Power distribution network distributed power supply in-situ operation optimization method based on dimension raising linear mapping

Abstract

The invention relates to a power distribution network distributed power supply local operation optimization method and device based on dimension raising linear mapping, computer equipment, a computer readable storage medium and a computer program product, and relates to the field of distributed power supply control. The method comprises the following steps: acquiring historical operation data of a power distribution network in a target area at different moments and partition information of the target area; taking the node power data corresponding to the power distribution network at each moment as training input data, taking each distributed power supply reactive power control strategy of the power distribution network at the corresponding moment as training output data, and training a pre-constructed distributed power supply local operation control model; and according to the partition information of the target area, decomposing the trained distributed power supply local operation control model to obtain sub-models corresponding to each sub-area in the target area. According to the method, distribution line parameters are not needed, and the reactive power control strategy of the distributed power supply can be quickly obtained only through on-site limited data and a simple matrix.

Description

Technical Field

[0001] The present application relates to the field of distributed power supply control technology, and in particular to a method for optimizing the on-site operation of distributed power supplies in a distribution network based on dimensionality-increasing linear mapping. Background Art

[0002] In recent years, an increasing number of distributed power sources have been connected to distribution networks. Due to the intermittent and uncertain nature of distributed power sources, large-scale access to these sources can easily lead to problems such as power reverse flow and voltage over-limit in the distribution system, seriously impacting the operation of the distribution network.

[0003] Physical-based methods for localized operation of distributed generation (DG) rely primarily on precise distribution line parameters to formulate DG reactive power control strategies. However, in actual operation, these parameters are often inaccurate. Machine learning-based methods for localized operation of DG often require lengthy training cycles, making rapid deployment difficult.

[0004] Therefore, how to accurately and quickly control the on-site operation of distributed power sources is an urgent problem to be solved. Summary of the Invention

[0005] Based on this, it is necessary to provide a distribution network distributed power on-site operation optimization method based on dimensionality-increasing linear mapping, which can accurately and quickly realize the control of distributed power on-site operation in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping, comprising:

[0007] Obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0008] The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment is used as training output data to train the pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on the dimensionality-increasing linear mapping method;

[0009] According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0010] In one embodiment, a pre-built on-site operation control model of a distributed power source is trained, including: for the training input data corresponding to each moment, according to the basis vector corresponding to at least one lifting dimension, the dimension expansion of the training input data is performed to obtain the dimension-lifted expansion vector of the training input data; data fusion is performed on the training input data and the dimension-lifted expansion vector corresponding to the corresponding training input data to obtain the fused input data of the training input data; the fused input data of the training input data is used as the model independent variable, and the training output data corresponding to the corresponding training input data is used as the model dependent variable, to determine the dimension-lifted linear mapping matrix of the pre-built on-site operation control model of the distributed power source, so as to train the pre-built on-site operation control model of the distributed power source.

[0011] In one embodiment, the zoning information of the target area includes the access location of the distributed power source and the zoning information of the distribution network; accordingly, according to the zoning information of the target area, the trained on-site operation control model of the distributed power source is decomposed to obtain sub-models corresponding to each sub-area in the target area, including: according to the access location of the distributed power source, splitting the increased-dimensional linear mapping matrix into increased-dimensional linear mapping sub-matrices that match the zoning information of the distribution network; wherein the increased-dimensional linear mapping sub-matrix is ​​the increased-dimensional linear mapping matrix of the sub-model corresponding to the sub-area where the distribution network zoning information is located.

[0012] In one embodiment, the training input data is dimensionally expanded according to a basis vector corresponding to at least one lifting dimension to obtain a dimension-upgraded expanded vector of the training input data, including: for each lifting dimension, the training input data is first dimensionally expanded according to the basis vector corresponding to the lifting dimension to obtain an initial expanded vector of the training input data in the lifting dimension; and the initial expanded vectors in different lifting dimensions are combined to obtain the dimension-upgraded expanded vector of the training input data.

[0013] In one embodiment, the node power data includes injected active power and load reactive power; the injected active power includes the active output of the distributed power source and the active power demand of the load.

[0014] In one of the embodiments, based on the node power data of each node in the sub-region at any target moment, the reactive power control strategy for the distributed power supply in the sub-region at the corresponding target moment is determined, including: for each sub-region, obtaining the real node power data of each node in the sub-region at the target moment and the estimated node power data of each node in other sub-regions; generating input data corresponding to the sub-region based on the real node power data and the estimated node power data of each node in other sub-regions; based on the basis vector corresponding to at least one lifted dimension, dimensionality expansion of the input data is performed to obtain a lifted dimension expansion vector of the input data; data fusion is performed on the input data and the lifted dimension expansion vector corresponding to the input data to obtain fused input data of the input data; the fused input data of the input data is input into the sub-model corresponding to the sub-region to determine the reactive power control strategy for the distributed power supply in the sub-region.

[0015] In a second aspect, the present application further provides a device for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping, comprising:

[0016] An acquisition module is used to obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0017] The training module is used to train a pre-built local operation control model for distributed power sources, using the node power data corresponding to the distribution network at each moment as training input data and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment as training output data. The local operation control model for distributed power sources is implemented based on the dimensionality-increasing linear mapping method.

[0018] The decomposition module is used to decompose the trained distributed power supply on-site operation control model according to the partition information of the target area to obtain the sub-model corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0019] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0020] Obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0021] The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment is used as training output data to train the pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on the dimensionality-increasing linear mapping method;

[0022] According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0023] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0024] Obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0025] The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment is used as training output data to train the pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on the dimensionality-increasing linear mapping method;

[0026] According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0027] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0028] Obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0029] The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment is used as training output data to train the pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on the dimensionality-increasing linear mapping method;

[0030] According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0031] The above-mentioned method, device, computer equipment, computer-readable storage medium and computer program product for optimizing the on-site operation of distributed power sources in the distribution network based on dimensionality-increasing linear mapping, on the one hand, utilize the historical operation data of the distribution network in the target area at different times and the partition information of the target area, that is, the limited on-site data, to train the sub-models corresponding to each sub-area in the target area, and then utilize the sub-models to obtain the reactive power control strategies of the distributed power sources in each sub-area. The above-mentioned process does not require the distribution line parameters, and therefore, it can avoid the problem of being unable to accurately control the on-site operation of the distributed power sources due to the poor accuracy of the distribution line parameters; on the other hand, the on-site operation control model of the distributed power sources is implemented based on the dimensionality-increasing linear mapping method. Therefore, by utilizing the on-site operation control model of the distributed power sources constructed based on the dimensionality-increasing linear mapping method, that is, by utilizing a simple matrix, the nonlinear coupling relationship between the training input data and the training output data can be quickly fitted, thereby quickly obtaining the reactive power control strategies of the distributed power sources in each sub-area. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 1 is a flow chart of a method for optimizing on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping in one embodiment;

[0034] Figure 2 A schematic flow chart of the training steps of a pre-built distributed power source on-site operation control model in one embodiment;

[0035] Figure 3Schematic diagram of a process for expanding the dimension of training input data in one embodiment;

[0036] Figure 4 1 is a flow chart of steps for determining a reactive power control strategy for a distributed power source in one embodiment;

[0037] Figure 5 Schematic diagram of a flow chart of a method for optimizing on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping in another embodiment;

[0038] Figure 6 Schematic diagram of the structure of each node in the target area distribution network in one embodiment;

[0039] Figure 7 is a day-ahead forecast curve of source-load output corresponding to the target area distribution network in one embodiment;

[0040] Figure 8 Schematic diagram showing comparison of node voltage amplitudes between solution 1 and solution 2 in one embodiment;

[0041] Figure 9 A schematic diagram comparing node voltage amplitudes of Solution 2 and Solution 3 in one embodiment;

[0042] Figure 10 1. A structural block diagram of an on-site operation optimization device for distributed power sources in a distribution network based on dimensionality-increasing linear mapping in one embodiment;

[0043] Figure 11 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0045] In an exemplary embodiment, Figure 1 As shown, a method for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping is provided, which is applied to a main computing device in a target area. The method includes the following steps S110 to S130.

[0046] S110: Acquire historical operation data of the distribution network in the target area at different times and zoning information of the target area; the historical operation data includes node power data of each node in the distribution network and reactive power control strategy for each distributed power source in the distribution network.

[0047] The target area may include multiple sub-areas. Each sub-area may include at least one distribution network node. The distribution network node may be connected to at least one distributed power source and / or at least one load.

[0048] The historical operation data at different historical moments can be understood as the node power data generated by each node in the distribution network at different historical moments and the reactive power control strategy of each distributed power source in the distribution network.

[0049] The different historical moments can be understood as different moments within the target historical time period. For example, the different historical moments can be different moments of the day.

[0050] The target historical time period can be selected from at least one candidate historical time period with the same or similar historical operating data. For example, assuming that the historical operating data for the past five consecutive days is similar, any day can be selected as the target historical time period. For example, the first day can be selected as the target historical time period.

[0051] The node power data can be understood as the injection data of active power and reactive power at each distribution network node in the distribution network.

[0052] The reactive power control strategy for each distributed power source in the distribution network can be understood as the reactive power output by each distributed power source in the distribution network for reducing the voltage fluctuation range of the distribution network.

[0053] The zoning information of the target area may include distribution network zoning information and the access location of the distributed power source in each sub-area.

[0054] Optionally, the distribution network partition information may include the number of sub-areas, and the number of sub-areas may be predetermined.

[0055] The access location of the distributed power source can be understood as the sub-area to which the distributed power source is connected and / or the node of the sub-area to which the distributed power source is connected. Optionally, the access location of the distributed power source can be represented by the identifier of the sub-area to which the distributed power source is connected and / or the identifier of the node of the sub-area to which the distributed power source is connected.

[0056] In an optional embodiment, the main computing device of the target area can obtain historical operating data and the partition information of the target area from a device information database or a power enterprise data platform. It should be noted that this application does not specifically limit the method of obtaining the historical operating data and the partition information of the target area.

[0057] S120: Using the node power data corresponding to the distribution network at each moment as training input data, and using the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment as training output data, a pre-built distributed power source on-site operation control model is trained; the distributed power source on-site operation control model is implemented based on the dimensionality-increasing linear mapping method.

[0058] Among them, the distributed power supply on-site operation control model can be understood as the mapping relationship between the node power data corresponding to the distribution network at each moment and the reactive power control strategy of each distributed power supply in the distribution network at the corresponding moment.

[0059] The node power data corresponding to the distribution network at each moment and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment can be used as a pair of training data. At least one pair of training data can be obtained based on the historical operation data at different moments.

[0060] Optionally, the number of pairs of training data may be preset. The preset number of pairs of training data may be set or adjusted by technicians based on needs or experience, or determined repeatedly through a large number of experiments, and this application does not impose any limitation on this.

[0061] Among them, the dimensionality-increasing linear mapping method can be understood as mapping data from low-dimensional space to high-dimensional space, enhancing the expressiveness of data by introducing new features to express the nonlinear relationship between input data and output data in high-dimensional space.

[0062] S130: Decompose the trained distributed power on-site operation control model according to the partition information of the target area to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0063] Among them, the trained distributed power supply on-site operation control model is decomposed according to the partition information of the target area. It can be understood that the mapping relationship between input data and output data is decomposed according to the partition information of the target area to obtain the sub-mapping relationship between sub-input data and sub-output data corresponding to each sub-area, that is, the sub-model corresponding to each sub-area.

[0064] In the above-mentioned method for optimizing the on-site operation of distributed power sources in the distribution network based on dimensionality-increasing linear mapping, on the one hand, after training the sub-models corresponding to each sub-region in the target area by utilizing the historical operation data of the distribution network in the target area at different times and the partition information of the target area, the distributed power reactive power control strategy of each sub-region is obtained by utilizing the sub-model. The above process does not require the distribution line parameters. Therefore, it can avoid the problem that the on-site operation of the distributed power source cannot be accurately controlled due to the poor accuracy of the distribution line parameters; on the other hand, the on-site operation control model of the distributed power source is implemented based on the dimensionality-increasing linear mapping method. Therefore, by utilizing the on-site operation control model of the distributed power source constructed based on the dimensionality-increasing linear mapping method, the linear coupling relationship between the training input data and the training output data can be quickly fitted, so that the distributed power reactive power control strategy of each sub-region can be quickly obtained, and the on-site operation of the distributed power source in the sub-region can be quickly controlled.

[0065] On the basis of the technical solutions of the above embodiments, the present application also provides an optional embodiment, in which the training step of the pre-built distributed power source on-site operation control model in S120 is refined.

[0066] See also Figure 2 The training steps for the pre-built distributed power generation local operation control model shown include:

[0067] S210: For the training input data corresponding to each moment, perform dimension expansion on the training input data according to the basis vector corresponding to at least one lifted dimension to obtain a lifted dimension expanded vector of the training input data.

[0068] The lifting dimensions may be preset. In an optional embodiment, the number of lifting dimensions may be preset. The preset number of lifting dimensions may be set or adjusted by technicians based on needs or experience, or determined repeatedly through extensive testing, and this application does not impose any limitation on this.

[0069] In an optional embodiment, the historical operation data at different moments may include historical operation data at H moments, where H is a positive integer greater than 1. The historical operation data at each moment includes the training input data and the training output data at that moment.

[0070] The training input data corresponding to each moment can include N n elements; where N n is a positive integer greater than 1.

[0071] For the training input data corresponding to each moment, the dimension of the training input data corresponding to each moment can be expanded based on the basis vector of at least one lifted dimension corresponding to each element in the training input data to obtain the lifted dimension expanded vector of the training input data corresponding to each moment.

[0072] S220: Perform data fusion on the training input data and the dimension-increased expansion vector corresponding to the corresponding training input data to obtain fused input data of the training input data.

[0073] In an optional embodiment, for each moment, the training input data at that moment and the corresponding dimensionality-increased expansion vector may be concatenated to obtain concatenated input data of the training input data, that is, fused input data at each moment.

[0074] S230: Using the fused input data of the training input data as the model independent variable and the training output data corresponding to the corresponding training input data as the model dependent variable, determine the dimensionality-increasing linear mapping matrix of the pre-built distributed power on-site operation control model to train the pre-built distributed power on-site operation control model.

[0075] In an optional embodiment, the fused input data at each moment may be used as the model independent variable, and the training output data corresponding to the corresponding training input data may be used as the model dependent variable to train the pre-built distributed power supply on-site operation control model.

[0076] In an embodiment of the present application, for the training data corresponding to each moment, the basis vector corresponding to at least one lifting dimension can be used to fully expand the dimension of the training, so as to map the data from the low-dimensional space to the high-dimensional space, thereby more accurately representing the nonlinear relationship between the input data and the output data in the high-dimensional space.

[0077] Based on the technical solutions of the above embodiments, the present application also provides an optional embodiment, in which the zoning information of the target area is refined into the distributed power supply access location and the distribution network zoning information. Accordingly, based on the zoning information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area within the target area, including: based on the distributed power supply access location, splitting the increased-dimensional linear mapping matrix into increased-dimensional linear mapping sub-matrices that match the distribution network zoning information; wherein the increased-dimensional linear mapping sub-matrix is ​​the increased-dimensional linear mapping matrix corresponding to the sub-model of the sub-area where the distribution network zoning information is located.

[0078] In an optional embodiment, the distribution network partition information may include the number of sub-regions. The distributed power supply access location may include identification information of the sub-region to which the distributed power supply is connected. Splitting the increased-dimensional linear mapping matrix into increased-dimensional linear mapping sub-matrices that match the distribution network partition information based on the distributed power supply access location may include: splitting the increased-dimensional linear mapping matrix into increased-dimensional linear mapping sub-matrices corresponding to each sub-region that matches the number of sub-regions based on the identification information of the sub-region to which the distributed power supply is connected.

[0079] For example, the three distributed generation (DG) sources are DG1, DG2, and DG3. DG1 is connected to partition 1; DG2 and DG3 are connected to partition 2. Assume that the resulting dimensionality-increasing linear mapping matrix is ​​a matrix with 3 rows and 100 columns. The first row of the dimensionality-increasing linear mapping matrix corresponds to DG1, the second row corresponds to DG2, and the third row corresponds to DG3. Therefore, based on the first row of the dimensionality-increasing linear mapping matrix, the first dimensionality-increasing linear mapping submatrix corresponding to partition 1 can be obtained. Based on the second and third rows of the dimensionality-increasing linear mapping matrix, the second dimensionality-increasing linear mapping submatrix corresponding to partition 2 can be obtained.

[0080] In an embodiment of the present application, by splitting the dimensionality-raising linear mapping matrix into dimensionality-raising linear mapping sub-matrices that match the distribution network partition information, a distributed power supply on-site operation control model corresponding to each sub-region can be obtained, so that each sub-region can control the on-site operation of the distributed power supply within the region separately, thereby improving the on-site operation optimization efficiency.

[0081] Based on the technical solutions of the above embodiments, the present application also provides an optional embodiment, in which the dimension expansion step of the training input data in S210 is refined.

[0082] See also Figure 3 The steps for expanding the dimension of the training input data shown include:

[0083] S310: For each lifting dimension, perform first-dimensional expansion on the training input data according to the basis vector corresponding to the lifting dimension to obtain an initial expansion vector of the training input data under the lifting dimension.

[0084] The above training input data can be understood as the training input data corresponding to each moment.

[0085] In an optional embodiment, for each lifting dimension, the difference between the jth element of the training input data and the jth element of the basis vector corresponding to the lifting dimension can be determined; j is greater than or equal to 1 and less than or equal to the number of elements N in the training input data. nThen determine N n The sum of squares of the differences, and N n Finally, N n The sum of squares of the differences, and N n The product of the logarithm of the sum of the differences is used as the initial expansion vector of the training input data under the lifting dimension.

[0086] S320: Combining the initial expansion vectors under different lifting dimensions to obtain the dimension-lifted expansion vector of the training input data.

[0087] In an optional embodiment, the initial expanded vectors under different lifting dimensions are spliced ​​and the spliced ​​vectors are transposed to obtain the dimension-lifted expanded vectors of the training input data.

[0088] In an embodiment of the present application, for the training data corresponding to each moment, the first dimension of the training input data is expanded according to the basis vector corresponding to the lifted dimension, thereby achieving accurate dimensionality increase and expansion of the training input data.

[0089] Based on the technical solutions of the above embodiments, the present application also provides an optional embodiment, in which the node power data can be refined into injected active power and load reactive power; the injected active power can be refined into distributed power source active output and load active demand.

[0090] The node power data may include the active power injected into the distribution network node and the reactive power of the load.

[0091] The active power injected into a distribution network node can be understood as the sum of the active power output of distributed generation (DGs) and the active power demand of the load. The active power output of DGs can be understood as the actual power output generated by DGs during operation. The active power demand of the load can be understood as the total amount of actual effective power that electrical devices in the power system require from the grid per unit time.

[0092] Load reactive power can be understood as the power supplied to the load by the power source in an AC circuit. Load reactive power is used to exchange electric and magnetic fields in the circuit, establishing and maintaining the magnetic field in electrical equipment, but it does not directly perform external work.

[0093] In the embodiment of the present application, based on the active output of the distributed power source, the active power demand of the load and the reactive power of the load, the operating status of the distribution network under the control of the historical distributed power source reactive power control strategy can be more comprehensively and accurately reflected, thereby training the constructed distributed power source on-site operation control model, and achieving a more accurate distributed power source on-site operation control model.

[0094] Based on the technical solutions of the above embodiments, the present application also provides an optional embodiment, in which the steps for determining the reactive power control strategy of the distributed power supply are refined.

[0095] See also Figure 4 The steps for determining the reactive power control strategy of the distributed generation shown include:

[0096] S410: For each sub-area, obtain the actual node power data of each node in the sub-area and the estimated node power data of each node in other sub-areas at the target time.

[0097] The actual node power data may be understood as the actual measured node power data.

[0098] In an optional embodiment, the real node power data can be obtained by the edge computing device corresponding to the sub-area by measuring the data of each node in the sub-area.

[0099] The estimated node power data can be understood as the estimated node power of each node at the target time. It should be understood that for data security reasons, the actual node power data of each node in each sub-area cannot be transmitted to each other. Therefore, the estimated node power data is used here instead of the actual node power data.

[0100] Optionally, the estimated node power data can be estimated by technical personnel based on needs, experience, or a large number of experiments, and this application does not impose any restrictions on this. Or optionally, the estimated node power data can be obtained by using a node power data estimation model. Among them, the node power data estimation model can be implemented based on a traditional machine learning model or a deep learning model, and this application does not impose any restrictions on the specific network structure of the node power data estimation model. Exemplarily, the node power data estimation model can be implemented based on the TiSASRec (Time Interval Aware Self-Attention for Sequential Recommendation) algorithm. It should be noted that this application does not impose any restrictions on the method of obtaining the estimated node power data.

[0101] In an optional embodiment, the estimated node power data corresponding to each sub-area may be obtained by the main computing device and then distributed to each sub-area.

[0102] S420: Generate input data corresponding to the sub-area based on the actual node power data and the estimated node power data of each node in other sub-areas.

[0103] In an optional embodiment, for each sub-region, the actual node power data of the sub-region and the estimated node power data of each node in other sub-regions may be fused to generate input data corresponding to the sub-region.

[0104] S430: Based on the basis vector corresponding to at least one lifted dimension, perform dimension expansion on the input data to obtain a dimension-lifted expanded vector of the input data.

[0105] In an optional embodiment, the input data may include N n elements; where N n is a positive integer greater than 1. For each lifting dimension, the difference between the jth element of the input data and the jth element of the basis vector corresponding to the lifting dimension can be determined; j is greater than or equal to 1 and less than or equal to N n Then determine N n The sum of squares of the differences, and N n Finally, N n The sum of squares of the differences, and N n The product of the logarithms of the sum of the differences is used as the initial expansion vector of the input data in the lifted dimension. The initial expansion vectors in different lifted dimensions are concatenated and the concatenated vectors are transposed to obtain the dimension-lifted expansion vector of the input data.

[0106] S440: Perform data fusion on the input data and the dimension-increased expansion vector corresponding to the input data to obtain fused input data of the input data.

[0107] In an optional embodiment, the input data and the corresponding dimension-increased expansion vector may be concatenated to obtain concatenated input data, that is, fused input data.

[0108] S450: Inputting the fused input data of the input data into the sub-model corresponding to the sub-region to determine the reactive power control strategy of the distributed power supply in the sub-region.

[0109] Among them, the reactive power control strategy of distributed power sources in a sub-area is only effective for reactive power control of distributed power sources in this sub-area.

[0110] In an embodiment of the present application, for each sub-area, the reactive power control strategy of the distributed power source in the sub-area is determined based on the actual node power data of each node in the sub-area at the target moment and the estimated node power data of each node in other sub-areas, which can ensure the coordination of the reactive power control of the distributed power source in each sub-area within the target area.

[0111] Based on the technical solutions of the above embodiments, the present application also provides an optional embodiment, in which a method for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping is described in detail.

[0112] See also Figure 5 The flowchart of the method for optimizing the local operation of distributed power generation in a distribution network based on dimensionality-increasing linear mapping is shown, including:

[0113] S501: For the distribution network in the selected target area, obtain the network parameter information of the target area, the historical operation data of the distribution network at different times, the real-time measurement data of each section of the distribution network, and the day-ahead source and load output forecast information of each section of the distribution network.

[0114] The network parameter information of the target area includes: the access location of the distributed power supply and the distribution network partition information.

[0115] The historical operating data of the distribution network at different times can be understood as the historical operating data at different times on multiple typical days. A typical day can be understood as any time period among multiple candidate time periods with the same or similar corresponding historical operating data. For example, a typical day can be any day among multiple days with the same or similar corresponding historical operating data.

[0116] Historical operating data includes the active power injected into distribution network nodes, the reactive power of loads, and the reactive power control strategies of each distributed generation in the distribution network. The active power injected into distribution network nodes includes the sum of the active power output of distributed generation and the active power demand of loads.

[0117] Among them, the day-ahead forecast information of source-load output in each section of the distribution network includes the output forecast curve of distributed photovoltaics and the load fluctuation forecast curve.

[0118] The real-time measurement data for each distribution network segment includes real-time measurement data of node-injected active power and real-time measurement data of load reactive power. Real-time measurement data of node-injected active power includes the sum of real-time measurement data of distributed generation active power output and real-time measurement data of load active power demand.

[0119] S502: Set the current time to t=0, the control step to △t, and the total control time to T.

[0120] S503: Constructing a training set corresponding to a distributed generation local operation control model based on historical operation data of the distribution network in the target area at different times.

[0121] The training set includes training objects corresponding to different time instants. Each training object includes training input data and training output data. The training input data is the node power data corresponding to the distribution network at each time instant; the training output object is the reactive power control strategy for each distributed generation in the distribution network at that time instant.

[0122] In an optional embodiment, the training set of the constructed distributed power supply on-site operation control model can be expressed as:

[0123] (1)

[0124] In formula (1), and Respectively represent the training input data and training output data in the training object; For the The input variables of the training object are as follows: The nodes in the distribution network inject active power and load reactive power at all times; express Time Node The nodes inject active power; and Respectively Time Node The active power output of distributed generation and the active power demand of load; express Time Node The reactive power of the load; For the The output variables of the group training object include Reactive power control strategy of each distributed power source at the moment; express The distribution network at the moment Reactive power control strategy for distributed power sources; is the number of training subjects; The number of input variables for the training object; is the number of distributed power sources.

[0125] S504: Using the training set, a pre-built distributed power source on-site operation control model is trained; the distributed power source on-site operation control model is implemented based on a dimensionality-increasing linear mapping method.

[0126] In an optional embodiment, the distributed power supply local operation control model constructed based on the dimensionality-increasing linear mapping method can be expressed as:

[0127] (2)

[0128] In formula (2), is the transpose of the matrix; For input data; To output data; is the dimension-raising linear mapping matrix; for The dimension-increasing expansion vector of .

[0129] In an optional embodiment, the trained distributed power supply on-site operation control model is expressed as:

[0130] (3)

[0131] In formula (3), is the transpose of the matrix; is the Moore-Penrose inverse of the matrix; and Represent the training input data and training output data respectively; Represents the dimension-increased expanded vector of the training input data; For the Group training output data; For the The dimension-increased expansion vector of the group training objects; Indicates the The first set of training input data Dimensionality increase and expansion of vector; For the The first set of training input data elements; It is The first of the basis vectors elements; is the number of training subjects; The number of elements of input data for training; To increase the number of dimensions.

[0132] S505: Decomposing the trained distributed power supply local operation control model according to the partition information of the target area to obtain sub-models corresponding to each sub-area in the target area.

[0133] In an optional embodiment, based on the partition information of the target area, the trained distributed power supply local operation control model is decomposed and executed according to the following formula:

[0134] (4)

[0135] In formula (4), is the transpose of the matrix; For input data; To output data; is the dimension-raising linear mapping matrix; for The dimension-increasing expansion vector of ; Sub-area of ​​the distribution network The dimension-raising linear mapping submatrix of ; Sub-area of ​​the distribution network The output matrix of the sub-region The output matrix consists of sub-regions Reactive power control strategy of distributed power generation; is the number of sub-regions.

[0136] S506: Based on the real-time measurement data of each distribution network partition at the current moment and the source-load output forecast information of each distribution network partition at the current moment, determine the reactive power control strategy for the distributed power sources in the sub-area at the current moment, and perform reactive power control on the distributed power sources in the sub-area according to the distributed power source reactive power control strategy.

[0137] In an optional embodiment, the reactive power control strategy of the distributed generation in the sub-region at the current moment can be expressed as:

[0138] (5)

[0139] In formula (5), is the transpose of the matrix, Zoning the distribution network The dimension-raising linear mapping matrix of ; for Time distribution network partition The output data, that is, Time distribution network partition Reactive power control strategy of distributed power generation; for Time distribution network partition Input data; for The dimension-increasing expansion vector of ; for Time distribution network partition The node injects real-time measurement data of active power and load reactive power; Determined based on the day-ahead forecast information of source and load output Time distribution network partition The node injected active power and node injected reactive power; among them, ; is the number of sub-regions.

[0140] S507: Update the current time according to the control step size.

[0141] S508: Determine whether the current time is greater than the total control time; if less than or equal to, repeat the step of S506; if greater than, end.

[0142] The present embodiment also provides a comparison of the on-site operation voltage optimization results with two other solutions. See Table 1 for the comparison of the on-site operation voltage optimization results of the distribution network under different solutions.

[0143] Table 1

[0144]

[0145] Among them, the first option is to obtain the distribution network operation data without optimizing the reactive power control strategy of the distributed power generation in the target area distribution network. The second option is to obtain the distribution network operation data after optimizing the reactive power control strategy of the distributed power generation in the target area distribution network using the distribution network distributed power local operation optimization method based on dimensionality-increasing linear mapping of this application. The third option is to obtain the theoretically optimal distribution network operation data after optimizing the reactive power control strategy of the distributed power generation in the target area distribution network using a real-time centralized method.

[0146] The computer hardware environment for performing the calculation is Intel(R) Core(TM) i7-12700 CPU with a main frequency of 2.10GHz and a memory of 32GB; the software environment is Windows 11 operating system.

[0147] Table 2 shows the distributed photovoltaic access location, capacity and partition information of each node in the distribution network of the target area.

[0148] Table 2

[0149]

[0150] Figure 6 A schematic diagram of the structure of each node in the target area distribution network is shown. Figure 7 The day-ahead forecast curve of source and load output corresponding to the target area distribution network is shown. Figure 8 A schematic diagram comparing the node voltage amplitudes of Scheme 1 and Scheme 2 is shown. Figure 9 A schematic diagram comparing the node voltage amplitudes of Scheme 2 and Scheme 3 is shown.

[0151] A comparison of Schemes 1 and 2 shows that, without any control strategy, a high proportion of distributed generation (DGs) will lead to significant system voltage fluctuations. By optimizing the DG reactive power control strategy using Scheme 2, each DG can adjust reactive power compensation in real time, maintaining the system voltage at a safe operating level. A comparison of Schemes 2 and 3 shows that Scheme 2 achieves an optimality rate of 98.61%, achieving comparable operational optimization results to centralized control methods. Formulating a DG reactive power control strategy in Scheme 3 requires collecting global information and solving complex optimization problems, and the effectiveness of the control relies on precise network parameters. Scheme 2, on the other hand, is data-driven, requiring only limited local data and simple matrix calculations to formulate a DG reactive power control strategy, making it more suitable for scenarios with rapid voltage fluctuations.

[0152] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0153] Based on the same inventive concept, an embodiment of the present application further provides a device for optimizing the on-site operation of distributed power sources in a distribution network based on an ascending-dimensional linear mapping, which is used to implement the above-mentioned method for optimizing the on-site operation of distributed power sources in a distribution network based on an ascending-dimensional linear mapping. The implementation solution provided by the device for solving the problem is similar to the implementation solution described in the above-mentioned method. Therefore, the specific limitations of one or more embodiments of the device for optimizing the on-site operation of distributed power sources in a distribution network based on an ascending-dimensional linear mapping provided below can be referred to the limitations of the method for optimizing the on-site operation of distributed power sources in a distribution network based on an ascending-dimensional linear mapping above, and will not be repeated here.

[0154] In an exemplary embodiment, Figure 10 As shown, a device for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping is provided, comprising: an acquisition module 1010, a training module 1020, and a decomposition module 1030, wherein:

[0155] An acquisition module 1010 is configured to acquire historical operating data of the distribution network in a target area at different times and zoning information of the target area; the historical operating data includes node power data of each node in the distribution network and reactive power control strategies for each distributed power source in the distribution network;

[0156] A training module 1020 is configured to train a pre-built on-site operation control model for distributed power sources using the node power data corresponding to the distribution network at each moment as training input data and the reactive power control strategy for each distributed power source in the distribution network at the corresponding moment as training output data. The on-site operation control model for distributed power sources is implemented based on a dimensionality-increasing linear mapping method.

[0157] The decomposition module 1030 is used to decompose the trained distributed power supply on-site operation control model according to the partition information of the target area to obtain a sub-model corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0158] In an optional embodiment, the training module 1020 is specifically used to: for the training input data corresponding to each moment, expand the dimension of the training input data according to the basis vector corresponding to at least one lifting dimension to obtain the lifted dimension expansion vector of the training input data; perform data fusion on the training input data and the lifted dimension expansion vector corresponding to the corresponding training input data to obtain the fused input data of the training input data; use the fused input data of the training input data as the model independent variable and the training output data corresponding to the corresponding training input data as the model dependent variable to determine the lifted dimension linear mapping matrix of the pre-built distributed power supply on-site operation control model to train the pre-built distributed power supply on-site operation control model.

[0159] In an optional embodiment, the zoning information of the target area includes the access location of the distributed power source and the distribution network zoning information; accordingly, the decomposition module 1030 is specifically used to: split the increased-dimensional linear mapping matrix into increased-dimensional linear mapping sub-matrices that match the distribution network zoning information according to the access location of the distributed power source; wherein the increased-dimensional linear mapping sub-matrix is ​​the increased-dimensional linear mapping matrix of the sub-model corresponding to the sub-area where the distribution network zoning information is located.

[0160] In an optional embodiment, the training module 1020 is specifically used to: for each lifting dimension, expand the training input data in the first dimension according to the basis vector corresponding to the lifting dimension to obtain the initial expansion vector of the training input data under the lifting dimension; combine the initial expansion vectors under different lifting dimensions to obtain the upgraded dimension expansion vector of the training input data.

[0161] In an optional embodiment, the node power data includes injected active power and load reactive power; the injected active power includes the active output of the distributed power source and the active power demand of the load.

[0162] In an optional embodiment, based on the node power data of each node in the sub-area at any target moment, a reactive power control strategy prediction for the distributed power supply in the sub-area at the corresponding target moment is performed, including: for each sub-area, obtaining the real node power data of each node in the sub-area at the target moment and the estimated node power data of each node in other sub-areas; generating input data corresponding to the sub-area based on the real node power data and the estimated node power data of each node in other sub-areas; based on the basis vector corresponding to at least one lifted dimension, dimensionality expansion of the input data is performed to obtain a lifted dimension expansion vector of the input data; data fusion is performed on the input data and the lifted dimension expansion vector corresponding to the input data to obtain fused input data of the input data; the fused input data of the input data is input into the sub-model corresponding to the sub-area to determine the reactive power control strategy of the distributed power supply in the sub-area.

[0163] Each module in the above-mentioned distribution network distributed power source on-site operation optimization device based on increasing-dimensional linear mapping can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0164] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping is implemented.

[0165] Those skilled in the art will understand that Figure 11The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0166] In an exemplary embodiment, a computer device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: obtaining historical operation data of a distribution network in a target area at different times and partition information of the target area; the historical operation data includes node power data of each node in the distribution network and reactive power control strategies of each distributed power source in the distribution network; using the node power data corresponding to the distribution network at each time as training input data, and using the reactive power control strategies of each distributed power source in the distribution network at the corresponding time as training output data, to train a pre-built on-site operation control model of distributed power sources; the on-site operation control model of distributed power sources is implemented based on a dimensionality-increasing linear mapping method; according to the partition information of the target area, the trained on-site operation control model of distributed power sources is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy of the distributed power sources in the sub-area at the corresponding target time based on the node power data of each node in the sub-area at any target time.

[0167] In one embodiment, when the processor executes the computer program, the following steps are also implemented: for the training input data corresponding to each moment, the training input data is dimensionally expanded according to the basis vector corresponding to at least one lifting dimension to obtain the lifted dimension expanded vector of the training input data; the training input data and the lifted dimension expanded vector corresponding to the corresponding training input data are data fused to obtain the fused input data of the training input data; the fused input data of the training input data is used as the model independent variable, and the training output data corresponding to the corresponding training input data is used as the model dependent variable to determine the lifted dimension linear mapping matrix of the pre-built distributed power supply on-site operation control model to train the pre-built distributed power supply on-site operation control model.

[0168] In one embodiment, the zoning information of the target area includes the access location of the distributed power source and the zoning information of the distribution network; when the processor executes the computer program, the following steps are also implemented: according to the access location of the distributed power source, the dimensionality-increasing linear mapping matrix is ​​split into dimensionality-increasing linear mapping sub-matrices that match the zoning information of the distribution network; wherein the dimensionality-increasing linear mapping sub-matrix is ​​the dimensionality-increasing linear mapping matrix of the sub-model corresponding to the sub-area where the distribution network zoning information is located.

[0169] In one embodiment, when the processor executes the computer program, the following steps are further implemented: for each lifting dimension, the training input data is expanded in the first dimension according to the basis vector corresponding to the lifting dimension to obtain the initial expanded vector of the training input data under the lifting dimension; the initial expanded vectors under different lifting dimensions are combined to obtain the dimension-upgraded expanded vector of the training input data.

[0170] In one embodiment, the node power data includes injected active power and load reactive power; the injected active power includes the active output of the distributed generation and the active power demand of the load.

[0171] In one embodiment, based on the node power data of each node in the sub-area at any target time, a reactive power control strategy prediction for distributed power sources in the sub-area at the corresponding target time is performed, including:

[0172] For each sub-area, the real node power data of each node in the sub-area and the estimated node power data of each node in other sub-areas at the target time are obtained; based on the real node power data and the estimated node power data of each node in other sub-areas, the input data corresponding to the sub-area is generated; based on the basis vector corresponding to at least one lifted dimension, the input data is dimensionally expanded to obtain a lifted dimension expanded vector of the input data; the input data and the lifted dimension expanded vector corresponding to the input data are data fused to obtain fused input data of the input data; the fused input data of the input data is input into the sub-model corresponding to the sub-area to determine the reactive power control strategy of the distributed power supply in the sub-area.

[0173] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0174] Obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0175] The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment is used as training output data to train the pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on the dimensionality-increasing linear mapping method;

[0176] According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0177] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0178] Obtain historical operating data of the distribution network in the target area at different times and the zoning information of the target area; the historical operating data includes the node power data of each node in the distribution network and the reactive power control strategy of each distributed power source in the distribution network;

[0179] The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source in the distribution network at the corresponding moment is used as training output data to train the pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on the dimensionality-increasing linear mapping method;

[0180] According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

[0181] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0182] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0183] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for optimizing the on-site operation of distributed power sources in a distribution network based on dimensionality-increasing linear mapping, characterized in that: The method comprises: Acquire historical operating data of the distribution network in the target area at different times and zoning information of the target area; the historical operating data includes node power data of each node in the distribution network and reactive power control strategies for each distributed power source in the distribution network; The node power data corresponding to the distribution network at each moment is used as training input data, and the reactive power control strategy of each distributed power source of the distribution network at the corresponding moment is used as training output data to train a pre-built distributed power source local operation control model; the distributed power source local operation control model is implemented based on a dimensionality-increasing linear mapping method; According to the partition information of the target area, the trained distributed power supply on-site operation control model is decomposed to obtain sub-models corresponding to each sub-area in the target area; the sub-models corresponding to the sub-areas are used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

2. The method according to claim 1, characterized in that The training of the pre-built distributed power source local operation control model includes: For the training input data corresponding to each moment, perform dimension expansion on the training input data according to the basis vector corresponding to at least one lifted dimension to obtain a lifted dimension expanded vector of the training input data; Performing data fusion on the training input data and the dimension-increased expansion vector corresponding to the training input data to obtain fused input data of the training input data; Taking the fused input data of the training input data as the model independent variable and the training output data corresponding to the corresponding training input data as the model dependent variable, the dimensionality-increasing linear mapping matrix of the pre-built distributed power supply on-site operation control model is determined to train the pre-built distributed power supply on-site operation control model.

3. The method according to claim 2, characterized in that The zoning information of the target area includes the access location of the distributed power supply and the zoning information of the distribution network; accordingly, the trained distributed power supply local operation control model is decomposed according to the zoning information of the target area to obtain sub-models corresponding to each sub-area in the target area, including: According to the access location of the distributed power supply, the dimensionality-increasing linear mapping matrix is ​​split into dimensionality-increasing linear mapping sub-matrices that match the distribution network partition information; wherein the dimensionality-increasing linear mapping sub-matrix is ​​the dimensionality-increasing linear mapping matrix of the sub-model corresponding to the sub-area where the distribution network partition information is located.

4. The method according to claim 3, characterized in that The step of performing dimension expansion on the training input data according to the basis vector corresponding to the at least one lifted dimension to obtain the lifted dimension expanded vector of the training input data includes: For each lifting dimension, performing a first dimension expansion on the training input data according to a basis vector corresponding to the lifting dimension to obtain an initial expansion vector of the training input data in the lifting dimension; The initial expanded vectors under different lifting dimensions are combined to obtain the dimension-lifted expanded vector of the training input data.

5. The method according to claim 1, wherein The node power data includes injected active power and load reactive power; the injected active power includes the active output of the distributed power source and the active power demand of the load.

6. The method according to any one of claims 1 to 5, characterized in that The determining of the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment includes: For each sub-area, obtaining the actual node power data of each node in the sub-area and the estimated node power data of each node in other sub-areas at the target time; generating input data corresponding to the sub-area based on the actual node power data and the estimated node power data of each node in each of the other sub-areas; Performing dimension expansion on the input data based on a basis vector corresponding to at least one lifted dimension to obtain a lifted dimension expanded vector of the input data; Performing data fusion on the input data and the dimension-increased expansion vector corresponding to the input data to obtain fused input data of the input data; The fused input data of the input data is input into the sub-model corresponding to the sub-region to determine the reactive power control strategy of the distributed power supply in the sub-region.

7. A device for optimizing the local operation of distributed power sources in a distribution network based on dimensional linear mapping, characterized in that: The device comprises: An acquisition module is configured to acquire historical operating data of the distribution network in a target area at different times and zoning information of the target area; the historical operating data includes node power data of each node in the distribution network and reactive power control strategies for each distributed power source in the distribution network; A training module is configured to train a pre-built on-site operation control model for distributed power sources using the node power data corresponding to the distribution network at each moment as training input data and the reactive power control strategy for each distributed power source in the distribution network at the corresponding moment as training output data; the on-site operation control model for distributed power sources is implemented based on a dimensionality-increasing linear mapping method; A decomposition module is used to decompose the trained distributed power supply on-site operation control model according to the partition information of the target area to obtain a sub-model corresponding to each sub-area in the target area; the sub-model corresponding to the sub-area is used to determine the reactive power control strategy for the distributed power supply in the sub-area at the corresponding target moment based on the node power data of each node in the sub-area at any target moment.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images