Load aggregation method and device based on space-time cooperative coupling, medium and equipment

Abstract

The application provides a load aggregation method and device based on space-time coupling, a medium and equipment. The load data of a target industry is obtained and corrected according to the industry characteristics to obtain load benchmark data as a unified reference. The load benchmark data is used to correct the deviation of the load data to construct a load feature matrix. Then, a space-time coupling matrix is extracted and the coupling coefficients between various loads are calculated. Next, the target load that can effectively participate in the aggregation adjustment is determined according to the coupling coefficient threshold, and the load object with the potential of collaborative adjustment is screened out. On this basis, the total aggregation adjustment power is calculated by combining the coupling coefficient and the load benchmark data. Finally, the total aggregation adjustment power, the space-time coupling matrix and the load benchmark data are collaboratively integrated to generate an aggregation data interaction result. Thus, the calculation of the aggregation adjustment capacity fully conforms to the actual operation law and the space-time correlation characteristics of various loads, and the multi-element load aggregation adjustment capacity is accurately quantified.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a load aggregation method, apparatus, medium and equipment based on spatiotemporal cooperative coupling. Background Technology

[0002] With the accelerated construction of new power systems, the demand for coordinated dispatch of power generation, grid, load and storage continues to increase. A large number of diverse loads, such as industrial, commercial, residential, charging stations and virtual power plants, are participating in grid regulation. The scenarios are characterized by multiple subjects and multiple spatiotemporal scales. The industry's requirements for the accuracy and efficiency of load aggregation data interaction have also increased significantly.

[0003] However, in existing technologies, the quantitative calculation of aggregated regulation capacity of multiple loads is usually completed by relying on simple load data. The resulting aggregated regulation capacity value has low accuracy and poor matching with the actual operating conditions of the power grid, making it difficult to adapt to the actual operating requirements of the current new power system. Summary of the Invention

[0004] The purpose of this application is to at least solve one of the above-mentioned technical defects, in particular the technical defect that the existing technology relies on simple load data to complete quantitative calculations, resulting in low accuracy of the final aggregate regulation capacity value and poor matching with the actual operating conditions of the power grid, making it difficult to adapt to the actual operating needs of the current new power system.

[0005] Firstly, this application provides a load aggregation method based on spatiotemporal cooperative coupling, the method comprising:

[0006] Obtain load data for the target industry, and correct the load data according to the industry characteristics of the target industry to obtain load baseline data;

[0007] Using the load baseline data as a reference, the load data is corrected for baseline deviation to obtain a load feature matrix, and a spatiotemporal coupling matrix is ​​extracted based on the load feature matrix.

[0008] The coupling coefficients of various types of loads are calculated based on the spatiotemporal coupling matrix, and the loads with coupling coefficients greater than a preset threshold are identified as target loads.

[0009] Based on the coupling coefficients of various target loads and the load baseline data, the total aggregated regulation power is calculated. Combining the total aggregated regulation power, the spatiotemporal coupling matrix, and the load baseline data, aggregated data interaction results are generated.

[0010] In one embodiment, the step of correcting the load data according to the industry characteristics of the target industry to obtain load baseline data includes:

[0011] Obtain the load fitting function corresponding to the target industry, and perform load fluctuation fitting on the load data according to the load fitting function;

[0012] The load correction data is determined based on the fitted parameter values, and the load data is corrected based on the load correction data to obtain the load baseline data.

[0013] In one embodiment, the step of using the load reference data as a reference to perform reference deviation correction on the load data to obtain a load characteristic matrix includes:

[0014] The load data is preprocessed to form a load matrix;

[0015] For each feature column in the load matrix, the sample mean of the feature column is calculated, and the target feature value of each feature value of the feature column after benchmark deviation correction is determined based on each feature value of the feature column, the load benchmark data, and the sample mean.

[0016] When the target feature values ​​of each feature column in the load matrix are determined, a load feature matrix is ​​generated based on the target feature values ​​of each feature column in the load matrix.

[0017] In one embodiment, the step of extracting the spatiotemporal coupling matrix based on the load feature matrix includes:

[0018] Determine the preset multi-scale spatiotemporal graph convolutional network;

[0019] The load feature matrix is ​​input into the multi-scale spatiotemporal graph convolutional network so that the multi-scale spatiotemporal graph convolutional network can capture the load correlation features of the load feature matrix under different time scales and spatial regions, and output the spatiotemporal coupling matrix.

[0020] In one embodiment, the step of calculating the coupling coefficient of various types of loads based on the spatiotemporal coupling matrix, and identifying loads with coupling coefficients greater than a preset threshold as target loads, includes:

[0021] Calculate the coupling strength between various types of loads in the spatiotemporal coupling matrix;

[0022] The average coupling strength between each type of load and other types of loads is determined as the coupling coefficient of that type of load. Loads with coupling coefficients greater than a preset threshold are selected from each type of load and determined as target loads.

[0023] In one embodiment, calculating the total aggregate regulation power based on the coupling coefficients of various target loads and the load baseline data includes:

[0024] Based on a multi-scale spatiotemporal attention mechanism, and combining the coupling coefficients of various target loads with the load baseline data, a spatiotemporal fusion feature matrix is ​​determined.

[0025] The aggregation weights of various target loads are calculated based on the coupling coefficients of various target loads and the spatiotemporal fusion feature matrix.

[0026] The individual adjustable power of each type of target load is obtained, and the individual adjustable power of each type of target load is weighted and summed according to the aggregation weight of each type of target load to obtain the total aggregated adjustable power.

[0027] In one embodiment, determining the spatiotemporal fusion feature matrix based on a multi-scale spatiotemporal attention mechanism, combining the coupling coefficients of various target loads and the load baseline data, includes:

[0028] The coupling coefficients of various target loads are concatenated with the load reference data to obtain the input dataset;

[0029] Based on the multi-scale spatiotemporal attention mechanism, the correlation features of different time granularities and different regional loads in the input dataset are extracted to obtain the attention features of the input dataset at each scale.

[0030] The attention features are weighted and fused to form a spatiotemporal fusion feature matrix.

[0031] Secondly, this application provides a load aggregation device based on spatiotemporal cooperative coupling, the device comprising:

[0032] The data correction module is used to acquire load data of the target industry and correct the load data according to the industry characteristics of the target industry to obtain load baseline data.

[0033] The spatiotemporal coupling module is used to perform benchmark deviation correction on the load data using the load benchmark data as a reference, to obtain the load feature matrix, and to extract the spatiotemporal coupling matrix based on the load feature matrix.

[0034] The load screening module is used to calculate the coupling coefficient of various types of loads based on the spatiotemporal coupling matrix, and to identify loads with coupling coefficients greater than a preset threshold as target loads.

[0035] The load aggregation module is used to calculate the total aggregated regulation power based on the coupling coefficients of various target loads and the load reference data, and to generate aggregated data interaction results by combining the total aggregated regulation power, the spatiotemporal coupling matrix and the load reference data.

[0036] Thirdly, this application provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the load aggregation method based on spatiotemporal cooperative coupling as described in any of the above embodiments.

[0037] Fourthly, this application provides a computer device, including: one or more processors, and a memory;

[0038] The memory stores computer-readable instructions, and when the one or more processors execute the computer-readable instructions, they perform the steps of the load aggregation method based on spatiotemporal cooperative coupling as described in any of the above embodiments.

[0039] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0040] This application provides a load aggregation method, apparatus, medium, and equipment based on spatiotemporal synergistic coupling. It acquires load data from the target industry and corrects it according to industry characteristics, using the resulting load benchmark data as a unified reference. This benchmark data is then used to correct deviations in the load data to construct a load characteristic matrix. Subsequently, a spatiotemporal coupling matrix is ​​extracted, and coupling coefficients between various load types are calculated. This fully explores the inherent correlations between load characteristics in time and space. Based on coupling coefficient thresholds, target loads capable of effectively participating in aggregation regulation are determined, and load objects with synergistic regulation potential are selected. On this basis, the total aggregation regulation power is calculated by combining the coupling coefficients and the load benchmark data. Finally, the total aggregation regulation power, the spatiotemporal coupling matrix, and the load benchmark data are synergistically integrated to generate aggregation data interaction results. This ensures that the calculation of aggregation regulation capability fully aligns with the actual operating patterns and spatiotemporal correlation characteristics of various loads, accurately quantifying the aggregation regulation capability of multi-load loads. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart illustrating a load aggregation method based on spatiotemporal cooperative coupling provided in this application embodiment;

[0043] Figure 2 An example diagram illustrating a load aggregation method based on spatiotemporal cooperative coupling provided in this application embodiment;

[0044] Figure 3 A schematic diagram of a load aggregation device based on spatiotemporal cooperative coupling provided in an embodiment of this application;

[0045] Figure 4 This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] In one embodiment, this application provides a load aggregation method based on spatiotemporal cooperative coupling. The following embodiments illustrate the application of this method to a server. It is understood that the load aggregation method based on spatiotemporal cooperative coupling can be performed on a single server or on a server cluster consisting of multiple servers, and this application does not impose any specific limitations on this.

[0048] like Figure 1 As shown, this application provides a load aggregation method based on spatiotemporal cooperative coupling, the method comprising:

[0049] S101: Obtain load data for the target industry and correct the load data according to the industry characteristics of the target industry to obtain load baseline data.

[0050] Load data refers to the original measured electricity consumption data collected from the target industry, including real-time power, electricity consumption periods, and equipment response parameters for various loads such as industrial and commercial loads, residential loads, and charging piles. Load reference data refers to load reference data obtained by correcting the original load data for industry characteristics, which can accurately reflect the standard electricity consumption patterns of the target industry under normal production and operation conditions.

[0051] In this step, when the power grid needs to perform source-grid-load-storage coordinated scheduling, it can send relevant instructions to the server. When the server receives the instruction, it identifies the target industry and collects real-time or timed data from various power-consuming terminals within the target industry through devices such as smart meters and IoT sensors. This data includes information such as timestamps, power values, and types of electrical equipment. Subsequently, feature-by-feature corrections are made based on the industry's specific attributes. For example, the industrial sector needs to consider the impact of seasonal production cycles and equipment maintenance downtime; the commercial sector needs to eliminate deviations caused by changes in business hours during holidays; and the residential sector needs to eliminate abnormal electricity consumption caused by special events. Ultimately, load benchmark data that characterizes the load fluctuation amplitude, periodic characteristics, and baseline operating level of the target industry under standard operating conditions is obtained.

[0052] Specifically, the load data includes electricity consumption data for various types of loads in the target industry, including but not limited to commercial air conditioning, industrial loads, charging piles, and residential loads.

[0053] S102: Using the load baseline data as a reference, the load data is corrected for baseline deviation to obtain the load characteristic matrix, and the spatiotemporal coupling matrix is ​​extracted based on the load characteristic matrix.

[0054] The load feature matrix refers to a two-dimensional matrix formed by arranging load samples and feature dimensions according to the multi-dimensional features of various loads after benchmark deviation correction. The spatiotemporal coupling matrix refers to the feature matrix output after extracting the temporal and spatial load correlation features through a multi-scale spatiotemporal graph convolutional network. Its matrix elements are used to quantify the spatiotemporal coupling feature values ​​of different regions in different feature dimensions.

[0055] In this step, after the server determines the load baseline data, it uses this baseline data as a reference to perform baseline deviation correction on each feature column of the load data. Specifically, the eigenvalues ​​of each feature column can be proportionally calibrated compared to the corresponding data in the load baseline data to eliminate the magnitude differences caused by different load types having different base values. After completing the baseline deviation correction, different load types can be used as row indices of the matrix, and feature dimensions can be used as column indices of the matrix. Then, the corresponding corrected eigenvalues ​​are filled into the corresponding positions in the matrix to obtain the load feature matrix. Finally, load correlation features under different time scales and spatial regions can be extracted from the load feature matrix to obtain the spatiotemporal coupling matrix.

[0056] S103: Calculate the coupling coefficient of each type of load based on the spatiotemporal coupling matrix, and identify the loads with coupling coefficients greater than a preset threshold as target loads.

[0057] The coupling coefficient measures the degree of coordination and correlation between the corresponding load and other loads in terms of electricity consumption behavior in both time and space. The target load refers to the type of load eligible for inclusion in the aggregated regulation scope.

[0058] In this step, the coupling coefficients of various load types are determined by calculating the correlation or other coupling indices between different row vectors in the spatiotemporal coupling matrix. Specifically, suitable coupling metric functions can be selected, such as Pearson correlation coefficient, cosine similarity, or other time series similarity indices; this application does not impose specific limitations on this. Then, the correlation between different load types is calculated based on the coupling metric function, and the coupling coefficient of each load type is calculated based on the correlation between each load type and other load types. After determining the coupling coefficient of each load type in the spatiotemporal coupling matrix, loads with coupling coefficients greater than a preset threshold can be identified as target loads.

[0059] Furthermore, when measuring the correlation between different types of loads, the adjacency matrix weights in graph neural networks can be used to reflect the dependencies between loads, thereby determining the correlation between different types of loads.

[0060] It is understandable that a larger coupling coefficient indicates that this type of load has a strong correlation with other loads in the spatiotemporal dimension and has high aggregation and regulation value.

[0061] Furthermore, after calculating the coupling coefficients of various load types, the coupling strength of each load can be classified according to these coefficients to determine its coupling level. Loads classified as strongly coupled are preferentially included in the aggregation scope to achieve efficient coordinated regulation and support rapid response in load scheduling. Loads classified as moderately coupled serve as secondary aggregation objects, supplementing aggregation regulation capabilities. Loads classified as weakly coupled are temporarily excluded from the aggregation scope to avoid affecting the stability of aggregation regulation. It should be noted that the granularity of the coupling level classification can be adjusted according to actual conditions; this application does not impose specific restrictions on this. In one example, the coupling strength classification formula can be expressed as:

[0062]

[0063] In the formula, The coupling coefficient represents the k-th type of load, and 1, 2, and 3 represent the coupling strength levels.

[0064] S104: Based on the coupling coefficients of various target loads and load baseline data, calculate the total aggregated regulation power, and combine the total aggregated regulation power, spatiotemporal coupling matrix and load baseline data to generate aggregated data interaction results.

[0065] The total aggregated regulation power refers to the sum of the regulation power of multiple target loads weighted by coupling coefficients, representing the upper limit of the effective regulation output that this multi-load aggregation can provide to fill the gap in the power grid. The aggregation data interaction result refers to a multi-dimensional data set generated by integrating information from the spatiotemporal coupling matrix, load baseline data, and total aggregated regulation power.

[0066] In this step, the server first determines the weight of each target load in the aggregation process, i.e., the aggregation weight, based on the coupling coefficients of various target loads and the load baseline data. Then, combining the individual adjustable power of each type of target load, the total aggregated adjustment power is calculated. In the above calculation process, the quantification of the aggregated adjustment power fully integrates the spatiotemporal correlation strength and baseline operating level of various loads, effectively improving the accuracy and adaptability of the calculation results. Finally, the total aggregated adjustment power, the spatiotemporal coupling matrix, and the load baseline data are fused to generate the aggregated data interaction result.

[0067] Furthermore, after generating the aggregated data interaction results, the results can be directly used for visualization, output to the power grid dispatching system for subsequent dispatching decisions, and transmitted to the metering system as the basis for demand response subsidy settlement.

[0068] In the above embodiments, load data of the target industry is acquired and corrected according to industry characteristics. The resulting load benchmark data serves as a unified reference. This benchmark data is then used to correct deviations in the load data to construct a load feature matrix. Subsequently, a spatiotemporal coupling matrix is ​​extracted, and coupling coefficients between various load types are calculated. This fully explores the inherent correlations between load characteristics in time and space. Based on coupling coefficient thresholds, target loads capable of effectively participating in aggregated regulation are determined, and load objects with collaborative regulation potential are selected. On this basis, the total aggregated regulation power is calculated by combining the coupling coefficients and the load benchmark data. Finally, the total aggregated regulation power, the spatiotemporal coupling matrix, and the load benchmark data are collaboratively integrated to generate aggregated data interaction results. This ensures that the calculation of aggregated regulation capability fully aligns with the actual operating patterns and spatiotemporal correlation characteristics of various loads, accurately quantifying the aggregated regulation capability of multiple loads.

[0069] In one embodiment, load data is corrected according to the industry characteristics of the target industry to obtain load baseline data, including:

[0070] S1: Obtain the load fitting function corresponding to the target industry, and fit the load data to the load fluctuation based on the load fitting function.

[0071] S2: Determine the load correction data based on the fitted parameter values, and correct the load data based on the load correction data to obtain the load baseline data.

[0072] The load fitting function refers to a pre-constructed mathematical analytical expression based on the inherent fluctuation patterns of loads in various industries. The load correction data refers to the set of correction parameters, consisting of the amplitude, angular frequency, initial phase, and reference value, obtained after fitting the original load data according to the load fitting function.

[0073] In this embodiment, a load fitting function corresponding to the industry can be matched first to complete the fluctuation fitting of the load data. Load correction data is generated based on the fitting parameters, and then the load data is calibrated using the load correction data to obtain the load baseline data. In this process, the original load data of each industry is subjected to parameter identification and fluctuation fitting through a targeted fitting function. The correction is completed with the guidance of quantified correction data, so that the output load baseline data fully eliminates random noise and atypical fluctuation interference in the original sampling, and accurately characterizes the inherent fluctuation amplitude, periodic characteristics and baseline operating level of each industry load under standard operating conditions.

[0074] In one example, the process of correcting load data can be represented as:

[0075] (1) Define the load data as follows:

[0076]

[0077] In the formula, Represents load data, Let represent the j-th load characteristic parameter of the i-th load type, n represent the number of load types, and d represent the load characteristic dimension.

[0078] (2) Obtain the load fitting function, which can be expressed as:

[0079]

[0080] In the formula, This represents the actual load operating value at time t. This indicates the amplitude of load fluctuations (characterizing the periodicity of load fluctuations). Indicates angular frequency (characterizing the periodic characteristics of load fluctuations). This indicates the initial phase (characterizing the initial offset of load fluctuations). This represents the load baseline value.

[0081] (3) Based on the above load fitting function, load fluctuation fitting is performed on the load data, and the result is obtained by the least squares method. , , , Four parameters are used to summarize the load correction data. Finally, the load data is corrected based on the load correction data to obtain the load baseline data. This process can be represented as:

[0082]

[0083] In the formula, This represents the load baseline data. This represents the fitting correction amount calculated from the load correction data.

[0084] In one embodiment, using load baseline data as a reference, a baseline deviation correction is performed on the load data to obtain a load characteristic matrix, including:

[0085] S1: Preprocess each feature column in the load data to form a load matrix.

[0086] S2: For each feature column in the load matrix, calculate the sample mean of the feature column, and determine the target feature value of each feature value after benchmark deviation correction based on each feature value of the feature column, the load benchmark data, and the sample mean.

[0087] S3: When the target eigenvalues ​​of each feature column in the load matrix are determined, a load feature matrix is ​​generated based on the target eigenvalues ​​of each feature column in the load matrix.

[0088] In this embodiment, a basic load matrix can be constructed by uniformly preprocessing each feature column in the load data. Based on this, a benchmark deviation correction is performed on each feature column in the load matrix. First, the arithmetic mean of all feature values ​​in the feature column is calculated as the sample mean of the feature column, which represents the typical level of the feature in the entire sample. Then, each feature value in the feature column is traversed, and the deviation between the feature value and the sample mean of its column is calculated. This deviation is then added to the corresponding value of the load benchmark data at that time to obtain the target feature value after benchmark deviation correction. Finally, a load feature matrix is ​​generated based on the target feature value obtained by deviation correction.

[0089] Specifically, various data processing methods can be adopted in the preprocessing of load data, such as data denoising, missing value imputation, and normalization. This application does not impose any specific restrictions on these methods.

[0090] It is understandable that the dimensional differences and benchmark offsets of various load characteristics are systematically eliminated. The multi-source heterogeneous data that were originally incomparable due to differences in acquisition equipment, industry characteristics and dimensionality are uniformly mapped to a standardized coordinate space centered on the load benchmark data. This makes the data of different characteristics and different industries comparable and consistent under the same reference system, so as to carry out subsequent calculations.

[0091] In one example, the preprocessed load matrix can be represented as:

[0092]

[0093] In the formula, Represents the load matrix. This represents the j-th preprocessed load characteristic parameter of the i-th type of load.

[0094] Using load baseline data as the industry load benchmark reference, for The feature columns in the data undergo deviation correction, and the process is as follows:

[0095]

[0096] In the formula, This represents the j-th characteristic value of the i-th type of load after deviation correction, i.e., the target characteristic value. This represents the sample mean of the j-th feature column.

[0097] Finally, based on the target eigenvalues ​​of each feature column in the load matrix, the generated load feature matrix can be represented as:

[0098]

[0099] In another embodiment, deviation correction can be performed only on the power feature column, while the preprocessed feature values ​​can be used as target feature values ​​for other feature columns.

[0100] In one embodiment, the spatiotemporal coupling matrix is ​​extracted based on the load feature matrix, including:

[0101] S1: Determine the preset multi-scale spatiotemporal graph convolutional network.

[0102] S2: Input the load feature matrix into the multi-scale spatiotemporal graph convolutional network so that the multi-scale spatiotemporal graph convolutional network can capture the load correlation features of the load feature matrix under different time scales and spatial regions, and output the spatiotemporal coupling matrix.

[0103] Among them, the multi-scale spatiotemporal graph convolutional network refers to a deep learning network architecture that integrates graph convolution and multi-scale convolution mechanisms. It simultaneously configures multiple convolutional kernels of different sizes to capture temporal dependencies at different time granularities, and introduces graph convolutional structures to model the mutual influence of loads between spatial regions. This network can jointly extract and fuse temporal and spatial information from the load feature matrix, outputting a high-dimensional data representation rich in spatiotemporal correlation features. Load correlation features refer to the temporal synchronicity of load fluctuations and the degree of mutual influence in the spatial dimension among different regions and categories of loads.

[0104] In this embodiment, a pre-trained multi-scale spatiotemporal graph convolutional network is first retrieved from a pre-defined model library. This network consists of an input layer, a multi-scale graph convolutional layer, a fusion layer, and an output layer. The multi-scale graph convolutional layer is configured with three convolutional kernels at different time scales, corresponding to 15-minute, 30-minute, and 60-minute time granularities, respectively, to capture temporal dependencies at different levels. This layer also incorporates a spatial adjacency matrix built based on geographical distance between regions to characterize the spatial correlation strength between regions. After the load feature matrix is ​​input into the network, it first calculates the spatial correlation weights between regions based on the spatial adjacency matrix. For any two regions, the closer their geographical distance, the greater the spatial correlation weight, and vice versa. Simultaneously, the network performs a sliding window calculation on the load feature matrix in the time dimension using convolutional kernels of each scale to extract the temporal change patterns at that scale. Small-scale convolutional kernels are adept at capturing rapid load fluctuations, such as the start-stop response of air conditioning equipment; medium-scale convolutional kernels can extract periodic changes such as those caused by shift changes; and large-scale convolutional kernels focus on the overall trend of daily electricity consumption patterns. Finally, the spatiotemporal coupling matrix is ​​output by fusing load correlation features from various scales.

[0105] Furthermore, the number and length of convolutional scales in the multi-scale spatiotemporal graph convolutional network can be set according to actual needs, and this application does not impose specific restrictions on this. It is understood that in the above process, the multi-scale spatiotemporal graph convolutional network mines and extracts spatiotemporal information from the load feature matrix, simultaneously considering short-term, medium- and long-term temporal patterns and regional spatial linkages, fully restoring the real spatiotemporal coordinated power consumption patterns of diverse loads, and effectively improving the fit between subsequent aggregated regulation capability quantification and grid coordinated dispatch.

[0106] In one example, a multi-scale spatiotemporal graph convolutional network and a spatiotemporal coupling computation method are used to achieve accurate analysis of coupling characteristics by first extracting spatiotemporal features and then quantizing coupling strength.

[0107] First, spatiotemporal feature extraction is performed. A multi-scale spatiotemporal graph convolutional network is used to extract spatiotemporal collaborative features. Three convolutional kernels of different scales (corresponding to 15min, 30min, and 60min time scales, denoted as s=1,2,3) are employed to capture load correlation features at different time scales and spatial regions (industrial parks, commercial areas, and residential areas). The output is a coupled feature matrix Y containing complete spatiotemporal information.

[0108]

[0109] In the formula, , This represents the spatiotemporal coupling feature value of region i in the k-th output feature dimension. This feature value integrates the region's own load characteristics, neighboring region association characteristics, and multi-timescale information. Represents the load characteristic matrix, Let represent the convolution weight matrix at scale s. Let represent the bias term at the s-th scale. Represents the ReLU activation function, used for efficient extraction and fusion of spatiotemporal features. A represents the load spatial correlation adjacency matrix, A = M represents the total number of spatial regions. This indicates the load correlation strength between region i and region j. , This represents the geographical distance between region i and region j. This represents the distance attenuation parameter. This represents the normalized load space association adjacency matrix.

[0110] In one embodiment, the coupling coefficients of various types of loads are calculated based on the spatiotemporal coupling matrix, and loads with coupling coefficients greater than a preset threshold are identified as target loads, including:

[0111] S1: Calculate the coupling strength between various types of loads in the spatiotemporal coupling matrix.

[0112] S2: Determine the average coupling strength between each type of load and other types of loads as the coupling coefficient of that type of load, and select loads with coupling coefficients greater than a preset threshold from each type of load, and determine the selected loads as target loads.

[0113] Among them, coupling strength is used to measure the degree of correlation and coordination of electricity consumption behavior between any two types of loads in the spatiotemporal dimension.

[0114] In this embodiment, the spatiotemporal correlation strength between each type of load and all other types of loads is quantified and summarized into a coupling coefficient. This allows for a precise assessment of the synergistic status and regulation potential of each type of load within the overall load system. Based on this, target loads with regulation potential can be selected from among the various types of loads. This ensures that strongly synergistic loads are prioritized to improve aggregation regulation efficiency, while effectively excluding weakly correlated loads to avoid their interference with aggregation stability.

[0115] In one example, the coupling strength between various types of loads can be calculated using the following expression:

[0116]

[0117] In the formula, This represents the coupling strength between the i-th type of load and the j-th type of load. , Taken from the spatiotemporal coupling matrix, Represents variance. Describing covariance, The larger the value, the stronger the coupling between the two types of loads.

[0118] In one embodiment, the total aggregate regulation power is calculated based on the coupling coefficients of various target loads and load baseline data, including:

[0119] S1: Based on the multi-scale spatiotemporal attention mechanism, the spatiotemporal fusion feature matrix is ​​determined by combining the coupling coefficients of various target loads and the load baseline data.

[0120] S2: Calculate the aggregation weight of various target loads based on the coupling coefficients and spatiotemporal fusion feature matrices of various target loads.

[0121] S3: Obtain the individual adjustable power of each type of target load, and perform a weighted summation of the individual adjustable power of each type of target load according to the aggregation weight of each type of target load to obtain the total aggregated adjustable power.

[0122] The multi-scale spatiotemporal attention mechanism calculates the relevance scores between the query vector and the key vector at different time scales for the input features, and then normalizes these scores to obtain attention features at each scale. This allows the model to adaptively focus on spatiotemporal feature dimensions that contribute more to the current aggregation task. The spatiotemporal fusion feature matrix is ​​a comprehensive feature representation output after weighted fusion of various target loads through the multi-scale attention mechanism. The aggregation weights are used to quantify the contribution ratio of each type of target load in the overall aggregation adjustment.

[0123] In this embodiment, the multi-scale spatiotemporal attention mechanism enables adaptive extraction and fusion of spatiotemporal features of various target loads, allowing key features at different time scales to be reasonably focused. Furthermore, the allocation of aggregation weights fully integrates both coupling strength and feature saliency, thereby ensuring that the quantification of total aggregated regulation power accurately matches the spatiotemporal coupling characteristics and actual regulation potential of various loads, thus improving the accuracy of aggregated regulation capacity calculation and the adaptability to operating conditions.

[0124] In one example, based on the coupling coefficients of various target loads and the spatiotemporal fusion feature matrix, the aggregate weights of various target loads can be calculated according to the following expression:

[0125]

[0126] In the formula, This represents the aggregate weight of the k-th type of load. This represents the coupling coefficient of the k-th type of load. This represents the feature corresponding to the k-th type of load in the spatiotemporal fusion feature matrix. It represents its second norm, used to characterize the multi-scale spatiotemporal correlation strength of this type of load.

[0127] Integrated individual regulating power of type k load Aggregate weights Coupling coefficient The total aggregate regulation power was calculated. :

[0128]

[0129] In the formula, m represents the quantity of the target load.

[0130] Finally, an aggregated data interaction model is constructed, and the aggregated data interaction results are output. :

[0131]

[0132] In the formula, , , These are the weighting coefficients; a set of optional weighting coefficients is: , , ,satisfy .

[0133] In one embodiment, based on a multi-scale spatiotemporal attention mechanism, and combining the coupling coefficients of various target loads and load baseline data, a spatiotemporal fusion feature matrix is ​​determined, including:

[0134] S1: The coupling coefficients of various target loads are concatenated with the load baseline data to obtain the input dataset.

[0135] S2: Based on the multi-scale spatiotemporal attention mechanism, the correlation features of different time granularities and different regional loads in the input dataset are extracted to obtain the attention features of the input dataset at each scale.

[0136] S3: Weighted fusion of each attention feature is performed to form a spatiotemporal fusion feature matrix.

[0137] Among them, attention features refer to the weighted representation vector output after processing by the multi-scale spatiotemporal attention mechanism.

[0138] In this embodiment, a multi-scale spatiotemporal attention mechanism adaptively evaluates and weights the feature dimensions of the input dataset at each temporal granularity, enhancing the most informative feature dimensions and effectively suppressing irrelevant features at different scales. The spatiotemporal fusion feature matrix is ​​then generated by fusing the attention features from each scale. This multi-scale attention mechanism improves the representation ability and extraction accuracy of complex spatiotemporally related features.

[0139] In one example, load baseline data Coupling coefficient with single type of load Feature concatenation is performed, and the resulting data serves as the original input to the model. Standardization is then performed using linear mapping to eliminate dimensional differences, ensuring stable and reliable model input, and the output dataset is the input dataset. :

[0140]

[0141] In the formula, The mean of the concatenated features. represents the standard deviation of the spliced ​​features.

[0142] A multi-scale spatiotemporal attention mechanism is constructed based on three time scales (s=1,2,3, corresponding to 15min, 30min, and 60min respectively). Using this as input, it captures the correlation features of loads at different time granularities and in different regions, overcoming the limitation of traditional models that only perform single-scale analysis, and outputs attention features at each scale. :

[0143]

[0144] In the formula, , Let represent the query weight matrix and the key weight matrix at the s-th scale, respectively.

[0145] Attention features at three scales are weighted and fused to balance real-time responsiveness and medium- to long-term scheduling stability, forming a unified spatiotemporal fusion feature matrix. :

[0146]

[0147] In the formula, This represents the weight coefficient at the s-th scale. denoted as the feature map weight at the s-th scale.

[0148] In one embodiment, such as Figure 2 As shown, Figure 2 This is an example diagram illustrating a load aggregation method based on spatiotemporal cooperative coupling, provided as an embodiment of this application. Figure 2 The first part is the fitting of industry-specific load benchmark data, the second part is the analysis of the spatiotemporal coupling characteristics of multi-dimensional loads, the third part is the construction of the adjustment model based on the multi-scale aggregation model, and the fourth part is the interaction of multi-dimensional load aggregation data.

[0149] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0150] The load aggregation device based on spatiotemporal cooperative coupling provided in the embodiments of this application is described below. The load aggregation device based on spatiotemporal cooperative coupling described below can be referred to in correspondence with the load aggregation method based on spatiotemporal cooperative coupling described above.

[0151] like Figure 3 As shown, this application provides a load aggregation device 200 based on spatiotemporal cooperative coupling, the device comprising:

[0152] The data correction module 201 is used to acquire load data of the target industry and correct the load data according to the industry characteristics of the target industry to obtain load baseline data.

[0153] The spatiotemporal coupling module 202 is used to correct the load data for reference based on the load reference data, obtain the load feature matrix, and extract the spatiotemporal coupling matrix based on the load feature matrix.

[0154] The load screening module 203 is used to calculate the coupling coefficient of various types of loads based on the spatiotemporal coupling matrix, and to identify the loads with coupling coefficients greater than a preset threshold as target loads.

[0155] The load aggregation module 204 is used to calculate the total aggregated regulation power based on the coupling coefficients of various target loads and load reference data, and to generate aggregated data interaction results by combining the total aggregated regulation power, the spatiotemporal coupling matrix and the load reference data.

[0156] In the above embodiments, load data of the target industry is acquired and corrected according to industry characteristics. The resulting load benchmark data serves as a unified reference. This benchmark data is then used to correct deviations in the load data to construct a load feature matrix. Subsequently, a spatiotemporal coupling matrix is ​​extracted, and coupling coefficients between various load types are calculated. This fully explores the inherent correlations between load characteristics in time and space. Based on coupling coefficient thresholds, target loads capable of effectively participating in aggregated regulation are determined, and load objects with collaborative regulation potential are selected. On this basis, the total aggregated regulation power is calculated by combining the coupling coefficients and the load benchmark data. Finally, the total aggregated regulation power, the spatiotemporal coupling matrix, and the load benchmark data are collaboratively integrated to generate aggregated data interaction results. This ensures that the calculation of aggregated regulation capability fully aligns with the actual operating patterns and spatiotemporal correlation characteristics of various loads, accurately quantifying the aggregated regulation capability of multiple loads.

[0157] In one embodiment, the data correction module includes:

[0158] The load fitting submodule is used to obtain the load fitting function corresponding to the target industry and to fit the load data to load fluctuations based on the load fitting function.

[0159] The baseline correction submodule is used to determine the load correction data based on the fitted parameter values, and to correct the load data based on the load correction data to obtain the load baseline data.

[0160] In one embodiment, the spatiotemporal coupling module includes:

[0161] The preprocessing submodule is used to preprocess the various feature columns in the load data to form a load matrix;

[0162] The deviation correction submodule is used to calculate the sample mean of each feature column in the load matrix, and determine the target feature value of each feature value after benchmark deviation correction based on each feature value of the feature column, the load benchmark data and the sample mean.

[0163] The matrix generation submodule is used to generate a load feature matrix based on the target feature values ​​of each feature column in the load matrix when the target feature values ​​of each feature column in the load matrix are determined.

[0164] In one embodiment, the spatiotemporal coupling module includes:

[0165] The network determination submodule is used to determine the preset multi-scale spatiotemporal graph convolutional network;

[0166] The matrix output submodule is used to input the load feature matrix into the multi-scale spatiotemporal graph convolutional network, so that the multi-scale spatiotemporal graph convolutional network can capture the load correlation features of the load feature matrix under different time scales and spatial regions, and output the spatiotemporal coupling matrix.

[0167] In one embodiment, the load screening module includes:

[0168] The coupling calculation submodule is used to calculate the coupling strength between various types of loads in the spatiotemporal coupling matrix;

[0169] The load selection submodule is used to determine the average coupling strength between each type of load and other types of loads as the coupling coefficient of that type of load, and select loads with coupling coefficients greater than a preset threshold from each type of load, and determine the selected loads as target loads.

[0170] In one embodiment, the load aggregation module includes:

[0171] The feature extraction submodule is used to determine the spatiotemporal fusion feature matrix based on a multi-scale spatiotemporal attention mechanism, combined with the coupling coefficients of various target loads and load baseline data;

[0172] The weight calculation submodule is used to calculate the aggregate weight of various target loads based on the coupling coefficients and spatiotemporal fusion feature matrices of various target loads.

[0173] The power determination submodule is used to obtain the individual adjustable power of each type of target load, and to perform a weighted summation of the individual adjustable power of each type of target load according to the aggregation weight of each type of target load to obtain the total aggregated adjustable power.

[0174] In one embodiment, the feature extraction submodule includes:

[0175] The feature splicing unit is used to splice the coupling coefficients of various target loads with the load baseline data to obtain the input dataset;

[0176] The attention computation unit is used to extract the correlation features of different temporal granularities and different regional loads in the input dataset based on the multi-scale spatiotemporal attention mechanism, so as to obtain the attention features of the input dataset at each scale.

[0177] The weighted fusion unit is used to weight and fuse various attention features to form a spatiotemporal fusion feature matrix.

[0178] The division of modules in the above-described load aggregation device based on spatiotemporal cooperative coupling is merely illustrative. In other embodiments, the load aggregation device based on spatiotemporal cooperative coupling can be divided into different modules as needed to complete all or part of its functions. Each module in the above-described load aggregation device based on spatiotemporal cooperative coupling can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0179] In one embodiment, this application also provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the load aggregation method based on spatiotemporal cooperative coupling as described in any of the above embodiments.

[0180] In one embodiment, this application also provides a computer device storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the load aggregation method based on spatiotemporal cooperative coupling as described in any of the above embodiments.

[0181] Indicatively, such as Figure 4 As shown, Figure 4 This is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. (Refer to...) Figure 4 The computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions, such as application programs, that can be executed by the processing component 302. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the load aggregation method based on spatiotemporal co-coupling of any of the above embodiments.

[0182] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0183] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0184] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. In this document, the singular forms "a," "an," and "the" may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising / including” or “having” specify the presence of the stated features, wholes, steps, operations, components, parts or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0185] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0186] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A load aggregation method based on spatiotemporal cooperative coupling, characterized in that, The method includes: Obtain load data for the target industry, and correct the load data according to the industry characteristics of the target industry to obtain load baseline data; Using the load baseline data as a reference, the load data is corrected for baseline deviation to obtain a load feature matrix, and a spatiotemporal coupling matrix is ​​extracted based on the load feature matrix. The coupling coefficients of various types of loads are calculated based on the spatiotemporal coupling matrix, and the loads with coupling coefficients greater than a preset threshold are identified as target loads. Based on the coupling coefficients of various target loads and the load baseline data, the total aggregated regulation power is calculated. Combining the total aggregated regulation power, the spatiotemporal coupling matrix, and the load baseline data, aggregated data interaction results are generated.

2. The load aggregation method based on spatiotemporal synergistic coupling according to claim 1, characterized in that, The step of correcting the load data according to the industry characteristics of the target industry to obtain load baseline data includes: Obtain the load fitting function corresponding to the target industry, and perform load fluctuation fitting on the load data according to the load fitting function; The load correction data is determined based on the fitted parameter values, and the load data is corrected based on the load correction data to obtain the load baseline data.

3. The load aggregation method based on spatiotemporal cooperative coupling according to claim 1, characterized in that, The step of using the load baseline data as a reference to perform baseline deviation correction on the load data to obtain the load characteristic matrix includes: The load data is preprocessed to form a load matrix; For each feature column in the load matrix, the sample mean of the feature column is calculated, and the target feature value of each feature value of the feature column after benchmark deviation correction is determined based on each feature value of the feature column, the load benchmark data, and the sample mean. When the target feature values ​​of each feature column in the load matrix are determined, a load feature matrix is ​​generated based on the target feature values ​​of each feature column in the load matrix.

4. The load aggregation method based on spatiotemporal cooperative coupling according to claim 1, characterized in that, The step of extracting the spatiotemporal coupling matrix based on the load feature matrix includes: Determine the preset multi-scale spatiotemporal graph convolutional network; The load feature matrix is ​​input into the multi-scale spatiotemporal graph convolutional network so that the multi-scale spatiotemporal graph convolutional network can capture the load correlation features of the load feature matrix under different time scales and spatial regions, and output the spatiotemporal coupling matrix.

5. The load aggregation method based on spatiotemporal cooperative coupling according to claim 1, characterized in that, The step of calculating the coupling coefficient of various types of loads based on the spatiotemporal coupling matrix, and identifying loads with coupling coefficients greater than a preset threshold as target loads, includes: Calculate the coupling strength between various types of loads in the spatiotemporal coupling matrix; The average coupling strength between each type of load and other types of loads is determined as the coupling coefficient of that type of load. Loads with coupling coefficients greater than a preset threshold are selected from each type of load and determined as target loads.

6. The load aggregation method based on spatiotemporal cooperative coupling according to any one of claims 1 to 5, characterized in that, The calculation of the total aggregate regulation power based on the coupling coefficients of various target loads and the load baseline data includes: Based on a multi-scale spatiotemporal attention mechanism, and combining the coupling coefficients of various target loads with the load baseline data, a spatiotemporal fusion feature matrix is ​​determined. The aggregation weights of various target loads are calculated based on the coupling coefficients of various target loads and the spatiotemporal fusion feature matrix. The individual adjustable power of each type of target load is obtained, and the individual adjustable power of each type of target load is weighted and summed according to the aggregation weight of each type of target load to obtain the total aggregated adjustable power.

7. The load aggregation method based on spatiotemporal cooperative coupling according to claim 6, characterized in that, The multi-scale spatiotemporal attention mechanism, combined with the coupling coefficients of various target loads and the load baseline data, determines the spatiotemporal fusion feature matrix, including: The coupling coefficients of various target loads are concatenated with the load reference data to obtain the input dataset; Based on the multi-scale spatiotemporal attention mechanism, the correlation features of different time granularities and different regional loads in the input dataset are extracted to obtain the attention features of the input dataset at each scale. The attention features are weighted and fused to form a spatiotemporal fusion feature matrix.

8. A load aggregation device based on spatiotemporal synergistic coupling, characterized in that, The device includes: The data correction module is used to acquire load data of the target industry and correct the load data according to the industry characteristics of the target industry to obtain load baseline data. The spatiotemporal coupling module is used to perform benchmark deviation correction on the load data using the load benchmark data as a reference, to obtain the load feature matrix, and to extract the spatiotemporal coupling matrix based on the load feature matrix. The load screening module is used to calculate the coupling coefficient of various types of loads based on the spatiotemporal coupling matrix, and to identify loads with coupling coefficients greater than a preset threshold as target loads. The load aggregation module is used to calculate the total aggregated regulation power based on the coupling coefficients of various target loads and the load reference data, and to generate aggregated data interaction results by combining the total aggregated regulation power, the spatiotemporal coupling matrix and the load reference data.

9. A storage medium, characterized in that: The storage medium stores computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the load aggregation method based on spatiotemporal cooperative coupling as described in any one of claims 1 to 7.

10. A computer device, characterized in that, include: One or more processors, and memory; The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of the load aggregation method based on spatiotemporal co-coupling as described in any one of claims 1 to 7.

Images