset jitter detection window, the number of alternations between power increase and decrease events is counted to characterize whether the direction of power change frequently reverses within this window. Multiple power jitter events of the charging pile are determined based on the number of alternations. For example, when the number of alternations meets a preset jitter judgment condition, it is marked as a power jitter event. The jitter judgment condition can be determined based on historical statistical characteristics, such as using an alternation number exceeding a preset threshold as the basis for jitter judgment. Power jitter events reflect the reciprocating adjustment or unstable change behavior of the charging pile during operation. It typically manifests as multiple alternations of upward and downward adjustments within a short time interval, resulting in a significant back-and-forth or fluctuating power curve.

[0028] After obtaining power increase events, power decrease events, and power fluctuation events, the events are sorted and organized according to the timestamps corresponding to the multiple power operation events to construct the power event sequence of the charging pile.

[0029] Step S3: Traverse multiple power event sequences through preset trajectory windows, extract multiple event state features of the target station within each preset trajectory window, and construct multiple window event state vectors of the target station.

[0030] In this embodiment, multiple power event sequences are aggregated and statistically analyzed using a preset trajectory window as a unit to obtain a state representation reflecting the group operation characteristics of the site within different time windows. Specifically, within the same trajectory window, event state characteristics reflecting the group behavior pattern, such as event density, event coverage range, and event synchronicity, can be statistically analyzed to construct a window event state vector, which is used to characterize the overall operational behavior structure of the site within that window, so that the overall operation process of the site can be structurally expressed in the form of a window state sequence.

[0031] For the specific construction process of the window event state vector, firstly, based on the operating status of the charging pile, determine multiple window target charging piles in the charging state within each preset trajectory window. This avoids interference from zero power fluctuations, plugging and unplugging transients, or sampling noise that may occur during power events in non-charging states, which could affect the stability of statistical features.

[0032] Multiple event state features are extracted from the power event sequences of multiple window target charging piles. These event state features include adjustment behavior density, adjustment behavior coverage, jitter behavior density, and behavior synchronization rate.

[0033] Specifically, the adjustment behavior density characterizes the activity level of station adjustment behavior within the trajectory window, determined by counting the total number of power increase and decrease events within the window. Adjustment behavior coverage characterizes the coverage area of ​​adjustment behavior within the window, i.e., the percentage of target charging piles within the window that have experienced at least one power increase / decrease within the window. Jitter behavior density characterizes the degree of repeated or cyclical adjustment of power change direction within the window, determined by the number of power jitter events within the window. Higher jitter behavior density usually indicates significant power cyclical changes or frequent back-and-forth within the window, potentially leading to decreased overall station operational stability and reduced adjustment efficiency. Behavior synchronization rate characterizes the degree of synchronization of adjustment behavior within the window in the time dimension, obtained by counting the percentage of target charging piles simultaneously experiencing power operation events in each preset differential window within the window. For example, it can be the ratio between the number of target charging piles experiencing power increase / decrease events in the same preset differential window and the number of target charging piles in the charging state within the window, and the ratios at multiple times within the window can be statistically aggregated to find the maximum value. A higher rate of behavioral synchronization usually indicates that there are more obvious synchronized or coordinated behaviors within the site, reflecting the degree of concentration of site-wide regulation.

[0034] Finally, the aforementioned event state features are combined to form the window event state vector corresponding to the trajectory window. By scrolling through multiple power event sequences according to the trajectory window, a series of window event state vectors for the target station within different preset trajectory windows can be obtained.

[0035] Step S4: Calculate the multiple state difference degrees of the target station based on the window event state vector and construct the state difference curve to identify multiple segment nodes to divide the target station into multiple running trajectory segments.

[0036] In this embodiment, by identifying the time boundaries at which the operating mechanism of a target site changes during operation, specifically by calculating the degree of difference between the event state vectors of adjacent windows to form a state difference curve, when a significant change in the degree of difference occurs, it can be identified as a segmentation node, thereby dividing the site's operating history into multiple operating trajectory segments. Within each operating trajectory segment, the group event characteristics of the site are relatively consistent.

[0037] As an optional implementation process, the specific process of constructing the state difference curve and segmenting multiple running trajectory segments includes: Based on the window event state vectors between two adjacent preset trajectory windows, the state difference degree is determined according to the distance between the two vectors. This is used to characterize the change amplitude of the event state characteristics of the station from a group perspective between adjacent windows. The distance can be characterized by Euclidean distance to obtain multiple state difference degree sequences arranged in chronological order, thereby forming the state difference curve of the target station, which is used to reflect the dynamic evolution of the intensity of window state changes during station operation.

[0038] After obtaining the state difference curve, multiple local peaks are identified within it. These are points of difference with significantly high values ​​in their vicinity. For example, the local maximum value among several state difference values ​​can be taken as the local peak to filter out potential segmentation candidate points from the full time range. Simultaneously, the target state difference threshold is determined based on the state difference curve. This can be done statistically, such as setting it based on the quantiles or mean of the state difference values ​​to highlight larger-amplitude state changes.

[0039] Local peaks where the state difference exceeds the target state difference threshold are marked as segment nodes. The state difference curve is then segmented based on the segment nodes to obtain multiple running trajectory segments of the target station. The changes in the window event state vector within each segment are relatively stable or have similar behavioral structures. However, there are significant state changes between segments, represented by the segment nodes.

[0040] Step S5: Extract the station operation status vector for each operation trajectory segment and identify multiple operation mechanism modes, generate the pile-to-pile response dataset for each operation mechanism mode, and construct a directed coupling graph.

[0041] In this embodiment, different operating mechanism modes are identified for multiple operating trajectory segments of the target site, and the response relationships between charging piles under different mechanism modes are further characterized. For the same operating mechanism mode, historical time period data corresponding to the mode can be summarized to construct an inter-pile response dataset, and response analysis between different charging piles can be performed to extract the inter-pile response relationship to construct a directed coupling spectrum, which is used to characterize the directional influence relationship between charging piles, such as which power change is more likely to trigger the response change of which other charging pile.

[0042] As an optional implementation process, the above process of identifying multiple operating mechanism modes and constructing a directed coupling graph specifically includes: The multiple window event state vectors contained in each running trajectory segment are fused to construct the station running status vector for each running trajectory segment. Specifically, each running trajectory segment consists of window event state vectors corresponding to multiple preset trajectory windows. The multiple window event state vectors within the trajectory segment are fused, for example, by using the mean fusion method to calculate the mean of the multiple window event state vectors as the station running status vector of the running trajectory segment.

[0043] After obtaining the station operation state vectors for each operation trajectory segment, the state similarity between any two operation trajectory segments is calculated based on the station operation state vectors. This can be obtained through methods such as vector distance or similarity functions, for example, calculating the Euclidean distance between two state vectors as the state similarity. Based on the state similarity, pattern recognition is performed on multiple operation trajectory segments, grouping those with high similarity into the same category. This can be done using clustering algorithms such as K-Means clustering or DBSCAN clustering, thereby generating multiple operation mechanism patterns for the target station. These patterns describe several typical group behavior structures that have appeared during the historical operation of the target station, with each pattern corresponding to a group of operation trajectory segments with similar station operation state vector characteristics.

[0044] Based on the operating mechanism mode to which each operating trajectory segment belongs, candidate response datasets for each operating mechanism mode are extracted from multiple sets of historical pile-level operating data. Specifically, for the same operating mechanism mode, historical time segments corresponding to multiple operating trajectory segments belonging to that mode can be aggregated to form a historical sample set for that mechanism mode. Pile-level power differential sequences and operating status information for multiple charging piles are then extracted from this sample set as the main content of the candidate response dataset.

[0045] Similar to the construction method of the pile-level smoothed power sequence, the station-level smoothed power sequence of the target station is extracted based on the historical station-level operation data. Based on the station-level smoothed power sequence, the station-driven stripping process is performed on each candidate response dataset to construct the pile-level power residual sequence for each charging pile.

[0046] It is worth noting that the station-level smoothed power sequence is used to characterize the overall power variation trend of the station. Its changes can reflect common driving factors at the station level, such as concurrent session changes, station-level power boundary changes, or unified station adjustment behavior. If the pile-level power differential sequence is directly used for inter-pile response analysis, synchronous changes caused by common station driving factors may be misidentified as inter-pile coupled responses, thus affecting the accuracy of the coupling spectrum. Therefore, this embodiment introduces station-driven stripping processing before response analysis to weaken common driving components and highlight the true inter-pile response relationship.

[0047] For the specific stripping process, differential features are first extracted from the station-level smoothed power sequence based on a preset differential window to construct the corresponding station-level power differential sequence. Then, the driving ratio coefficient between the station-level power differential sequence and the power differential sequence of each pile in the candidate response dataset is calculated. For example, least-squares fitting is performed on the station-level power differential sequence and each pile-level power differential sequence to obtain the driving ratio coefficient between the two sequences, which describes the following ratio of a single charging pile to the common changes of the station.

[0048] The system performs site-driven stripping processing on the pile-level power differential sequence based on the driving ratio coefficient, and calculates the residual features of each power differential feature in the pile-level power differential sequence. Specifically, the driving differential feature is obtained by multiplying the driving ratio coefficient by each power differential feature in the station-level power differential sequence. This product represents the impact of joint driving by different stations within different windows on a single charging pile. The difference between each power differential feature in the pile-level power differential sequence and its corresponding driving differential feature is calculated to obtain the residual feature corresponding to the power differential feature. In this way, the residual feature representing the pile power change rate after joint driving stripping is obtained. Based on multiple residual features, a pile-level power residual sequence for the charging pile is constructed. Through site-driven stripping processing, the synchronous changes of joint driving and the linkage changes of inter-pile response can be distinguished at the data level.

[0049] Based on the pile-level power residual sequence, a directed response analysis is performed on any two charging piles within the candidate response dataset to calculate the directed response weights between the two charging piles. Specifically, for any two charging piles, the moment a specific power change event, such as a power reduction event, occurs at the first charging pile can be used as the trigger point. The changing trend of the pile-level power residual sequence of the second charging pile is analyzed within a preset response period to characterize the response strength of the second charging pile to the power change of the first charging pile. For example, the residual characteristics of multiple pile power change rates of the second charging pile within the preset response period are determined, and the cumulative residual power increment represents the response strength of a single power operation event. Specifically, for a power reduction event of the first charging pile, the power increase behavior of the second charging pile is used as the response analysis object. The sum of multiple residual characteristics greater than 0 is calculated as the cumulative residual power increment, which is more consistent with the operating logic of the charging station, i.e., the phenomenon that after a power reduction of a single charging pile, the power is redistributed to other charging piles while maintaining the total output power unchanged.

[0050] The above method is used to calculate multiple directed response intensities between any two charging piles, and these intensities are aggregated. For example, the mean is calculated as the directed response weight between the two charging piles to characterize the directionality and intensity of the influence between the piles.

[0051] After obtaining the directed response weights among multiple charging piles, the multiple charging piles are used as nodes of the directed coupling graph, and the directed response weights are used as the directed edge weights between the nodes to construct the directed coupling graph corresponding to this operating mechanism mode. The directional coupling relationship between charging piles in the station under this operating mechanism mode is expressed in graph structure form.

[0052] Step S6: Based on the directed coupling graph, determine multiple behavioral coupling partitions of the operation mechanism mode, extract the adjustment upper limit features of the operation mechanism mode and construct a collaborative control set, and optimize the online operation control of multiple charging piles based on the collaborative control set.

[0053] In this embodiment, a set of charging piles with strong mutual influence or linkage response under the same operating mechanism mode can be identified based on the directed coupling graph, and multiple behavioral coupling partitions are defined accordingly. Adjustment upper limit features corresponding to different operating mechanism modes are extracted based on historical data to limit the size or intensity of the coupling partitions participating in adjustment at the same time, thereby employing different collaborative control boundaries under different operating mechanism modes. Finally, the behavioral coupling partitions and adjustment upper limit features are combined to form a collaborative control set. When the site is running online, the corresponding collaborative control set is invoked according to the current operating mechanism mode to optimize the operation control of multiple charging piles within the site, thereby suppressing unnecessary linkage adjustments and improving the overall operational stability and service capabilities of the site.

[0054] As an optional implementation process, the process of extracting the adjustment upper limit characteristics of the operating mechanism mode specifically includes: For the identification process of behavioral coupling partitions, if there is a significant bidirectional response relationship between any two charging piles, i.e., the directed response weights in both the first direction (A→B) and the opposite direction (B→A) are greater than a preset response strength threshold, then the two charging piles are considered to have a mutual coupling relationship. Based on this mutual coupling relationship, the two directed edges between the two charging piles are merged into one undirected coupling edge, thereby constructing an undirected coupling graph for multiple charging piles. The preset response strength threshold can be determined based on the historical data distribution of the operating mechanism mode, for example, the statistical quantile of the directed response weights can be taken. Threshold screening can reduce weak coupling edges caused by random fluctuations.

[0055] After obtaining the undirected coupling graph, multiple connected components are extracted from the graph based on multiple undirected coupling edges. The set of charging piles corresponding to each connected component is then determined as the behavioral coupling partition under this operating mechanism mode. The connected components are used to reflect the transitivity of the coupling relationship: when charging piles A and B are mutually coupled, and B and C are mutually coupled, A, B, and C together constitute a mutually influential coupled subsystem.

[0056] After obtaining multiple behavioral coupling partitions of the operating mechanism mode, in order to effectively constrain the linkage scale during online control, this embodiment further extracts the adjustment upper limit feature of the operating mechanism mode, which is used to limit the number of behavioral coupling partitions that are allowed to enter the adjustment state at the same time under the mechanism mode, thereby reducing the risk of linkage diffusion and power fluctuation caused by too many partitions adjusting at the same time.

[0057] Specifically, the number of behavioral coupling partitions in the candidate response dataset that are in an adjustment state within each preset trajectory window is counted. A behavioral coupling partition in an adjustment state refers to a situation where any charging pile within that partition experiences a power increase or decrease event within the trajectory window and is in a charging state. This method yields a sequence of partition adjustment numbers that varies over a time window, characterizing the concurrent scale of in-station adjustment behavior under this operating mechanism.

[0058] By combining the jitter behavior density corresponding to the preset trajectory window, the number of behavior coupling partitions of the preset trajectory window is combined with the jitter behavior density to construct a partition linkage risk curve, which is used to depict how the jitter phenomenon of the target site changes when the number of coupling partitions involved in the adjustment increases under a specific operating mechanism mode.

[0059] In the partition-linked risk curve, critical inflection points are identified, and the number of behaviorally coupled partitions corresponding to these inflection points is used as the upper limit feature for the adjustment of the operating mechanism mode. A critical inflection point can be understood as the point where, after the number of adjustment partitions increases to a certain value, the jitter phenomenon significantly intensifies. By extracting the number of partitions corresponding to this critical point, the maximum acceptable concurrent adjustment scale under a specific operating mechanism mode can be obtained.

[0060] During the online control phase, when a site is identified as being in this operating mechanism mode, the upper limit feature of the adjustment can be used as the collaborative control boundary to limit the number of behavior coupling partitions that are simultaneously in the adjustment state to no more than the upper limit. This allows for differentiated stability constraints under different operating mechanism modes, avoiding the problem of a single fixed threshold failing in different mechanism phases.

[0061] As an optional implementation process, online operation control optimization of multiple charging piles based on a collaborative control set includes: Collect online operation monitoring data for the target site, including online station-level operation data and online pile-level operation data. Using the windowed statistical method employed in historical data identification, extract event state features from the online data within a preset trajectory window to form a window event state vector. By matching this window event state vector with the historically identified operation mechanism modes regarding different features, determine the target operation mechanism mode currently in which the site operates.

[0062] For each operating mechanism mode, the average of the operating status vectors of multiple sites can be used as the mode status vector. The distance between the window event status vector of the online operation monitoring data and the mode status vector of different operating mechanism modes can be calculated, for example, by calculating the Euclidean distance, and the operating mechanism mode with the closest distance can be selected as the target operating mechanism mode of the target site.

[0063] During online operation, each behavioral coupling partition can be used as a basic adjustment unit for status determination and adjustment control. Specifically, under the target operating mechanism mode, the real-time operating status of each behavioral coupling partition with respect to online operating monitoring data is identified, i.e., the number of behavioral coupling partitions currently in the adjustment state is determined. The number of behavioral coupling partitions currently in the adjustment state at the monitoring target site is obtained by counting them. The adjustment upper limit characteristic corresponding to the target operating mechanism mode is then read from the collaborative control set.

[0064] When real-time observed values ​​exceed the adjustment limit, adjustment constraint processing is triggered. This includes measures such as restricting new partitions from entering the adjustment state, switching some partitions to the anchoring state, or queuing and staggering the execution of partition adjustment actions. Specific constraint methods can be reasonably set according to actual needs, aiming to avoid a large number of behaviorally coupled partitions simultaneously being in the adjustment state, leading to the risk of chain reaction and power fluctuations caused by excessive partition adjustments. When real-time observed values ​​do not exceed the adjustment limit, partitions are allowed to adjust normally within the predetermined cooperative control boundaries; that is, no intervention is taken, and normal operation control is maintained. By employing different adjustment limit constraints under different operating mechanism modes, both site safety constraints and throughput capacity and operational stability can be considered, achieving intelligent control optimization that is more closely aligned with the operational phase.

[0065] Figure 2 This is a schematic diagram of the intelligent control system for electric vehicle charging piles provided in an embodiment of the present invention. Please refer to [link / reference]. Figure 2 The system includes: The operation data acquisition module is used to acquire historical operation monitoring data of the target site. The historical operation monitoring data includes historical station-level operation data of the target site, as well as historical pile-level operation data corresponding to multiple charging piles within the target site. The event extraction module is used to determine multiple power operation events for each charging pile based on historical charging pile operation data, and to construct a power event sequence for each charging pile. The event state analysis module traverses multiple power event sequences through preset trajectory windows, extracts multiple event state features of the target station within each preset trajectory window, and constructs multiple window event state vectors for the target station. The trajectory segmentation module is used to calculate the degree of difference between multiple states of the target station based on the window event state vector and construct the state difference curve, and identify multiple segmentation nodes to segment multiple trajectory segments of the target station. The operation mechanism identification module is used to extract the station operation status vector of each operation trajectory segment and identify multiple operation mechanism modes, generate the pile response dataset of each operation mechanism mode and construct a directed coupling map; The operation control optimization module is used to determine multiple behavioral coupling partitions of the operation mechanism mode based on the directed coupling graph, extract the adjustment upper limit features of the operation mechanism mode and construct a collaborative control set, and perform online operation control optimization for multiple charging piles based on the collaborative control set.

[0066] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A smart control method for the operation of a car charging station, characterized in that, include: Obtain historical operation monitoring data for the target site. The historical operation monitoring data includes the historical station-level operation data of the target site, as well as the historical pile-level operation data corresponding to multiple charging piles within the target site. Based on historical charging pile operation data, multiple power operation events for each charging pile are determined, and a power event sequence for each charging pile is constructed. By traversing multiple power event sequences through preset trajectory windows, multiple event state features of the target station within each preset trajectory window are extracted, and multiple window event state vectors of the target station are constructed. Calculate the multiple state difference degrees of the target station based on the window event state vector and construct the state difference curve to identify multiple segment nodes to segment multiple running trajectory segments of the target station. Extract the station operation status vector for each operation trajectory segment and identify multiple operation mechanism modes, generate the pile-to-pile response dataset for each operation mechanism mode, and construct a directed coupling graph; Based on the directed coupling graph, multiple behavioral coupling partitions of the operation mechanism mode are determined, the adjustment upper limit features of the operation mechanism mode are extracted, and a collaborative control set is constructed. Based on the collaborative control set, online operation control optimization is performed on multiple charging piles.

2. The intelligent control method for the operation of a car charging pile according to claim 1, characterized in that, Based on historical charging pile operation data, multiple power operation events for each charging pile are determined, and a power event sequence for each charging pile is constructed, including: Based on historical pile-level operation data, the pile-level output power sequence and operation status of multiple charging piles are determined, and the output power sequence of each pile-level is smoothed to obtain the pile-level smooth power sequence. Calculate multiple power differential features of each pile-level smooth power sequence on a preset differential window, construct a pile-level power differential sequence for each charging pile, determine the change threshold of each pile-level smooth power sequence based on the pile-level power differential sequence, and determine the power increase and power decrease events of the charging pile based on the change threshold. Within a preset jitter detection window, the number of alternations between power increase and power decrease events is counted. Based on the number of alternations, multiple power jitter events of the charging pile are determined. A power event sequence of the charging pile is constructed based on the timestamps of multiple power operation events.

3. The intelligent control method for the operation of a car charging pile according to claim 2, characterized in that, Based on the window event state vector, the multiple state difference degrees of the target station are calculated and a state difference curve is constructed. Multiple segmentation nodes are identified to segment multiple running trajectory segments of the target station, including: For the window event state vector, multiple target charging piles in the charging state within each preset trajectory window are determined according to the operating status of the charging piles. Multiple event state features are extracted based on the power event sequence of the multiple target charging piles. The event state features include adjustment behavior density, adjustment behavior coverage, jitter behavior density, and behavior synchronization rate. The window event state vector of the target station in different preset trajectory windows is constructed based on the multiple event state features. Calculate the state difference degree of the target station between any two adjacent preset trajectory windows based on the window event state vector, and construct a state difference curve based on multiple state difference degrees. Multiple local peaks in the state difference curve are identified, the target state difference threshold is determined based on the state difference curve, local peaks with state difference greater than the target state difference threshold are marked as segmentation nodes, and the state difference curve is segmented according to multiple segmentation nodes to obtain multiple running trajectory segments of the target station.

4. The intelligent control method for the operation of a car charging pile according to claim 3, characterized in that, Extract the station operation status vector for each operation trajectory segment and identify multiple operation mechanism modes. Generate the inter-stake response dataset for each operation mechanism mode and construct a directed coupling graph, including: The state vectors of multiple window events contained in each running trajectory segment are fused to construct the station running state vector of each running trajectory segment. The state similarity between any two running trajectory segments is calculated based on the station running state vector. Based on the state similarity, pattern recognition is performed on multiple running trajectory segments to generate multiple running mechanism modes of the target station. Based on the operating mechanism mode to which each operating trajectory segment belongs, candidate response datasets for each operating mechanism mode are extracted from multiple sets of historical pile-level operating data, including pile-level power differential sequences and operating status of multiple charging piles; Based on historical station-level operation data, the station-level smoothed power sequence of the target station is extracted. Based on the station-level smoothed power sequence, the station-driven stripping process is performed on each candidate response dataset. This includes constructing the station-level power difference sequence corresponding to the station-level smoothed power sequence, calculating the driving ratio coefficient between the station-level power difference sequence and each pile-level power difference sequence in the candidate response dataset, performing the station-driven stripping process on the pile-level power difference sequence based on the driving ratio coefficient, calculating the residual feature of each power difference feature in the pile-level power difference sequence, and constructing the pile-level power residual sequence for each charging pile. Based on the pile-level power residual sequence, a directed response analysis is performed on any two charging piles in the candidate response dataset to calculate the directed response weight between any two charging piles. Using multiple charging piles as nodes of the directed coupling graph, a directed coupling graph of the operation mechanism mode is constructed by combining multiple directed response weights.

5. The intelligent control method for the operation of a car charging pile according to claim 4, characterized in that, The adjustment upper limit characteristics of the extracted operating mechanism mode include: For behavioral coupling partitions, if the directed response weights between any two charging piles are greater than the preset response strength threshold, the directed edges between the two charging piles are merged into undirected coupling edges, and an undirected coupling graph about multiple charging piles is constructed. Based on the multiple undirected coupling edges, multiple connected components in the undirected coupling graph are extracted to obtain multiple behavioral coupling partitions of the operation mechanism mode. The number of behavioral coupling partitions in the adjustment state within each preset trajectory window of the candidate response dataset is counted. The number of behavioral coupling partitions in the preset trajectory window and the jitter behavior density are combined to construct a partition linkage risk curve. The critical inflection point of the partition linkage risk curve is identified, and the number of behavioral coupling partitions corresponding to the critical inflection point is used as the adjustment upper limit feature of the operation mechanism mode.

6. The intelligent control method for the operation of a car charging pile according to claim 5, characterized in that, Optimization of online operation control for multiple charging piles based on a collaborative control set includes: Collect online operation monitoring data of the target site, identify the target operation mechanism mode of the target site based on the online operation monitoring data, identify the real-time operation status of each behavior coupling partition with respect to the online operation monitoring data according to the multiple behavior coupling partitions of the target operation mechanism mode, monitor the number of behavior coupling partitions of the target site in the adjustment state in real time, and impose adjustment constraints based on the adjustment upper limit characteristics of the target operation mechanism mode.

7. An intelligent control system for the operation of a car charging station, characterized in that, The system is used to implement the intelligent control method for the operation of a car charging pile as described in any one of claims 1-6, including: The operation data acquisition module is used to acquire historical operation monitoring data of the target site. The historical operation monitoring data includes historical station-level operation data of the target site, as well as historical pile-level operation data corresponding to multiple charging piles within the target site. The event extraction module is used to determine multiple power operation events for each charging pile based on historical charging pile operation data, and to construct a power event sequence for each charging pile. The event state analysis module traverses multiple power event sequences through preset trajectory windows, extracts multiple event state features of the target station within each preset trajectory window, and constructs multiple window event state vectors for the target station. The trajectory segmentation module is used to calculate the degree of difference between multiple states of the target station based on the window event state vector and construct the state difference curve, and identify multiple segmentation nodes to segment multiple trajectory segments of the target station. The operation mechanism identification module is used to extract the station operation status vector of each operation trajectory segment and identify multiple operation mechanism modes, generate the pile response dataset of each operation mechanism mode and construct a directed coupling map; The operation control optimization module is used to determine multiple behavioral coupling partitions of the operation mechanism mode based on the directed coupling graph, extract the adjustment upper limit features of the operation mechanism mode and construct a collaborative control set, and perform online operation control optimization for multiple charging piles based on the collaborative control set.

Images