Virtual power plant resource scheduling method and system based on multi-source data and knowledge guidance

Abstract

The present application relates to the technical field of virtual power plant resource scheduling, and provides a virtual power plant resource scheduling method and system based on multi-source data and knowledge guidance, which comprises the following steps: encoding and fusing the data of the same flexible resource in the resource state mode set obtained by identifying the resource state and response mode based on the preset state mode rule library after dimension reduction of the region resource operation data set obtained in real time to generate a resource dimension reduction feature set, to obtain a semantic enhanced data set; generating an event collaborative response scheme in combination with the resource state mode set and the resource correlation matrix according to the event type to be responded obtained by using a preset machine learning model to identify the autonomous abnormal events of the region; and verifying and optimizing the target collaborative response scheme based on digital twinning technology to obtain a target collaborative response scheme for executing the region resource response scheduling. The present application can effectively improve the autonomous operation abnormal efficient collaborative response capability and supply-demand self-balancing capability of the resource aggregation region, and provide reliable guarantee for the long-term stable operation of the virtual power plant.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual power plant resource scheduling, in particular to a virtual power plant resource scheduling method and system based on multi-source data and knowledge guidance. BACKGROUND

[0002] A virtual power plant aggregates distributed energy resources, energy storage devices and controllable loads to form a collaborative optimization power dispatch system and provide flexible support for the electricity market. However, the existing virtual power plant resource scheduling method usually relies on a single data source to analyze the characteristics of the resources, which results in incomplete characteristic analysis and makes it difficult to adapt to the dynamic changes of flexible resources in different working conditions. At the same time, the current virtual power plant lacks research on resource coordination and interaction in the self-organizing operation scenario of the resource aggregation area in the dynamic environment, and the existing resource coordination analysis only relies on electrical association and short-term economy, without considering the functional substitution degree and response coordination degree between flexible resources, which cannot guarantee the comprehensiveness and reliability of resource coordination analysis and cannot guarantee the effective coordination and scheduling of multiple resources and the self-balancing of supply and demand in the resource aggregation area in complex abnormal response scenarios. SUMMARY

[0003] The purpose of the present application is to provide a virtual power plant resource scheduling method based on multi-source data and knowledge guidance, which realizes real-time perception of autonomous operation abnormal events in the resource aggregation area through a resource characteristic analysis mechanism based on multi-dimensional data and a semantic fusion mechanism based on knowledge guidance, and realizes efficient coordinated response to abnormal events based on a coordinated response analysis mechanism of the multi-dimensional association relationship of resources, which can effectively improve the autonomous operation abnormal efficient coordinated response capability and the self-balancing capability of the resource aggregation area, and thus provide reliable protection for the long-term effective operation of the virtual power plant.

[0004] In order to achieve the above-mentioned purpose, it is necessary to provide a virtual power plant resource scheduling method and system based on multi-source data and knowledge guidance in view of the above-mentioned technical problems.

[0005] In the first aspect, the embodiments of the present application provide a virtual power plant resource scheduling method based on multi-source data and knowledge guidance, which is applied to a resource aggregation area in a virtual power plant, and the resource aggregation area includes a plurality of flexible resources; the method comprises:

[0006] Real-time acquisition of regional resource operation data set, and dimension reduction processing of the regional resource operation data set to generate corresponding resource dimension reduction feature set; the regional resource operation data set includes multi-dimensional operation data of each flexible resource;

[0007] Resource state and response mode identification according to the regional resource operation data set and a preset state mode rule library to obtain a corresponding resource state mode set; the resource state mode set includes the operation state and response mode of each flexible resource;

[0008] Fusing data of the same flexible resource in the resource state mode set and the resource dimension reduction feature set to obtain a corresponding semantic enhancement data set;

[0009] According to the semantic enhancement data set, performing regional autonomous abnormal event identification based on a preset machine learning model to obtain a corresponding event type to be responded;

[0010] According to the event type to be responded, the resource state mode set and a preset resource association matrix, generating a corresponding event collaborative response scheme; the preset resource association matrix is constructed based on economic association degree, functional substitution degree and response collaboration degree between resources;

[0011] Based on digital twin technology, verifying and optimizing the event collaborative response scheme to obtain a target collaborative response scheme, and according to the target collaborative response scheme, performing regional resource response scheduling.

[0012] Further, the step of performing dimension reduction processing on the regional resource running data set to generate a corresponding resource dimension reduction feature set comprises:

[0013] Based on the flexible resource type, the regional resource running data set is divided into a plurality of resource class running sub-data sets; the resource class running sub-data set includes an energy running sub-data set, an energy storage running sub-data set and a load running sub-data set;

[0014] Obtaining a resource class data matrix corresponding to each resource class running sub-data set, and performing standardization processing on the resource class data matrix to obtain a corresponding standardized resource class data matrix;

[0015] Obtaining a first covariance matrix corresponding to each standardized resource class data matrix, and performing eigenvalue decomposition on the first covariance matrix to obtain a corresponding resource class dimension reduction feature matrix;

[0016] Splicing each resource class dimension reduction feature matrix to obtain a corresponding global resource feature matrix;

[0017] Obtaining a second covariance matrix corresponding to the global resource feature matrix, and performing eigenvalue decomposition on the second covariance matrix to obtain a corresponding global resource dimension reduction feature matrix;

[0018] Performing hierarchical reverse mapping on the global resource dimension reduction feature matrix to obtain principal component features of each flexible resource;

[0019] According to the principal component features of all flexible resources, obtaining the resource dimension reduction feature set.

[0020] Further, the preset state mode rule base includes a resource state evaluation rule set and a response mode evaluation rule set corresponding to different flexible resource types.

[0021] The resource state and response mode recognition according to the regional resource operation data set and the preset state mode rule base includes:

[0022] The multi-dimensional operation data of each flexible resource in the regional resource operation data set is matched with the resource state evaluation rule set of the corresponding flexible resource type in the preset state mode rule base to obtain the corresponding operation state.

[0023] The multi-dimensional operation data of each flexible resource in the regional resource operation data set is matched with the response mode evaluation rule set of the corresponding flexible resource type in the preset state mode rule base to obtain the corresponding response mode.

[0024] Further, the encoding and fusion of the resource state mode set and the data of the same flexible resource in the resource dimension reduction feature set to obtain the corresponding semantic enhancement data set includes:

[0025] The operation state and the response mode of each flexible resource in the resource state mode set are respectively one-hot encoded to obtain the corresponding operation state code and response mode code.

[0026] The dimension reduction features of each flexible resource in the resource dimension reduction feature set are combined with the corresponding operation state code and response mode code to obtain the semantic enhancement data of each flexible resource.

[0027] The semantic enhancement data of all flexible resources is summarized to obtain the semantic enhancement data set.

[0028] Further, the construction of the preset resource association matrix includes:

[0029] According to the unit power adjustment cost of each flexible resource, economic correlation analysis is performed to obtain the economic correlation degree of each resource pair, and according to the economic correlation degree of all resource pairs, a first association matrix is obtained.

[0030] According to the resource type and adjustable power range of each flexible resource, functional substitution analysis is performed to obtain the functional substitution degree of each resource pair, and according to the functional substitution degree of all resource pairs, a second association matrix is obtained.

[0031] According to the historical response curve of each flexible resource, dynamic time warping analysis is performed to obtain the response coordination degree of each resource pair, and according to the response coordination degree of all resource pairs, a third association matrix is obtained.

[0032] The first correlation matrix, the second correlation matrix and the third correlation matrix are weightedly fused, and a resultant fusion matrix is normalized to obtain the preset resource correlation matrix.

[0033] Further, the to-be-responded event type includes a power up coordination response event and a power down coordination response event.

[0034] The step of generating the event coordination response scheme corresponding to the to-be-responded event type, the resource state mode set and the preset resource correlation matrix includes:

[0035] According to the resource state mode set, a current available resource set is obtained.

[0036] According to the to-be-responded event type, available resources in the current available resource set are screened to obtain a response event resource set, and current adjustable capabilities of flexible resources in the response event resource set are obtained.

[0037] Based on the preset resource correlation matrix, the flexible resources in the response event resource set are subjected to cluster analysis to obtain a plurality of resource clusters.

[0038] According to the current adjustable capabilities of all flexible resources in each resource cluster, the economic correlation degree and the response coordination degree among the flexible resources, a cluster capability of each resource cluster is obtained; the cluster capability includes a cluster total adjustment capability, a cluster economic score and a cluster response coordination degree.

[0039] According to the cluster capability of each resource cluster, a power adjustment amount corresponding to the to-be-responded event type is distributed to obtain the event coordination response scheme.

[0040] Further, the step of obtaining a plurality of resource clusters based on the preset resource correlation matrix and the flexible resources in the response event resource set includes:

[0041] According to the total number of flexible resources in the response event resource set and a preset upper limit of the number of clusters, a target number of clusters is obtained.

[0042] According to the preset resource correlation matrix, a corresponding Laplacian matrix is obtained.

[0043] According to the target number of clusters, a feature vector matrix of the Laplacian matrix is obtained.

[0044] The row vectors of the feature vector matrix are subjected to K-means clustering to obtain a plurality of resource clusters.

[0045] Further, the step of obtaining the cluster capability of each resource cluster according to the current adjustable capacity of all flexible resources in the resource cluster, the economic correlation degree between the flexible resources, and the response coordination degree between the flexible resources comprises:

[0046] adding up the current adjustable capacity of all flexible resources in each resource cluster to obtain a corresponding cluster total adjustment capacity;

[0047] obtaining a corresponding cluster economic score according to the economic correlation degree between the flexible resources in each resource cluster and the number of cluster resources;

[0048] obtaining a corresponding cluster response coordination degree according to the response coordination degree between the flexible resources in each resource cluster and a preset response coordination coefficient.

[0049] Further, the step of distributing the power adjustment amount corresponding to the event type to be responded to according to the cluster capability of each resource cluster to obtain the event collaborative response scheme comprises:

[0050] sorting all the resource clusters based on a cluster priority sorting rule to generate a corresponding cluster sequence; the cluster priority sorting rule is to sort in descending order of the cluster economic score, and then to sort the clusters with the same cluster economic score in descending order of the cluster response coordination degree;

[0051] sequentially traversing each resource cluster in the cluster sequence and distributing the power adjustment amount corresponding to the event type to be responded to according to the cluster total adjustment capacity of each resource cluster to obtain a cluster power adjustment distribution amount of each resource cluster;

[0052] distributing the cluster power adjustment distribution amount of each resource cluster according to the proportion of the current adjustable capacity of each flexible resource in the corresponding resource cluster to obtain a power adjustment distribution amount of each flexible resource in the corresponding resource cluster;

[0053] generating the event collaborative response scheme according to the cluster power adjustment distribution amount of each resource cluster and the power adjustment distribution amount of all flexible resources in each resource cluster.

[0054] In a second aspect, an embodiment of the present application provides a virtual power plant resource scheduling system based on multi-source data and knowledge guidance, which is applied to a resource aggregation area in a virtual power plant, the resource aggregation area includes a plurality of flexible resources, and the system comprises:

[0055] a preprocessing module configured to acquire a regional resource operation data set in real time and perform dimension reduction processing on the regional resource operation data set to generate a corresponding resource dimension reduction feature set; the regional resource operation data set includes multi-dimensional operation data of each flexible resource;

[0056] a data analysis module configured to perform resource state and response mode identification according to the regional resource operation data set and a preset state mode rule library, to obtain a corresponding resource state mode set; the resource state mode set includes operation states and response modes of each flexible resource;

[0057] a semantic enhancement module configured to encode and fuse data of the same flexible resource in the resource state mode set and the resource dimension reduction feature set, to obtain a corresponding semantic enhancement data set;

[0058] an anomaly perception module configured to perform regional autonomous anomaly event identification based on a preset machine learning model according to the semantic enhancement data set, to obtain a corresponding to-be-responded event type;

[0059] a collaborative analysis module configured to generate an event collaborative response scheme according to the to-be-responded event type, the resource state mode set and a preset resource correlation matrix; the preset resource correlation matrix is constructed based on economic correlation degrees, functional substitution degrees and response collaboration degrees among resources;

[0060] a response scheduling module configured to verify and optimize the event collaborative response scheme based on digital twinning technology, to obtain a target collaborative response scheme, and to perform regional resource response scheduling according to the target collaborative response scheme.

[0061] The application provides a virtual power plant resource scheduling method and system based on multi-source data and knowledge guidance, which is applied to a resource aggregation area in a virtual power plant and includes a plurality of flexible resources, realizes real-time acquisition of an area resource operation data set including multi-dimensional operation data of each flexible resource, performs dimension reduction processing on the area resource operation data set, generates a corresponding resource dimension reduction feature set, identifies the operation state and response mode of each flexible resource according to the resource operation data set and a preset state mode rule library to obtain a resource state mode set, and after encoding and fusing the data of the same flexible resource in the resource state mode set and the resource dimension reduction feature set to obtain a corresponding semantic enhancement data set, identifies an event type to be responded to according to the semantic enhancement data set based on a preset machine learning model, generates a corresponding event collaborative response scheme according to the event type to be responded to, the resource state mode set and a preset resource correlation matrix constructed based on the economic correlation degree, the functional substitution degree and the response synergy degree between resources, verifies and optimizes the event collaborative response scheme based on digital twinning technology to obtain a target collaborative response scheme, and executes the technical scheme of regional resource response scheduling according to the target collaborative response scheme. Compared with the prior art, the virtual power plant resource scheduling method based on multi-source data and knowledge guidance realizes real-time perception of autonomous operation abnormal events in the resource aggregation area based on the resource characteristic analysis mechanism of multi-dimensional data and the semantic fusion mechanism based on knowledge guidance, and realizes efficient collaborative response to abnormal events in combination with the collaborative response analysis mechanism based on the multi-dimensional correlation between resources, which can effectively improve the autonomous operation efficient collaborative response capability and the supply-demand self-balancing capability of the resource aggregation area, and further provides reliable protection for the long-term effective operation of the virtual power plant. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 is a flowchart of the virtual power plant resource scheduling method based on multi-source data and knowledge guidance in the embodiment of the application;

[0063] Figure 2 is a structural schematic diagram of the virtual power plant resource scheduling system based on multi-source data and knowledge guidance in the embodiment of the application;

[0064] Among them, the reference signs are:

[0065] 1, preprocessing module; 2, data analysis module; 3, semantic enhancement module; 4, abnormal perception module; 5, collaborative analysis module; 6, response scheduling module. DETAILED DESCRIPTION

[0066] In order to make the purposes, technical solutions and beneficial effects of the present application clearer, further detailed description will be given to the present application in combination with the drawings and examples. Obviously, the following described examples are a part of the embodiments of the present application, and are only used to illustrate the present application, but not to limit the scope of the present application. Based on the examples in the present application, all other examples obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0067] In one embodiment, as shown in Figure 1 A multi-source data and knowledge guided virtual power plant resource scheduling method is provided, which is applied to a resource aggregation area in a virtual power plant. The resource aggregation area can be understood as an aggregation of multiple flexible resources in close geographical locations, which is divided according to the scheduling and management requirements of the virtual power plant. The virtual power plant can include multiple resource aggregation areas, and each resource aggregation area includes several flexible resources. The flexible resources include distributed new energy, distributed energy storage and controllable load. The controllable load includes interruptible load (industrial air conditioning system, industrial enterprise non-continuous production line, landscape lighting, advertising lamp box, etc.), translatable load (electric vehicle charging pile, non-emergency process in industrial production, etc.), and adjustable load (intelligent temperature control system, flexible production line, etc.).

[0068] In actual application, each resource aggregation area can receive resource scheduling instructions from the virtual power plant management layer, and meet the application requirements of the entire virtual power plant through collaborative scheduling between resource aggregation areas. At the same time, each resource aggregation area can also select not to accept the resource scheduling instructions from the external virtual power plant management layer, but to select local autonomous operation to realize internal supply and demand complementation. However, the content resource scheduling in the local autonomous operation process of the existing resource aggregation area is usually directly allocated to the adjustment target according to the adjustable capacity of each flexible resource in the area, ignoring the running state and response mode of each flexible resource, as well as the multi-dimensional correlation between flexible resources, resulting in poor resource collaborative effect in complex abnormal response scenarios. The multi-source data and knowledge guided virtual power plant resource scheduling method provided by the present application can be applied to efficient and reliable collaborative scheduling of multiple resources in the local autonomous operation abnormal event response scenario of each resource aggregation area in the virtual power plant. The specific method includes:

[0069] S11, real-time acquisition of regional resource operation data set, and dimension reduction processing is carried out on the regional resource operation data set, and corresponding resource dimension reduction feature set is generated;Wherein, the regional resource operation data set can be understood as the operation state data of all flexible resources in the resource aggregation area to be analyzed, which is obtained according to the actual application demand, based on SCADA (Supervisory Control And Data Acquisition) system and other related virtual power plant operation information monitoring system, and according to the preset data sampling frequency;In order to ensure the comprehensiveness and reliability of the analysis of the characteristics of flexible resources, the multi-source data related to the running state and response mode analysis of each flexible resource is preferably collected, that is, the regional resource operation data set includes multi-dimensional operation data of each flexible resource.It should be noted that the data items in the multi-dimensional operation data of each flexible resource type are not the same, the data items in the multi-dimensional operation data of the same flexible resource type are the same, and the data items in the multi-dimensional operation data of each flexible resource type can be selected according to the actual application demand, such as the multi-dimensional operation data of distributed energy (photovoltaic / wind power), which can include cumulative power, output power, fluctuation characteristics (fluctuation amplitude), availability, inverter temperature and component DC voltage, etc.;The multi-dimensional operation data of distributed energy storage can include charging and discharging efficiency, charging and discharging power, state of charge, health status and response rate (climbing rate), etc.;The multi-dimensional operation data of controllable load can include real-time power, adjustable power, adjustable rate, historical response rate and response time average, etc., which are not limited here.

[0070] In order to improve the efficiency and accuracy of subsequent resource operation state and response mode analysis and collaborative optimization analysis based on regional resource operation data set, principal component analysis is preferably used to reduce the dimension of multi-dimensional operation data of each flexible resource;Specifically, the step of dimension reduction processing of the regional resource operation data set to generate corresponding resource dimension reduction feature set includes:

[0071] Based on the type of flexible resource, the regional resource operation data set is divided into a plurality of resource class operation sub data sets;The resource class operation sub data set includes energy operation sub data set, energy storage operation sub data set and load operation sub data set;It should be noted that data cleaning (including outlier deletion and missing value completion) is required before classification of regional resource operation data set, in order to avoid invalid data distortion of principal component direction, and to provide guarantee for the reliability of subsequent principal component analysis result.

[0072] Get the resource class data matrix corresponding to each resource class operation sub data set, and standardize the resource class data matrix to get the corresponding standardized resource class data matrix;Wherein, the standardization processing can use existing standardization processing method to eliminate the dimension difference of different data types, which is not described in detail here.

[0073] obtain a first covariance matrix corresponding to each of the standardized resource category data matrices, and perform eigenvalue decomposition on the first covariance matrix to obtain a corresponding resource category dimension-reduced feature matrix; wherein the resource category dimension-reduced feature matrix can be understood as follows: after performing eigenvalue decomposition on each of the first covariance matrices, a principal component dimension is obtained based on the Kaiser criterion and a cumulative variance contribution rate range (such as greater than or equal to 80%), a corresponding projection matrix is obtained based on the principal component dimension, and the required resource category dimension-reduced feature matrix is obtained based on the product of the standardized resource category data matrix and the projection matrix; it should be noted that the specific calculation process of the resource category dimension-reduced feature matrix can be obtained by referring to the existing principal component analysis method, and the cumulative variance contribution rate range used to obtain the principal component dimension of each type of flexible resource can be the same or different, which will not be described in detail here.

[0074] splicing each of the resource category dimension-reduced feature matrices to obtain a corresponding global resource feature matrix; wherein the global resource feature matrix can be understood as a matrix obtained by splicing each of the resource category dimension-reduced feature matrices in a predetermined arrangement order, serving as basic data for subsequent secondary dimension reduction.

[0075] obtain a second covariance matrix corresponding to the global resource feature matrix, and perform eigenvalue decomposition on the second covariance matrix to obtain a corresponding global resource dimension-reduced feature matrix; wherein the obtaining process of the global resource dimension-reduced feature matrix can refer to the related description of the resource category dimension-reduced feature matrix described above, which will not be described here.

[0076] perform hierarchical reverse mapping on the global resource dimension-reduced feature matrix to obtain principal component features of each flexible resource; wherein the hierarchical reverse mapping can be understood as follows: first, the global resource dimension-reduced feature matrix is subjected to principal component inverse transformation to obtain a plurality of type feature estimation matrices, and then each of the type feature estimation matrices is subjected to principal component inverse transformation to obtain the principal component features of each flexible resource in each type of flexible resource. The principal component inverse transformation process can be understood as follows: according to the dimension-reduced feature matrix corresponding eigenvector matrix and mean vector , based on the formula project the dimension-reduced feature matrix back to the original feature space to obtain the corresponding feature estimation matrix the process of the inverse transformation of principal components; in practical applications, after the global resource dimensionality reduction feature matrix is subjected to the inverse transformation of principal components to obtain the corresponding feature estimation matrix, the feature estimation matrix is split into various feature estimation matrices based on the positions of various resources in the feature estimation matrix, the inverse transformation of principal components is performed on each feature estimation matrix to obtain the corresponding resource class feature estimation matrix, and then each resource class feature estimation matrix is split into the principal component features of each flexible resource based on the positions of various flexible resource features in the resource class feature estimation matrix, that is, the process of restoring the individual features of each flexible resource with high precision is realized; it should be noted that the specific implementation of the hierarchical reverse mapping can refer to the related prior art, and the principal component features of each flexible resource obtained through the above dimensionality reduction process vary with actual analysis data.

[0077] According to the principal component features of all flexible resources, the resource dimensionality reduction feature set is obtained.

[0078] The hierarchical dimensionality reduction mechanism of the embodiment based on principal component analysis first reduces the dimensionality of the data of the same type of flexible resource, then jointly performs secondary dimensionality reduction on the resource class dimensionality reduction feature matrices, and then performs hierarchical inverse transformation on the obtained global resource dimensionality reduction feature matrix to obtain the principal component features of each flexible resource, which can reduce the dimensionality of data analysis and data redundancy while revealing the hidden association between the same type of resource and the different types of resources, effectively capturing the common features of the same type of flexible resource and the collaborative features between different types of flexible resources, and providing a reliable data basis for subsequent resource state pattern acquisition and resource collaborative analysis.

[0079] S12, resource state and response mode identification is performed according to the regional resource operation data set and the preset state mode rule library to obtain a corresponding resource state mode set; the resource state mode set includes the operation state and the response mode of each flexible resource, and the operation state and the response mode can be determined based on actual analysis results, for example, the operation state of the distributed energy resource can be one or more of resource overload (downward adjustment), resource underload (upward adjustment), low efficiency, communication interruption, adjustment capacity saturation, and response delay, and the corresponding response mode can be one mode of normal response, response delay, response deviation, response failure, and active adjustment limitation.

[0080] The preset state mode rule library in the embodiment can be understood as a database composed of matching rules for reliably identifying different running states and different response modes, which are formulated in advance based on analysis of running state data of various flexible resources in different running states and different response modes. That is, the preset state mode rule library includes resource state evaluation rule sets and response mode evaluation rule sets corresponding to different flexible resource types, each resource state evaluation rule set includes running data matching conditions corresponding to different running states, and each response mode evaluation rule set includes running data matching conditions corresponding to different response modes. The specific construction method can be realized by referring to the construction method of an existing rule library, and will not be described in detail here.

[0081] Specifically, the step of identifying resource states and response modes according to the regional resource running data set and the preset state mode rule library to obtain a corresponding resource state mode set includes:

[0082] Iterative matching of the multi-dimensional running data of each flexible resource in the regional resource running data set with the resource state evaluation rule set of the corresponding flexible resource type in the preset state mode rule library is performed to obtain a corresponding running state.

[0083] Iterative matching of the multi-dimensional running data of each flexible resource in the regional resource running data set with the response mode evaluation rule set of the corresponding flexible resource type in the preset state mode rule library is performed to obtain a corresponding response mode.

[0084] It should be noted that the process of matching resource state evaluation rules and response mode evaluation rules based on the multi-dimensional running data of each flexible resource to obtain running states and response modes can be realized based on the existing Rete algorithm of the rule engine, and will not be described in detail here. The design of the embodiment based on the combination of data-driven decision-making and rule-based knowledge engine can dynamically and real-timely perceive changes in resource states and response modes, and provide reliable guarantee for the efficiency and effectiveness of subsequent regional autonomous abnormal event identification.

[0085] S13, encode and fuse the data of the same flexible resource in the resource state mode set and the resource dimension reduction feature set to obtain a corresponding semantic enhancement data set. The semantic enhancement data set can be understood as a set of enhanced features obtained by combining the principal component features of each flexible resource with the corresponding running state and response mode. Specifically, the step of encoding and fusing the data of the same flexible resource in the resource state mode set and the resource dimension reduction feature set to obtain a corresponding semantic enhancement data set includes:

[0086] The running state and response mode of each flexible resource in the resource state mode set are respectively one-hot encoded to obtain corresponding running state codes and response mode codes; the running state codes and response mode codes can be obtained by referring to existing one-hot encoding technology, and details are not described herein.

[0087] The dimension-reduced features of each flexible resource in the resource dimension-reduced feature set are combined with the corresponding running state codes and response mode codes to obtain semantic enhancement data of each flexible resource; the semantic enhancement data can be understood as data obtained by sequentially concatenating the running state codes and response mode codes of each flexible resource behind the corresponding dimension-reduced features.

[0088] The semantic enhancement data of all flexible resources are summarized to obtain the semantic enhancement data set.

[0089] The embodiment realizes the effect of attaching business labels to physical features by encoding and fusing the low-dimensional numerical features (representing physical states) and semantic running state / response mode codes (representing business logic) of each flexible resource, can greatly improve the information density and interpretability of data, and can significantly improve the efficiency, accuracy, robustness and interpretability of the preset machine learning model in identifying regional autonomous abnormal events.

[0090] S14, based on the preset machine learning model, regional autonomous abnormal event identification is performed according to the semantic enhancement data set to obtain corresponding response event types; the preset machine learning model can be understood as a dataset obtained by processing a large amount of historical regional resource running data set of a resource aggregation area based on the above-mentioned semantic enhancement data set acquisition method and performing regional autonomous abnormal event annotation, and a network model trained and constructed to predict regional autonomous abnormal events based on the semantic enhancement data of all flexible resources in the aggregation area, such as a classification model (based on gradient boosting decision tree ensemble learning algorithm, support vector machine, convolutional neural network, etc.), an anomaly detection model (long short-term memory autoencoder), etc. trained and constructed based on machine learning technology, and the specific network structure and model training method can be selected based on actual application requirements, which is not limited herein. It should be noted that the abnormal event types that can be identified by the preset machine learning model in actual application can be determined by the abnormal event types involved in the historical regional resource running data set used for training, and the corresponding response event types obtained are different due to the abnormal event types supported by the preset machine learning model. In order to ensure the continuous effectiveness of the application of the preset machine learning model, online updating training can be performed based on the real-time collected regional resource running data set.

[0091] S15, generating a corresponding event collaborative response scheme according to the type of the event to be responded, the set of resource state modes and a preset resource correlation matrix; wherein the preset resource correlation matrix can be understood as a matrix for quantifying the collaborative relationship between flexible resources in the resource aggregation area. In order to ensure the comprehensiveness and reliability of the quantification of the collaborative relationship between flexible resources, and to provide a reliable analysis basis for event collaborative response analysis, the embodiment preferably constructs the preset resource correlation matrix based on the economic correlation degree, the functional substitution degree and the response collaboration degree between resources. Specifically, the construction steps of the preset resource correlation matrix include:

[0092] According to the economic correlation analysis of the unit power regulation cost of each flexible resource, the economic correlation degree of each group of resource pairs is obtained, and according to the economic correlation degree of all resource pairs, a first correlation matrix is obtained; wherein the unit power regulation cost of each flexible resource can be obtained by using existing calculation methods, such as the unit power regulation cost of distributed new energy, which can include the opportunity cost of abandoned electricity or the cost of increased electricity, the unit power regulation cost of energy storage equipment, which can include the cost of single charging and discharging and the cost of loss, and the unit power regulation cost of controllable load, which includes the compensation cost of translatable, transferable and reducible load, etc. Here, it is not described in detail; the economic correlation degree corresponding to each group of resource pairs can be calculated based on the following economic correlation degree index, and the economic correlation degree index can be understood as a mathematical expression constructed based on the principle that the smaller the cost difference, the higher the economic correlation degree, and selected based on the nonlinear similarity decay characteristic in economics, expressed as:

[0093]

[0094] In the formula, Eij represents the economic correlation degree of flexible resource i and flexible resource j in the resource aggregation area; and respectively represent the unit power regulation cost of flexible resource i and flexible resource j; and respectively represent the maximum unit power regulation cost and the minimum unit power regulation cost of all flexible resources in the resource aggregation area; represents a very small number to avoid a denominator of 0; represents a decay coefficient for controlling the sensitivity of the cost difference, which can be set according to the actual application scenario.

[0095] The first correlation matrix in the embodiment can be understood as a matrix composed of the economic correlation degrees between all flexible resources in the aggregation resource area, which can provide a reference basis for the economic correlation level of subsequent resource collaborative scheduling.

[0096] According to the resource type and the adjustable power range of each flexible resource, a functional substitution analysis is performed to obtain a functional substitution degree of each resource pair, and according to the functional substitution degrees of all resource pairs, a second correlation matrix is obtained; wherein the functional substitution degree can be understood as an index for quantifying the overlapping degree of the adjustable power range of two flexible resources, so as to reflect the functional complementarity of the two flexible resources in the power adjustment capability, and is expressed as:

[0097]

[0098] In the formula, denotes the functional substitution degree of the flexible resource i and the flexible resource j in the resource aggregation area; and denote the minimum adjustable power and the maximum adjustable power of the flexible resource i respectively; and denote the minimum adjustable power and the maximum adjustable power of the flexible resource j respectively; denotes the length of the overlapping interval of the adjustable power range of the flexible resource i and the adjustable power range of the flexible resource j.

[0099] The second correlation matrix in the embodiment can be understood as a matrix composed of the functional substitution degrees between all flexible resources in the aggregation resource area, which can provide a reference basis for the functional substitution related level for subsequent resource cooperative scheduling.

[0100] According to the historical response curve of each flexible resource, a dynamic time warping analysis is performed to obtain a response cooperation degree of each resource pair, and according to the response cooperation degrees of all resource pairs, a third correlation matrix is obtained; wherein the response cooperation degree can be understood as an index for quantifying the similarity between the response modes of the flexible resources, so as to reflect the cooperation of the dynamic response behavior of the two flexible resources under the fluctuation constraint, and is expressed as:

[0101]

[0102] In the formula, denotes the response cooperation degree of the flexible resource i and the flexible resource j in the resource aggregation area; and denote the historical power response sequence of the flexible resource i and the flexible resource j respectively; T denotes the number of data points of the historical power response sequence; and denote the standard deviation corresponding to the historical power response sequence of the flexible resource i and the flexible resource j respectively; denotes the dynamic time warping distance of the historical power response sequence of the flexible resource i and the flexible resource j, which is used to measure the shape similarity of the historical power sequence of the flexible resources (allowing nonlinear alignment of the time axis), and the smaller the value is, the more similar the resource response trajectories are.

[0103] The third correlation matrix in the embodiment can be understood as a matrix composed of the response synergy degrees between all flexible resources in the aggregated resource area, and can provide a reference basis for the response synergy related level for subsequent resource synergy scheduling.

[0104] The first correlation matrix, the second correlation matrix and the third correlation matrix are weighted and fused, and the obtained fusion matrix is normalized to obtain the preset resource correlation matrix; wherein the fusion weight coefficients of the first correlation matrix, the second correlation matrix and the third correlation matrix can be adjusted based on the actual application scene, which is not limited here.

[0105] The embodiment performs multi-dimensional correlation analysis on the flexible resources from three levels of economic correlation degree, functional substitution degree and response synergy degree, which can effectively improve the comprehensiveness and accuracy of resource correlation analysis in the resource aggregation area, and provide reliable decision basis for subsequent resource synergy analysis, and provide reliable guarantee for improving the flexibility and robustness of autonomous operation in the resource aggregation area.

[0106] Considering that abnormal situations that may occur in actual resource aggregation area autonomous operation include regional supply-demand balance abnormality (regional power surplus, regional power shortage, power fluctuation exceeding limit, etc.), regional voltage abnormality (voltage upper limit, voltage lower limit, voltage fluctuation too large, etc.), regional reactive power abnormality (reactive power shortage, reactive power surplus, reactive power sending, etc.) and regional safety abnormality (line / transformer overload, communication interruption, etc.), and most of the abnormal situations can be attributed to the power synergy scheduling scene solution, the embodiment preferably sets the to-be-responded event types to include power up-regulation synergy response event and power down-regulation synergy response event; specifically, the step of generating a corresponding event synergy response scheme according to the to-be-responded event type, the resource state mode set and the preset resource correlation matrix includes:

[0107] According to the resource state mode set, a current available resource set is obtained; wherein the current available resource set can be understood as a set of normally operating flexible resources that can participate in synergy response, which is obtained by screening based on the operating states of the flexible resources in the resource state mode set, and the specific screening process is not described here.

[0108] According to the type of the event to be responded, available resources in the set of currently available resources are filtered to obtain a set of response event resources, and current adjustable capacities of each flexible resource in the set of response event resources are obtained; in actual application, when the type of the event to be responded is a power up-regulation cooperative response event, based on the operating states of each available resource in the set of currently available resources, resources with up-regulation capacity are filtered from the set of currently available resources to form the set of response event resources, and when the type of the event to be responded is a power down-regulation cooperative response event, resources with down-regulation capacity are filtered from the set of currently available resources to form the set of response event resources; the current adjustable capacities of each flexible resource in the set of response event resources can be calculated by using an existing evaluation method of adjustable capacities of various flexible resources in a virtual power plant, and details are not described herein.

[0109] Based on the preset resource association matrix, clustering analysis is performed on the flexible resources in the set of response event resources to obtain a plurality of resource clusters; wherein, the clustering analysis can be understood as clustering analysis on the similarity between resource pairs by taking the preset resource association matrix as a similarity matrix, so as to ensure that the cooperative capacity of different resources in the resource cooperation process can be fully utilized, and thus the efficiency of the response cooperation is ensured. Specifically, the step of performing clustering analysis on the flexible resources in the set of response event resources based on the preset resource association matrix to obtain a plurality of resource clusters includes:

[0110] According to the total number of flexible resources in the set of response event resources and a preset upper limit of clustering number, a target clustering number is obtained; wherein, the preset upper limit of clustering number can be set based on actual demand, and details are not limited herein; considering that in actual application, the clustering number is closely related to the total number of flexible resources, and when the total number of flexible resources increases, the clustering number slowly increases (sub-linearly), in order to avoid excessive calculation overhead caused by using existing iterative evaluation algorithms such as elbow rule and silhouette coefficient, improve the efficiency of setting the target clustering number, and effectively prevent the resources from being excessively dispersed in the clustering analysis, and ensure that each resource class is allocated with appropriate amount of resources, the embodiment preferably automatically calculates a reasonable target clustering number based on the total number of flexible resources and the preset upper limit of clustering number by using the following calculation formula:

[0111]

[0112] In the formula, n represents the target clustering number; N represents the preset upper limit of clustering number; and M represents the total number of flexible resources.

[0113] ​​​According to the preset resource association matrix, a corresponding Laplacian matrix is obtained; wherein, the Laplacian matrix can be calculated based on the preset resource association matrix and its corresponding degree matrix (a matrix composed of diagonal elements in the preset resource association matrix), and the specific calculation process can refer to the prior art implementation, which will not be described in detail here.

[0114] According to the target cluster number, a feature vector matrix of the Laplacian matrix is obtained; wherein, the feature vector matrix is obtained by first solving the eigenvalues of the Laplacian matrix, then solving the corresponding eigenvectors according to each eigenvalue, and then combining the normalized eigenvectors to obtain the matrix.

[0115] The row vectors of the feature vector matrix are subjected to K-means clustering to obtain a plurality of resource clusters.

[0116] The embodiment can realize more balanced classification based on the difference in resource association degree, and can also ensure that the resources in the same resource cluster are highly complementary in economy, function and response, thereby facilitating the rapid allocation of resources in the resource cluster.

[0117] According to the current adjustable capacity of all flexible resources in each resource cluster, the economic association degree and the response synergy degree between the flexible resources, the cluster capacity of each resource cluster is obtained; the cluster capacity includes cluster total adjustment capacity, cluster economy score and cluster response coordination degree; specifically, the step of obtaining the cluster capacity of each resource cluster according to the current adjustable capacity of all flexible resources in each resource cluster, the economic association degree and the response synergy degree between the flexible resources includes:

[0118] The current adjustable capacity of all flexible resources in each resource cluster is accumulated to obtain the corresponding cluster total adjustment capacity; that is, the cluster total adjustment capacity is the sum of the current adjustable capacity of all flexible resources in the resource cluster.

[0119] According to the economic association degree between the flexible resources in each resource cluster and the number of cluster resources, the corresponding cluster economy score is obtained; wherein, the cluster economy score can be understood as an index for measuring the average level of the economic association degree between the resources in a resource cluster, so as to evaluate the economic efficiency of the entire resource cluster, the higher the score, the better the economy, the higher the resource cooperation efficiency in the cluster, and the lower the synergy cost, which can be expressed as:

[0120]

[0121] In the formula, represents the kth resource cluster; represents the number of resources in the kth resource cluster; an economic correlation degree of flexible resource l and flexible resource m in the kth resource cluster, obtained from the economic correlation degrees of each pair of resources described above; a cluster economic score of the kth resource cluster.

[0122] According to the response synergy degrees between the flexible resources in each of the resource clusters and a preset response synergy coefficient, a corresponding cluster response coordination degree is obtained; the cluster response coordination degree can be understood as an index for measuring the overall response capability of the resource cluster; considering that the actual synergy performance of the resource cluster is limited by the weakest synergy relationship, the embodiment preferably selects the weakest synergy relationship between the resources in the cluster to quantify the cluster response performance, so as to ensure the overall synergy response efficiency, and can be expressed as:

[0123]

[0124] In the formula, the kth resource cluster; the preset response synergy coefficient, which can be set according to the actual application scenario; a response synergy degree of flexible resource l and flexible resource m in the kth resource cluster, obtained from the response synergy degrees of each pair of resources described above; a cluster response coordination degree of the kth resource cluster.

[0125] The embodiment comprehensively evaluates the synergy response capability of the resource cluster based on the total regulation capability of the cluster, the cluster economic score and the cluster response coordination degree, thereby providing effective support for formulating an efficient and reliable event synergy response scheme based on the synergy response capability of each resource cluster.

[0126] According to the cluster capabilities of each of the resource clusters, the power regulation amount corresponding to the event type to be responded to is allocated, and an event synergy response scheme is obtained; the event synergy response scheme can be understood as a resource cluster coordination scheduling strategy that can guarantee efficient and reliable completion of event synergy response, which is obtained by reasonably allocating cluster scheduling priorities based on the cluster capabilities of each of the resource clusters; specifically, the step of allocating the power regulation amount corresponding to the event type to be responded to according to the cluster capabilities of each of the resource clusters to obtain the event synergy response scheme includes:

[0127] The cluster priority sorting rule is based on descending order of cluster economy score, and for the clusters with the same cluster economy score, descending order of cluster response coordination degree; that is, the cluster sequence is sorted according to the cluster economy score from high to low, and for the resource clusters with the same cluster economy score, the cluster response coordination degree is sorted from high to low, so as to improve the coordination efficiency as much as possible under the condition of ensuring the total coordination ability, and ensure the timeliness of the abnormal event response.

[0128] The cluster power adjustment allocation amount of each resource cluster is obtained by sequentially traversing each resource cluster in the cluster sequence and allocating the power adjustment amount corresponding to the event type to be responded according to the cluster total adjustment ability of each resource cluster; wherein the power adjustment amount corresponding to the event type to be responded can be determined according to the supply and demand imbalance of the actual power up or power down cooperative response event; in actual application, the power adjustment amount corresponding to the event type to be responded is sequentially disassembled according to the sorting and corresponding cluster total adjustment ability of the resource cluster in the cluster sequence, until the power adjustment amount is completely disassembled, that is, the cluster power adjustment allocation amount of the resource cluster is obtained; it should be noted that there may be a case that part of the resource clusters sorted later are not allocated to the cluster power adjustment allocation amount, then they do not need to participate in the cooperative response.

[0129] The cluster power adjustment allocation amount of each resource cluster is obtained by sequentially traversing each resource cluster in the cluster sequence and allocating the power adjustment amount corresponding to the event type to be responded according to the cluster total adjustment ability of each resource cluster; wherein the power adjustment amount corresponding to the event type to be responded can be determined according to the supply and demand imbalance of the actual power up or power down cooperative response event; in actual application, the power adjustment amount corresponding to the event type to be responded is sequentially disassembled according to the sorting and corresponding cluster total adjustment ability of the resource cluster in the cluster sequence, until the power adjustment amount is completely disassembled, that is, the cluster power adjustment allocation amount of the resource cluster is obtained; it should be noted that there may be a case that part of the resource clusters sorted later are not allocated to the cluster power adjustment allocation amount, then they do not need to participate in the cooperative response.

[0130] The event cooperative response scheme is generated according to the cluster power adjustment allocation amount of each resource cluster and the power adjustment allocation amount of all flexible resources in each resource cluster.

[0131] The embodiment reasonably determines the cluster scheduling priority based on the cluster ability of each resource cluster, allocates the cluster power adjustment allocation amount based on the cluster scheduling priority, and allocates the resources in the cluster based on the current adjustable capacity of the flexible resource, which can effectively avoid local optimization, ensure the generation of a globally optimal cooperative scheme, effectively meet the cooperative response demand, ensure the efficiency of the cooperative response, and maintain the efficient and stable autonomous operation of the resource aggregation area.

[0132] S16, verifying and optimizing the event collaborative response scheme based on the digital twin technology to obtain a target collaborative response scheme, and performing regional resource response scheduling according to the target collaborative response scheme.

[0133] The process of verifying and optimizing the event collaborative response scheme based on the digital twin technology can include:

[0134] Based on the pre-constructed resource aggregation area running twin model, the event collaborative response scheme is simulated to obtain simulation data. The resource aggregation area running twin model can be obtained by using the existing digital twin model construction method of a virtual power plant, which will not be described in detail here. In actual application, the regional resource running data set of the identified event type to be responded to is injected into the resource aggregation area running twin model to simulate the corresponding regional autonomous abnormal event. Then, the resource collaborative scheduling of the resource aggregation area is performed according to the event collaborative response scheme, and the data such as regional power generation, regional load power, actual power regulation value and regulation time length of the collaborative response process are collected for verifying and evaluating the running effect of the event collaborative response scheme. It should be noted that the specific simulation data obtained can be determined according to actual evaluation requirements, which is not limited here.

[0135] The collaborative response effect is evaluated according to the simulation data to obtain a corresponding evaluation result. The evaluation result can be understood as an evaluation result obtained by analyzing the simulation data based on pre-set evaluation indicators such as power balance indicators (calculated based on the difference between regional power generation and regional load power), regulation target deviation rate (calculated based on the difference between the actual power regulation value and the power adjustment amount corresponding to the event type to be responded to), and regulation time length deviation rate (calculated based on the difference between the regulation time length and the target regulation time length) and corresponding indicator thresholds.

[0136] According to the evaluation result, the event collaborative response scheme is adjusted and optimized to obtain a target collaborative response scheme. In actual application, based on the compliance of each evaluation indicator in the evaluation result, the Apriori algorithm can be used to mine the association rules between each non-compliant indicator and economic correlation degree, functional substitution degree and response collaboration degree, and based on these rules, the optimization of the weighting coefficients used when constructing the pre-set resource association matrix based on economic correlation degree, functional substitution degree and response collaboration degree can be guided. It should be noted that the implementation process of the association reasoning analysis based on the Apriori algorithm can refer to the related prior art implementation, and the adjustment method of the weighting coefficients of the economic correlation degree, the functional substitution degree and the response collaboration degree can be determined based on the actual obtained association rules, which will not be described in detail here.

[0137] The embodiment can ensure that a safe, feasible and efficient target collaborative response scheme is obtained, and the actual trial and error risk and cost are greatly reduced.

[0138] The application is applied to a resource aggregation area including a plurality of flexible resources in a virtual power plant, and a region resource running data set including multi-dimensional running data of each flexible resource is obtained in real time, dimension reduction processing is performed on the region resource running data set, a corresponding resource dimension reduction feature set is generated, resource state and response mode recognition is performed according to the resource running data set and a preset state mode rule library to obtain a resource state mode set including the running state and the response mode of each flexible resource, and after the resource state mode set and the data of the same flexible resource in the resource dimension reduction feature set are encoded and fused to obtain a corresponding semantic enhancement data set, a preset machine learning model is used to perform region autonomous abnormal event recognition according to the semantic enhancement data set to obtain an event type to be responded, a corresponding event collaborative response scheme is generated according to the event type to be responded, the resource state mode set and a preset resource correlation matrix constructed based on the economic correlation degree, the functional substitution degree and the response collaboration degree among resources, and a target collaborative response scheme is obtained by verifying and optimizing the event collaborative response scheme based on digital twinning technology, and the technical scheme of region resource response scheduling is executed according to the target collaborative response scheme, the real-time perception of autonomous running abnormal events in the resource aggregation area is realized based on the resource characteristic analysis mechanism of multi-dimensional data and the semantic fusion mechanism guided by knowledge, and the efficient collaborative response of abnormal events is realized in combination with the collaborative response analysis mechanism based on the multi-dimensional correlation among resources, so that the autonomous running efficient collaborative response capability and the supply-demand self-balancing capability of the resource aggregation area are effectively improved, and reliable protection is provided for the long-term effective operation of the virtual power plant.

[0139] It should be noted that although each step in the above flowchart is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps has no strict order limitation, and these steps can be executed in other orders.

[0140] In one embodiment, as shown in Figure 2 A virtual power plant resource scheduling system based on multi-source data and knowledge guidance is provided, which is applied to a resource aggregation area in a virtual power plant, the resource aggregation area includes a plurality of flexible resources, and the system includes:

[0141] The preprocessing module 1 is configured to obtain a region resource running data set in real time, and perform dimension reduction processing on the region resource running data set to generate a corresponding resource dimension reduction feature set; the region resource running data set includes multi-dimensional running data of each flexible resource.

[0142] a data analysis module 2 configured to perform resource state and response mode identification according to the regional resource operation data set and a preset state mode rule library, to obtain a corresponding resource state mode set; the resource state mode set comprises operation states and response modes of each flexible resource;

[0143] a semantic enhancement module 3 configured to encode and fuse data of the same flexible resource in the resource state mode set and the resource dimension reduction feature set, to obtain a corresponding semantic enhancement data set;

[0144] an anomaly perception module 4 configured to perform regional autonomous anomaly event identification based on a preset machine learning model according to the semantic enhancement data set, to obtain a corresponding event type to be responded to;

[0145] a collaborative analysis module 5 configured to generate an event collaborative response scheme according to the event type to be responded to, the resource state mode set and a preset resource correlation matrix; the preset resource correlation matrix is constructed based on economic correlation degrees, functional substitution degrees and response collaborative degrees among resources;

[0146] a response scheduling module 6 configured to verify and optimize the event collaborative response scheme based on digital twinning technology, to obtain a target collaborative response scheme, and to perform regional resource response scheduling according to the target collaborative response scheme.

[0147] The specific definitions of the virtual power plant resource scheduling system based on multi-source data and knowledge guidance can refer to the definitions of the virtual power plant resource scheduling method based on multi-source data and knowledge guidance, and the corresponding technical effects can also be obtained equally, which will not be repeated here. Each module in the virtual power plant resource scheduling system based on multi-source data and knowledge guidance can be realized by software, hardware and their combinations. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to call and execute the operations of the above modules by the processor.

[0148] In summary, the virtual power plant resource scheduling method and system based on multi-source data and knowledge guidance provided by the embodiments of the present application realize real-time perception of autonomous operation anomaly events in the resource aggregation area based on the resource characteristic analysis mechanism of multi-dimensional data and the semantic fusion mechanism based on knowledge guidance, and realize efficient collaborative response to anomaly events in combination with the collaborative response analysis mechanism based on the multi-dimensional correlation relationship of resources, which can effectively improve the autonomous operation anomaly efficient collaborative response ability and supply-demand self-balancing ability of the resource aggregation area, and further provide reliable protection for the long-term effective operation of the virtual power plant.

[0149] Various embodiments are described herein with reference to the drawings, wherein each embodiment is described in a progressive manner, and each embodiment directly or indirectly refers to each other, and each embodiment focuses on the differences from other embodiments. In particular, the system embodiments are described more simply because they are basically similar to the method embodiments, and the relevant parts can be referred to the description of the method embodiments. It should be noted that the technical features of the above embodiments can be combined in any manner, and in order to make the description simple, not all possible combinations of the technical features of the above embodiments are described, but as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the description.

[0150] The above-described embodiments only express several preferred embodiments of the present application, which are described in a more specific and detailed manner, but should not be understood as limiting the scope of the patent. It should be noted that for ordinary skilled in the art, several improvements and replacements can be made without departing from the technical principles of the present application, and these improvements and replacements should be considered as the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the protection scope of the claims.

Claims

1. A virtual power plant resource scheduling method based on multi-source data and knowledge guidance, characterized in that, The method, applied to a resource aggregation area within a virtual power plant, wherein the resource aggregation area includes several flexible resources, comprises: The system acquires a regional resource operation dataset in real time and performs dimensionality reduction processing on the dataset to generate a corresponding resource dimensionality reduction feature set. The regional resource operation dataset includes multidimensional operation data of various flexible resources. Based on the regional resource operation dataset and the preset state pattern rule base, resource status and response patterns are identified to obtain the corresponding resource status pattern set; the resource status pattern set includes the operation status and response pattern of each flexible resource; The data of the same flexible resources in the resource state pattern set and the resource dimensionality reduction feature set are encoded and fused to obtain the corresponding semantically enhanced dataset. Based on the semantically enhanced dataset, regional autonomous abnormal events are identified using a preset machine learning model to obtain the corresponding event types to be responded to. Based on the event type to be responded to, the resource status pattern set, and the preset resource association matrix, a corresponding event collaborative response scheme is generated; the preset resource association matrix is ​​constructed based on the economic correlation, functional substitutability, and response synergy between resources. The event collaborative response scheme is verified and optimized based on digital twin technology to obtain the target collaborative response scheme, and regional resource response scheduling is executed according to the target collaborative response scheme.

2. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 1, characterized in that, The step of performing dimensionality reduction processing on the regional resource operation dataset to generate a corresponding resource dimensionality reduction feature set includes: Based on flexible resource types, the regional resource operation dataset is divided into multiple resource-type operation sub-datasets; the resource-type operation sub-datasets include energy operation sub-datasets, energy storage operation sub-datasets, and load operation sub-datasets; Obtain the resource class data matrix corresponding to each of the resource class running subsets, and perform standardization processing on the resource class data matrix to obtain the corresponding standardized resource class data matrix; Obtain the first covariance matrix corresponding to each of the standardized resource class data matrices, and perform eigenvalue decomposition on the first covariance matrix to obtain the corresponding resource class dimensionality reduction feature matrix; The dimensionality-reduced feature matrices of each resource class are concatenated to obtain the corresponding global resource feature matrix; Obtain the second covariance matrix corresponding to the global resource feature matrix, and perform eigenvalue decomposition on the second covariance matrix to obtain the corresponding global resource dimensionality reduction feature matrix; The global resource dimensionality reduction feature matrix is ​​subjected to hierarchical inverse mapping to obtain the principal component features of each flexible resource; Based on the principal component characteristics of all flexible resources, the resource dimensionality reduction feature set is obtained.

3. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 1, characterized in that, The preset state mode rule base includes resource state assessment rule sets and response mode assessment rule sets corresponding to different flexible resource types; The step of identifying resource status and response patterns based on the regional resource operation dataset and the preset status pattern rule base to obtain the corresponding resource status pattern set includes: The multidimensional operation data of each flexible resource in the regional resource operation dataset is used to traverse and match the resource status evaluation rule set of the corresponding flexible resource type in the preset state mode rule base to obtain the corresponding operation status. The multidimensional operational data of each flexible resource in the regional resource operation dataset is traversed and matched with the response mode evaluation rule set of the corresponding flexible resource type in the preset state mode rule base to obtain the corresponding response mode.

4. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 1, characterized in that, The step of encoding and fusing data of the same flexible resources in the resource state pattern set and the resource dimensionality reduction feature set to obtain the corresponding semantically enhanced dataset includes: The operating status and response mode of each flexible resource in the resource status mode set are individually encoded to obtain the corresponding operating status code and response mode code. The dimensionality reduction features of each flexible resource in the resource dimensionality reduction feature set are combined with the corresponding running state code and response mode code to obtain the semantic enhancement data of each flexible resource. The semantic enhancement dataset is obtained by aggregating the semantic enhancement data of all flexible resources.

5. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 1, characterized in that, The steps for constructing the preset resource association matrix include: An economic correlation analysis is performed based on the unit power adjustment cost of each flexible resource to obtain the economic correlation degree of each resource pair, and a first correlation matrix is ​​obtained based on the economic correlation degree of all resource pairs. Functional substitutability analysis is performed based on the resource type and adjustable power range of each flexible resource to obtain the functional substitutability degree of each resource pair, and a second correlation matrix is ​​obtained based on the functional substitutability degree of all resource pairs. Dynamic time-correction analysis is performed based on the historical response curves of each flexible resource to obtain the response synergy of each resource pair, and a third correlation matrix is ​​obtained based on the response synergy of all resource pairs. The first association matrix, the second association matrix, and the third association matrix are weighted and fused, and the resulting fused matrix is ​​standardized to obtain the preset resource association matrix.

6. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 1, characterized in that, The types of events to be responded to include power up-adjustment coordinated response events and power down-adjustment coordinated response events; The step of generating a corresponding event collaborative response scheme based on the event type to be responded to, the resource status pattern set, and the preset resource association matrix includes: Based on the resource status pattern set, obtain the currently available resource set; Based on the type of event to be responded to, the available resources in the current available resource set are filtered to obtain the response event resource set, and the current adjustable capability of each flexible resource in the response event resource set is obtained. Based on the preset resource association matrix, cluster analysis is performed on the flexible resources in the response event resource set to obtain several resource clusters; The cluster capability of each resource cluster is obtained based on the current adjustable capability of all flexible resources within each resource cluster, the economic correlation between flexible resources, and the response coordination degree; the cluster capability includes the total cluster adjustment capability, the cluster economic score, and the cluster response coordination degree. Based on the cluster capabilities of each resource cluster, the power adjustment amount corresponding to the event type to be responded to is allocated to obtain the event collaborative response scheme.

7. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 6, characterized in that, The step of performing cluster analysis on the flexible resources in the response event resource set based on the preset resource association matrix to obtain several resource clusters includes: The target number of clusters is obtained based on the total number of flexible resources in the response event resource set and the preset upper limit of the number of clusters; Based on the preset resource association matrix, the corresponding Laplace matrix is ​​obtained; Based on the target cluster number, obtain the eigenvector matrix of the Laplacian matrix; K-means clustering is performed on the row vectors of the feature vector matrix to obtain several resource clusters.

8. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 6, characterized in that, The step of obtaining the cluster capability of each resource cluster based on the current adjustable capability of all flexible resources within each resource cluster, the economic correlation between flexible resources, and the response synergy includes: The current adjustable capabilities of all flexible resources within each resource cluster are summed to obtain the corresponding total cluster adjustment capability. Based on the economic correlation between flexible resources within each resource cluster and the number of cluster resources, the corresponding cluster economic score is obtained; The corresponding cluster response coordination degree is obtained based on the response coordination degree between flexible resources within each resource cluster and the preset response coordination coefficient.

9. The virtual power plant resource scheduling method based on multi-source data and knowledge guidance as described in claim 6, characterized in that, The step of allocating power adjustment amounts corresponding to the event types to be responded to based on the cluster capabilities of each of the resource clusters to obtain the event collaborative response scheme includes: Based on the cluster priority ranking rule, all the resource clusters are sorted to generate a corresponding cluster sequence; the cluster priority ranking rule is based on sorting in descending order of cluster economic score, and then sorting clusters with the same cluster economic score in descending order of cluster response coordination degree. The power adjustment amount corresponding to the event type to be responded to is allocated according to the total adjustment capacity of each resource cluster in the cluster sequence, so as to obtain the cluster power adjustment allocation amount of each resource cluster. The power adjustment allocation of each resource cluster is allocated according to the current adjustable capacity ratio of each flexible resource in the corresponding resource cluster, thus obtaining the power adjustment allocation of each flexible resource in the corresponding resource cluster. The event coordination response scheme is generated based on the cluster power adjustment allocation of each resource cluster and the power adjustment allocation of all flexible resources within each resource cluster.

10. A virtual power plant resource scheduling system based on multi-source data and knowledge guidance, characterized in that, A resource aggregation area applied within a virtual power plant, the resource aggregation area comprising several flexible resources, the system comprising: The preprocessing module is used to acquire the regional resource operation dataset in real time and perform dimensionality reduction processing on the regional resource operation dataset to generate the corresponding resource dimensionality reduction feature set; the regional resource operation dataset includes multidimensional operation data of each flexible resource; The data analysis module is used to identify resource status and response patterns based on the regional resource operation dataset and the preset status pattern rule base to obtain the corresponding resource status pattern set; the resource status pattern set includes the operation status and response pattern of each flexible resource; The semantic enhancement module is used to encode and fuse data of the same flexible resources in the resource state pattern set and the resource dimensionality reduction feature set to obtain the corresponding semantic enhancement dataset; An anomaly detection module is used to identify regional autonomous anomalies based on the semantically enhanced dataset and a preset machine learning model to obtain the corresponding event types to be responded to. The collaborative analysis module is used to generate a collaborative response plan for the event based on the type of the event to be responded to, the set of resource status patterns, and a preset resource association matrix; the preset resource association matrix is ​​constructed based on the economic correlation, functional substitutability, and response synergy between resources. The response scheduling module is used to verify and optimize the event collaborative response scheme based on digital twin technology, obtain the target collaborative response scheme, and execute regional resource response scheduling according to the target collaborative response scheme.

Images