Regional load prediction method and novel regional load capacity optimization layout method

Abstract

According to the regional load prediction method and the novel regional load capacity optimization layout method provided by the invention, the improved Informer model is trained based on the characteristics of dimension reduction processing and a plurality of historical regional loads, and high-precision and high-robustness load prediction is realized; based on the load prediction result and the novel load locating and sizing and operation scheme, obtaining a bilevel programming model, and combining reliability and economical efficiency to obtain a novel locating and sizing scheme of the load single-point access regional power grid; and by taking the maximum reliability benefit as the target, the line and switch planning under the multipoint access topology is optimized, and the reliability and the economical efficiency are balanced, so that the global optimal effect of regional power grid resource allocation is realized.

Description

Technical Field

[0001] This invention relates to the field of power grid layout, specifically to regional load forecasting methods and novel regional load capacity optimization layout methods. Background Technology

[0002] With the construction of new power systems, a large number of diverse new loads, such as distributed photovoltaic power and electric vehicle charging piles, are being connected to regional distribution networks. This optimizes the energy structure and improves the utilization rate of renewable energy, making the grid load more flexible and adaptable. However, the volatility, intermittency, and uncertainty of new energy output pose significant challenges to peak shaving and voltage regulation. To ensure the efficient and stable operation of the grid, capacity optimization of diverse new loads in grid planning has become a key link in ensuring stable system operation. Load forecasting is an important reference value for determining the capacity configuration level of new loads in grid dispatch. By combining the aggregation and interaction characteristics of diverse loads for precise modeling, the accuracy and timeliness of load forecasting can be improved, enabling optimized grid operation dispatch and reasonable allocation of new load capacity, thereby ensuring the safe and stable operation of the grid and enhancing the absorption capacity of renewable energy.

[0003] However, the diverse new load power is closely related to various electricity consumption scenarios such as commercial charging piles and residential charging piles, as well as various geographical and climatic characteristics such as temperature, atmospheric pressure, and light intensity. The uncertainty of these characteristics leads to a decrease in the accuracy of load forecasting, especially under complex climate and electricity consumption pattern changes, posing a significant challenge to power grid dispatching, load balancing, and capacity optimization planning. In addition, incorporating diverse new loads into power grid planning requires addressing the high investment and operation and maintenance costs. How to balance the economy and reliability of power grid operation is an urgent problem to be solved.

[0004] Traditional feature extraction methods neglect the nonlinear and non-stationary characteristics of load data under different application scenarios and geographical climates, resulting in high model computational complexity and low prediction accuracy. Furthermore, traditional load capacity optimization research ignores the positive impact of load forecasting, making it difficult to adapt to the complex and ever-changing power grid environments under different application scenarios and geographical climates. Existing methods only consider optimizing the economics or reliability of power grid operation in isolation, neglecting the trade-offs and comprehensive improvement between the two, leading to low economic efficiency and poor reliability in the regional diversified new load capacity layout. Summary of the Invention

[0005] Based on this, the present invention provides a regional load forecasting method and a novel regional load capacity optimization layout method. Based on the characteristics of dimensionality reduction processing and training an improved Informer model using several historical regional loads, high-precision and robust load forecasting is achieved. Based on the above load forecasting results and novel load location, capacity determination, and operation schemes, a two-layer planning model is obtained. Combining reliability and economy, a novel location and capacity determination scheme for single-point load access to the regional power grid is derived. Furthermore, with the goal of maximizing reliability benefits, the line and switch planning under the multi-point access topology is optimized, balancing reliability and economy to achieve the globally optimal allocation of regional power grid resources, thereby solving the problem in existing technologies where both power grid economy and reliability cannot be simultaneously considered.

[0006] In a first aspect, the present invention provides a regional load forecasting method, comprising:

[0007] Obtain historical electricity consumption scenarios, historical geographical and climatic characteristics, historical network topology, and historical regional load of the regional power grid;

[0008] By fusing historical electricity consumption scenarios, historical geographical and climatic characteristics, and historical network topology at the corresponding time of the historical region's load, a multi-dimensional load aggregation and interaction historical feature is obtained.

[0009] Based on the historical load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics, the multivariate fast maximum information coefficient is obtained;

[0010] The multi-load aggregation interaction history features are subjected to dimensionality reduction processing to obtain the dimensionality-reduced multi-load aggregation interaction history features.

[0011] The reduced multivariate load aggregation interaction historical features, multivariate fast maximum information coefficient and historical regional load are input into the improved Informer model for training to obtain the trained improved Informer model.

[0012] The system acquires real-time power consumption scenarios, real-time geographic climate features, and real-time network topology, and integrates them to obtain multi-dimensional load aggregation and interaction features.

[0013] The multi-variable load aggregation interaction features are input into the trained improved Informer model to obtain the real-time regional load.

[0014] Furthermore, the process of obtaining multivariate fast maximum information coefficients based on the load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics includes:

[0015] Based on the historical load of each historical region and the historical characteristics of multi-dimensional load aggregation and interaction at the corresponding time, the mutual information between the historical load of the historical region and the historical characteristics of multi-dimensional load aggregation and interaction is obtained.

[0016] The multivariate fast maximum information coefficients are obtained based on the mutual information normalization.

[0017] Furthermore, the multi-dimensional load aggregation interaction history features are subjected to dimensionality reduction processing to obtain dimensionality-reduced multi-dimensional load aggregation interaction history features, specifically including the following steps:

[0018] Obtain the eigenvalue magnitude and cumulative variance contribution rate of each feature factor in the historical features of multivariate load aggregation interaction;

[0019] When the cumulative variance contribution rate of any combination of multiple feature factors exceeds a preset threshold, the combination of these multiple feature factors is set as the dimensionality-reduced multivariate load aggregation interaction historical feature.

[0020] Furthermore, the improved Informer model includes a dimension segmentation mechanism and a two-stage attention mechanism;

[0021] The data of each feature dimension in the dimensionality-reduced multivariate composite aggregation interaction history features are divided into sequence segments of a preset length. Each sequence segment is normalized to obtain the normalized-dimensionality-reduced multivariate load aggregation interaction history features.

[0022] Each sequence segment in the normalized-dimensional reduction multivariate composite aggregation interaction history feature is embedded into a preset position vector to obtain the sequence segment embedded at the time step;

[0023] By embedding sequence segments at each time step in each feature dimension and combining them with a multi-head probabilistic sparse self-attention mechanism, the correlation between sequence segments at different time steps can be captured.

[0024] A routing vector mechanism is adopted to capture the correlation between various reduced-dimensional multivariate load aggregation interaction feature sequences by separating information from the routing vector.

[0025] Secondly, the present invention also provides a novel method for optimizing the layout of regional load capacity, comprising:

[0026] Taking the location, capacity, and operation scheme under the new single-point load access topology as the planning object, and combining the real-time regional load obtained by the regional load forecasting method described in any one of the first aspects, a two-level planning model including reliability and economy is constructed to obtain the location and capacity scheme of the regional power grid under the new single-point load access topology.

[0027] Based on the site selection and capacity determination scheme for the new type of load single-point access to the regional power grid, with the goal of maximizing the benefits of the regional power grid after adding load cables under the new type of load multi-point access topology, the location and capacity of the newly added lines and equipment under the new type of load multi-point access topology are adjusted to obtain the optimized site selection and capacity determination scheme for the new type of load in the region.

[0028] Furthermore, the bi-level programming model includes an upper-level objective function and a lower-level objective function;

[0029] The specific expression for the upper-level objective function is as follows:

[0030] ,

[0031] in, For the upper-level decision variables in the bilevel programming model, For the lower-level transformation coefficients of the bilevel programming model, For the investment cost of new loads, To incur penalties for power outages, For network line loss costs, Cost of power quality deviation, Penalty costs for load fluctuations For power grid nodes Capacity to accommodate new loads, For nodes New load installation capacity limit, This represents the upper limit for the installation capacity of new loads in the regional power grid.

[0032] The specific expression for the lower-level objective function is as follows:

[0033] ,

[0034] in, For the lower-level decision variables in a bilevel programming model, For the lower-level equality constraints of the bilevel programming model This represents the lower-level inequality constraints in a bilevel programming model.

[0035] Furthermore, the objective of maximizing the system reliability improvement benefit after adding load cables under the new single-point load access topology is specifically expressed as follows:

[0036] ,

[0037] in To increase revenue through reliability, , The power outage penalty coefficient per unit time. Nodes under new single-point load access conditions The duration of power outage at the load, This represents the expected power outage time for the regional power grid after the integration of new loads. For the investment cost of new load, For the number of nodes in the regional power grid, branch road The number of nodes.

[0038] Furthermore, the reliability of a regional power distribution network is specifically expressed as follows:

[0039] ,

[0040] in, For the reliability of the regional power grid, This represents the expected power outage time for the regional power grid after the integration of new loads. , For regional power grid nodes under the new single-point load access situation Power outage duration.

[0041] Thirdly, the present invention provides a regional load forecasting device, comprising:

[0042] The historical data acquisition module is used to acquire historical electricity consumption scenarios, historical geographical and climatic characteristics, historical network topology, and historical regional load of the regional power grid.

[0043] The feature fusion module is used to fuse the historical electricity consumption scenarios, historical geographical and climatic characteristics and historical network topology of the historical region at the corresponding time to obtain multi-dimensional load aggregation and interaction historical features.

[0044] The correlation calculation module is used to obtain the multivariate fast maximum information coefficient based on the historical load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics.

[0045] The feature dimensionality reduction module is used to perform dimensionality reduction processing on the multi-dimensional load aggregation interaction history features to obtain the dimensionality-reduced multi-dimensional load aggregation interaction history features.

[0046] The model training module is used to input the dimensionality-reduced multivariate load aggregation interaction historical features, multivariate fast maximum information coefficient and historical regional load into the improved Informer model for training, so as to obtain the trained improved Informer model.

[0047] The real-time feature acquisition module is used to acquire real-time power consumption scenarios, real-time geographic climate features, and real-time network topology, and fuse them to obtain multi-dimensional load aggregation interaction features.

[0048] The real-time load prediction module is used to input the multi-variable load aggregation interaction features into the trained improved Informer model to obtain the real-time regional load.

[0049] Fourthly, the present invention also provides a novel regional load capacity optimization layout device, comprising:

[0050] The single-point access optimization module is used to take the location, capacity and operation scheme under the new load single-point access topology as the planning object, and combine the real-time regional load obtained by the regional load forecasting device mentioned in the third aspect to construct a two-level planning model that includes reliability and economy, so as to obtain the location and capacity scheme of the regional power grid under the new load single-point access topology.

[0051] The regional power grid optimization module is used to adjust the location and capacity of newly added lines and equipment under the new load multi-point access topology, based on the site selection and capacity determination scheme of the new load single-point access regional power grid, with the goal of maximizing the regional power grid benefits after adding load cables under the new load multi-point access topology, to obtain the regional new load site selection and capacity determination optimization scheme.

[0052] Fifthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of any one of the regional load forecasting methods in the first aspect or any one of the regional novel load capacity optimization layout methods in the second aspect.

[0053] In a sixth aspect, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that, when the processor executes the computer program, it executes the steps of any one of the regional load forecasting methods in the first aspect or any one of the regional novel load capacity optimization layout methods in the second aspect.

[0054] The beneficial effects of adopting the above technical solution are as follows: This embodiment considers different scenarios such as commercial charging piles, residential charging piles, and medium-voltage energy storage power stations, different geographical and climatic characteristics, and different network topologies. Based on the multivariate fast maximum information coefficient, it analyzes the correlation between the multivariate load aggregation interaction characteristics and the regional load sequence, enabling the prediction model to better adapt to multidimensional input data. Then, it performs dimensionality reduction processing based on the extracted highly correlated feature sequences. Finally, it improves the traditional Informer model by introducing a dimensional segmentation embedding mechanism and a two-stage attention mechanism, thereby improving the prediction model's ability to analyze the correlation of different feature sequences. Thus, this embodiment obtains a high-precision and robust improved Informer model, which can be used to perform high-precision load forecasting for the regional power grid.

[0055] Furthermore, based on the above load forecasting results, this embodiment optimizes the location and capacity determination of new loads in the regional power grid in two stages. In the first stage, the location and capacity determination under the single-point access topology of new loads is used as the planning object, and a two-level planning model is constructed to obtain the location and capacity determination scheme of the regional power grid under the single-point access topology of new loads. In the second stage, based on the location and capacity determination scheme of the single-point access topology of new loads in the first stage, the location, equipment location, and equipment capacity of the newly added lines and equipment under the multi-point access topology of new loads are adjusted with the goal of maximizing the benefits of the regional power grid after adding new load cables under the multi-point access topology of new loads. This yields an optimized location and capacity determination scheme for new loads in the region, so that the regional power grid with access to new loads obtains the most efficient location and capacity determination scheme for new loads, achieving the highest overall efficiency. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0057] Figure 1 This is a schematic diagram of a regional load forecasting method in one embodiment of this application;

[0058] Figure 2 This is a schematic diagram of the regional load forecasting method architecture in one embodiment of this application;

[0059] Figure 3 This is a schematic diagram of a novel regional load capacity optimization layout method in one embodiment of this application;

[0060] Figure 4 This is a schematic diagram of the architecture of a novel regional load capacity optimization layout method in one embodiment of this application. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. To describe the present invention in more detail, the regional load forecasting method and the new regional load capacity optimization layout method provided by the present invention will be specifically described below with reference to the accompanying drawings.

[0062] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms "an," "a," or "the" do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. The terms "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up," "down," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0063] The diverse new load power is closely related to various electricity consumption scenarios such as commercial and residential charging piles, as well as geographical factors and climatic characteristics such as temperature, atmospheric pressure, and light intensity. The uncertainty of these characteristics leads to a decrease in the accuracy of load forecasting, especially under complex climate and electricity consumption pattern changes, posing a significant challenge to power grid dispatching, load balancing, and capacity optimization planning. In addition, incorporating diverse new loads into power grid planning requires addressing the high investment and operation and maintenance costs. How to balance the economy and reliability of power grid operation is an urgent problem to be solved.

[0064] Traditional feature extraction methods neglect the nonlinear and non-stationary characteristics of load data under different application scenarios and geographical climates, resulting in high model computational complexity and low prediction accuracy. Furthermore, traditional load capacity optimization research ignores the positive impact of load forecasting, making it difficult to adapt to the complex and ever-changing power grid environments under different application scenarios and geographical climates. Existing methods only consider optimization from the perspective of power grid operation economy or reliability alone, ignoring the trade-offs and comprehensive improvement between the two, leading to low economic efficiency and poor reliability in the regional diversified new load capacity layout.

[0065] Furthermore, with the construction of new power systems, a large number of distributed photovoltaic (PV) and electric vehicle charging piles, among other diverse new loads, are being connected to regional distribution networks, optimizing the energy structure, improving the utilization rate of renewable energy, and making the grid load more flexible and variable. The volatility, intermittency, and uncertainty of new energy processing also pose significant challenges to grid peak shaving and voltage regulation. To ensure the efficient and stable operation of the grid, capacity optimization of diverse new loads is crucial in grid planning. Load forecasting is an important means to improve grid dispatching and new capacity allocation. By combining the aggregation and interaction characteristics of diverse loads for precise modeling, the accuracy and timeliness of load forecasting can be improved, enabling optimized grid operation dispatching and rational allocation of new load capacity, thereby ensuring the safe and stable operation of the grid and enhancing the absorption capacity of renewable energy.

[0066] Existing technical solutions only consider the relationship between a single feature and the load sequence, failing to cover the load aggregation and interaction characteristics under diverse load access, different application scenarios, and different geographical and climatic environments. This results in high overlap of feature information and low model prediction capability. Furthermore, existing technologies fail to consider the positive impact of different devices processing predictions, making it difficult to adapt to the complex and ever-changing power grid environment under different application scenarios and geographical and climatic conditions. Consequently, this affects the flexible scheduling and efficient utilization of diverse new loads, leading to poor economic efficiency and reliability in the regional distribution of diverse new load capacity.

[0067] Based on this, combined with the appendix Figure 1 The diagram and appendix show the regional load forecasting method. Figure 2 The diagram shown illustrates the architecture of the regional load forecasting method. This invention is based on the appendix. Figure 2 The regional load forecasting method architecture shown proposes a method that improves the Informer model based on dimensionality-reduced features and several historical regional loads, achieving high-precision and robust load forecasting. Specifically, it includes the following steps:

[0068] Step S101: Obtain historical electricity consumption scenarios, historical geographical and climatic characteristics, historical network topology, and historical regional load of the regional power grid.

[0069] Specifically, the electricity usage scenarios considered in the embodiments of the present invention include, but are not limited to, commercial charging piles, residential charging piles, medium-voltage energy storage power stations, and low-voltage residential energy storage. Various electricity usage scenarios; geographical and climatic characteristics including but not limited to temperature, atmospheric pressure, and light intensity. Geographical and climatic characteristics; network topology including but not limited to distributed photovoltaic, charging piles and other new load access. A network topology.

[0070] In the training phase of improving the Informer model, it is necessary to obtain the load of several historical regions and the historical power consumption scenario, historical geographical and climatic characteristics, and historical network topology of the regional power grid at the corresponding time for each historical load.

[0071] Step S102: Based on the historical electricity consumption scenarios, historical geographical and climatic characteristics, and historical network topology of the historical region at the corresponding time, the multi-dimensional load aggregation and interaction historical characteristics are obtained by fusing them together.

[0072] Specifically, each of the above-mentioned historical electricity consumption scenarios, historical geographical and climatic characteristics, and historical network topology can be combined to obtain multi-dimensional load aggregation and interaction characteristics, denoted as... , Indicates the first Under this network topology, the first The first application environment This involves considering various climatic characteristics. Specifically, for each historical region's load at a given time, there exists a specific combination of historical electricity consumption scenarios, historical geographical and climatic characteristics, and historical network topology. This combination can be denoted as the multivariate load aggregation and interaction historical characteristics at the corresponding time in the historical region. Based on this, combining the historical regional load and the multivariate load aggregation and interaction characteristics at the corresponding time, a multivariate dataset of the regional power grid is constructed, denoted as... , This is the regional load sequence.

[0073] Step S103: Based on the historical load of each historical region and the historical characteristics of multivariate load aggregation and interaction at the corresponding time, obtain the multivariate fast maximum information coefficient.

[0074] Specifically, step S103 includes the following steps:

[0075] Step S1301: Based on the historical load of each historical region and the historical characteristics of multi-dimensional load aggregation interaction at the corresponding time, obtain the mutual information between the historical load of the historical region and the historical characteristics of multi-dimensional load aggregation interaction.

[0076] Specifically, respectively Divided into axis, Classified to Axis, pair shaft and The axes are divided into grids, with each grid representing a subset of data points; for the gridded dataset... The ratio of the number of points falling into a given grid to the total number of points is defined as the approximate probability density of that grid. This allows us to obtain the mutual information between the regional load sequence and the multi-dimensional load aggregation interaction characteristics, denoted as... The specific expression for mutual information is: ,in, To The number of grids for axis division, , To The number of grids for axis division, , For dataset Classified into Joint probability density under each grid; for Classified into Edge probability density under each grid for exist Edge probability density under each grid division

[0077] Step S1302: Obtain the multivariate fast maximum information coefficients based on the mutual information normalization.

[0078] Specifically, the mutual information between regional load sequences and multi-source load aggregation interaction characteristics is normalized to obtain the MRMIC value, which is expressed as: ,when The larger the value, the more likely it is to indicate the interaction characteristics of multi-load aggregation. With regional load sequence The stronger the correlation.

[0079] Step S104: Perform dimensionality reduction processing on the multi-dimensional load aggregation interaction history features to obtain the dimensionality-reduced multi-dimensional load aggregation interaction history features.

[0080] To reduce the dimensionality of the input features in the regional load forecasting model and decrease computational complexity, exploratory factor analysis is used to reduce the dimensionality of the model's input features, merging numerous multivariate load aggregation and interaction features into several representative common factors. Specifically, this includes the following steps:

[0081] Step S1401: Obtain the eigenvalue magnitude and cumulative variance contribution rate of each feature factor in the multivariate load aggregation interaction historical features.

[0082] Step S1402: When the cumulative variance contribution rate of any combination of multiple feature factors exceeds a preset threshold, the combination of these multiple feature factors is set as the dimensionality-reduced multivariate load aggregation interaction historical feature.

[0083] In this embodiment, the preset threshold can be set to 85%. If the cumulative variance contribution rate of the combination of G feature factors exceeds 85%, it means that the combination of the G feature factors can replace the original total number of features. The multivariate load aggregation interaction features have less than 15% unexplained variance. Dimensionality reduction will not lose much information but can improve the processing speed of the load forecasting model. Therefore, it can be used... The number of feature factors is reduced from 1 to G, achieving the goal of dimensionality reduction. The combination of G feature factors is set as the dimensionality-reduced multivariate load aggregation interaction history feature.

[0084] Step S105: Input the dimensionality-reduced multivariate load aggregation interaction historical features, multivariate fast maximum information coefficient, and historical regional load into the improved Informer model for training to obtain the trained improved Informer model.

[0085] Specifically, the improved Informer model mentioned in step S105 is defined as including a dimension segmentation mechanism and a two-stage attention mechanism.

[0086] The dimension segmentation mechanism includes:

[0087] The data of each feature dimension in the dimensionality-reduced multivariate composite aggregation interaction history features are divided into sequence segments of a preset length. Each sequence segment is normalized to obtain normalized-dimensionality-reduced multivariate load aggregation interaction history features. Each sequence segment in the normalized-dimensionality-reduced multivariate composite aggregation interaction history features is embedded into a preset position vector to obtain a sequence segment embedded at a time step.

[0088] The specific expression for the normalized-dimensional reduction multivariate load aggregation interaction historical features is as follows: , For normalized-dimensionality-reduced multivariate load aggregation interaction feature sequences, For feature dimensions Upper Dimensionally reduced multivariate load aggregation interaction feature sequence at each time step For time-based sliding windows, For time steps, The length of each sequence segment, For input dimensions.

[0089] The specific expression for the sequence segment with embedded time steps is: ,in, For feature dimensions Upper A sequence segment at each time step, The parameter matrix has a dimension of [missing information]. , This is a position vector.

[0090] The two-stage attention mechanism includes:

[0091] By embedding time-step sequence segments in each feature dimension and combining them with a multi-head probabilistic sparse self-attention mechanism, the correlation between sequence segments at different time steps is captured. A routing vector mechanism is adopted to capture the correlation between various dimensionality-reduced multivariate load aggregation interaction feature sequences by separating information from the routing vector.

[0092] During the cross-time phase, the output of the dimensionality-segmented embedding layer is used as input. A multi-head probabilities-based self-attention (MPSA) mechanism is employed along the time series on each feature dimension of the dimensionality-reduced multivariate load aggregation interaction features to capture the correlation between different time series, expressed as:

[0093] ,

[0094] ,

[0095] For the first The output of the double-layer feedforward network (DLF) layer at any given time; For the first The output of the two-layer feedforward network at any given time. For layer normalization operation, It is a two-layer feedforward network. For the first The output of the MPSA layer at time t, its dimension is , The feature dimension after vector embedding. Output dimension for dimension segmentation embedding layer The vector of all time steps in the vector. For dimension A random vector at all time steps; For the first MPSA layer operations at specific times.

[0096] In the cross-feature stage, a routing vector mechanism is adopted to capture the correlation between various dimensionality-reduced multivariate load aggregation interaction feature sequences by separating information from the routing vector. The specific expression is as follows:

[0097] ,

[0098] ,

[0099] in, For time steps The intermediate routing mechanism outputs, The routing vector is a vector with dimension 1. A fixed-dimensional science department parameter matrix is ​​set. For time steps a two-dimensional vector, This is the first-level MPSA operation for the feature sequence. This is the second-level MPSA operation for the feature sequence.

[0100] Step S106: Obtain real-time power consumption scenarios, real-time geographic climate features, and real-time network topology, and fuse them to obtain multi-load aggregation and interaction features.

[0101] It should be noted that in this embodiment, the purpose of dimensionality reduction of the multi-load aggregation interaction features during the preliminary model training phase is to reduce computational complexity while maintaining prediction accuracy. However, in the real-time regional load prediction phase, although the input data has a larger dimensionality, further dimensionality reduction is not necessary. The dimensionality reduction during the training phase has already extracted and mapped the main relevant features to a low-dimensional space. Based on this, the real-time input multi-load aggregation interaction features have been standardized and have the same dimensionality as the data obtained during the training phase. Therefore, the multi-load aggregation interaction features obtained from the real-time electricity consumption scenario, real-time geographic climate features, and real-time topology fusion do not require further dimensionality reduction; instead, the standardized features are directly input into the trained model.

[0102] Step S107: Input the multi-variable load aggregation interaction features into the trained improved Informer model to obtain the real-time regional load.

[0103] The specific expression for real-time regional load obtained by improving the Informer model is as follows:

[0104] ,

[0105] For the improved Informer model trained in regional load forecasting, For the first Load value at any given time The historical window size for a time step, i.e., using past... Historical information from each time step is used to predict the load at the current time step.

[0106] In this embodiment of the regional load forecasting method, firstly, different scenarios such as commercial charging piles, residential charging piles, and medium-voltage energy storage power stations, as well as different geographical and climatic characteristics and network topologies, are considered. Based on the multivariate fast maximum information coefficient, correlation analysis is performed between multivariate load aggregation interaction features and regional load sequences, enabling the forecasting model to better adapt to multidimensional input data. Secondly, exploratory factor analysis is used to reduce the dimensionality of the extracted highly correlated feature sequences. Finally, the traditional Informer model is improved by introducing a dimensionality segmentation embedding mechanism and a two-stage attention mechanism, enhancing the forecasting model's ability to analyze the correlation of different feature sequences. The improved Informer model is trained using a regional historical load dataset, resulting in higher accuracy and robustness, providing high-precision data support for the subsequent quantitative optimization of new load access site selection in the regional power grid.

[0107] Furthermore, traditional methods for optimizing the location and capacity of diverse new loads neglect the positive impact of load forecasting, reducing the applicability of these methods. They also fail to consider the trade-off and comprehensive improvement of grid operation economy and reliability, resulting in poor economic efficiency and low reliability in optimizing the location and capacity of new loads. Considering the inherent connections and complementarities among diverse new loads, a reliability and economic model for the integration of new loads such as distributed photovoltaics and electric vehicle charging piles is constructed based on analytical methods.

[0108] To ensure accurate optimization of equipment location and capacity under multi-source load access conditions, based on the regional power grid load forecast results, this study comprehensively improves the economy and reliability of regional power grid operation, and combines this with... Figure 3 Schematic diagram and appendix of the new regional load capacity optimization layout method Figure 4 This embodiment presents a two-stage method for optimizing the layout of new load capacity in a region, considering both reliability and economy. The optimization of the location and capacity of new load equipment is divided into two stages. In the first stage, the location and capacity of equipment under a single-point access topology for new loads are used as the planning objects, and a two-layer planning model based on multivariate new load forecasting is constructed. In the second stage, a multivariate new load multi-point access model is constructed under different scenarios, with the ultimate goal of comprehensively improving economy and reliability. Multi-point redundant access of new loads is achieved by adding new power supply lines, and the location and capacity of new lines, switches, and other equipment are optimized, thereby further improving system reliability while optimizing operational economy and effectively improving the accuracy of the model.

[0109] Step S201: Taking the location, capacity, and operation scheme under the new single-point load access topology as the planning object, and combining the real-time regional load obtained in step S107, a two-layer planning model including reliability and economy is constructed to obtain the location and capacity scheme of the regional power grid under the new single-point load access topology.

[0110] In the first phase, the site selection, capacity determination, and operation schemes for new loads with computer access topologies are the planning objects. A two-layer planning model is established based on the real-time regional load obtained from the load forecasting model. The upper layer of the two-layer planning model is the planning layer, which optimizes with the goal of minimizing the expected comprehensive cost over a large time scale, using the installed power, capacity, and location of the new loads as decision variables. Once the upper layer model determines the site selection and capacity determination scheme, it is passed to the lower layer of the two-layer planning model, the operation layer. The lower layer model optimizes at regular intervals based on the real-time regional load obtained from the load forecasting model, aiming to minimize the operating cost under the time-series power flow characteristics, considering decision variables such as demand response schemes and energy storage charging and discharging.

[0111] In this embodiment, the two-level programming model includes an upper-level objective function and a lower-level objective function; the specific expression of the upper-level objective function is as follows:

[0112] ,

[0113] For the upper-level decision variables in the bilevel programming model, For the lower-level transformation coefficients of the bilevel programming model, For the investment cost of new loads, To incur penalties for power outages, For network line loss costs, Cost of power quality deviation, Penalty costs for load fluctuations For power grid nodes Capacity to accommodate new loads, For nodes New load installation capacity limit, This represents the upper limit for the installed capacity of new loads in the regional power grid. as well as These are the constraints for the upper-level objective function.

[0114] The specific expression for the lower-level objective function is:

[0115] ,

[0116] in, For the lower-level decision variables in a bilevel programming model, For the lower-level equality constraints of the bilevel programming model For the lower-level inequality constraints of the bilevel programming model, and These are the constraints for the lower-level objective function.

[0117] Furthermore, step S201 mentions that the bi-level programming model constructed in this embodiment has reliability and economy.

[0118] Power supply reliability is a key indicator for assessing regional power supply quality, and it is negatively correlated with the average annual power outage time. Most components in the distribution network are repairable; therefore, the regional power grid status can be divided into normal operating status and fault repair status. The power supply reliability at load points is related to cable length, the number of circuit breakers, the number of distribution transformers, as well as the failure rate and repair time.

[0119] consider For each regional power grid node, the reliability of the region can be measured by the power supply failure rate, repair time, outage duration, and power supply availability to the load at the j-th node, using the regional main power source.

[0120] Step S301: Obtain the cable length between the load and the main power source at the regional power grid node, the total number of circuit breakers and switches between the load and the main power source, the cable failure rate, the switch failure rate, the distribution network transformer failure rate, the cable repair time, the switch repair time, and the distribution transformer repair time.

[0121] in, For nodes in the regional power grid The length of the cable between the load and the main power supply. This refers to the total number of circuit breakers and switches between the load and the main power source at a regional power grid node. For cable failure rate, For switch failure rate, For the failure rate of distribution network transformers, For cable repair time, For switch repair time, This refers to the transformer repair time.

[0122] Step S302: The power supply failure rate of the regional power grid node is obtained based on the cable length between the load and the main power supply at the node, the total number of circuit breakers and switches between the load and the main power supply, the cable failure rate, the switch failure rate, and the distribution transformer failure rate.

[0123] The specific expression for the regional power grid node power supply failure rate in step S302 is as follows:

[0124] ,

[0125] in, For regional power grid nodes Power supply failure rate.

[0126] Step S303: The repair time of the regional power grid node is obtained based on the cable length between the load and the main power supply at the node, the cable failure rate, the cable repair time, the total number of circuit breakers and switches between the load and the main power supply, the switch failure rate, the switch repair time, the distribution network transformer failure rate, and the distribution transformer repair time.

[0127] The specific expression for the regional power grid node repair time in step S303 is as follows:

[0128] ,

[0129] in, For regional power grid nodes Repair time.

[0130] Step S304: Obtain the power outage duration of the regional power grid node based on the power supply failure rate and repair time of the regional power grid node.

[0131] The specific expression for the power outage duration of the power grid node in step S304 is as follows:

[0132] ,

[0133] in, For regional power grid nodes The duration of the power outage.

[0134] Step S305: Obtain the power supply reliability rate based on the power outage duration of the power grid node.

[0135] The specific expression for the power supply reliability rate in step S305 is as follows:

[0136] ,

[0137] For the main power source to the node in the regional power grid Power supply reliability.

[0138] Furthermore, in the case of multiple new loads accessing the network, continuous power supply to multiple faulty nodes can be operationally achieved by controlling the opening and closing of individual switches. The average power supply failure rate, repair time, and power outage duration of each new load are related to the relevant location of the source / load nodes. The specific calculation method of the power supply reliability model for the multi-point access topology of multiple new loads is as follows:

[0139] Step S401: Obtain the cable length between the new load at each node and the load at other nodes in the regional power grid, as well as the total number of circuit breakers and switches.

[0140] in, For nodes in the regional power grid New types of loads and nodes Cable length between loads, For nodes in the regional power grid New types of loads and nodes The total number of circuit breakers and switches between loads.

[0141] Step S402: Based on the cable length between the new load at the node and the loads at other nodes, the total number of circuit breakers and switches between the new load at the node and the loads at other nodes, the cable failure rate, the switch failure rate, and the distribution transformer failure rate, the power supply failure rate of the new load at the node to the loads at other nodes in the regional power grid is obtained.

[0142] The specific expression for the power supply failure rate of the new load at a node in the regional power grid to the load at other nodes in step S402 is as follows:

[0143] ,

[0144] in, For nodes in the regional power grid New type of load pair node Power supply failure rate of the load.

[0145] Step S403: Based on the cable length between the new load at the node and the loads at other nodes, the total number of circuit breakers and switches between the new load at the node and the loads at other nodes, the cable failure rate, the switch failure rate and the distribution transformer failure rate, the cable repair time, the switch repair time and the distribution transformer repair time, the repair time of the cable, the repair time of the switch and the repair time of the distribution transformer are obtained, the repair time of the new load at the node to the loads at other nodes in the regional power grid is obtained.

[0146] The specific expression for the repair time of the new load at a node in the regional power grid to the load at other nodes in step S403 is as follows:

[0147] ,

[0148] in, For nodes in the regional power grid New type of load pair node Repair time for the load.

[0149] Step S404: Based on the power supply failure rate and repair time of the new load at a node in the regional power grid to the loads at other nodes, obtain the power outage duration of the new load at a node in the regional power grid to the loads at other nodes.

[0150] The specific expression for the power outage duration of the new load at a node in the regional power grid to loads at other nodes in step S404 is as follows:

[0151] ,

[0152] in, For nodes in the regional power grid New type of load pair node The duration of power outage at the load.

[0153] Step S405: Determine the power supply reliability of the new load at a node in the regional power grid to the loads at other nodes based on the power outage duration of the new load at a node to the loads at other nodes.

[0154] The specific expression for the power supply reliability of the new load at a node in the regional power grid to the loads at other nodes in step S405 is as follows:

[0155] ,

[0156] For nodes in the regional power grid New type of load pair node Power supply reliability for the load.

[0157] When the main power supply fails, the required power can be supplied to the load point through a new load, achieving continuous power supply for a certain period and reducing the power outage duration caused by the main power supply failure. This applies to nodes under the condition of single-point connection of the new load. The power restoration time and power outage time of the load can be expressed as follows:

[0158] ,

[0159] ,

[0160] in, For nodes in the case of new single-point load access The time required to restore power to the load. For nodes in the case of new single-point load access The duration of power outage at the load, For nodes The upper limit of the state of charge for connecting new loads. For nodes The limit of the state of charge when a new type of load is connected. For nodes The success rate of transferring new loads to the site. For nodes The capacity to connect new loads, For nodes The active power of the load.

[0161] Based on the above, the reliability of the regional power distribution network in this embodiment is specifically represented as follows:

[0162] ,

[0163] in, For the reliability of the regional power grid, This represents the expected power outage time for the regional power grid after the integration of new loads. , For regional power grid nodes under the new single-point load access situation Power outage duration.

[0164] In addition, the economic model under the new load single-point access topology mainly considers the investment in new loads, the cost of system power outage faults, the cost of network line loss, the cost of power quality deviation, and the cost of load fluctuation penalty.

[0165] The investment cost of new loads includes initial investment cost and operation and maintenance cost, as specifically expressed as:

[0166] ,

[0167] For the investment cost of new loads, The initial investment cost for the new load, , This is the discount factor for new load investments, which is related to the lifespan of the new load. The unit capacity investment cost of new loads, For the operation and maintenance costs of new loads, , This is the equivalent coefficient for the operation and maintenance costs of new loads. When the operation and maintenance cost is measured in annual units, it can be approximately estimated as a certain percentage of the initial investment.

[0168] The system outage cost is related to the expected outage value of the regional power grid after the new load is connected at a single point, and the specific expression is as follows:

[0169] ,

[0170] in, To incur penalties for power outages, The penalty price for power outages per unit of time. This refers to the number of nodes in the regional power grid.

[0171] Network line loss costs mainly come from the line losses of each branch line, as expressed in the following formula:

[0172] ,

[0173] For network line loss costs, The unit line loss electricity price, For branch road collection, branch road The resistance, For the first Time Branch The current.

[0174] The cost of power quality deviation mainly comes from voltage deviation penalties, specifically expressed as:

[0175] ,

[0176] Cost of power quality deviation, The price is the penalty for unit voltage deviation. This refers to the node voltage rating. For the first Time Node The voltage.

[0177] The specific expression for the load fluctuation penalty is:

[0178] ,

[0179] Penalty costs for load fluctuations The price is the penalty for unit load fluctuation. for Time Node Load demand at the location, for Time Node Photovoltaic treatment at the site, for Time Node The charging power of the energy storage device for Time Node The discharge power of the energy storage device. It is the set of photovoltaic access nodes in the regional power grid. This refers to the set of nodes for energy storage devices to be connected to the regional power grid. The average load power during the scheduling period. .

[0180] Compared to the investment cost of a single-point access topology for new loads, a multi-point access structure for new loads requires additional equipment and thus additional investment costs. These additional costs mainly include the investment and construction costs of cables, and operation and maintenance costs. The specific investment cost for new loads in a regional distribution network is as follows:

[0181] ,

[0182] in, The investment cost for new loads in the regional power distribution network, This is the discount factor for investment. The equivalent coefficient for the operation and maintenance costs of newly added cables. The additional cable cost per unit, To increase the length of the power supply cable, The equivalent coefficient for the operation and maintenance costs of the new type of load. To add a new set of load access nodes, The unit capacity investment cost of new loads, For regional power grid nodes The capacity to connect new loads.

[0183] Step S202: Based on the site selection and capacity determination scheme of the new load single-point access to the regional power grid, with the goal of maximizing the benefits of the regional power grid after adding load cables under the new load multi-point access topology, adjust the location and capacity of the new lines and equipment under the new load multi-point access topology to obtain the regional new load site selection and capacity determination optimization scheme.

[0184] The aforementioned economic model is a comprehensive optimization objective based on multiple factors, including the investment cost of new loads, the penalty cost of power outages, network line loss costs, power quality deviation costs, and load fluctuation penalty costs. These cost factors are integrated into the objective function through a weighted approach, with each cost factor having a corresponding weight coefficient indicating its importance in the overall economic optimization. By setting appropriate weights, the influence of each cost factor can be balanced, ensuring the rationality and practical feasibility of the economic optimization.

[0185] The second phase, based on the site selection, capacity determination, and operation plan for new loads in the first phase, aims to maximize the system reliability improvement benefits after adding new load cables under the new load multi-point access topology. The planning of new lines, switches, and other equipment under the new load multi-point access topology is used as the decision variable. Specifically, the objective of maximizing the regional power grid benefits after adding new load cables under the new load multi-point access topology is expressed as follows:

[0186] ,

[0187] in To increase revenue through reliability, , The power outage penalty coefficient per unit time. Nodes under new single-point load access conditions The duration of power outage at the load, This represents the expected power outage time for the regional power grid after the integration of new loads. The cost of investing in new load.

[0188] For model solving, the first stage adopts a single-point access structure for multiple new loads as the planning object, constructing a two-layer planning model considering reliability and economy. The upper-layer model passes the multi-load location and capacity determination scheme to the lower layer, and the lower-layer model optimizes the operation based on the upper-layer planning. It can be solved using the Gurobi optimizer and MATLAB's Yalmip toolkit, and the operation scheme is returned to the upper layer. The upper and lower-layer models influence each other, and the final multi-load location and capacity determination scheme is obtained through iterative solution. The second stage considers the introduction of a multi-point access structure, realizing multi-point redundant access of new loads by adding power supply lines, thereby further improving system reliability while optimizing operation economy. This stage aims to maximize the benefits of reliability improvement, optimizing the location and capacity of new lines, switches, and other equipment.

[0189] It should be understood that, although attached Figure 1 The steps in the flowchart are shown sequentially according to the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders. Furthermore, [the following is a list of steps]. Figure 1-4 At least some of the steps in the process may include multiple sub-steps or sub-stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0190] The above-described embodiments of the present invention describe in detail the regional load forecasting method and the regional novel load capacity optimization layout method. The above-described methods of the present invention can be implemented by various forms of equipment. Therefore, the present invention also discloses a regional load forecasting device and a regional novel load capacity optimization layout device. Specific embodiments are given below for detailed description.

[0191] The regional load forecasting device includes the following modules:

[0192] The historical data acquisition module is used to acquire historical electricity consumption scenarios, historical geographical and climatic characteristics, historical network topology, and historical regional load of the regional power grid.

[0193] The feature fusion module is used to fuse the historical electricity consumption scenarios, historical geographical and climatic characteristics and historical network topology of the historical region at the corresponding time to obtain multi-dimensional load aggregation and interaction historical features.

[0194] The correlation calculation module is used to obtain the multivariate fast maximum information coefficient based on the historical load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics.

[0195] The feature dimensionality reduction module is used to perform dimensionality reduction processing on the multivariate load aggregation interaction history features based on the multivariate fast maximum information coefficient, so as to obtain the dimensionality-reduced multivariate load aggregation interaction history features.

[0196] The model training module is used to input the dimensionality-reduced multivariate load aggregation interaction historical features and historical regional loads into the improved Informer model for training, so as to obtain the trained improved Informer model.

[0197] The real-time feature acquisition module is used to acquire real-time power consumption scenarios, real-time geographic climate features, and real-time network topology, and fuse them to obtain multi-dimensional load aggregation interaction features.

[0198] The real-time load prediction module is used to input the multi-variable load aggregation interaction features into the trained improved Informer model to obtain the real-time regional load.

[0199] Furthermore, based on the aforementioned regional load forecasting device, the regional novel load capacity optimization layout device includes the following modules:

[0200] The single-point access optimization module is used to plan the location, capacity, and operation scheme under the new load single-point access topology. It combines the real-time regional load obtained by the regional load forecasting device to construct a two-level planning model that includes reliability and economy, and obtains the location and capacity scheme of the regional power grid under the new load single-point access topology.

[0201] The regional power grid optimization module is used to adjust the location and capacity of newly added lines and equipment under the new load multi-point access topology, based on the site selection and capacity determination scheme of the new load single-point access regional power grid, with the goal of maximizing the regional power grid benefits after adding load cables under the new load multi-point access topology, to obtain the regional new load site selection and capacity determination optimization scheme.

[0202] For details regarding the regional load forecasting device and the regional new load capacity optimization layout device, please refer to the above-described method limitations; they will not be repeated here. Each module in the aforementioned device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the terminal device in hardware form or independently of it, or stored in the memory of the terminal device in software form, so that the processor can call and execute the corresponding operations of each module.

[0203] In one embodiment, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described regional load forecasting method or regional novel load capacity optimization layout method.

[0204] The computer-readable storage medium may be an electronic storage device such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory), hard disk, or ROM. Optionally, the computer-readable storage medium includes a non-transitory computer-readable storage medium. The computer-readable storage medium has storage space for program code that performs any of the method steps described above. This program code can be read from or written to one or more computer program products, and the program code may be compressed in an appropriate form.

[0205] In one embodiment, the present invention provides a computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program performing the above-described regional load forecasting method or regional novel load capacity optimization layout method.

[0206] The computer device includes a memory, a processor, and one or more computer programs, wherein the one or more computer programs may be stored in the memory and configured to be executed by one or more processors, and the one or more application programs are configured to perform the aforementioned regional load forecasting method or regional novel load capacity optimization layout method.

[0207] A processor may include one or more processing cores. The processor connects to various parts of the computer device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory, and by calling data stored in memory. Optionally, the processor may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem may also be implemented separately as a communication chip, without being integrated into the processor.

[0208] The memory may include random access memory (RAM) or read-only memory (ROM). The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), and instructions for implementing the various method embodiments described above. The data storage area may also store data created by the terminal device during use.

[0209] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A regional load forecasting method, characterized in that, include: Obtain historical electricity consumption scenarios, historical geographical and climatic characteristics, historical network topology, and historical regional load of the regional power grid; By fusing historical electricity consumption scenarios, historical geographical and climatic characteristics, and historical network topology at the corresponding time of the historical region's load, a multi-dimensional load aggregation and interaction historical feature is obtained. Based on the historical load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics, the multivariate fast maximum information coefficient is obtained; The multi-load aggregation interaction history features are subjected to dimensionality reduction processing to obtain the dimensionality-reduced multi-load aggregation interaction history features. The reduced multivariate load aggregation interaction historical features, multivariate fast maximum information coefficient and historical regional load are input into the improved Informer model for training to obtain the trained improved Informer model. The system acquires real-time power consumption scenarios, real-time geographic climate features, and real-time network topology, and integrates them to obtain multi-dimensional load aggregation and interaction features. The multi-variable load aggregation interaction features are input into the trained improved Informer model to obtain the real-time regional load.

2. The regional load forecasting method as described in claim 1, characterized in that, The process of obtaining multivariate fast maximum information coefficients based on the load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics includes: Based on the historical load of each historical region and the historical characteristics of multi-dimensional load aggregation and interaction at the corresponding time, the mutual information between the historical load of the historical region and the historical characteristics of multi-dimensional load aggregation and interaction is obtained. The multivariate fast maximum information coefficients are obtained based on the mutual information normalization.

3. The regional load forecasting method as described in claim 1, characterized in that, The dimensionality reduction process is performed on the multi-dimensional load aggregation interaction history features to obtain the dimensionality-reduced multi-dimensional load aggregation interaction history features. Specifically, this includes the following steps: Obtain the eigenvalue magnitude and cumulative variance contribution rate of each feature factor in the historical features of multivariate load aggregation interaction; When the cumulative variance contribution rate of any combination of multiple feature factors exceeds a preset threshold, the combination of these multiple feature factors is set as the dimensionality-reduced multivariate load aggregation interaction historical feature.

4. The regional load forecasting method as described in claim 1, characterized in that, The improved Informer model includes a dimension segmentation mechanism and a two-stage attention mechanism; The data of each feature dimension in the dimensionality-reduced multivariate composite aggregation interaction history features are divided into sequence segments of a preset length. Each sequence segment is normalized to obtain the normalized-dimensionality-reduced multivariate load aggregation interaction history features. Each sequence segment in the normalized-dimensional reduction multivariate composite aggregation interaction history feature is embedded into a preset position vector to obtain the sequence segment embedded at the time step; By embedding sequence segments at each time step in each feature dimension and combining them with a multi-head probabilistic sparse self-attention mechanism, the correlation between sequence segments at different time steps can be captured. A routing vector mechanism is adopted to capture the correlation between various reduced-dimensional multivariate load aggregation interaction feature sequences by separating information from the routing vector.

5. A novel method for optimizing regional load capacity layout, characterized in that, include: Taking the location, capacity, and operation scheme under the new single-point load access topology as the planning object, and combining the real-time regional load obtained by the regional load forecasting method in any one of claims 1-4, a two-level planning model including reliability and economy is constructed to obtain the location and capacity scheme of the regional power grid under the new single-point load access topology. Based on the site selection and capacity determination scheme for the new type of load single-point access to the regional power grid, with the goal of maximizing the benefits of the regional power grid after adding load cables under the new type of load multi-point access topology, the location and capacity of the newly added lines and equipment under the new type of load multi-point access topology are adjusted to obtain the optimized site selection and capacity determination scheme for the new type of load in the region.

6. The regional novel load capacity optimization layout method as described in claim 5, characterized in that, The bi-level programming model includes an upper-level objective function and a lower-level objective function; The specific expression for the upper-level objective function is as follows: ， in, For the upper-level decision variables in the bilevel programming model, For the lower-level transformation coefficients of the bilevel programming model, For the investment cost of new loads, To incur penalties for power outages, For network line loss costs, Cost of power quality deviation, Penalty costs for load fluctuations For power grid nodes Capacity to accommodate new loads, For nodes New load installation capacity limit, This represents the upper limit for the installation capacity of new loads in the regional power grid. The specific expression for the lower-level objective function is as follows: ， in, For the lower-level decision variables in a bilevel programming model, For the lower-level equality constraints of the bilevel programming model This represents the lower-level inequality constraints in a bilevel programming model.

7. The method for optimizing the layout of new regional load capacity as described in claim 6, wherein the objective is to maximize the regional power grid benefits after adding load cables under the new multi-point access topology, is specifically expressed as follows: ， in To increase revenue through reliability, , The power outage penalty coefficient per unit time. Nodes under new single-point load access conditions The duration of power outage at the load, This represents the expected power outage time for the regional power grid after the integration of new loads. For the investment cost of new load, For the number of nodes in the regional power grid, branch road The number of nodes.

8. The regional novel load capacity optimization layout method as described in claim 7, characterized in that, The reliability of a regional power distribution network is specifically expressed as follows: ， in, For the reliability of the regional power grid, This represents the expected power outage time for the regional power grid after the integration of new loads. , For regional power grid nodes under the new single-point load access situation Power outage duration.

9. A regional load forecasting device, characterized in that, include: The historical data acquisition module is used to acquire historical electricity consumption scenarios, historical geographical and climatic characteristics, historical network topology, and historical regional load of the regional power grid. The feature fusion module is used to fuse the historical electricity consumption scenarios, historical geographical and climatic characteristics and historical network topology of the historical region at the corresponding time to obtain multi-dimensional load aggregation and interaction historical features. The correlation calculation module is used to obtain the multivariate fast maximum information coefficient based on the historical load of each historical region and the corresponding time-based multivariate load aggregation interaction historical characteristics. The feature dimensionality reduction module is used to perform dimensionality reduction processing on the multi-dimensional load aggregation interaction history features to obtain the dimensionality-reduced multi-dimensional load aggregation interaction history features. The model training module is used to input the dimensionality-reduced multivariate load aggregation interaction historical features, multivariate fast maximum information coefficient and historical regional load into the improved Informer model for training, so as to obtain the trained improved Informer model. The real-time feature acquisition module is used to acquire real-time power consumption scenarios, real-time geographic climate features, and real-time network topology, and fuse them to obtain multi-dimensional load aggregation interaction features. The real-time load prediction module is used to input the multi-variable load aggregation interaction features into the trained improved Informer model to obtain the real-time regional load.

10. A novel regional load capacity optimization layout device, characterized in that, include: The single-point access optimization module is used to take the location, capacity and operation scheme under the new load single-point access topology as the planning object, and combine the real-time regional load obtained by the regional load forecasting device as described in claim 9 to construct a two-level planning model that includes reliability and economy, so as to obtain the location and capacity scheme of the regional power grid under the new load single-point access topology. The regional power grid optimization module is used to adjust the location and capacity of newly added lines and equipment under the new load multi-point access topology, based on the site selection and capacity determination scheme of the new load single-point access regional power grid, with the goal of maximizing the regional power grid benefits after adding load cables under the new load multi-point access topology, to obtain the regional new load site selection and capacity determination optimization scheme.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of any one of the regional load forecasting methods of claims 1-4 or any one of the regional novel load capacity optimization layout methods of claims 5-8.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it performs the steps of any one of the regional load forecasting methods of claims 1-4 or any one of the regional novel load capacity optimization layout methods of claims 5-8.

Images