Load spatio-temporal distribution prediction method and device based on multi-source spatio-temporal information hierarchical association, storage medium and electronic equipment

Abstract

The invention relates to the technical field of power system load prediction, in particular to a load spatio-temporal distribution prediction method and device based on multi-source spatio-temporal information hierarchical association, a storage medium and electronic equipment, and the method comprises the steps: obtaining a multi-source data set in a power supply range of a to-be-predicted area, and obtaining power cells; inputting each power cell into the multi-source spatio-temporal information hierarchical association model to obtain corresponding inherent attributes and external attributes; and inputting the inherent attributes and the external attributes of the power cells to the load prediction model to obtain a load prediction value of each power cell. According to the method, the relevance of the multi-source data is analyzed based on the multi-source spatio-temporal information hierarchical association model, a solid big data basis is provided for load spatial distribution feature analysis, and the load spatio-temporal distribution prediction result of the to-be-predicted region is obtained based on the relevance of the multi-source data through the clustering algorithm and the sparse least square support vector regression network model. And the load space-time distribution prediction result is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of load forecasting for an electric power system, and is a method, device, storage medium and electronic equipment for forecasting the spatiotemporal distribution of loads based on hierarchical association of multi-source spatiotemporal information. Background Art

[0002] The accuracy of the temporal and spatial distribution information of regional energy Internet load has important guiding significance for the decision-making of planning departments and energy management departments. With the rapid development of social informatization, especially the advent of the big data era, the interconnection of a large amount of social development data, economic data, meteorological data, geographic information, taxi operation trajectory data, distributed photovoltaic power data and other data resources has become a reality, which provides the possibility of mining the correlation between electric vehicle operation data, photovoltaic power installed capacity data and other types of data in power grid big data. In particular, a large amount of economic and social data is closely related to photovoltaic power installed capacity data. The correlation and coupling of this data contains a lot of knowledge and economic operation laws, and has important social and economic value.

[0003] The basis for the prediction of the spatiotemporal distribution of load and photovoltaic power is data, especially multi-source data with obvious spatial characteristics and interrelatedness. However, the current storage architecture of multi-source data of regional energy Internet is mostly based on the data model for distribution network, and the logical association of these multi-source data in the database is centered on the equipment ledger. In this storage architecture, the distribution network operation data, topological data, etc. are logically close to the equipment ledger, and their storage and access efficiency is high. However, user electricity consumption data, geographic information, taxi operation data, etc., which are larger in scale and more closely related to the prediction of the spatiotemporal distribution of load and photovoltaic power, are in a relatively marginal area in the logical architecture. When such data have large-scale access requirements, the efficiency of the application may be significantly reduced. In addition, the multi-source data of regional energy Internet is divided into two categories: structured data and semi- / unstructured data. The information extraction of structured data is convenient and the value density is usually high, while the information extraction of semi- / unstructured data is usually relatively complex and the value density is low. This difference in data format and application value is a barrier to the effective integration of multi-source information of regional energy Internet. In addition, the existing distribution network geographic information system usually expresses and stores geographic spatial data and its attribute data separately, which will make it difficult to meet the integrated access requirements of regional energy Internet spatiotemporal data for load and photovoltaic power spatiotemporal distribution prediction. Summary of the invention

[0004] The present invention provides a load spatiotemporal distribution prediction method, device, storage medium and electronic device based on hierarchical association of multi-source spatiotemporal information, which overcomes the shortcomings of the above-mentioned prior art and can effectively solve the problem that the scattered data sources in the existing regional energy Internet big data are difficult to directly and efficiently apply to load spatiotemporal distribution prediction.

[0005] One of the technical solutions of the present invention is achieved by the following measures: a method for predicting the spatiotemporal distribution of load based on hierarchical association of multi-source spatiotemporal information, comprising:

[0006] Obtain a multi-source data set within the power supply range of the area to be predicted, grid the power supply range of the area to be predicted into equal areas, and record each grid as a power cell, wherein the multi-source data includes power enterprise measurement data, power enterprise operation data, and power enterprise external data;

[0007] Input each power cell into the multi-source spatiotemporal information hierarchical association model to obtain the corresponding inherent attributes and external attributes, wherein the inherent attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell; the external attributes include the annual GDP, population and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of the user in the power cell during the monthly forecast, the prosperity index of the main upstream industry and the monthly average temperature in the region; the multi-source spatiotemporal information hierarchical association model includes dividing the multi-source data into corresponding data layers according to their data types, storing the data in each data layer in grids or network nodes, and associating different data layers through location correlation to obtain a multi-source spatiotemporal information hierarchical association data set;

[0008] One or more corresponding load forecasting models are selected from the load forecasting model set, and the inherent attributes and external attributes of the power cell are input to obtain the load forecast value of each power cell, thereby obtaining the load spatiotemporal distribution prediction result of the area to be predicted, wherein each load forecasting model in the load forecasting model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

[0009] The following are further optimizations and / or improvements to the above technical solutions:

[0010] The execution process of the multi-source spatiotemporal information hierarchical association model includes:

[0011] Identify each multi-source data and determine the attributes of each multi-source data, wherein the attributes include source, format, collection time, geographical coordinates of collection location and related information;

[0012] Map the attributes of each multi-source data into a standard and unified metadata format to form corresponding metadata;

[0013] Each metadata is flattened into the corresponding data layer according to the geographic coordinates of the collection location in the attributes, and each data layer is abstracted in a network or grid manner, and the corresponding geographic coordinates of the collection location or time series data are stored in each node, edge or grid.

[0014] The above-mentioned process flattens each metadata into the corresponding data layer according to the geographic coordinates of the collection location in the attribute, and abstracts each data layer into a network or grid, and stores the corresponding geographic coordinates of the collection location or time series data in each node, edge or grid, including:

[0015] Each metadata is flattened into the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and each data layer is networked or gridded for abstraction;

[0016] The horizontal and vertical associations between different data nodes, edges or grids are established through the three-dimensional spatial coordinate system, where the X-axis and Y-axis describe the spatial position corresponding to the data, and the Z-axis describes the data's membership level. For data nodes, edges or grids in the same data layer, horizontal associations are achieved through the same Z-axis coordinates, and they are distinguished by the differences in the geographic coordinates of the collection locations. For data nodes, edges or grids in different data layers, vertical associations between data are achieved with the geographic coordinates of the collection location as the core. If data drilling is performed on different data layers along the Z-axis.

[0017] The construction process of the above load forecasting model set includes:

[0018] Obtain historical load data and related data within the power supply range of the power company, grid the power supply range of the power company in equal areas, and record each grid as a power cell;

[0019] Extract the attributes of each power cell, where the attributes include intrinsic attributes and external attributes. The intrinsic attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell. The external attributes include the annual GDP, population and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of users in the power cell, the prosperity index of the main upstream industry and the monthly average temperature in the region during the monthly forecast;

[0020] The inherent attributes are used as the feature vectors of multi-level clustering, and the k-means algorithm is used to perform cluster analysis on the power cells to obtain multiple categories of cell sub-sets;

[0021] The sparse least squares support vector regression network model is trained based on each cell set to obtain the corresponding load forecasting model.

[0022] Based on the uncertainty of the external attributes of the power cell, the confidence condition is set, and the inherent attributes and external attributes of the power cell are input into the load forecasting model under the confidence condition to forecast the load of the power cell, and finally obtain the load spatiotemporal distribution forecast result of the area to be predicted.

[0023] The second technical solution of the present invention is achieved by the following measures: a load spatiotemporal distribution prediction device based on hierarchical association of multi-source spatiotemporal information, comprising:

[0024] The data resource layer obtains a multi-source data set within the power supply range of the area to be predicted, grids the power supply range of the area to be predicted into equal areas, and records each grid as a power cell. The multi-source data includes the measurement data of the power enterprise, the operation data of the power enterprise, and the external data of the power enterprise;

[0025] The data processing layer inputs each power cell into the multi-source spatiotemporal information hierarchical association model to obtain the corresponding inherent attributes and external attributes, wherein the inherent attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell; the external attributes include the annual GDP, population, and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of the user in the power cell during the monthly forecast, the prosperity index of the main upstream industry, and the monthly average temperature in the region; the multi-source spatiotemporal information hierarchical association model includes dividing the multi-source data into corresponding data layers according to their data types, storing the data in each data layer in a grid or network node, and associating different data layers through location correlation to obtain a multi-source spatiotemporal information hierarchical association data set;

[0026] The data warehouse layer stores the processing results of the data processing layer;

[0027] At the data application layer, one or more corresponding load forecasting models are selected from the load forecasting model set, and the inherent attributes and external attributes of the power cell are input to obtain the load forecast value of each power cell, thereby obtaining the load spatiotemporal distribution forecast result of the area to be predicted. Each load forecasting model in the load forecasting model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

[0028] The following are further optimizations and / or improvements to the above technical solutions:

[0029] The above data processing layer includes:

[0030] An identification module identifies each multi-source data and determines the attributes of each multi-source data, wherein the attributes include source, format, collection time and geographical coordinates of collection location;

[0031] The metadata construction module maps the attributes of each multi-source data into a standard and unified metadata format to form the corresponding metadata;

[0032] The grid / networking module flattens each metadata into the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and performs network or grid abstraction on each data layer;

[0033] The association module establishes horizontal and vertical associations between different data nodes, edges or grids through a three-dimensional spatial coordinate system, where the X-axis and Y-axis describe the spatial position corresponding to the data, and the Z-axis describes the data's hierarchy. For data nodes, edges or grids in the same data layer, horizontal associations are achieved through the same Z-axis coordinates, and they are distinguished by the differences in the geographic coordinates of the collection locations. For data nodes, edges or grids in different data layers, vertical associations between data are achieved with the geographic coordinates of the collection location as the core, if data drilling is performed on different data layers along the Z-axis.

[0034] The above data application layer includes:

[0035] The model building module includes: obtaining historical load data and related data within the power supply range of the power company, and gridding the power supply range of the power company in equal areas, and recording each grid as a power cell; extracting the attributes of each power cell, where the attributes include inherent attributes and external attributes; using the inherent attributes as the feature vector of multi-level clustering, using the k-means algorithm to perform cluster analysis on the power cells, and obtaining multiple categories of cell sub-sets; training the sparse least squares support vector regression network model based on each cell sub-set to obtain the corresponding load forecasting model;

[0036] The prediction module selects one or more corresponding load prediction models from the load prediction model set, inputs the inherent properties and external properties of the power cell, obtains the load prediction value of each power cell, and thus obtains the load spatiotemporal distribution prediction result of the area to be predicted, wherein each load prediction model in the load prediction model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

[0037] The present invention analyzes the correlation of multi-source data based on a multi-source spatiotemporal information hierarchical association model, and correlates user power load data with regional power grid topology, equipment inventory, geographic information, meteorological data and other data to varying degrees, providing a solid big data foundation for load spatial distribution feature analysis. The load forecasting model is further constructed through a clustering algorithm and a sparse least squares support vector regression network model, and the load spatiotemporal distribution prediction result of the to-be-predicted area is obtained based on the correlation of multi-source data, making the load spatiotemporal distribution prediction result more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Attached Figure 1 It is a schematic diagram of the flow chart of the load spatiotemporal distribution prediction method of the present invention.

[0039] Attached Figure 2 It is a flowchart of the execution method of the multi-source spatiotemporal information hierarchical association model in the present invention.

[0040] Attached Figure 3 This is a schematic diagram of metadata association relationships with geographic information as the core in the present invention.

[0041] Attached Figure 4 It is a schematic diagram of the structure of a multi-source spatiotemporal information hierarchical association model in the present invention.

[0042] Attached Figure 5 It is a schematic diagram of the flow of the method for constructing a load forecasting model set in the present invention.

[0043] Attached Figure 6 It is a schematic diagram of the structure of the load spatiotemporal distribution prediction device of the present invention. DETAILED DESCRIPTION

[0044] The present invention is not limited by the following embodiments, and specific implementation methods can be determined based on the technical solution of the present invention and actual conditions.

[0045] The present invention will be further described below in conjunction with embodiments and drawings:

[0046] Embodiment 1: As attached Figure 1 As shown, the embodiment of the present invention discloses a method for predicting the spatiotemporal distribution of load based on hierarchical association of multi-source spatiotemporal information, including:

[0047] Step S110, obtaining a multi-source data set within the power supply range of the area to be predicted, gridding the power supply range of the area to be predicted into equal areas, and recording each grid as a power cell, wherein the multi-source data includes power enterprise measurement data, power enterprise operation data, and power enterprise external data.

[0048] In the above steps, the multi-source data set within the power supply range of the area to be predicted is obtained from the regional energy Internet big data system, among which the external data of the power enterprise refers to the data other than the measurement data and operation data of the power enterprise, such as taxi operation trajectory data, geographic information, meteorological data, etc.

[0049] Step S120, input each power cell into the multi-source spatiotemporal information hierarchical association model to obtain corresponding inherent attributes and external attributes, wherein the inherent attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell; the external attributes include the annual GDP, population and annual average temperature in the area where the power cell is located during the annual forecast, the prosperity index of the main industry of the user in the power cell during the monthly forecast, the prosperity index of the main upstream industry and the monthly average temperature in the area; the multi-source spatiotemporal information hierarchical association model includes dividing the multi-source data into corresponding data layers according to their data types, storing the data in each data layer in a grid or network node, and associating different data layers through position correlation to obtain a multi-source spatiotemporal information hierarchical association data set.

[0050] In the above steps, as shown in the attached Figure 2 As shown in FIG. 1 , the execution process of the multi-source spatiotemporal information hierarchical association model includes:

[0051] Step S121, identifying each multi-source data, determining the attributes of each multi-source data, wherein the attributes include source, format, collection time, geographical coordinates of the collection location and related information;

[0052] Step S122, mapping the attributes of each multi-source data into a standard unified metadata format to form corresponding metadata;

[0053] The above-mentioned mapping of the attributes of each multi-source data into a standard unified metadata format is mainly used to solve the heterogeneous problem of regional energy Internet big data. Here, the mapping of the attributes of each multi-source data into a standard unified metadata format can be performed by discretization, dimension unification, and attribute screening. This embodiment uses XML format metadata to describe the actual data. Although the data format and type of each type of regional energy Internet data may be different, their common feature is that they all have spatiotemporal attributes, that is, each actual data object has a corresponding collection location geographic coordinates and collection time identifier. Therefore, for each actual data object, the marking field information such as device ID identifier, collection time and collection location geographic coordinates can be mapped into a standard unified metadata format, as shown in the attached figure. Figure 3 shown.

[0054] Step S123, flattens each metadata into a corresponding data layer according to the geographic coordinates of the collection location in the attribute, and abstracts each data layer into a network or grid, and stores the corresponding geographic coordinates of the collection location or time series data in each node, edge or grid.

[0055] The above step S123 specifically includes:

[0056] (1) Flatten each metadata into the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and perform network or grid abstraction on each data layer;

[0057] Here, data layers are set up according to the composition of regional energy Internet big data and the needs of related business applications. They may include but are not limited to geographic information data layer, power grid information data layer, load and photovoltaic power data layer, time series data layer, traffic information data layer and socio-economic data layer, etc. Multi-source data are divided into each data layer respectively, and all data are flattened to each data layer according to the geographic coordinate identifier in the metadata, and each data layer is networked or gridded abstracted, that is, a data network composed of nodes and edges is generated in each data layer, or the data layer is divided into data grids of equal size.

[0058] (2) Establish horizontal and vertical associations between different data nodes, edges or grids through a three-dimensional spatial coordinate system, where the X-axis and Y-axis describe the spatial position of the data, and the Z-axis describes the data's hierarchy. For data nodes, edges or grids in the same data layer, horizontal associations are achieved through the same Z-axis coordinates, and they are distinguished by the differences in the geographic coordinates of the collection locations. For data nodes, edges or grids in different data layers, vertical associations between data are achieved with the geographic coordinates of the collection locations as the core. If data is drilled along the Z-axis for different data layers.

[0059] The multi-source spatiotemporal information hierarchical association model corresponding to the above steps 121 to 123 can be shown in the attached figure. Figure 4 shown.

[0060] Step S130, select one or more corresponding load forecasting models from the load forecasting model set, input the inherent attributes and external attributes of the power cell, obtain the load forecast value of each power cell, and thereby obtain the load spatiotemporal distribution forecast result of the area to be predicted, wherein each load forecasting model in the load forecasting model set is obtained by training a sparse least squares support vector regression network model using historical load data and related data.

[0061] As attached Figure 5 As shown in FIG. 1 , the construction process of the above load forecasting model set includes:

[0062] (1) Obtain historical load data and related data within the power supply range of the power company, grid the power supply range of the power company in equal areas, and record each grid as a power cell;

[0063] Each cell can be represented as C(i,j)(i=1,2,…,N r ,j=1,2,…,N c ), where N r and N cThe total number of rows and columns of the grid respectively.

[0064] (2) Extract the attributes of each power cell, where the attributes include intrinsic attributes and external attributes. The intrinsic attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell. The external attributes include the annual GDP, population, and annual average temperature in the region where the power cell is located during the annual forecast, and the prosperity index of the main industry of users in the power cell, the prosperity index of the main upstream industry, and the monthly average temperature in the region during the monthly forecast;

[0065] The change of power cell load is related to many factors, including urban planning, historical load, GDP and population. Considering that the power cell historical load data and related data used for modeling should have the same or similar time granularity and spatial scale. The embodiment of the present invention divides the power cell attributes into two categories: intrinsic attributes and external attributes, which are respectively expressed as follows:

[0066] x in =[X1,X2,X3] (1)

[0067] x ex =[x ex1 ,x ex2 ,x ex3 ] (2)

[0068] Where: x in is the data vector of the inherent attributes of the power cell, X1, X2 and X3 are the distance from the power cell to the surrounding public facilities, the proportion of users in various industries in the power cell, and the historical load of the power cell; x ex is the data vector of the external attributes of the cell. When making annual forecasts, x ex1 ,x ex2 ,x ex3 are the annual GDP, population and annual average temperature of the region where the power cell is located; when making monthly forecasts, x ex1 ,x ex2 ,x ex3 They are respectively the prosperity index of the main industries of users within the power cell, the prosperity index of the main upstream industries, and the monthly average temperature in the region.

[0069] (3) Taking the inherent attributes as the feature vectors of multi-level clustering, the k-means algorithm is used to perform cluster analysis on the power cells to obtain multiple categories of cell sub-sets;

[0070] (4) Based on each cell set, the sparse least squares support vector regression network model is trained to obtain the corresponding load forecasting model.

[0071] Considering that different power cells have different inherent properties, and the inherent properties of power cells have a great influence on the load of power cells, the present invention uses the inherent property data of power cells as the feature vector of multi-level clustering, and uses the k-means algorithm to perform cluster analysis on power cells to obtain multiple categories of cell subsets, as follows:

[0072] First, the embodiment of the present invention is As feature vectors, the k-means algorithm is used to cluster cells into c1 classes that are close in space or have similar spatial functions; then, based on the first-level clustering, As the feature vector, we conduct secondary clustering in the c1 class cells, and further cluster the cells into c2 class (c2>c1) with similar industry characteristics; finally, As the feature vector, three-level clustering is carried out in the c2 class cells, and the cells are further clustered into c3 classes with similar load characteristics. Finally, all cells are divided into c3 classes based on multi-level spatial clustering, and the cells in each class have a high degree of similarity in spatial distribution, industry characteristics, load characteristics, etc. On this basis, load forecasting modeling is carried out for each type of cell, which helps to balance the diversity and aggregation of loads in different cells and better grasp the spatiotemporal distribution characteristics and change trends of loads.

[0073] Furthermore, for each type of cell obtained by multi-level clustering, a load forecasting model is established. A total of c3 different forecasting models need to be constructed. The cell intrinsic attributes and external attributes are used as input, and the cell load is used as output to train each forecasting model. Assume that there are n p cells, and the duration of the sample data corresponding to the cell is T (T>3) time periods (years, months), then the load y of cell C(i,j) in the tkth time period is ij (t k ) and the associated factor x ij (t k )=[x in (t k ),x ex (t k )] together constitute the data set containing (T-3)·n p training samples. The sample set of the load forecasting model corresponding to the p-th cell can be expressed as:

[0074] S p ={[x ij (t k ),y ij (t k )]C(i,j)}∈Categoryp;t k=4,5,6,…,T (3)

[0075] Considering that the model training algorithm must be able to handle high-dimensional data modeling problems in both small sample and large sample cases, the embodiment of the present invention adopts the SLS-SVRNs algorithm (sparse least squares support vector regression network model) to model the load change trends of various types of power cells.

[0076] Based on the sample set of SLS-SVRNs and c3 types of power cells, c3 load forecasting models for specific power cell types can be obtained through training, namely, f p (x in ,x ex ),p=1,2,…,c3. Without considering the uncertainty of input data, these models can be used to obtain the load spatiotemporal distribution prediction results.

[0077] Furthermore, based on the uncertainty of the external attributes of the power cell, a confidence condition is set, and the inherent attributes and external attributes of the power cell are input into the load forecasting model under the confidence condition, and the load of the power cell is predicted, and finally the load spatiotemporal distribution prediction result of the area to be predicted is obtained.

[0078] Specifically, assuming that the prediction results of each cell external attribute include m sampling samples and m confidence levels, and there are a total of p input variables corresponding to the cell external attributes, then each prediction model will have H = m p There are H input scenarios. Correspondingly, for the load of the same cell at a certain time point in the future, there will be H point prediction results. So far, the prediction results of the cell load can be converted into prediction intervals under variable confidence conditions.

[0079]

[0080] in, and Py,h are the sample value and confidence level of the h-th sample respectively.

[0081] It should also be noted that the above step S130 obtains the load prediction value of each power cell, and the load prediction value of each power cell is a point value, which is combined to obtain the load spatiotemporal distribution prediction result of the area to be predicted.

[0082] The present invention discloses a load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information. The method analyzes the correlation of multi-source data based on a hierarchical association model of multi-source spatiotemporal information, correlates user power load data with regional power grid topology, equipment ledger, geographic information, meteorological data and other data to varying degrees, and provides a solid big data foundation for load spatial distribution feature analysis. The load prediction model is further constructed through a clustering algorithm and a sparse least squares support vector regression network model, and the load spatiotemporal distribution prediction result of the to-be-predicted area is obtained based on the correlation of multi-source data, so that the load spatiotemporal distribution prediction result is more accurate.

[0083] Embodiment 2: As attached Figure 6 As shown, the embodiment of the present invention discloses a load spatiotemporal distribution prediction device based on hierarchical association of multi-source spatiotemporal information, including:

[0084] The data resource layer obtains a multi-source data set within the power supply range of the area to be predicted, grids the power supply range of the area to be predicted into equal areas, and records each grid as a power cell. The multi-source data includes the measurement data of the power company, the operation data of the power company, and the external data of the power company.

[0085] The above-mentioned data resource layer configuration interface obtains multi-source data sets from the regional energy Internet big data system, among which the big data platform intelligent distribution network system includes geographic information system, distribution transformer load monitoring system, electricity consumption information collection system, marketing business application system, meteorological information system, social and economic database, etc.

[0086] The data processing layer inputs each power cell into the multi-source spatiotemporal information hierarchical association model to obtain the corresponding inherent attributes and external attributes, wherein the inherent attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell; the external attributes include the annual GDP, population, and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of the user in the power cell during the monthly forecast, the prosperity index of the main upstream industry, and the monthly average temperature in the region; the multi-source spatiotemporal information hierarchical association model includes dividing the multi-source data into corresponding data layers according to their data types, storing the data in each data layer in a grid or network node, and associating different data layers through location correlation to obtain a multi-source spatiotemporal information hierarchical association data set;

[0087] The above data processing layer includes:

[0088] An identification module identifies each multi-source data and determines the attributes of each multi-source data, wherein the attributes include source, format, collection time and geographical coordinates of collection location;

[0089] The metadata construction module maps the attributes of each multi-source data into a standard and unified metadata format to form the corresponding metadata;

[0090] The grid / networking module flattens each metadata into the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and performs network or grid abstraction on each data layer;

[0091] The association module establishes horizontal and vertical associations between different data nodes, edges or grids through a three-dimensional spatial coordinate system, where the X-axis and Y-axis describe the spatial position corresponding to the data, and the Z-axis describes the data's hierarchy. For data nodes, edges or grids in the same data layer, horizontal associations are achieved through the same Z-axis coordinates, and they are distinguished by the differences in the geographic coordinates of the collection locations. For data nodes, edges or grids in different data layers, vertical associations between data are achieved with the geographic coordinates of the collection location as the core, if data drilling is performed on different data layers along the Z-axis.

[0092] The data warehouse layer stores the processing results of the data processing layer.

[0093] At the data application layer, one or more corresponding load forecasting models are selected from the load forecasting model set, and the inherent attributes and external attributes of the power cell are input to obtain the load forecast value of each power cell, thereby obtaining the load spatiotemporal distribution forecast result of the area to be predicted. Each load forecasting model in the load forecasting model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

[0094] The above data application layer includes:

[0095] The model building module includes: obtaining historical load data and related data within the power supply range of the power company, and gridding the power supply range of the power company in equal areas, and recording each grid as a power cell; extracting the attributes of each power cell, where the attributes include inherent attributes and external attributes; using the inherent attributes as the feature vector of multi-level clustering, using the k-means algorithm to perform cluster analysis on the power cells, and obtaining multiple categories of cell sub-sets; training the sparse least squares support vector regression network model based on each cell sub-set to obtain the corresponding load forecasting model;

[0096] The prediction module selects one or more corresponding load prediction models from the load prediction model set, inputs the inherent properties and external properties of the power cell, obtains the load prediction value of each power cell, and thus obtains the load spatiotemporal distribution prediction result of the area to be predicted, wherein each load prediction model in the load prediction model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

[0097] Embodiment 3: The embodiment of the present invention discloses a storage medium, on which a computer program readable by a computer is stored, and the computer program is configured to execute a load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information when running.

[0098] The above storage medium may include, but is not limited to: a USB flash drive, a read-only memory, a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

[0099] Embodiment 4: The embodiment of the present invention discloses an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement a load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information.

[0100] The processor may be a central processing unit (CPU), a general purpose processor, a digital signal processor (DSP), an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. It may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like. The memory may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory, a mobile hard disk, a magnetic disk or an optical disk.

[0101] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0102] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0104] The above technical features constitute the best embodiment of the present invention, which has strong adaptability and best implementation effect. Non-essential technical features can be added or reduced according to actual needs to meet the requirements of different situations.

Claims

1. A load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information, It is characterized in that include: Obtain a multi-source data set within the power supply range of the area to be predicted, grid the power supply range of the area to be predicted into equal areas, and record each grid as a power cell, wherein the multi-source data includes power enterprise measurement data, power enterprise operation data, and power enterprise external data; Input each power cell into the multi-source spatiotemporal information hierarchical association model to obtain the corresponding inherent attributes and external attributes, wherein the inherent attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell; the external attributes include the annual GDP, population and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of the user in the power cell during the monthly forecast, the prosperity index of the main upstream industry and the monthly average temperature in the region; the multi-source spatiotemporal information hierarchical association model includes dividing the multi-source data into corresponding data layers according to their data types, storing the data in each data layer in grids or network nodes, and associating different data layers through location correlation to obtain a multi-source spatiotemporal information hierarchical association data set; One or more corresponding load forecasting models are selected from the load forecasting model set, and the inherent attributes and external attributes of the power cell are input to obtain the load forecast value of each power cell, thereby obtaining the load spatiotemporal distribution prediction result of the area to be predicted, wherein each load forecasting model in the load forecasting model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

2. The load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information according to claim 1, It is characterized in that The execution process of the multi-source spatiotemporal information hierarchical association model includes: Identify each multi-source data and determine the attributes of each multi-source data, wherein the attributes include source, format, collection time, geographical coordinates of collection location and related information; Map the attributes of each multi-source data into a standard and unified metadata format to form corresponding metadata; Each metadata is flattened into the corresponding data layer according to the geographic coordinates of the collection location in the attributes, and each data layer is abstracted in a network or grid manner, and the corresponding geographic coordinates of the collection location or time series data are stored in each node, edge or grid.

3. The load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information according to claim 2, It is characterized in that The method of tiling each metadata to the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and abstracting each data layer in a network or grid manner, and storing the corresponding geographical coordinates of the collection location or time series data in each node, edge or grid, includes: Each metadata is flattened into the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and each data layer is networked or gridded for abstraction; The horizontal and vertical associations between different data nodes, edges or grids are established through the three-dimensional spatial coordinate system, where the X-axis and Y-axis describe the spatial position corresponding to the data, and the Z-axis describes the data's hierarchy. For data nodes, edges or grids in the same data layer, horizontal associations are achieved through the same Z-axis coordinates, and they are distinguished by the differences in the geographic coordinates of the collection locations. For data nodes, edges or grids in different data layers, vertical associations between data are achieved with the geographic coordinates of the collection location as the core. If data drilling is performed on different data layers along the Z-axis.

4. The load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information according to any one of claims 1 to 3, It is characterized in that The process of constructing the load forecasting model set includes: Obtain historical load data and related data within the power supply range of the power company, grid the power supply range of the power company in equal areas, and record each grid as a power cell; Extract the attributes of each power cell, where the attributes include intrinsic attributes and external attributes. The intrinsic attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell. The external attributes include the annual GDP, population and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of users in the power cell, the prosperity index of the main upstream industry and the monthly average temperature in the region during the monthly forecast; The inherent attributes are used as the feature vectors of multi-level clustering, and the k-means algorithm is used to perform cluster analysis on the power cells to obtain multiple categories of cell sub-sets; The sparse least squares support vector regression network model is trained based on each cell set to obtain the corresponding load forecasting model.

5. The load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information according to any one of claims 1 to 3, It is characterized in that Based on the uncertainty of the external attributes of the power cell, a confidence condition is set, and the inherent attributes and external attributes of the power cell are input into the load forecasting model under the confidence condition to forecast the load of the power cell, and finally obtain the load spatiotemporal distribution forecast result of the area to be predicted.

6. A load spatiotemporal distribution prediction device based on hierarchical association of multi-source spatiotemporal information, using the method as claimed in any one of claims 1 to 5, It is characterized in that include: The data resource layer obtains a multi-source data set within the power supply range of the area to be predicted, grids the power supply range of the area to be predicted into equal areas, and records each grid as a power cell. The multi-source data includes the measurement data of the power enterprise, the operation data of the power enterprise, and the external data of the power enterprise; The data processing layer inputs each power cell into the multi-source spatiotemporal information hierarchical association model to obtain the corresponding inherent attributes and external attributes, wherein the inherent attributes include the distance from the power cell to the surrounding public facilities, the proportion of users in each industry in the power cell, and the historical load of the power cell; the external attributes include the annual GDP, population, and annual average temperature in the region where the power cell is located during the annual forecast, the prosperity index of the main industry of the user in the power cell during the monthly forecast, the prosperity index of the main upstream industry, and the monthly average temperature in the region; the multi-source spatiotemporal information hierarchical association model includes dividing the multi-source data into corresponding data layers according to their data types, storing the data in each data layer in a grid or network node, and associating different data layers through location correlation to obtain a multi-source spatiotemporal information hierarchical association data set; The data warehouse layer stores the processing results of the data processing layer; At the data application layer, one or more corresponding load forecasting models are selected from the load forecasting model set, and the inherent attributes and external attributes of the power cell are input to obtain the load forecast value of each power cell, thereby obtaining the load spatiotemporal distribution forecast result of the area to be predicted. Each load forecasting model in the load forecasting model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

7. The load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information according to claim 6, It is characterized in that The data processing layer includes: An identification module identifies each multi-source data and determines the attributes of each multi-source data, wherein the attributes include source, format, collection time and geographical coordinates of collection location; The metadata construction module maps the attributes of each multi-source data into a standard and unified metadata format to form the corresponding metadata; The grid / networking module flattens each metadata into the corresponding data layer according to the geographical coordinates of the collection location in the attribute, and performs network or grid abstraction on each data layer; The association module establishes horizontal and vertical associations between different data nodes, edges or grids through a three-dimensional spatial coordinate system, where the X-axis and Y-axis describe the spatial position corresponding to the data, and the Z-axis describes the data's hierarchy. For data nodes, edges or grids in the same data layer, horizontal associations are achieved through the same Z-axis coordinates, and they are distinguished by the differences in the geographic coordinates of the collection locations. For data nodes, edges or grids in different data layers, vertical associations between data are achieved with the geographic coordinates of the collection location as the core, if data drilling is performed on different data layers along the Z-axis.

8. The load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information according to claim 6 or 7, It is characterized in that The data application layer includes: The model building module includes: obtaining historical load data and related data within the power supply range of the power company, and gridding the power supply range of the power company in equal areas, and recording each grid as a power cell; extracting the attributes of each power cell, where the attributes include inherent attributes and external attributes; using the inherent attributes as the feature vector of multi-level clustering, using the k-means algorithm to perform cluster analysis on the power cells, and obtaining multiple categories of cell sub-sets; training the sparse least squares support vector regression network model based on each cell sub-set to obtain the corresponding load forecasting model; The prediction module selects one or more corresponding load prediction models from the load prediction model set, inputs the inherent properties and external properties of the power cell, obtains the load prediction value of each power cell, and thus obtains the load spatiotemporal distribution prediction result of the area to be predicted, wherein each load prediction model in the load prediction model set is obtained by training the sparse least squares support vector regression network model using historical load data and related data.

9. A storage medium, It is characterized in that The storage medium stores a computer program that can be read by a computer, and the computer program is configured to execute the load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information as described in any one of claims 1 to 5 when running.

10. An electronic device, It is characterized in that The method comprises a processor and a memory, wherein a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the load spatiotemporal distribution prediction method based on hierarchical association of multi-source spatiotemporal information as described in any one of claims 1 to 5.

Images