Methods, devices and systems for virtual power plant load forecasting and control for multi-microgrid systems

Abstract

This invention provides a method, device, and system for load forecasting and control of virtual power plants with multiple microgrids. It continuously collects load time-series records of each microgrid within the coverage area of ​​the virtual power plant, obtaining a set of load time-series records for each microgrid. Load difference calculations are performed on adjacent collection periods, and load fluctuation records with load differences exceeding a preset fluctuation range are selected. This identifies overlapping periods where load fluctuations occur synchronously among the microgrids. For each overlapping period, the order of occurrence and amplitude of load fluctuations in each microgrid are analyzed to determine the causal relationship between the loads of each microgrid, obtaining causal relationship pairs and records for each microgrid load. Load record entries corresponding to the causal relationship pairs are selected from the load time-series records of each microgrid, and load trend extrapolation calculations are performed to obtain the load extrapolation results for each microgrid, determining the load control method. This invention can improve the accuracy of load control.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method, apparatus, and system for load forecasting and control of virtual power plants for multi-microgrid systems. Background Technology

[0002] With the development of distributed energy and smart grid technologies, virtual power plant load forecasting and control technology for multiple microgrids has become crucial for supporting the coordinated operation of power systems. This technology focuses on the multiple microgrids covered by virtual power plants, achieving global load balancing through load sensing, trend prediction, and strategy execution. Currently, the industry generally collects time-series load data from each microgrid, conducts trend extrapolation based on the historical load patterns of individual microgrids, and then formulates independent control strategies for each microgrid and distributes them to the corresponding terminals for execution. However, this approach struggles to accurately capture the inherent logic of load changes between microgrids, and the extrapolation relies solely on historical data from individual microgrids, which can easily lead to discrepancies between the extrapolation results and the actual operating conditions. The formulated control strategies are also difficult to adapt to the global coordinated operation requirements of virtual power plants, failing to provide effective support for power system stability optimization. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide at least one method, apparatus and system for virtual power plant load forecasting and control for multi-microgrids.

[0004] According to a first aspect of the present invention, a virtual power plant load forecasting and control method for multiple microgrids is provided. The method includes: continuously collecting load time-series records of each microgrid within the coverage area of ​​the virtual power plant to obtain a load time-series record set for each microgrid, wherein the load time-series record set contains load record entries for each microgrid within a continuous collection period, and each load record entry corresponds to a unique collection time marker; performing load difference calculation on the load time-series record sets of each microgrid for adjacent collection periods, filtering out load fluctuation record entries whose load difference exceeds a preset fluctuation range, locating overlapping periods where load fluctuations occur synchronously among the microgrids based on the load fluctuation record entries, and obtaining a set of load synchronous fluctuation periods; and targeting the load synchronous fluctuation periods... For each overlapping time period in the segment set, the order of occurrence and the corresponding amplitude of load fluctuations in each microgrid are analyzed to determine the causal relationship between the loads of each microgrid, resulting in causal relationship pairs and relationship records for each microgrid load. The relationship records include the corresponding load fluctuation amplitudes of the associated time periods. Based on the causal relationship pairs and relationship records, load record entries corresponding to the causal relationship pairs are selected from the load time sequence record set of each microgrid to obtain a load relationship record set. Based on the load relationship record set, load trend extrapolation calculations are performed to obtain the load extrapolation results for each microgrid. Based on the load extrapolation results, the load control method of the virtual power plant is determined, and corresponding load control instructions are generated and sent to the load control terminals of each microgrid.

[0005] According to a second aspect of the present invention, a virtual power plant load forecasting and control device is provided, comprising: a data acquisition module, configured to continuously acquire load time-series records of each microgrid within the coverage area of ​​the virtual power plant, thereby obtaining a load time-series record set for each microgrid, wherein the load time-series record set includes load record entries for each microgrid within a continuous acquisition period, and each load record entry corresponds to a unique acquisition time marker; a load calculation module, configured to perform load difference calculation on the load time-series record sets of each microgrid for adjacent acquisition periods, filter out load fluctuation record entries whose load difference exceeds a preset fluctuation range, locate overlapping periods of synchronous load fluctuations between microgrids based on the load fluctuation record entries, thereby obtaining a set of synchronous load fluctuation periods; and a relationship analysis module, configured to analyze the synchronous load fluctuation periods. For each overlapping time period in the set, the order of occurrence and the corresponding amplitude of load fluctuations in each microgrid are analyzed to determine the causal relationship between the loads of each microgrid, resulting in causal relationship pairs and relationship records for each microgrid load. The relationship records include the corresponding amplitude of load fluctuations in the associated time periods. An item filtering module is used to filter load record items corresponding to the causal relationship pairs from the load time sequence record set of each microgrid based on the causal relationship pairs and relationship records, resulting in a load relationship record set. A load control module is used to perform load trend extrapolation calculations based on the load relationship record set, obtain the load extrapolation results for each microgrid, determine the load control method of the virtual power plant based on the load extrapolation results, generate corresponding load control instructions, and issue them to the load control terminals of each microgrid.

[0006] According to a third aspect of the present invention, a computer system is provided, comprising: a processor; and a memory, wherein the memory stores computer-readable code that, when executed by the processor, causes the processor to perform the method as described above.

[0007] This invention establishes a unified data benchmark across microgrids by continuously collecting load time-series records from each microgrid within a virtual power plant and binding them with a unique collection time stamp, eliminating the interference of data dimensional differences between different microgrids on subsequent analysis. By filtering fluctuation records based on load differences between adjacent time periods and locating overlapping periods of synchronous load fluctuations across microgrids, it transcends the limitations of local monitoring of a single microgrid and accurately pinpoints the core research interval for coordinated microgrid fluctuations. Analyzing the sequence and amplitude correspondence of load fluctuations in each microgrid within synchronous fluctuation periods overcomes the conventional limitation of merely identifying synchronicity, clarifying the inherent influence direction and corresponding logic of load fluctuations between microgrids. Based on causal relationships, it filters corresponding load record entries from the full load data, selectively extracting core correlated data and significantly reducing the interference of irrelevant data on subsequent inferences. Based on correlated load records, it infers load trends and determines the global control method, integrating microgrid causal relationships throughout the entire inference and decision-making process, making control commands more adaptable to the global coordinated needs of the virtual power plant and improving the accuracy of load control. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of an application scenario provided by the present invention; Figure 2 This is a flowchart illustrating a virtual power plant load forecasting and control method for multi-microgrid systems provided by the present invention. Figure 3 This is a schematic diagram of the structure of a virtual power plant load forecasting and control device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer system provided in an embodiment of the present invention. Detailed Implementation

[0009] like Figure 1 As shown, the application scenario of this invention includes a computer system 10 and a microgrid cluster. The microgrid cluster may include multiple microgrids; the number of microgrids will not be limited here. Figure 1 As shown, a microgrid cluster can specifically include microgrid 1, microgrid 2, ..., microgrid n; it can be understood that microgrid 1, microgrid 2, microgrid 3, ..., microgrid n can all be connected to computer system 10 via network so that each microgrid can interact with computer system 10 via network connection.

[0010] It is understood that computer system 10 can refer to a device that executes the virtual power plant load forecasting and control method for multiple microgrids provided in the embodiments of the present invention. Computer system 10 can be, for example, a server, a single physical server, or a server cluster or distributed system consisting of at least two physical servers. Microgrids can specifically refer to smartphones, tablets, laptops, desktop computers, etc., but are not limited to these. Various microgrids and computer systems 10 can be directly or indirectly connected via wired or wireless communication. Furthermore, the number of microgrids and computer systems 10 can be one or at least two; the present invention does not impose any limitations on this.

[0011] Further, please see Figure 2 This is a flowchart illustrating a virtual power plant load forecasting and control method for multi-microgrid systems provided in an embodiment of the present invention. Figure 2 As shown, this method can be derived from... Figure 1 The method is executed by a computer system 10, wherein the virtual power plant load forecasting and control method for multi-microgrids may include the following steps: Step S100: Continuously collect load time-series records of each microgrid within the coverage area of ​​the virtual power plant to obtain a load time-series record set for each microgrid. The load time-series record set contains load record entries for each microgrid during the continuous collection period, and each load record entry corresponds to a unique collection time marker.

[0012] Load time-series records are chronologically ordered records of microgrid loads, reflecting load changes over time. A load time-series record set is a collection of load time-series records from all microgrids. A continuous acquisition period is a time interval during which load data is collected uninterruptedly. A load record entry is a specific record within the load time-series record set, containing microgrid load data at a specific point in time. Acquisition timestamps are used to identify the acquisition time of each load record entry, determining the order in which load data was collected. This step utilizes a distributed data acquisition system, deploying load sensors across various microgrids within the virtual power plant's coverage area. These sensors continuously collect microgrid load data at regular time intervals, such as every minute or every five minutes. The collected data is stored on local data storage devices and simultaneously transmitted over a network to a data center. In the data center, this data is processed and categorized, sorted according to microgrid identifiers and acquisition times, forming load time-series record sets for each microgrid. For example, in a virtual power plant comprising multiple commercial and residential microgrids, load data is collected by installing smart meters at the incoming line of each microgrid. The smart meters collect load data every three minutes and upload it to a data center. The data center then categorizes this data according to the identifiers of the commercial and residential microgrids, and sorts it by collection time, ultimately obtaining a set of load time-series records for each microgrid.

[0013] Step S200: Perform load difference calculation on the load time sequence record set of each microgrid for adjacent collection periods, filter out load fluctuation record entries whose load difference exceeds the preset fluctuation range, locate the overlapping periods of load fluctuation between microgrids based on the load fluctuation record entries, and obtain the load synchronous fluctuation period set.

[0014] In one implementation, step S200 may specifically include the following steps S210 to S260: Step S210: Perform global time base alignment on the load time series record sets of each microgrid, convert the collection time stamps of the load record entries of all microgrids into a unified time scale, eliminate the time deviation caused by the difference in the collection system of different microgrids, and generate a standardized load time series with consistent time base.

[0015] Specifically, a global time base is first determined, such as Coordinated Universal Time (UTC). Then, the acquisition timestamps in the load time series records of each microgrid are converted. This can be done by adding a time calibration module to each microgrid's acquisition system to compare and correct the acquired timestamps with the global time base. During the correction process, a time synchronization protocol, such as Network Time Protocol (NTP), can be used to ensure the consistency of the acquired timestamps with the global time base. For already acquired load record entries, conversion can be performed by calculating the time offset. For example, assuming a microgrid's acquisition system has a +5-minute time deviation, then the acquisition timestamps of all load record entries for that microgrid need to be subtracted by 5 minutes to align with the global time base. Finally, all microgrid load record entries are sorted according to a unified time scale to generate a standardized load time series with a consistent time base. By using the NTP protocol to calibrate the time of each microgrid's acquisition system and calculating and converting the time offsets of the acquired load record entries, a standardized load time series with a consistent time base is finally obtained.

[0016] Step S220: Perform state chain association between adjacent time periods for the load record entries of each microgrid in the standardized load time series, take the load record entries of the previous collection time period as the preceding association items of the next collection time period, establish the load status transmission link between time periods, and generate a load status transmission link set.

[0017] Adjacent time period status chain association establishes a chain relationship by associating load record entries from adjacent data collection periods to analyze the transmission of load status between time periods. Precedence association items are load record entries from the previous data collection period used as association items for the next data collection period. Load status transmission links are established through adjacent time period status chain associations, reflecting the links through which load status is transmitted between time periods. The load status transmission link set is a collection that aggregates the load status transmission links of each microgrid.

[0018] For each microgrid load record entry in the standardized load time series, load record entries from previous collection periods are associated with load record entries from subsequent collection periods according to the chronological order of collection time. This can be achieved by adding an association pointer to the load record entry data structure, pointing the pointer of the load record entry from the previous collection period to the load record entry from the subsequent collection period, thus establishing a chain-like association between adjacent time periods. For example, for load record entries in microgrid A, according to the collection times t1, t2, t3, etc., the load record entry at time t1 is associated with the load record entry at time t2, the load record entry at time t2 is associated with the load record entry at time t3, and so on, forming a load status transmission link. This operation is performed for each microgrid, ultimately generating a set of load status transmission links. In practical implementation, a linked list data structure can be used to store the load status transmission links, with each linked list node storing a load record entry and its association pointer.

[0019] Step S230: Perform change correlation calculation on each link in the load status transmission link set, bind the load status of the preceding and following time periods in the link in pairs, generate load status change correlation pairs for adjacent collection time periods, and ensure that each pair of correlation items corresponds to a unique time scale interval.

[0020] In one implementation, step S230 may specifically include the following steps S231 to S236: Step S231: Node splitting is performed on each link in the load status transmission link set. Each load record entry in the link is split into an independent status node. Each status node contains the corresponding microgrid identifier, time scale and load status information, generating a link status node set.

[0021] Node splitting involves separating each load record entry from the link structure in the load status transmission link, forming an independent status node. A microgrid identifier uniquely identifies each microgrid, the time scale is the collection time corresponding to the load record entry, and load status information can include load value, load change trend, etc. The link status node set is a collection that aggregates all status nodes obtained after splitting each link.

[0022] During this step, for each link in the load status transmission link set, the load record entries in the link are traversed. For each load record entry, the microgrid identifier, time scale, and load status information are extracted and combined into an independent status node. For example, for a load record entry containing the information (microgrid ID: A, collection time: t, load value: P), it is split into a status node (microgrid identifier: A, time scale: t, load status information: load value P). This splitting operation is performed on all load record entries in each link, ultimately generating a link status node set. In implementation, object-oriented programming can be used to define a status node class, which contains attributes such as microgrid identifier, time scale, and load status information. Each load record entry is converted into an object of this class and stored in the link status node set.

[0023] Step S232: Perform adjacent time period matching on the nodes in the link status node set. According to the order of the time scale, bind each status node to the next adjacent time scale status node to generate adjacent time period status node pairs, so that the microgrid identifier of each pair of nodes remains consistent.

[0024] Adjacent time period matching involves identifying the state node immediately following the next time scale in the link state node set for each state node based on the chronological order of the time scale, and then pairing them up. An adjacent time period state node pair consists of two state nodes at adjacent time scales, and these two nodes must have the same microgrid identifier to ensure that the comparison is of the same microgrid in adjacent time periods.

[0025] First, sort all state nodes in the link state node set according to their time scale. Sorting algorithms such as quicksort and mergesort can be used to ensure the state nodes are arranged in chronological order. Then, iterate through the sorted state node set. For each state node, find its immediately adjacent time scale. If the microgrid identifier of the found state node is the same as the current node's microgrid identifier, pair these two nodes together to form a pair of state nodes for adjacent time periods. For example, suppose the link state node set contains state nodes (microgrid identifier: C, time scale: t1) and (microgrid identifier: C, time scale: t2), where t2 is the immediately adjacent time scale of t1. Then, pair these two nodes together as a pair of state nodes for adjacent time periods: ((microgrid identifier: C, time scale: t1), (microgrid identifier: C, time scale: t2)). Perform this operation on all state nodes in the link state node set, ultimately generating a set of pairs of state nodes for adjacent time periods.

[0026] Step S233: Perform state association coding on adjacent time period state node pairs, convert the microgrid identifier, time scale and load status information of each pair of nodes into standard association codes, the coding content includes the time difference between nodes and the state association mark, and generate state association code pairs.

[0027] Standard association coding includes the microgrid identifier of a node, time scale, load status information, time difference between nodes, and status association markers. The time difference is the difference between the time scales of adjacent nodes, and the status association marker is used to indicate the association relationship between the load status of two nodes, such as whether the load is increasing, decreasing, or remaining unchanged. A status association code pair is a code pair obtained by encoding adjacent time period state node pairs.

[0028] For each pair of nodes in the set of adjacent time period state node pairs, extract the microgrid identifier, time scale, and load status information. Calculate the time difference between the two nodes, i.e., the time scale of the latter node minus the time scale of the former node. Based on the load status information of the two nodes, determine their state association relationship. For example, if the load value of the latter node is greater than that of the former node, the state association is marked as "load increase"; if the load value of the latter node is less than that of the former node, the state association is marked as "load decrease"; if they are equal, the state association is marked as "load unchanged". Encode this information according to certain encoding rules to generate standard association codes. For example, binary encoding can be used to convert the microgrid identifier, time scale, time difference, and state association mark into binary strings, and then combine the codes of the two nodes in the adjacent time period state node pair into a state association code pair. Perform this encoding operation on all pairs in the set of adjacent time period state node pairs, finally generating a set of state association code pairs.

[0029] Step S234: Extract the change amount from the state association code pair, parse the load state information of the previous and next time periods from the code content, associate and bind the two state information to generate load state change association items.

[0030] During this step, for each pair of codes in the state-associated code set, the load state information for the preceding and following time periods is parsed according to the encoding rules. For example, if the code is stored as a binary string, information such as the load value or load change rate needs to be extracted according to predefined bit segmentation rules. The parsed load state information for the preceding and following time periods is compared to calculate the change in load state. For example, the load change is obtained by subtracting the load value of the preceding time period from the load value of the following time period, or by calculating the load change rate (the load change divided by the load value of the preceding time period). The load state information for the preceding and following time periods and the calculated load state change are combined into a load state change association item. For example, for a state-associated code pair, the parsed load values ​​for the preceding and following time periods are P1 and P2, respectively. The load change ΔP = P2 - P1 is calculated, and (P1, P2, ΔP) is combined into a load state change association item. This operation is performed on all pairs in the state-associated code pair set, ultimately generating a set of load state change association items.

[0031] Step S235: Perform validity verification on load state change correlation items, verify whether there are conflicts between microgrid identifiers and time scales in the correlation items, eliminate invalid correlation items with conflicts, and generate valid load state change correlation items.

[0032] Validity verification checks the information in load state change correlation items to ensure the consistency and accuracy of microgrid identifiers and time scales. Conflicts occur when there are errors or inconsistencies in the microgrid identifiers or time scales; for example, the microgrid identifier may not match the actual situation, or the time scale sequence may be out of order. Effective load state change correlation items are those that have passed validity verification.

[0033] For each associated item in the load state change associated item set, the consistency of the microgrid identifier is checked first. This can be verified by querying a pre-established microgrid identifier database to confirm the existence and consistency of the microgrid identifier in the associated item with the actual situation. Next, the order and rationality of the time scale are checked. For example, the time scale of a later time period should be greater than that of an earlier time period; if this condition is not met, a time scale conflict is considered to exist. Associated items with microgrid identifier or time scale conflicts are excluded from the set. Finally, the remaining associated items form the effective load state change associated item set. For example, suppose a load state change associated item (microgrid identifier: D, time scale 1: t1, time scale 2: t0, load change: ΔP) has t0 > t1, indicating a time scale order disorder; this associated item is excluded. This verification operation is performed on all associated items in the load state change associated item set to obtain the effective load state change associated item set.

[0034] Step S236: Integrate the effective load status change correlation items in pairs, classify each effective correlation item according to the corresponding link, and generate load status change correlation pairs for adjacent collection periods.

[0035] For the set of load status change correlation items, each correlation item is first labeled with the load status transmission link identifier it belongs to. This can be done by adding a link identifier field to the correlation item data structure, or by querying the load status transmission link set based on the microgrid identifier and time scale information in the correlation item to determine its corresponding link. Then, the valid correlation items are classified according to the link identifier, and valid correlation items belonging to the same link are grouped together. For valid correlation items within the same link, they are sorted according to the chronological order of the time scale, and correlation items with adjacent time scales are paired and bound together to generate load status change correlation pairs for adjacent data collection periods. For example, for valid correlation items (P1, P2, ΔP1), (P2, P3, ΔP2), etc., belonging to link E, after arranging them in chronological order, (P1, P2, ΔP1) and (P2, P3, ΔP2), etc., are sequentially bound into load status change correlation pairs for adjacent data collection periods. This operation is performed on all valid correlation items for all links, ultimately generating a set of load status change correlation pairs for adjacent data collection periods.

[0036] Step S240: Perform cross-microgrid pre-matching of load state change association pairs for fluctuation range, bind each microgrid load state change association pair to the boundary of the preset fluctuation range one by one, generate fluctuation range binding results, and exclude invalid association pairs that do not correspond to the preset fluctuation range.

[0037] The preset fluctuation range is a pre-defined range of load changes, typically represented by upper and lower boundary values. Correspondence binding compares each load state change correlation pair with the boundaries of the preset fluctuation range to determine if the pair falls within it. The fluctuation range binding result is a set of matching records documenting the relationship between each load state change correlation pair and the preset fluctuation range. Invalid correlation pairs are those that do not correspond to the preset fluctuation range, meaning their load changes do not exceed the preset range. These pairs may not need to be considered in subsequent analyses and are therefore excluded.

[0038] For each load state change pair in the set of associated pairs, the load state change information is extracted. This load state change is then compared to the upper and lower boundaries of a preset fluctuation range. For example, if the preset fluctuation range is [-ΔPmin, ΔPmax], and the load state change ΔP of the associated pair satisfies -ΔPmin ≤ ΔP ≤ ΔPmax, then the associated pair is considered within the preset fluctuation range, marked as a valid associated pair, and its correspondence with the preset fluctuation range is recorded. If ΔP < -ΔPmin or ΔP > ΔPmax, then the associated pair is considered outside the preset fluctuation range, marked as an invalid associated pair, and excluded. This matching operation is performed for each load state change associated pair in each microgrid, ultimately generating a fluctuation range binding result set.

[0039] Step S250: Perform cross-microgrid time clustering on load state change association pairs that exceed the preset fluctuation range in the fluctuation range binding results, and cluster and integrate all microgrid fluctuation association pairs within the same time scale interval to generate cross-microgrid fluctuation clusters.

[0040] Cross-microgrid time clustering is a method of clustering load state change association pairs that exceed a preset fluctuation range in different microgrids based on time scale intervals. The same time scale interval is a time period with the same start and end times. Fluctuation association pairs are load state change association pairs that exceed the preset fluctuation range. A cross-microgrid fluctuation cluster is a set obtained by clustering and integrating all microgrid fluctuation association pairs within the same time scale interval, reflecting the load fluctuation situation of different microgrids within a certain time period. For load state change association pairs exceeding the preset fluctuation range in the fluctuation range binding result set, the time scale interval information of each association pair is first extracted. Then, these association pairs are classified according to the time scale interval, and all microgrid fluctuation association pairs with the same time scale interval are clustered together. For example, data structures such as hash tables or dictionaries can be used, with the time scale interval as the key, storing the fluctuation association pairs in the corresponding values. For each time scale interval, the clustered fluctuation association pairs are integrated to form a cross-microgrid fluctuation cluster. For example, within the time scale interval [t1, t2], microgrids G, H, and I all have load state change correlation pairs that exceed the preset fluctuation range. These correlation pairs are integrated together to form a cross-microgrid fluctuation cluster. This clustering operation is performed on fluctuation correlation pairs for all time scale intervals, ultimately generating a set of cross-microgrid fluctuation clusters.

[0041] Step S260: Extract time period boundaries for cross-microgrid fluctuation clusters, set the earliest time scale corresponding to each cluster as the start point of the time period and the latest time scale as the end point of the time period, bind the start point and the end point to form a complete time period, and organize to obtain the load synchronization fluctuation time period set.

[0042] In one implementation, step S260 may specifically include the following steps S261 to S266: Step S261: Perform cross-microgrid synchronization of time scale for cross-microgrid fluctuation clusters. Extract the time scale nodes of all involved microgrids from the load state change association pairs of each cluster. Align the time scale nodes according to a preset unified time benchmark to eliminate time benchmark deviations caused by differences in the acquisition systems of different microgrids. Generate a time scale synchronization set corresponding to each cluster. Each time scale in the set is bound to the load fluctuation association marker of at least one microgrid.

[0043] Cross-microgrid synchronization of time scales unifies the time scales of different microgrids within a cluster to a preset time base, eliminating time deviations caused by differences in data acquisition systems. The microgrid in question is the one exhibiting load fluctuation correlation pairs within a specific cluster. Time scale nodes are time scale information extracted from load state change correlation pairs. The preset unified time base is a pre-defined standard time representation method, and all microgrid time scales must be aligned according to this base. The time scale synchronization set is the time scale set corresponding to each cluster after time scale alignment, where each time scale is bound to at least one microgrid load fluctuation correlation marker, indicating which microgrids experienced load fluctuations at that time scale.

[0044] For each cluster in the cross-microgrid fluctuation cluster set, firstly, extract the time scale nodes of all involved microgrids from the load state change association pairs of that cluster. This can be done by traversing each association pair in the cluster, extracting the time scale information, and recording the corresponding microgrid identifier. Then, align these time scale nodes according to a preset unified time benchmark. A method similar to step S210 can be used, employing a time synchronization protocol (such as NTP) to calibrate the time scale, or calculating and adjusting the time offset to eliminate time benchmark deviations caused by differences in the acquisition systems of different microgrids. For each aligned time scale, mark the microgrids that experience load fluctuations at that time scale, forming load fluctuation association markers. For example, if microgrids J and K experience load fluctuations at time scale t, then the load fluctuation association marker bound to that time scale is (microgrid J, microgrid K). Combine all aligned time scales and their corresponding load fluctuation association markers into a time scale synchronization set. This time-scale synchronization operation is performed on all cross-micronetwork fluctuation clusters, ultimately generating a time-scale synchronization set for each cluster.

[0045] Step S262: Perform continuous state association processing on the time scale synchronization set corresponding to each cluster. Concatenate the time scales in the set in sequence according to the order of precedence, add continuous association markers to the load fluctuation records corresponding to adjacent time scales. The marker includes the load state transfer relationship between the front and rear scales, and generate a continuous association chain of load states within the cluster. Each node corresponds to a unique time scale and load state transfer information.

[0046] For the time scale synchronization set corresponding to each cluster, first sort the time scales in the set according to the order of precedence. Sorting algorithms (such as quicksort, mergesort, etc.) can be used to ensure that the time scales are arranged in ascending order. Then, traverse the sorted time scales in sequence. For two adjacent time scales, analyze the load fluctuation records corresponding to them. If the load fluctuation corresponding to the latter time scale is a continuation of the load fluctuation corresponding to the previous time scale, for example, the load continuously increases or continuously decreases, then mark it with continuous association markers such as "load continuously increasing" or "load continuously decreasing"; if the load state turns, for example, from increasing to decreasing, then mark it with markers such as "load turning point". Add these continuous association markers to the load fluctuation records corresponding to adjacent time scales to form a continuous association chain of load states within the cluster. For example, for time scales t1 and t2 (t1 < t2), if the load of microgrid L is increasing at time t1 and the load of microgrid L continues to increase at time t2, the continuous association marker is "load continuously increasing"; if the load of microgrid L is increasing at time t1 and the load of microgrid L starts to decrease at time t2, the continuous association marker is "load turning point". Perform such continuous state association processing on the time scale synchronization set of each cluster, and finally generate a continuous association chain of load states within each cluster.

[0047] Step S263: Perform cross-cluster association comparison processing on the continuous association chains of load states within all clusters. Pairwise bind the continuous association chains of load states within the clusters with common microgrid identifiers, compare the time scale transfer relationships of the associated chains after binding, mark the pairs of clusters with mutual influence, and generate a set of cross-cluster association influence markers. Each marker in the set corresponds to two mutually associated clusters and specific influence dimension information.

[0048] Cross-cluster association comparison involves comparing continuous load state association chains within different clusters to identify clusters sharing a common micronet identifier and analyzing their relationships. A common micronet identifier is the identifier of a micronet that appears in two or more clusters. Pair binding combines continuous load state association chains within clusters sharing a common micronet identifier into association chain pairs. Time scale transmission relationship refers to the chronological order of time scales and the transmission of load states within the association chains. Mutually influencing cluster pairs indicate situations where load fluctuations between two clusters are mutually correlated or influence each other. The cross-cluster association influence label set is a collection summarizing all mutually influencing cluster pairs and their specific influence dimension information. Specific influence dimension information may include the synchronicity of load fluctuations, the magnitude of load changes, etc.

[0049] For all consecutive load state association chains within a cluster, each association chain is first traversed, and the microgrid identifier information is extracted. Then, association chain pairs with common microgrid identifiers are identified. For each pair of association chains, their time scale transmission relationship is compared. By comparing the changes in the load state of the common microgrid in two association chains over time, it can be determined whether there is a synchronous or sequential relationship between them. For example, if the load of the common microgrid in two association chains shows similar fluctuations at the same time scale, or if the load fluctuation of the common microgrid in one association chain precedes the load fluctuation of the common microgrid in another association chain, then the two clusters are considered to have mutual influence. Based on the comparison results, cluster pairs with mutual influence are marked, and the specific influence dimension information is recorded. For example, if the load fluctuations of the common microgrid in two clusters are highly synchronous in time and the load change amplitudes are similar, they are marked as "highly synchronous and similar amplitudes". All cluster pairs with mutual influence and their specific influence dimension information are combined into a cross-cluster association influence label set. This cross-cluster association comparison process is performed on all continuous association chains of load states within clusters, ultimately generating a set of cross-cluster association influence markers.

[0050] Step S264: Record the association chain information for the cross-cluster association impact label set, and record the specific impact dimension of the label as additional information in the continuous association chain of load status within the corresponding cluster, thereby generating a load status association chain with association impact information.

[0051] For each tag in the cross-cluster association impact tag set, identify the two consecutive load state association chains corresponding to that tag within the same cluster. Add the specific impact dimension information from the tag as supplementary information to these two association chains. A new field can be added to the association chain data structure to store the association impact information. For example, for a tag (cluster M, cluster N, specific impact dimension: highly synchronized and similar magnitude), add the specific impact dimension information "highly synchronized and similar magnitude" to the consecutive load state association chains corresponding to clusters M and N. Perform this association chain information recording operation for all tags in the cross-cluster association impact tag set, ultimately generating a load state association chain set with association impact information.

[0052] Step S265: Perform time period boundary extraction processing on the load status association chain with associated impact information, set the starting time node of the association chain as the time period start point, set the last actual load status node of the association chain as the time period end point, bind the start point and end point as a complete time period range, generate a set of time period boundaries corresponding to the cluster, and each set corresponds to a unique cluster and the actual load fluctuation range.

[0053] The time period boundary extraction process determines the boundaries of the time period corresponding to a load state association chain containing related impact information. The start time node is the earliest time scale node in the association chain, which is set as the start point of the time period. The last actual load state node is the last node in the association chain containing actual load state information, and its corresponding time scale is set as the end point of the time period. The complete time period range is the time period consisting of the start point and the end point, which accurately represents the time range of load fluctuations corresponding to the cluster. The cluster-corresponding time period boundary set is a set that summarizes the complete time period ranges corresponding to all clusters. Each set corresponds to a unique cluster and the actual load fluctuation range of that cluster.

[0054] For each association chain in the load state association chain set containing correlation impact information, first identify the start time node and the last actual load state node of the association chain. This can be done by traversing the association chain, recording the earliest time scale and the last node containing actual load state information. Set the time scale of the start time node as the start point of the time period, and the time scale of the last actual load state node as the end point of the time period. Combine the start point and end point of the time period to form a complete time period range. For example, if the start point of the time period is t_start and the end point is t_end, then the complete time period range is [t_start, t_end]. Label each complete time period range with a corresponding cluster identifier, and combine this with the actual load fluctuation of the cluster to form the time period boundary set corresponding to the cluster. Perform this time period boundary extraction process on all association chains in the load state association chain set containing correlation impact information, ultimately generating the time period boundary set corresponding to the cluster.

[0055] Step S266: Standardize and organize the set of time period boundaries corresponding to the clusters, bind the start point and end point of each time period boundary to the microgrid identifier of the corresponding cluster, arrange all time period boundaries in the order of the start time node, and finally obtain the set of load synchronization fluctuation time periods. Each time period in the set contains the actual load fluctuation range of the corresponding microgrid and the recorded related impact information.

[0056] For each time period boundary in the set of time period boundaries corresponding to a cluster, its start and end points are bound to the identifier of the microgrid involved in the corresponding cluster. A field can be added to the data structure of the time period boundaries to store the identifier of the microgrid involved. Then, all time period boundaries are sorted according to the order of their start times. Sorting algorithms (such as quicksort, mergesort, etc.) can be used to ensure that the time period boundaries are arranged in ascending order of their start times. For each time period boundary, the actual load fluctuation range of the corresponding microgrid is determined based on the actual load fluctuation of the cluster, and the recorded correlation impact information is added to that time period. Finally, all processed time period boundaries are organized into a set of load synchronization fluctuation time periods. For example, for a time period boundary (start point: t1, end point: t2, cluster identifier: O), the involved microgrid identifiers for this cluster are (microgrid P, microgrid Q), the actual load fluctuation range is (P1-P2, Q1-Q2), and the associated impact information is "highly synchronized and with similar amplitudes." This information is combined into a complete time period (start point: t1, end point: t2, involved microgrid identifiers: (microgrid P, microgrid Q), actual load fluctuation range: (P1-P2, Q1-Q2), associated impact information: "highly synchronized and with similar amplitudes"). This process is performed on all time period boundaries to ultimately obtain a set of load synchronization fluctuation time periods.

[0057] Step S300: For each overlapping period in the set of load synchronous fluctuation periods, analyze the order of occurrence and the corresponding relationship of the fluctuation amplitude of each microgrid load, determine the causal relationship between each microgrid load, and obtain the causal relationship pairs and relationship records of each microgrid load. The relationship records include the corresponding relationship of the load fluctuation amplitude of the related periods.

[0058] In one implementation, step S300 may specifically include the following steps S310 to S360: Step S310: Extract microgrid fluctuation data for each overlapping period in the load synchronization fluctuation period set. Extract all load record entries within the overlapping period from the load time sequence record set of each microgrid, classify them according to the microgrid identifier, and generate a subset of microgrid load data within the period.

[0059] For each overlapping period in the set of load synchronization fluctuation periods, the load time-series record sets of each microgrid are traversed. For each microgrid's load time-series record set, all load record entries within that period are filtered based on the start and end times of the overlapping period. For example, if the overlapping period is [t1, t2], load record entries with collection timestamps within the range of [t1, t2] are filtered. Then, the filtered load record entries are categorized according to the microgrid identifier, grouping load record entries belonging to the same microgrid together to form a subset of microgrid load data within the period. Data structures such as dictionaries or hash tables can be used, with the microgrid identifier as the key and the load record entries of that microgrid within the overlapping period as the value. This microgrid fluctuation data extraction and categorization operation is performed for all overlapping periods, ultimately generating a subset of microgrid load data within each overlapping period.

[0060] Step S320: Decompose the microgrid load data subset within the time period into timestamps, and split the collection time mark of each load record entry into the timestamp sequence with the smallest time granularity to generate a set of microgrid load timestamp sequences, so that the time granularity of each sequence is consistent.

[0061] For each microgrid load data subset within a given time period, iterate through the load record entries for each microgrid. For each load record entry, split its collection timestamp into a timestamp sequence with the smallest time granularity. For example, if the smallest time granularity is minutes and the collection timestamp is "2024-01-01 10:30:00", then split it into a timestamp sequence with minute intervals from "2024-01-01 10:00:00" to "2024-01-01 10:30:00". Perform this splitting operation on the collection timestamps of each microgrid load record entry, ultimately generating a set of microgrid load timestamp sequences. To ensure that the time granularity of each sequence remains consistent, a preset smallest time granularity can be used for partitioning during the splitting process. Perform this timestamp splitting operation on all microgrid load data subsets within all time periods to obtain the final set of microgrid load timestamp sequences.

[0062] Step S330: Perform time-series correlation on the microgrid load timestamp sequence set, compare the timestamp of the first load fluctuation in the timestamp sequences of different microgrids, arrange each microgrid according to the order of the timestamps, and generate a microgrid fluctuation time-series sequence.

[0063] In one implementation, step S330 may specifically include the following steps S331 to S336: Step S331: Identify the starting point of load fluctuation in the microgrid load timestamp sequence set. Extract the first timestamp of load fluctuation from each microgrid timestamp sequence, bind the timestamp to the corresponding microgrid identifier, and generate a microgrid fluctuation start timestamp record.

[0064] The load fluctuation start point identification involves finding the earliest time point in the timestamp sequence of each microgrid where a load fluctuation occurs. The timestamp of the first occurrence of a load fluctuation is the timestamp corresponding to that time point. Binding this timestamp to the corresponding microgrid identifier records the start time of each microgrid's load fluctuation. The microgrid fluctuation start timestamp record is a set of records combining each microgrid's microgrid identifier and its first timestamp of a load fluctuation. For the set of microgrid load timestamp sequences, the timestamp sequence of each microgrid is traversed. For each timestamp sequence, the load values ​​corresponding to adjacent timestamps are compared sequentially. Assuming a preset load fluctuation threshold of ΔP, if the absolute value of the difference between the load value corresponding to the later timestamp and the load value corresponding to the earlier timestamp is greater than ΔP, then the later timestamp is considered the first timestamp of a load fluctuation. This timestamp is then bound to the corresponding microgrid identifier, for example, forming a tuple (microgrid identifier, timestamp). This fluctuation start point identification and binding operation is performed on all microgrid timestamp sequences, ultimately generating a set of microgrid fluctuation start timestamp records.

[0065] Step S332: Perform global time sorting on the microgrid fluctuation start timestamp records, arrange the fluctuation start timestamps of all microgrids in chronological order, and generate a global fluctuation start time sorting sequence. Each element in the sequence contains the microgrid identifier and the corresponding timestamp.

[0066] For the set of microgrid fluctuation start timestamp records, a sorting algorithm (such as quicksort, mergesort, etc.) is used to sort the timestamps. During sorting, the timestamp is used as the sorting key, while maintaining the correspondence between the microgrid identifier and the timestamp. For example, if the set of microgrid fluctuation start timestamp records contains (microgrid T, t3), (microgrid U, t1), and (microgrid V, t2), the resulting global fluctuation start time sorting sequence is ((microgrid U, t1), (microgrid V, t2), (microgrid T, t3)). Sorting the set of microgrid fluctuation start timestamp records ultimately generates the global fluctuation start time sorting sequence.

[0067] Step S333: Perform adjacent microgrid association on the global fluctuation start time sorting sequence, bind microgrid identifiers at adjacent positions in the sequence in pairs, and generate adjacent microgrid time series association pairs.

[0068] For the global fluctuation start time sorting sequence, starting from the first element of the sequence, the microgrid identifiers in adjacent elements are paired and bound together. For example, for the global fluctuation start time sorting sequence ((microgrid W, t1), (microgrid X, t2), (microgrid Y, t3)), adjacent microgrid time-series association pairs ((microgrid W, microgrid X), (microgrid X, microgrid Y)) are generated. This binding operation is performed on all adjacent elements in the global fluctuation start time sorting sequence, ultimately generating a set of adjacent microgrid time-series association pairs.

[0069] Step S334: Perform time difference association on adjacent microgrid time series association pairs. Calculate the time difference between the fluctuation start timestamps of the two microgrids in each association pair to generate a time difference association result. The result includes the microgrid pair identifier and the corresponding time difference.

[0070] For a set of adjacent microgrid time-series association pairs, iterate through each pair. For each pair, find the fluctuation start timestamps of the two microgrids in the pair from the global fluctuation start time sorting sequence. Then, calculate the time difference between these two timestamps. For example, if the association pair is (microgrid Z1, microgrid Z2), and their fluctuation start timestamps are t4 and t5 respectively, then the time difference Δt = t5 - t4. Combine the microgrid pair identifier (microgrid Z1, microgrid Z2) and the time difference Δt into a time difference association result, such as ((microgrid Z1, microgrid Z2), Δt). Perform this time difference calculation and combination operation on all adjacent microgrid time-series association pairs, finally generating a set of time difference association results.

[0071] Step S335: Mark the time difference association results with time-series attributes. Based on the calculation results of the time difference, mark the pre- and post-attributes of the micro-networks in the association pair to generate micro-network time-series association pairs with attribute tags.

[0072] For the set of time difference association results, iterate through each result. For each association pair in the result, if the time difference is greater than 0, it indicates that the load fluctuation of the previous microgrid occurred first, so mark the previous microgrid as the "preceding" attribute and the subsequent microgrid as the "following" attribute; if the time difference is less than 0, swap the tags. For example, for the time difference association result ((microgrid A1, microgrid A2), Δt = 5 minutes), mark microgrid A1 as the "preceding" attribute and microgrid A2 as the "following" attribute, forming a microgrid time-series association pair with attribute tags ((microgrid A1 (preceding), microgrid A2 (following))). Perform this time-series attribute tagging operation on all time difference association results, finally generating a set of microgrid time-series association pairs with attribute tags.

[0073] Step S336: Perform sequence integration on the microgrid time series correlation pairs with attribute tags, and concatenate all correlation pairs in the order of the global fluctuation start time sorting sequence to generate the microgrid fluctuation time series sequence.

[0074] Sequence integration involves connecting attribute-tagged microgrid time-series correlation pairs according to the order of the global fluctuation start time sequence to form a complete sequence. For a set of attribute-tagged microgrid time-series correlation pairs, the order of the global fluctuation start time sequence is first determined. Then, the attribute-tagged microgrid time-series correlation pairs are concatenated according to this order. For example, if the global fluctuation start time sorting sequence is ((microgrid B1, t1), (microgrid B2, t2), (microgrid B3, t3)), and the set of attribute-labeled microgrid time-series association pairs is (((microgrid B1 (preceding), microgrid B2 (following)), (microgrid B2 (preceding), microgrid B3 (following))), then the association pairs are concatenated according to the global fluctuation start time sorting sequence to generate the microgrid fluctuation time-series sequence (microgrid B1 (preceding) - microgrid B2 (following) - microgrid B3 (following)). During the concatenation process, it is essential to ensure that the order of each association pair is consistent with the global fluctuation start time sorting sequence to accurately reflect the chronological order of load fluctuations in each microgrid. Performing this sequence integration operation on the attribute-labeled microgrid time-series association pair set ultimately generates the complete microgrid fluctuation time-series sequence.

[0075] Step S340: Correlate the fluctuation state of the microgrids according to the time sequence of fluctuations, compare the load state change process of different microgrids in the overlapping period, and pair the load state change process of each microgrid with the change process of other microgrids to generate microgrid load state correspondence pairs.

[0076] For the chronological sequence of microgrid fluctuations, focusing on overlapping periods, the load state change process of each microgrid is extracted from the subset of microgrid load data within each period. This load state change process can be represented as a time series, where each element is the load value at the corresponding time point. Then, these load state change processes are compared pairwise. For example, for microgrids C and D, the load state change time series of microgrid C and microgrid D are compared point-by-point to analyze the direction and magnitude of load value changes at the same time points. The load state change processes of microgrids C and D are then bound into a microgrid load state correspondence pair. This pairwise binding operation is performed on all microgrids in the chronological sequence of microgrid fluctuations, ultimately generating a set of microgrid load state correspondence pairs. During the comparison process, the Dynamic Time Warping (DTW) algorithm can be used to measure the similarity between two load state change processes. This algorithm can handle time series of different lengths and can, to some extent, eliminate the influence of time axis scaling and offset on similarity calculation. Specifically, the DTW algorithm constructs a distance matrix and finds the optimal alignment path between two time series by calculating the cumulative distance of the elements in the matrix, thereby obtaining their similarity.

[0077] Step S350: Determine the causal association of the microgrid load status corresponding pairs, and causally bind the status corresponding pairs of the preceding and subsequent microgrids in the microgrid fluctuation time sequence to generate causal association candidate pairs.

[0078] In one implementation, step S350 may specifically include the following steps S351 to S356: Step S351: Extract the state change links for the corresponding association pairs of microgrid load states. Extract the load state change process of two microgrids in the overlapping time period from each association pair. The link contains load state nodes under continuous timestamps to generate microgrid load state change link pairs.

[0079] State change link extraction separates the load state change process of two microgrids within an overlapping time period from the microgrid load state correspondence pairs and represents it as a link containing load state nodes at consecutive timestamps. A load state node is the load state information of a microgrid at a specific timestamp, such as load value or load change rate. For each correlation pair in the set of microgrid load state correspondence pairs, the load state change process of the two microgrids within the overlapping time period is extracted from the correlation pair. These load state change processes are then arranged in chronological order of timestamps, with the load state information at each timestamp treated as a node, constructing a load state change link. For example, for a correlation pair of microgrids E and F, within the overlapping time period [t1, t2], the load state information of microgrid E at timestamps t1, t1+Δt, t1+2Δt…t2 is sequentially connected into a link, and similarly, the load state information of microgrid F is also constructed into a link, forming a microgrid load state change link pair (microgrid E link, microgrid F link). Perform this state change link extraction operation on all microgrid load state corresponding association pairs, and finally generate a set of microgrid load state change link pairs.

[0080] Step S352: Perform link node association comparison on the link pairs of microgrid load status change links, bind the timestamp nodes in the two links one-to-one, generate timestamp node association pairs, and ensure that each pair of nodes corresponds to the same timestamp.

[0081] For each link pair in the microgrid load state change link pair set, iterate through the timestamp nodes of one of the links. For each timestamp node, find a node with the same timestamp in the other link and bind them together to form a timestamp node association pair. For example, in the microgrid load state change link pair (microgrid G ​​link, microgrid H link), if there is a node with timestamp t in microgrid G ​​link, find a node with timestamp t in microgrid H link and bind them together to form a timestamp node association pair (microgrid G ​​node (t), microgrid H node (t)). Perform this link node association comparison operation on all link pairs in the microgrid load state change link pair set, ultimately generating a set of timestamp node association pairs.

[0082] Step S353: Perform state change synchronization association on the timestamp node association pairs, compare the load state change direction of the two microgrids in each pair of nodes, and generate state change synchronization association results, which include the synchronization marker of the corresponding timestamp.

[0083] For each timestamp node association pair in the set, the load status information of the two microgrids in each pair is extracted, and their load status change is calculated (i.e., the load value at the current timestamp minus the load value at the previous timestamp). The direction of load status change is determined by the sign of the load status change. If the load status changes of the two microgrids have the same sign (both positive or both negative), they are considered to be synchronized and marked as "synchronized"; if they have different signs (one positive and one negative), they are marked as "asynchronous"; if the load status change of one of the microgrids is zero, synchronization is determined based on the specific situation. For example, for the timestamp node association pair (microgrid node I (t), microgrid node J (t)), the load status change of microgrid I is ΔPI, and the load status change of microgrid J is ΔPJ. If ΔPI > 0 and ΔPJ > 0, it is marked as "synchronized". This state change synchronization association operation is performed on all pairs in the timestamp node association pair set, and finally, a set of state change synchronization association results is generated.

[0084] Step S354: Extract continuous synchronization intervals from the state change synchronization association results. Extract intervals from the association results where multiple consecutive timestamp nodes show synchronization markers, and generate continuous synchronization interval records.

[0085] A continuous synchronization interval is a time period in which the load state changes of two microgrids remain synchronized. A continuous synchronization interval record is a collection of start and end timestamps and synchronization information for these intervals. These records allow for further analysis of the duration and stability of the load state change synchronization between the two microgrids. The synchronization result set is traversed in chronological order of timestamps. When multiple consecutive timestamps are marked as "synchronized," the start and end timestamps of these continuous synchronization intervals are recorded. For example, in the synchronization result set, if the synchronization markers from timestamp t1 to t5 are all "synchronized," then the continuous synchronization interval is recorded as [t1, t5]. This extraction operation is performed on all continuous synchronization intervals in the synchronization result set, ultimately generating a continuous synchronization interval record set.

[0086] Step S355: Perform causal weight association on the continuous synchronization interval records, associate the length of the continuous synchronization interval with the corresponding microgrid time sequence attributes, assign causal association weights to the preceding microgrid, and generate causal weight association results.

[0087] Causal weight association combines the length of a continuous synchronization interval with the temporal sequence of the microgrids to assign a causal association weight to the preceding microgrid. The length of the continuous synchronization interval reflects the duration of the synchronization of load state changes between the two microgrids; the longer the duration, the greater the potential influence of the preceding microgrid on the following microgrid. The causal association weight is a numerical value used to measure the degree of causal influence of the preceding microgrid on the load state changes of the following microgrid. The causal weight association result is a set recording the preceding microgrid, the following microgrid, and the causal association weight corresponding to each continuous synchronization interval. These results can be used to preliminarily determine the strength of the causal relationship between the two microgrids. For each continuous synchronization interval record in the continuous synchronization interval record set, the preceding and following microgrids corresponding to that interval are determined based on the temporal sequence of microgrid fluctuations. The length of the continuous synchronization interval is calculated as the end timestamp minus the start timestamp. A linear mapping method can be used to map the length of the continuous synchronization interval to a causal association weight range, such as [0,1]. The longer the continuous synchronization interval, the greater the causal association weight assigned to the preceding microgrid. For example, for a continuous synchronization interval [t3, t7], the corresponding preceding micronet is micronet K, the following micronet is micronet L, and the length of the continuous synchronization interval is Δt = t7 - t3. Δt is mapped to a causal association weight wK, forming a causal weight association result (micronet K, micronet L, wK). This causal weight association operation is performed on all records in the continuous synchronization interval record set, ultimately generating a causal weight association result set.

[0088] Step S356: Confirm the causality of the causal weight association results. Based on the distribution of weights in the causal weight association results, confirm the causal relationship in the association pairs and generate causal association candidate pairs.

[0089] If the causal association weight reaches a certain threshold, the preceding microgrid is considered to have a causal impact on the load state change of the following microgrid, and this association pair is confirmed as a causal association candidate pair. Causal association candidate pairs are combinations of preceding and following microgrids that may have a causal relationship after causal confirmation; these candidate pairs will be further processed and their association records organized. For the causal weight association result set, a causal association weight threshold w_threshold is set. For each causal weight association result (preceding microgrid M, following microgrid N, wM), if wM > w_threshold, the preceding microgrid M is considered to have a causal relationship with the load state change of the following microgrid N, and (preceding microgrid M, following microgrid N) is considered a causal association candidate pair. This causal confirmation operation is performed on all results in the causal weight association result set, ultimately generating a causal association candidate pair set.

[0090] Step S360: Organize the association records of the causal association candidate pairs, bind the microgrid identifier, overlapping time period and status correspondence in the candidate pairs, and obtain the causal association pairs and association records of each microgrid load.

[0091] For each causal candidate pair in the causal association candidate pair set, the load state change information of the two microgrids in the candidate pair during the overlapping time period is extracted from the microgrid load data subset within the time period, and their state correspondence is analyzed (such as the synchronicity of load changes, amplitude ratio, etc.). The microgrid identifier, overlapping time period, and state correspondence in the candidate pair are bound together to form a complete association record. For example, for the causal candidate pair (microgrid O, microgrid P), the overlapping time period is [t4, t5], and the state correspondence is "synchronous load changes with an amplitude ratio of 1:2", then the association record is (microgrid O, microgrid P, [t4, t5], "synchronous load changes with an amplitude ratio of 1:2"). This association record organization operation is performed on all candidate pairs in the causal association candidate pair set, and finally, the causal association pairs and association relationship record sets of each microgrid load are obtained.

[0092] Step S400: Based on the causal relationship pairs and the relationship records, select the load record entries corresponding to the causal relationship pairs from the load time sequence record set of each microgrid to obtain the load relationship record set.

[0093] In one implementation, step S400 may specifically include the following steps S410 to S460: Step S410: Parse the causal relationship pairs and relationship records of each microgrid load, extract the identifiers of the two microgrids from each causal relationship pair, extract the corresponding overlapping time periods from the relationship records, and generate microgrid identifier pairs and corresponding time period sets.

[0094] Association information parsing processes the causal relationship records of each microgrid load, extracting key information. A microgrid identifier uniquely identifies each microgrid, and overlapping periods are the time intervals during which load fluctuations of two microgrids occur synchronously and are causally related. The microgrid identifier pair and corresponding time interval set is a set that combines the microgrid identifiers from each causal relationship into an identifier pair and binds it to the corresponding overlapping time interval. This set clearly identifies the microgrids and time range involved in each causal relationship.

[0095] For each microgrid load's causal relationship pairs and related relationship record set, iterate through each causal relationship pair and its corresponding related relationship record. Extract the identifiers of the two microgrids from the causal relationship pairs to form microgrid identifier pairs. Extract the corresponding overlapping time period information from the related relationship records, and combine all microgrid identifier pairs and their corresponding time periods to generate a set of microgrid identifier pairs and their corresponding time periods.

[0096] Step S420: Index the load time series records of each microgrid using the microgrid identifier, establish a retrieval directory indexed by the microgrid identifier for each microgrid load time series record set, and generate a microgrid load record retrieval index.

[0097] In one implementation, step S420 may specifically include the following steps S421 to S426: Step S421: Extract the microgrid identifier from the load time sequence record set of each microgrid, parse the corresponding microgrid identifier from each load record entry, bind the microgrid identifier to the load record entry, and generate a set of load record entries with identifiers.

[0098] For each microgrid's load time-series record set, iterate through each load record entry. Parse the value of the microgrid identifier field from the load record entry's data structure and bind the microgrid identifier to the load record entry. For example, if a load record entry is a triple (microgrid S, t8, P) containing the microgrid identifier, collection time, and load value, then retain it in the set of identified load record entries. Perform this microgrid identifier extraction and binding operation on all load record entries in the load time-series record sets of all microgrids, ultimately generating a set of identified load record entries.

[0099] Step S422: Perform deduplication association on the set of identified load record entries, classify all load record entries corresponding to the same microgrid identifier, eliminate duplicate load record entries under the same microgrid, and generate a deduplicated subset of microgrid load records.

[0100] For a set of identifiable load record entries, a hash table or dictionary is used, with the microgrid identifier as the key, to store load record entries corresponding to the same microgrid identifier in the same list. During storage, a uniqueness check is performed on each load record entry, for example, by comparing key information such as collection time and load value to determine if there are duplicates. If duplicate load record entries are found, only one is retained. For example, if the set of identifiable load record entries contains (microgrid T, t9, P1) and (microgrid T, t9, P1), only one of the entries is retained. This identifier deduplication and association operation is performed on all entries in the set of identifiable load record entries, ultimately generating a deduplicated subset of microgrid load records.

[0101] Step S423: Sort the deduplicated microgrid load record subsets by time scale, and arrange the load record entries in each subset according to the order of the collection time markers to generate a time-sorted microgrid load record sequence.

[0102] For the deduplicated subset of microgrid load records, iterate through each subset. For each subset, use a sorting algorithm (such as quicksort, mergesort, etc.) to sort the load record entries according to their acquisition time markers. For example, if the subset contains load record entries (microgrid U, t11, P3), (microgrid U, t10, P2), and (microgrid U, t12, P4), the sorted result is a time-ordered sequence of microgrid load records: ((microgrid U, t10, P2), (microgrid U, t11, P3), (microgrid U, t12, P4)). Perform this time-scale sorting operation on all subsets in the deduplicated subset of microgrid load records, ultimately generating a time-ordered set of microgrid load record sequences.

[0103] Step S424: Generate index entries for the time-sorted microgrid load record sequence. Create an index entry for each microgrid identifier. The entry content includes the microgrid identifier and the storage location of the corresponding load record sequence, and generate a set of microgrid index entries.

[0104] For a time-sorted set of microgrid load record sequences, iterate through each sequence. Create an index entry for each sequence's corresponding microgrid identifier, recording the microgrid identifier and the sequence's storage location in the storage medium (e.g., file path, memory address). For example, for the time-sorted microgrid load record sequences ((microgrid V, t13, P5), (microgrid V, t14, P6)), create an index entry (microgrid V, storage location X). Perform this index entry generation operation on all sequences in the time-sorted microgrid load record sequence set, ultimately generating a set of microgrid index entries.

[0105] Step S425: Construct an index directory for the microgrid index entry set, arrange all microgrid index entries in dictionary order of microgrid identifiers, establish an index directory with microgrid identifiers as search keys, and generate an initial microgrid load record retrieval index.

[0106] For the microgrid index entry set, a sorting algorithm (such as quicksort, mergesort, etc.) is used to sort the microgrid index entries in lexicographical order according to their microgrid identifiers. A hash table or dictionary is created, using the microgrid identifier as the key, and the sorted microgrid index entries are stored as values ​​within it. For example, for the microgrid index entry set ((microgrid W, storage location Y), (microgrid X, storage location Z)), after sorting in lexicographical order, an index directory is constructed ({microgrid W:(microgrid W, storage location Y), microgrid X:(microgrid X, storage location Z)}). Performing this index directory construction operation on the microgrid index entry set ultimately generates the initial microgrid load record retrieval index.

[0107] Step S426: Perform validity verification on the initial microgrid load record retrieval index, verify whether the load record sequence corresponding to each index entry is complete, exclude invalid index entries, and obtain the microgrid load record retrieval index.

[0108] For the initial microgrid load record retrieval index, each index entry is traversed. Based on the storage location information in the index entry, the corresponding load record sequence is accessed. The load record sequence is checked for missing load record entries, correct time order, etc. If an incomplete load record sequence or other problems are found, the index entry is marked as invalid and excluded from the initial microgrid load record retrieval index. For example, if the load record sequence corresponding to index entry (microgrid Y, storage location A) is missing load record entries for a certain time period, the index entry is excluded. This validity check is performed on all index entries in the initial microgrid load record retrieval index, ultimately resulting in the microgrid load record retrieval index.

[0109] Step S430: Generate retrieval requests for microgrid identifier pairs and corresponding time period sets. Convert each microgrid identifier and corresponding time period into a retrieval request. The request content includes the microgrid identifier and the start and end time scales of the time period, and generate a microgrid load retrieval request set.

[0110] For each microgrid identifier pair and its corresponding time period set, iterate through each microgrid identifier pair and its corresponding time period. For each microgrid identifier in each microgrid identifier pair, generate a retrieval request by combining the corresponding time period information. The retrieval request includes the microgrid identifier, the start time scale, and the end time scale of the time period. For example, for the microgrid identifier pair (microgrid Z, microgrid A1) and the corresponding time period [t15, t16], generate retrieval requests (microgrid Z, t15, t16) and (microgrid A1, t15, t16). Summarize all generated retrieval requests to generate a microgrid load retrieval request set.

[0111] Step S440: Perform batch retrieval of the microgrid load retrieval request set, match each retrieval request with the microgrid load record retrieval index, extract the load record entries within the corresponding time period from the load time series record set of the corresponding microgrid, and generate the retrieved load record entry set.

[0112] For each microgrid load retrieval request set, iterate through each request. Extract the microgrid identifier and time period information (start and end time scales) from the request. Based on the microgrid identifier, search for the corresponding index entry in the microgrid load record retrieval index to obtain the storage location of the load record sequence for that microgrid. Access the load record sequence at that storage location and filter the load record entries within the corresponding time period based on the time period information. For example, for the retrieval request (microgrid B1, t17, t18), find the load record sequence corresponding to microgrid B1 in the microgrid load record retrieval index and filter out load record entries whose collection time markers are in the range [t17, t18]. Summarize the load record entries retrieved from all retrieval requests to generate a set of retrieved load record entries.

[0113] Step S450: Perform causal association marking on the retrieved load record entry set, bind each load record entry with its corresponding microgrid identifier and overlapping time period, and generate load record entries with causal marking.

[0114] In one implementation, step S450 may specifically include the following steps S451 to S456: Step S451: Parse the item attributes of the retrieved load record item set, and parse the collection time stamp and microgrid identifier from each load record item to generate a load item attribute information set.

[0115] For the retrieved set of load record entries, iterate through each entry. Parse the values ​​of the data acquisition timestamp and microgrid identifier fields from the data structure of each entry, and combine them into a load entry attribute information. For example, for the load record entry (microgrid C1, t19, P7), parse the data acquisition timestamp t19 and the microgrid identifier microgrid C1 to form the load entry attribute information (microgrid C1, t19). Summarize all the load entry attribute information to generate a load entry attribute information set.

[0116] Step S452: Parse the time period attributes of the microgrid identifier pairs and the corresponding time period sets, parse the start and end time scales from each time period, bind the time period attributes with the corresponding microgrid identifier pairs, and generate a time period-microgrid identifier pair binding set.

[0117] For each microgrid identifier pair and its corresponding time period set, iterate through each microgrid identifier pair and its corresponding time period. Parse the start and end time scales from the time period and bind them to the microgrid identifier pair. For example, for the microgrid identifier pair (microgrid D1, microgrid E1) and its corresponding time period [t20, t21], parse the start time scale t20 and the end time scale t21, forming a time period-microgrid identifier pair binding ((microgrid D1, microgrid E1), [t20, t21]). Summarize all the time period-microgrid identifier pair bindings to generate a time period-microgrid identifier pair binding set.

[0118] Step S453: Perform attribute matching on the set of load item attribute information and the time period-microgrid identifier binding set, and match the collection time mark of each load item with the start and end time scale of the time period to generate attribute matching results.

[0119] For each load item attribute in the load item attribute information set, iterate through each time period-microgrid identifier pair in the time period-microgrid identifier pair binding set. Compare the collection time stamp in the load item attribute information with the start and end time scales in the time period-microgrid identifier pair binding. If the collection time stamp is within the range of the start and end time scales, the load item is considered to match the time period-microgrid identifier pair binding. For example, for the load item attribute information (microgrid F1, t22) and the time period-microgrid identifier pair binding ((microgrid F1, microgrid G1), [t22, t23]), since t22 is within the range [t22, t23], the load item is considered to match the binding. Perform this attribute matching operation on all items in the load item attribute information set and the time period-microgrid identifier pair binding set, finally generating an attribute matching result set.

[0120] Step S454: Perform causal association binding on the attribute matching results, bind the successfully matched load entries with the corresponding microgrid identifiers and time periods, and generate causal association binding records.

[0121] For the attribute matching result set, iterate through each matching result. If the matching result is successful, find the corresponding load record entry from the load entry attribute information set, and find the corresponding microgrid identifier pair and time period from the time period-microgrid identifier pair binding set. Bind the load record entry, microgrid identifier pair, and time period to form a causal association binding record. For example, for a successfully matched result (load entry (microgrid H1, t24, P8), time period-microgrid identifier pair binding ((microgrid H1, microgrid I1), [t24, t25])), generate a causal association binding record ((microgrid H1, t24, P8), (microgrid H1, microgrid I1), [t24, t25]). Perform this causal association binding operation on all matching results in the attribute matching result set, and finally generate a causal association binding record set.

[0122] Step S455: Mark and embed the causal association binding records, embed the microgrid identifier and time period information in the binding records into the corresponding load record entries, and generate load record entries with embedding marks.

[0123] For each causal binding record set, iterate through it. Extract the load record entry, microgrid identifier pair, and time period information from the binding record. Embed the microgrid identifier pair and time period information into the load record entry's data structure; for example, add microgrid identifier pair and time period information fields to the end of the load record entry. Perform this tagging and embedding operation on all binding records in the causal binding record set, ultimately generating a set of load record entries with embedding tags.

[0124] Step S456: Perform tag verification on the load record entries with embedded tags, verify whether the tag information embedded in each entry is consistent with the attribute matching result, exclude entries with incorrect tags, and generate load record entries with causal tags.

[0125] Tag verification ensures the accuracy of tag information in load record entries with embedded tags. Verifying that the embedded tag information for each entry matches the attribute matching results involves checking whether the embedded microgrid identifier pairs and time period information conform to the previous attribute matching results. Excluding incorrectly tagged entries improves data accuracy and ensures the quality of load record entries with causal tags. The final set of load record entries with causal tags is the result of tag verification, after removing incorrectly tagged entries.

[0126] For a set of load record entries with embedded tags, iterate through each entry. Extract the embedded microgrid identifier pair and time period information from the entry, and simultaneously find the corresponding attribute matching result from the attribute matching result set. Compare the embedded tag information with the microgrid identifier pair and time period information in the attribute matching result. If they do not match, the entry is considered incorrectly tagged and is excluded. For example, if the microgrid identifier pair in the attribute matching result for the load record entry with embedded tags (microgrid L1, t28, P10, (microgrid L1, microgrid M1), [t28, t29]) is (microgrid L1, microgrid N1), then the entry is considered incorrectly tagged and is excluded. Perform this tag verification operation on all entries in the set of load record entries with embedded tags, ultimately generating a set of load record entries with causal tags.

[0127] Step S460: Classify and organize the load record entries with causal tags, categorize them according to microgrid identifiers and overlapping time periods, and generate a load association record set.

[0128] For a set of load record entries with causal tags, a data structure such as a hash table or dictionary is used, with the combination of microgrid identifier pairs and overlapping time periods as keys, to store load record entries with the same key in the same list. For example, for load record entries with causal tags (microgrid O1, t30, P11, (microgrid O1, microgrid P1), [t30, t31]) and (microgrid O1, t31, P12, (microgrid O1, microgrid P1), [t30, t31]), they are stored in a list with the key ((microgrid O1, microgrid P1), [t30, t31]). This sorting operation is performed on all entries in the set of load record entries with causal tags, ultimately generating a load-related record set.

[0129] Step S500: Perform load trend extrapolation calculations based on the load association record set to obtain the load extrapolation results of each microgrid. Determine the load control mode of the virtual power plant based on the load extrapolation results, generate corresponding load control instructions, and send them to the load control terminals of each microgrid.

[0130] In one implementation, step S500 may specifically include the following steps S510-S560: Step S510: Perform causal grouping on the load association record set. Group the load association records according to each causal pair and the corresponding overlapping time period to generate a causal group set. Each group contains the load record entries and association information of the corresponding microgrid.

[0131] Causal grouping categorizes load-related records in a load-related record set according to causal pairs and corresponding overlapping time periods, grouping records with the same causal relationship and overlapping time periods together. For the load-related record set, a data structure such as a hash table or dictionary is used, with the combination of causal pairs and corresponding overlapping time periods as the key, storing load-related records with the same key in the same list. For example, for load-related record sets containing ((microgrid Q1, microgrid R1), [t32, t33], (microgrid Q1, t32, P13)) and ((microgrid Q1, microgrid R1), [t32, t33], (microgrid Q1, t33, P14)), they are stored in a list with the key ((microgrid Q1, microgrid R1), [t32, t33]). This causal grouping operation is performed on all records in the load-related record set, ultimately generating a causal grouping set.

[0132] Step S520: Perform time-series extrapolation of load trends for the causal association group set, extract continuous load state change processes from the load record entries of each group, extrapolate the load state of subsequent time scales based on the process, and generate single-group load trend extrapolation results.

[0133] In one implementation, step S520 may specifically include the following steps S521 to S526: Step S521: Extract the load status sequence for each group in the causal association grouping set. Extract the collection time stamp and corresponding load status from the load record entries of the group to generate a load status sequence sorted by time. Each element in the sequence contains a time scale and load status.

[0134] Load state sequence extraction involves separating the acquisition time stamps and corresponding load state information from the load record entries of each group in a causal association grouping set. The acquisition time stamps determine the temporal order of load states, which can be represented by indicators such as load value and load change rate. The time-sorted load state sequence is formed by arranging the extracted acquisition time stamps and load state information in chronological order. Each element contains a time scale and a corresponding load state, clearly showing how the load state changes over time.

[0135] For each group in the causal association grouping set, iterate through the load record entries within that group. Extract the acquisition time stamp and load status information from each load record entry, and combine them into an element. For example, for the load record entry (microgrid S1, t34, P15), extract the acquisition time stamp t34 and load status P15 to form the element (t34, P15). Sort all elements according to the chronological order of their acquisition time stamps to generate a time-sorted load status sequence. Perform this load status sequence extraction operation on all groups in the causal association grouping set, ultimately generating a time-sorted load status sequence for each group.

[0136] Step S522: Construct a state change link for the load state sequence sorted by time, and link consecutive load states in the sequence to generate a load state change link, which contains the transmission relationship between the previous and subsequent states.

[0137] The load state change link construction involves associating consecutive load states in a time-ordered load state sequence to form a link reflecting the transmission relationship of load state changes. The transmission relationship between preceding and subsequent states can be represented by indicators such as the amount and rate of change of load states, reflecting the changing pattern of load states between adjacent time scales. The load state change link connects consecutive load states in chronological order, allowing for a clearer analysis of load state change trends and patterns. For a time-ordered load state sequence, starting from the first element of the sequence, adjacent elements are linked together sequentially. For example, for a load state sequence ((t35, P16), (t36, P17), (t37, P18)) sorted by time, (t35, P16) and (t36, P17) are bound together, and the transmission relationships between them, such as the load state change amount (e.g., ΔP = P17 - P16) and the rate of change (e.g., ΔP / P16), are recorded. Then (t36, P17) and (t37, P18) are bound together, and so on, to generate a load state change link. This state change link construction operation is performed on all sequences in the time-sorted load state sequence set, ultimately generating the load state change link corresponding to each sequence.

[0138] Step S523: Extend the load state change link by performing trend extrapolation. Based on the last load state in the link and the state transmission relationship, generate the predicted load state for subsequent time scales and generate the extended load state change link.

[0139] For a load state change link, find the last load state element in the link. Based on the state transfer relationship between this element and the previous element (such as the amount and rate of change of load state), predict the load state for subsequent time scales. For example, if the last load state element is (t38, P19), and the rate of change of load state from the previous element to this element is r, then the predicted load state for the next time scale t39 is P20 = P19 * (1 + r). Add the predicted load state element (t39, P20) to the load state change link to generate an extended load state change link. Perform this trend-based link extension operation on all links in the load state change link set, ultimately generating the extended load state change link for each link.

[0140] Step S524: Perform a state simulation verification on the extended load state change link, verify the transmission relationship between the simulated load state and the preceding state in the link, ensure that the simulated state conforms to the transmission law, and generate the verified simulated load state.

[0141] The simulation state verification involves checking the simulated load state in the extended load state change link to ensure it conforms to the transmission relationship of the preceding states in the link. If the simulated load state does not conform to the transmission law, it may be due to inaccurate prediction methods or abnormal situations, requiring adjustment of the simulation results. The verified simulated load state is the simulated load state that conforms to the transmission relationship of the preceding states after verification, which can improve the accuracy of load trend prediction. For the extended load state change link, the simulated load state elements and the preceding load state elements in the link are extracted. Based on the transmission relationship of the preceding states (such as the amount of load state change, the rate of change, etc.), the load state value that should be obtained under this transmission relationship is calculated. The calculated load state value is compared with the simulated load state value. If the difference between the two is within the allowable error range, the simulated load state is considered to conform to the transmission law and is used as the verified simulated load state; if the difference exceeds the error range, the simulated load state needs to be adjusted, such as recalculating or using other prediction methods. This inference state verification operation is performed on all links in the extended load state change link set, and finally the verified inference load state corresponding to each link is generated.

[0142] Step S525: Perform time-series binding on the verified projected load status, binding the projected load status with the corresponding subsequent time scale to generate a load trend projection result with time scale.

[0143] Time-series binding associates verified projected load states with corresponding subsequent time scales, clarifying the time point corresponding to each projected load state. For the set of verified projected load states, each verified projected load state is iterated over. It is then bound to its corresponding subsequent time scale, forming an element. For example, for the verified projected load state P21 and the subsequent time scale t40, it is bound to the element (t40, P21). All bound elements are then aggregated to generate a load trend projection result with a time scale. This time-series binding operation is performed on all states in the set of verified projected load states, ultimately generating a load trend projection result with a time scale for each group.

[0144] Step S526: Group and correlate the load trend projection results with time scale, bind the results with the corresponding causal correlation groups, and generate a single group of load trend projection results.

[0145] Group association involves mapping time-scaled load trend projection results to corresponding causal association groups, clearly defining which causal association group each projection result is derived from. For a set of time-scaled load trend projection results, each time-scaled load trend projection result is bound to its corresponding causal association group based on the group's identifier information. For example, for the time-scaled load trend projection results ((t41, P22), (t42, P23)) and the causal association groups ((microgrid T1, microgrid U1), [t43, t44]), they are bound together to generate single-group load trend projection results (((microgrid T1, microgrid U1), [t43, t44]), ((t41, P22), (t42, P23))). This group association operation is performed on all results in the time-scaled load trend projection result set, ultimately generating a single-group load trend projection result corresponding to each causal association group.

[0146] Step S530: Perform cross-microgrid collaborative correction on the load trend projection results of a single group. Combine the microgrid correlation relationships between different causal groups, and collaboratively adjust the projection results of each group to generate cross-microgrid collaborative load trend projection results.

[0147] For a single set of load trend projection results, the microgrid relationships between different causal groups are analyzed. These relationships can be obtained from previous causal relationship analyses, such as load synchronicity and causal relationship strength between microgrids. Based on these relationships, the single-set load trend projection results for each group are adjusted collaboratively. For example, if there is strong load synchronicity between microgrids in two causal groups, when the projection results of one group show an increase in load, the projection results of the other group can also be appropriately adjusted upwards. This cross-microgrid collaborative correction operation is performed on all groups in the single-set load trend projection result set, ultimately generating cross-microgrid collaborative load trend projection results.

[0148] Step S540: Perform control logic matching on the cross-microgrid collaborative load trend projection results, parse the direction and range of load changes from the projection results, match the corresponding load control logic, and generate a preliminary load control logic set.

[0149] In one implementation, step S540 may specifically include the following steps S541 to S546: Step S541: Analyze the trend direction of the cross-microgrid collaborative load trend projection results. Extract the direction of load change from the projection results of each microgrid and generate a set of microgrid load trend directions. Each element in the set contains the microgrid identifier and the trend direction.

[0150] Trend direction analysis extracts load change direction information for each microgrid from the cross-microgrid collaborative load trend projection results. Load change direction can be categorized into three cases: increase, decrease, and no change. The microgrid load trend direction set is a collection that combines the microgrid identifier and the corresponding load change direction for each microgrid. This set provides a clear understanding of the load development trend for each microgrid.

[0151] For the set of cross-microgrid collaborative load trend projection results, iterate through the projection results of each microgrid. Extract load status information for subsequent time scales from the projection results and compare it with the load status at the current time. If the load status at a subsequent time scale is greater than the load status at the current time, the load change direction is considered to be increasing; if the load status at a subsequent time scale is less than the load status at the current time, the load change direction is considered to be decreasing; if the two are equal, the load change direction is considered to be unchanged. Combine the microgrid identifier and the corresponding load change direction into an element. For example, for microgrid V, its current load status is P_current, and the load status at subsequent time scales is P_future. If P_future > P_current, then generate the element (microgrid V, increasing). Summarize the elements of all microgrids to generate a set of microgrid load trend directions.

[0152] Step S542: Perform pre-matching of the control logic on the set of microgrid load trend directions, bind each trend direction to the preset load control logic, and generate the initial control logic binding result.

[0153] The pre-matching of control logic involves associating the load change direction of each microgrid in the microgrid load trend direction set with pre-set load control logic. The pre-set load control logic is formulated based on the virtual power plant's operational requirements and actual conditions. For example, when the load change direction is increasing, possible control logic includes starting standby generators and increasing power generation; when the load change direction is decreasing, possible control logic includes reducing power generation and adjusting load distribution. The initial control logic binding result is a set of results combining the load change direction of each microgrid with the corresponding pre-set control logic. This set allows for a preliminary determination of the control measures that each microgrid may need to take. For each element in the microgrid load trend direction set, the corresponding control logic is searched in the pre-set load control logic library based on the load change direction information. For example, if the element is (microgrid W, increase), the corresponding control logic when the load increases is searched in the preset control logic library, such as "start standby generator A" and "increase the power generation of main generator B by 10%". Microgrid W is then bound to these control logics to form (microgrid W, [start standby generator A, increase the power generation of main generator B by 10%]). This pre-matching and binding operation is performed on all elements in the microgrid load trend direction set, ultimately generating the initial control logic binding result set.

[0154] Step S543: Perform cross-microgrid collaborative verification on the initial control logic binding result to verify whether there are conflicts in the control logic of different microgrids. If there are conflicts, adjust the control logic to generate a set of collaboratively adjusted control logic.

[0155] For the initial control logic binding result set, analyze the control logic of different microgrids. For control logic involving resource sharing, such as the startup of standby generators and the use of power transmission lines, check whether multiple microgrids simultaneously request the same resource. Regarding power balance, calculate the sum of power generation and power consumption changes across different microgrids to determine if it will lead to an imbalance in the overall power of the virtual power plant. If conflicts are found, adjustments are made according to the specific circumstances. For example, if multiple microgrids simultaneously request the startup of the same standby generator, the standby generator can be prioritized for one microgrid based on its importance and load change rate, while other standby generators are selected or other control logics are adjusted for other microgrids. Such adjustments are performed on all conflicting control logics to generate a coordinated control logic set.

[0156] Step S544: Perform adaptability verification on the coordinated adjustment set of control logics, match each control logic with the current load state of the corresponding microgrid, verify the adaptability of the logic, and generate adaptability verification results.

[0157] For each control logic in the coordinated adjustment set, the current load status information of the corresponding microgrid is obtained. The feasibility and effectiveness of the control logic are evaluated based on the current load status. For example, if a control logic requires a microgrid to increase its power generation by 20%, but the microgrid is already operating close to full load, increasing power generation by 20% may not be achievable; in this case, the control logic is considered unsuitable. This suitability assessment is performed on each control logic, marking it as "suitable" or "unsuitable," ultimately generating a set of suitability verification results.

[0158] Step S545: Perform effective control logic screening on the adaptability verification results, select control logics that meet the adaptability requirements, and generate a set of effective control logics.

[0159] For the adaptability verification result set, iterate through each result. If a result is marked as "adaptable," extract the corresponding control logic. For example, for the result (microgrid X, control logic Y, adaptable), add control logic Y to the set of valid control logics. Perform this filtering operation on all results in the adaptability verification result set to finally generate the set of valid control logics.

[0160] Step S546: Perform microgrid association binding on the effective control logic set, binding each effective control logic with the corresponding microgrid identifier to generate a preliminary load control logic set.

[0161] Microgrid association and binding involves associating each control logic in the effective control logic set with its corresponding microgrid identifier, clarifying which microgrid each control logic targets. For the effective control logic set, each effective control logic is traversed. Based on the previous adaptability verification results, the microgrid identifier corresponding to that control logic is found. The microgrid identifier and control logic are bound together to form an element, for example, (microgrid Z, control logic Z1). All bound elements are then aggregated to generate the initial load control logic set.

[0162] Step S550: Perform instruction format conversion on the preliminary load control logic set, converting each control logic into an instruction format that conforms to the load control terminal identification specification, and generating load control instructions corresponding to each microgrid.

[0163] Command format conversion involves converting the control logic in the initial load control logic set into a format that the load control terminal can recognize and execute. The load control terminal is a device installed in each microgrid to perform load control operations; different terminals may have different command format requirements. By converting the control logic into a compliant command format, it ensures that the control logic is accurately conveyed to the load control terminal and correctly executed. The load control commands corresponding to each microgrid are a specific set of control commands for each microgrid after format conversion. For the initial load control logic set, the command format specifications of the load control terminal are understood, including the syntax structure of the commands and parameter settings. For each control logic, conversion is performed according to the command format specifications. For example, if the initial load control logic set contains (Microgrid A1, increase the power generation of main generator unit B by 10%), and the command format of the load control terminal is "[Microgrid Identifier]:[Operation Type]:[Equipment Name]:[Parameter Value]", then the control logic is converted to "[Microgrid A1]:[Increase Power Generation]:[Main Generator Unit B]:[10%]". All control logic in the initial load control logic set undergoes this instruction format conversion operation, ultimately generating the load control instruction set corresponding to each microgrid.

[0164] Step S560: Distribute the load control instructions to the terminal, sending the load control instructions corresponding to each microgrid to the load control terminal of that microgrid, thus completing the instruction distribution operation.

[0165] Terminal distribution involves accurately sending load control commands corresponding to each microgrid to the load control terminal of each microgrid. By sending the commands to the load control terminals, each microgrid can execute corresponding load control operations according to the commands, thereby achieving the load forecasting and control objectives of the virtual power plant. The completion of the command distribution operation marks the successful completion of the final stage of the entire virtual power plant load forecasting and control method for multiple microgrids, and the actual load control phase begins.

[0166] During this step, for each microgrid's load control command set, the corresponding load control command is sent to the appropriate terminal based on the communication address and protocol of each microgrid's load control terminal. Data transmission can be performed using wired communication (such as Ethernet) or wireless communication (such as Wi-Fi, 4G / 5G, etc.). Before sending the command, it needs to be encrypted and verified to ensure its security and integrity during transmission. For example, encryption algorithms (such as AES) can be used to encrypt the command, and cyclic redundancy check (CRC) can be used to verify the command data. During the command transmission process, the transmission status of the command is monitored to ensure that each command is successfully sent to the corresponding load control terminal. If a transmission failure occurs, a retry operation or other fault handling measures need to be taken, such as checking the communication line or reconfiguring the terminal parameters. This terminal distribution operation is performed on all microgrid load control commands, ultimately completing the command issuance operation, enabling each microgrid's load control terminal to begin executing the corresponding load control operations, thereby achieving stable operation of the virtual power plant and rational allocation of power resources.

[0167] Please see Figure 3 This is a schematic diagram of a virtual power plant load forecasting and control device provided in an embodiment of the present invention. The aforementioned virtual power plant load forecasting and control device can be a computer program (including program code) running on a network device; for example, the virtual power plant load forecasting and control device is an application software. This device can be used to execute corresponding steps in the method provided in the embodiments of the present invention. Figure 3 As shown, the virtual power plant load forecasting and control device 300 may include: The data acquisition module 310 is used to continuously collect the load time sequence records of each microgrid within the coverage area of ​​the virtual power plant, and obtain the load time sequence record set of each microgrid. The load time sequence record set contains the load record entries of each microgrid during the continuous acquisition period, and each load record entry corresponds to a unique acquisition time marker. The load calculation module 320 is used to perform load difference calculation on the load time sequence record set of each microgrid for adjacent collection periods, filter out load fluctuation record entries whose load difference exceeds the preset fluctuation range, locate the overlapping periods of load fluctuation between microgrids based on the load fluctuation record entries, and obtain the load synchronous fluctuation period set. The relationship analysis module 330 is used to analyze the order of occurrence and amplitude of load fluctuations in each microgrid for each overlapping period in the set of load synchronous fluctuation periods, determine the causal relationship between the loads of each microgrid, and obtain the causal relationship pairs and relationship records of each microgrid load. The relationship records include the corresponding relationship of load fluctuation amplitude in the associated periods. The item filtering module 340 is used to filter load record items corresponding to causal pairs from the load time series record set of each microgrid based on causal relationship pairs and related relationship records, and obtain a load related record set; The load control module 350 is used to perform load trend extrapolation calculations based on the load association record set, obtain the load extrapolation results of each microgrid, determine the load control mode of the virtual power plant based on the load extrapolation results, generate corresponding load control instructions and send them to the load control terminals of each microgrid.

[0168] This invention also provides a computer system, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the steps in the virtual power plant load forecasting and control method for multi-microgrids provided in this invention. See details below. Figure 4 This is a schematic diagram of the structure of a computer system provided in an embodiment of the present invention. Figure 4 As shown, the computer system 1000 described above may include: a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the computer system 1000 may also include: a user interface 1003, and at least one communication bus 1002. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as at least one disk storage device. Optionally, the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. Figure 4 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application program. Figure 4 In the computer system 1000 shown, the network interface 1004 provides network communication functions; the user interface 1003 is mainly used to provide an input interface; and the processor 1001 can be used to call the device control application stored in the memory 1005 to implement the methods provided in the above embodiments.

Claims

1. A virtual power plant load forecasting and control method for multi-microgrid systems, characterized in that, The method includes: The load time sequence records of each microgrid within the coverage area of ​​the virtual power plant are continuously collected to obtain a load time sequence record set for each microgrid. The load time sequence record set contains load record entries of each microgrid during the continuous collection period, and each load record entry corresponds to a unique collection time marker. The load time sequence records of each microgrid are processed by calculating the load difference between adjacent collection periods. Load fluctuation record entries with load differences exceeding the preset fluctuation range are selected. Based on the load fluctuation record entries, the overlapping periods of load fluctuations between microgrids are located to obtain a set of load synchronous fluctuation periods. For each overlapping period in the set of load synchronous fluctuation periods, the order of occurrence and the corresponding relationship of fluctuation amplitude of each microgrid load are analyzed to determine the causal relationship between each microgrid load, and the causal relationship pairs and relationship records of each microgrid load are obtained. The relationship records include the corresponding relationship of load fluctuation amplitude in the associated periods. Based on the causal relationship pairs and the relationship records, load record entries corresponding to the causal relationship pairs are selected from the load time-series record sets of each microgrid to obtain a load relationship record set; Based on the load association record set, load trend extrapolation calculations are performed to obtain the load extrapolation results for each microgrid. The load control mode of the virtual power plant is determined according to the load extrapolation results, and corresponding load control instructions are generated and sent to the load control terminals of each microgrid.

2. The method as described in claim 1, characterized in that, The method involves performing load difference calculations on the load time-series records of each microgrid for adjacent data collection periods, filtering out load fluctuation record entries whose load differences exceed a preset fluctuation range, and locating overlapping periods of synchronous load fluctuations between microgrids based on these load fluctuation record entries, thereby obtaining a set of synchronous load fluctuation periods, including: The load time series records of each microgrid are aligned with a global time base. The collection time stamps of the load record entries of all microgrids are converted into a unified time scale to eliminate the time deviation caused by the difference in the collection system of different microgrids and generate a standardized load time series with a consistent time base. For each microgrid load record entry in the standardized load time series, the state chain association between adjacent time periods is performed. The load record entry of the previous collection period is used as the preceding association item of the next collection period to establish the load state transmission link between time periods and generate a load state transmission link set. Perform change correlation calculation on each link in the load status transmission link set, bind the load status of the preceding and following time periods in the link in pairs, generate load status change correlation pairs for adjacent collection time periods, and ensure that each pair of correlation items corresponds to a unique time scale interval. The load state change association pairs are pre-matched across microgrids to determine their fluctuation range. Each microgrid load state change association pair is bound to the boundary of a preset fluctuation range one by one to generate a fluctuation range binding result. Invalid association pairs that do not correspond to the preset fluctuation range are excluded. For load state change association pairs that exceed the preset fluctuation range in the fluctuation range binding results, cross-microgrid time clustering is performed, and all microgrid fluctuation association pairs within the same time scale interval are clustered and integrated to generate cross-microgrid fluctuation clusters. The time period boundaries of the cross-microgrid fluctuation clusters are extracted. The earliest time scale corresponding to each cluster is set as the start point of the time period and the latest time scale is set as the end point of the time period. The start point and the end point are bound to a complete time period, and the set of load synchronization fluctuation time periods is obtained.

3. The method as described in claim 2, characterized in that, The step of performing change correlation calculations on each link in the load status transmission link set, binding the load status of consecutive time periods in the link into pairs, and generating load status change correlation pairs for adjacent data collection time periods includes: Each link in the load status transmission link set is split into nodes, and each load record entry in the link is split into an independent status node. Each status node contains the corresponding microgrid identifier, time scale and load status information, generating a link status node set. The nodes in the link status node set are matched for adjacent time periods. According to the order of the time scale, each status node is paired with the status node of the next immediately adjacent time scale to generate adjacent time period status node pairs, so that the micro-network identifier of each pair of nodes is consistent. The adjacent time period state node pairs are encoded with state association. The microgrid identifier, time scale and load status information of each pair of nodes are converted into standard association codes. The encoding content includes the time difference between nodes and the state association mark, and a state association code pair is generated. The change amount is extracted from the state association code pair, the load state information of the previous and next time periods is parsed from the code content, the two state information are associated and bound to generate a load state change association item; The validity of the load state change correlation items is verified. The microgrid identifier and time scale in the correlation items are checked for conflicts. Invalid correlation items with conflicts are excluded, and valid load state change correlation items are generated. The effective load status change correlation items are integrated in pairs, and each effective correlation item is classified according to the corresponding link to generate the load status change correlation pairs for adjacent collection periods.

4. The method as described in claim 3, characterized in that, The process involves extracting time-period boundaries for the cross-microgrid fluctuation clusters, setting the earliest time scale corresponding to each cluster as the start point of the time period and the latest time scale as the end point, binding the start and end points into a complete time period, and organizing this into the set of load synchronization fluctuation time periods, including: The cross-microgrid fluctuation clusters are synchronized across microgrids using time scale. The time scale nodes of all involved microgrids are extracted from the load state change association pairs of each cluster. The time scale nodes are aligned according to a preset unified time base to eliminate the time base deviation caused by the difference in the acquisition system of different microgrids. A time scale synchronization set corresponding to each cluster is generated. Each time scale in the set is bound to the load fluctuation association tag of at least one microgrid. For each cluster, the time scale synchronization set is processed to perform continuous state association. The time scales in the set are connected in sequence. Continuous association marks are added to the load fluctuation records corresponding to adjacent time scales. The marks contain the load state transmission relationship between the preceding and following scales. A continuous association chain of load states within the cluster is generated, where each node corresponds to a unique time scale and load state transmission information. Cross-cluster association comparison processing is performed on the continuous association chains of load status within all clusters. The continuous association chains of load status within clusters with common micronet identifiers are paired and bound. The time scale transmission relationship of the bound association chains is compared, and cluster pairs with mutual influence are marked. A set of cross-cluster association influence markers is generated. Each marker in the set corresponds to two mutually related clusters and specific influence dimension information. Record the association chain information of the cross-cluster association influence label set, and record the specific influence dimension of the label as additional information into the continuous association chain of load status within the corresponding cluster, thereby generating a load status association chain with association influence information. The load status association chain with associated impact information is processed by extracting time period boundaries. The starting time node of the association chain is set as the start point of the time period, and the last actual load status node of the association chain is set as the end point of the time period. The start point and the end point are bound to the complete time period range to generate a set of time period boundaries corresponding to the clusters. Each set corresponds to a unique cluster and the actual load fluctuation range. The set of time period boundaries corresponding to the clusters is standardized and organized. The start and end points of each time period boundary are bound to the microgrid identifier of the corresponding cluster. All time period boundaries are arranged in chronological order of the start time nodes. Finally, the set of load synchronization fluctuation time periods is obtained. Each time period in the set contains the actual load fluctuation range of the corresponding microgrid and the recorded related impact information.

5. The method as described in claim 1, characterized in that, For each overlapping period in the set of load synchronization fluctuation periods, the sequence of occurrence and amplitude of load fluctuations in each microgrid are analyzed to determine the causal relationship between the loads in each microgrid, resulting in causal pairs and relationship records for each microgrid load, including: For each overlapping period in the load synchronization fluctuation period set, microgrid fluctuation data is extracted. All load record entries within the overlapping period are extracted from the load time sequence record set of each microgrid, categorized according to microgrid identifier, and a subset of microgrid load data within the period is generated. The microgrid load data subset within the specified time period is decomposed into timestamps. The collection time stamp of each load record entry is split into a timestamp sequence with the smallest time granularity, generating a set of microgrid load timestamp sequences to ensure that the time granularity of each sequence remains consistent. The microgrid load timestamp sequence set is correlated in time sequence. The timestamp of the first load fluctuation in the timestamp sequence of different microgrids is compared. The microgrids are arranged in the order of the timestamps to generate a microgrid fluctuation time sequence. The fluctuation time sequence of the microgrid is correlated with the corresponding fluctuation state. The load state change process of different microgrids in the overlapping period is compared. The load state change process of each microgrid is paired with the change process of other microgrids to generate microgrid load state corresponding correlation pairs. A causal association determination is performed on the corresponding association pairs of the microgrid load status, and the corresponding association pairs of the state of the preceding microgrid and the following microgrid in the sequence of microgrid fluctuations are causally bound to generate causal association candidate pairs. The candidate causal association pairs are organized and associated records are compiled. The microgrid identifiers, overlapping time periods, and status correspondences in the candidate pairs are bound together to obtain the causal association pairs and association records of each microgrid load.

6. The method as described in claim 5, characterized in that, The step of performing time-series correlation on the microgrid load timestamp sequence set, comparing the timestamp of the first load fluctuation in the timestamp sequences of different microgrids, and arranging the microgrids according to the chronological order of these timestamps to generate a microgrid fluctuation time-series sequence includes: The set of microgrid load timestamp sequences is used to identify the starting point of the fluctuation. The first timestamp of load fluctuation is extracted from the timestamp sequence of each microgrid, and the timestamp is bound to the corresponding microgrid identifier to generate a microgrid fluctuation starting timestamp record. The microgrid fluctuation start timestamp records are sorted globally by time, and the fluctuation start timestamps of all microgrids are arranged in chronological order to generate a global fluctuation start time sorting sequence. Each element in the sequence contains a microgrid identifier and a corresponding timestamp. The global fluctuation start time sorting sequence is associated with adjacent microgrids, and the microgrid identifiers of adjacent positions in the sequence are bound in pairs to generate adjacent microgrid time sequence association pairs; The adjacent microgrid time-series association pairs are associated with time difference. The time difference between the fluctuation start timestamps of the two microgrids in each association pair is calculated to generate a time difference association result. The result includes the microgrid pair identifier and the corresponding time difference. The time difference association results are labeled with time-series attributes. Based on the calculation results of the time difference, the micro-networks in the association pair are labeled with the preceding and following attributes, and attribute-labeled micro-network time-series association pairs are generated. The microgrid time series correlation pairs with attribute tags are sequence integrated, and all correlation pairs are concatenated according to the order of the global fluctuation start time sorting sequence to generate the microgrid fluctuation time series sequence. The step of determining the causal association of the microgrid load state corresponding pairs involves causally binding the state corresponding pairs of the preceding and subsequent microgrids in the time sequence of microgrid fluctuations to generate causal association candidate pairs, including: For the microgrid load status corresponding association pairs, the state change link is extracted. The load status change process of two microgrids in the overlapping time period is extracted from each association pair. The link contains load status nodes under continuous timestamps, and microgrid load status change link pairs are generated. The link nodes of the microgrid load status change link pairs are compared and the timestamp nodes in the two links are bound one-to-one to generate timestamp node association pairs to ensure that each pair of nodes corresponds to the same timestamp. The state change synchronization association is performed on the timestamp node association pairs. The load state change direction of the two microgrids in each pair of nodes is compared to generate the state change synchronization association result, which includes the synchronization mark of the corresponding timestamp. The state change synchronization correlation results are used to extract continuous synchronization intervals. From the correlation results, intervals in which multiple consecutive timestamp nodes show synchronization marks are extracted to generate continuous synchronization interval records. A causal weight association is performed on the continuous synchronization interval records, and the length of the continuous synchronization interval is associated with the corresponding microgrid time sequence attribute. A causal association weight is assigned to the preceding microgrid, and a causal weight association result is generated. The causal weight association results are used to confirm causality. Based on the distribution of weights in the causal weight association results, the causal relationship in the association pair is confirmed, and the causal association candidate pair is generated.

7. The method as described in claim 1, characterized in that, Based on the causal relationship pairs and related records, load record entries corresponding to the causal relationship pairs are selected from the load time-series record sets of each microgrid to obtain a load association record set, including: The causal relationship pairs and relationship records of each microgrid load are parsed to extract the identifiers of two microgrids from each causal relationship pair and the corresponding overlapping time periods are extracted from the relationship records to generate microgrid identifier pairs and corresponding time period sets. A microgrid identifier index is created for the load time series record set of each microgrid, and a retrieval directory indexed by the microgrid identifier is established for the load time series record set of each microgrid, generating a microgrid load record retrieval index; A retrieval request is generated for the microgrid identifier pairs and their corresponding time periods. Each microgrid identifier and its corresponding time period is converted into a retrieval request. The request content includes the microgrid identifier and the start and end time scales of the time period, and a microgrid load retrieval request set is generated. A batch search is performed on the microgrid load retrieval request set. Each retrieval request is matched with the microgrid load record retrieval index. Load record entries within the corresponding time period are extracted from the load time series record set of the corresponding microgrid to generate a set of retrieved load record entries. The retrieved load record entry set is causally associated, and each load record entry is bound to its corresponding microgrid identifier and overlapping time period to generate load record entries with causal tags. The load record entries with causal tags are classified and organized, and then categorized according to microgrid identifiers and overlapping time periods to generate the load association record set.

8. The method as described in claim 7, characterized in that, The process of indexing the load time-series record sets of each microgrid using microgrid identifiers, establishing a retrieval directory indexed by the microgrid identifier for each microgrid's load time-series record set, and generating a microgrid load record retrieval index includes: Microgrid identifiers are extracted from the load time sequence record sets of each microgrid. The corresponding microgrid identifier is parsed from each load record entry, and the microgrid identifier is bound to the load record entry to generate a set of load record entries with identifiers. The set of identified load record entries is deduplicated and associated by identification. All load record entries corresponding to the same microgrid identifier are classified, duplicate load record entries under the same microgrid are eliminated, and a deduplicated subset of microgrid load records is generated. The deduplicated microgrid load record subsets are sorted by time scale, and the load record entries in each subset are arranged in the order of their acquisition time markers to generate a time-sorted microgrid load record sequence. Index entries are generated for the time-sorted microgrid load record sequence. An index entry is created for each microgrid identifier. The entry content includes the microgrid identifier and the storage location of the corresponding load record sequence, and a set of microgrid index entries is generated. An index directory is constructed for the set of microgrid index entries. All microgrid index entries are arranged in lexicographical order according to the microgrid identifier, and an index directory with the microgrid identifier as the retrieval key is established to generate an initial microgrid load record retrieval index. The initial microgrid load record retrieval index is validated to verify whether the load record sequence corresponding to each index entry is complete. Invalid index entries are excluded to obtain the microgrid load record retrieval index. The step of performing causal association tagging on the retrieved load record entry set, binding each load record entry with its corresponding microgrid identifier and overlapping time period, and generating load record entries with causal tags includes: The retrieved load record entry set is parsed to extract the collection time stamp and microgrid identifier from each load record entry, thereby generating a load entry attribute information set. The microgrid identifier pairs and their corresponding time period sets are parsed for time period attributes. The start and end time scales are parsed from each time period, and the time period attributes are bound to the corresponding microgrid identifier pairs to generate a time period-microgrid identifier pair binding set. The attribute information set of the load items is matched with the binding set of the time period-microgrid identifier. The collection time mark of each load item is matched with the start and end time scale of the time period to generate attribute matching results. The attribute matching results are causally linked, and the successfully matched load entries are bound to the corresponding microgrid identifiers and time periods to generate causally linked binding records. The causal association binding records are marked and embedded, and the microgrid identifier and time period information in the binding records are embedded into the corresponding load record entries to generate load record entries with embedded tags; The load record entries with embedded tags are validated to verify whether the embedded tag information of each entry matches the attribute matching result. Entries with incorrect tags are excluded, and the load record entries with causal tags are generated.

9. A virtual power plant load forecasting and control device, characterized in that, include: The data acquisition module is used to continuously collect load time-series records of each microgrid within the coverage area of ​​the virtual power plant, and obtain a load time-series record set for each microgrid. The load time-series record set contains load record entries of each microgrid during the continuous acquisition period, and each load record entry corresponds to a unique acquisition time marker. The load calculation module is used to perform load difference calculation on the load time sequence record set of each microgrid for adjacent collection periods, filter out load fluctuation record entries whose load difference exceeds the preset fluctuation range, locate the overlapping periods of load fluctuation between microgrids based on the load fluctuation record entries, and obtain a set of load synchronous fluctuation periods. The relationship analysis module is used to analyze the order of occurrence and amplitude of load fluctuations in each microgrid for each overlapping period in the set of load synchronous fluctuation periods, determine the causal relationship between the loads of each microgrid, and obtain the causal relationship pairs and relationship records of each microgrid load. The relationship records include the load fluctuation amplitude correspondence of the associated periods. The item filtering module is used to filter load record items corresponding to the causal relationship pairs from the load time series record set of each microgrid based on the causal relationship pairs and the relationship records, so as to obtain the load relationship record set. The load control module is used to perform load trend extrapolation calculations based on the load association record set, obtain the load extrapolation results of each microgrid, determine the load control mode of the virtual power plant based on the load extrapolation results, generate corresponding load control instructions, and send them to the load control terminals of each microgrid.

10. A computer system, characterized in that, include: processor; And a memory, wherein the memory stores computer-readable code that, when executed by the processor, causes the processor to perform the method as described in any one of claims 1 to 8.

Images