An intelligent power grid adaptive hierarchical cooperative optimization scheduling method and system based on regional coupling strength

Abstract

This invention discloses an adaptive hierarchical cooperative optimization scheduling method and system for smart grids based on regional coupling strength. The method first divides the power grid network into regions according to the network topology and extracts the power flow interaction characteristics between regions as input to a prediction model, dynamically outputting adaptive cooperative control parameters including cooperative window size, hierarchical coordination number, and update delay interval. Based on these parameters, a scheduling matrix is ​​constructed, and parallel scheduling batches and synchronization states are divided accordingly. Within each batch, active sub-regions are aligned with local boundary records, and local enhanced coordination and internal parallel solution are executed alternately layer by layer. The latest conflict-free boundary state is used to update the global operating state of the entire network until power flow convergence. This invention effectively overcomes the cross-regional conflicts and global instability problems caused by information lag in traditional parallel scheduling, significantly improving the computational efficiency and convergence stability of large-scale power grid scheduling.

Description

Technical Field

[0001] This invention relates to the field of smart grid operation optimization and large-scale collaborative control technology, specifically to a smart grid adaptive hierarchical collaborative optimization scheduling method and system based on regional coupling strength. Background Technology

[0002] With the advancement of Industry 4.0, the high penetration of new energy sources, and the integration of large-scale distributed power sources, the scale and network topology of new smart grids are becoming increasingly large and complex. In existing grid operation optimization, optimal power flow calculation (OPF) and unit combination scheduling are core components for ensuring the safe and economical operation of the grid. However, with the rapid increase in the number of control variables, traditional centralized scheduling methods face an extremely large computational dimension, leading to an exponential increase in computational complexity and falling into the so-called "curse of dimensionality." This massive computational load makes it difficult for centralized scheduling to meet the timeliness requirements of real-time, high-frequency grid scheduling.

[0003] To alleviate computational pressure and overcome dimensionality limitations, a partitioned collaborative scheduling mechanism based on the "divide and conquer" approach has been widely adopted in power system optimization. This mechanism decomposes the massive global network model into multiple smaller sub-regions of the power grid, performs dimensionality reduction solutions, and finally combines them into a globally optimal scheduling command. In this process, the temporal coordination and cooperation among the sub-regions (i.e., the collaborative scheduling architecture) directly determines the computational efficiency of the optimization and the quality of the final solution.

[0004] Currently, existing regional collaborative scheduling architectures are mainly limited to two extremes: a completely serial mode and a completely parallel mode. In the completely serial mode, each sub-region of the power grid is optimized one by one in a strict sequence. Although this mode can ensure that each calculation utilizes the latest global network state and effectively avoids cross-regional power flow conflicts, it is still essentially a sequential calculation, which greatly consumes the overall scheduling time and completely violates the original intention of regional scheduling to improve efficiency. Conversely, in the completely parallel scheduling mode, all sub-regions start calculations simultaneously. Although this mode maximizes computational efficiency, it seriously ignores the extremely complex physical coupling relationships between power grid regions (such as power flow constraints and power interactions of cross-regional transmission lines). In the process of blindly parallelizing, there is a serious lag in the state information of adjacent regions referenced by each region during optimization.

[0005] In reality, large-scale smart grids are neither completely decoupled independent systems nor tightly coupled individual units, but rather exhibit a typical "locally overlapping coupling" characteristic. Under this characteristic, pure parallel scheduling lacks an effective information synchronization and coordination mechanism, which easily leads to severe cross-regional power flow conflicts at overlapping boundaries, causing power oscillations across the entire network or even preventing the scheduling results from converging.

[0006] In summary, the academic and industrial communities urgently need an adaptive optimization framework that can effectively control the degree of information lag between regions while ensuring parallel computing efficiency, and can quickly resolve cross-regional conflicts in overlapping areas of the power grid with different coupling strengths, so as to adapt to the new smart grid with complex structure and variable operating status. Summary of the Invention

[0007] The purpose of this invention is to provide a smart grid optimization scheduling method based on regional coupling strength adaptive hierarchical coordination. In view of the problems of high computational complexity of existing centralized scheduling and easy cross-regional conflict in pure parallel partition scheduling, this invention proposes a smart grid adaptive hierarchical coordination optimization scheduling method based on regional coupling strength. Through adaptive optimization combining adjustable coordination window and hierarchical coordination mechanism, it achieves grid scheduling with fast convergence speed, high computational efficiency and strong system stability.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A smart grid optimization scheduling method based on regional coupling strength adaptive hierarchical coordination includes the following steps: S1. Power Grid Regional Division and Coupling Feature Extraction: The power grid is divided into regions according to the network topology, and the power flow interaction feature data between each region is extracted and input into the prediction model. The adaptive cooperative control parameters for the current power grid operating state are dynamically output. The cooperative control parameters include the cooperative window size, the number of hierarchical coordinations, and the update delay interval. S2. Initialization and scheduling matrix construction of adaptive hierarchical collaborative architecture: Based on the obtained collaborative control parameters, a scheduling matrix is ​​constructed, all the divided power grid sub-regions are allocated to the scheduling matrix, and the total number of parallel scheduling batches is set. S3. Parallel scheduling batch division and state synchronization based on cooperative window: The scheduling tasks are executed sequentially according to the total number of parallel scheduling batches. In the initial stage of each batch, the set of currently active regions is determined according to the scheduling matrix and the size of the cooperative window, and the local boundary records of each active region in this batch are synchronously initialized using the current global state of the power grid. S4. Alternating advancement of multi-layer local coordination and regional parallelism: Within the current scheduling batch, based on the scheduling matrix and the hierarchical coordination number, local enhanced coordination is performed layer by layer. While ensuring internal power balance, each active region aligns with the local boundary record and calls the underlying solver for parallel solution, and then updates its local boundary state. S5. Cross-regional conflict enhancement and resolution and global power grid state update: After all layer coordination in the current batch is completed, the latest conflict-free local boundary state generated by the local enhanced coordination is used to uniformly update the global operating state of the entire network, and then proceed to the next scheduling batch until the power flow of the entire network meets the convergence condition and the optimal scheduling instruction is output.

[0009] Furthermore, in step S1, the power flow interaction characteristic data between each region is extracted, specifically including: using the power flow sensitivity matrix, calculating the proportion of completely independent nodes within a region that do not exchange power flow with external regions, and the proportion of nodes with cross-regional connecting lines.

[0010] Further, in step S2, constructing the scheduling matrix specifically involves: constructing a scheduling matrix with a size equal to the number of layers multiplied by the total number of columns; placing all the divided power grid sub-regions sequentially in each row of the matrix; shifting each row backward by the corresponding update delay interval of control cycles as a buffer delay for information transmission to avoid power oscillations caused by strong coupling; and calculating the total number of parallel scheduling batches based on the real-time requirements of a single scheduling operation.

[0011] Furthermore, in step S3, at the initial stage of each batch, the column boundary interval corresponding to the current batch is calculated based on the scheduling matrix and the collaborative window size, thereby accurately locking the set of active regions corresponding to the current batch.

[0012] Furthermore, in step S4, when calling the underlying solver for parallel solution, the original dual interior point method or mixed integer linear programming algorithm is used to solve the optimal solution of unit output and power flow distribution while satisfying the internal operating constraints of the region and aligning with the cross-regional interactive power recorded by the local boundary.

[0013] Furthermore, in step S5, the convergence condition of the entire network power flow means that the change in the total system operating cost or network loss and over-limit penalty value of the entire network within a consecutive preset number of scheduling batches is less than the set convergence accuracy threshold.

[0014] The present invention also provides an adaptive hierarchical collaborative optimization scheduling system for smart grids based on regional coupling strength.

[0015] The present invention also provides a computer device.

[0016] The present invention also provides a computer-readable storage medium.

[0017] The present invention has the following advantages and effects compared with the prior art: 1) High computational efficiency and near-linear acceleration: By decomposing the massive power grid scheduling task into multiple regions and introducing a collaborative window, large-scale distributed parallel computing is achieved, which significantly reduces the solution time of a single scheduling operation and achieves near-linear parallel acceleration while ensuring system stability.

[0018] 2) Strong system stability and fast conflict resolution: In response to the frequent power fluctuations caused by the high penetration of new energy sources, this invention uses a hierarchical coordination mechanism and dynamic update delay to trigger high-frequency local enhanced coordination when over-limit or imbalance is detected, which effectively overcomes the global instability problem caused by information lag in traditional pure parallel scheduling.

[0019] 3) Good adaptability and scenario compatibility: The present invention can dynamically adjust the scale and update frequency of regional coordination according to changes in the power grid operation mode, without manual intervention, and is particularly suitable for new smart grid scenarios with complex structures. Attached Figure Description

[0020] Figure 1 This is a flowchart of an adaptive hierarchical collaborative optimization scheduling method for smart grids based on regional coupling strength, provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of the execution timeline of the adaptive hierarchical collaborative architecture scheduling and dispatching provided in the embodiments of the present invention. Figure 3 This is a structural block diagram of a smart grid adaptive hierarchical collaborative optimization scheduling system based on regional coupling strength, provided by an embodiment of the present invention. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments. The described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] This invention, taking large-scale optimal power flow (OPF) and generator unit scheduling in smart grids as its background, illustrates an adaptive hierarchical collaborative optimization scheduling method for smart grids based on regional coupling strength. With the expansion of the power grid and the integration of new energy sources, grid scheduling is modeled as a large-scale global optimization problem with extremely stringent physical constraints (such as node voltage balance and line thermal stability limits). This invention extracts the power flow interaction characteristics between power grid regions, predicts the optimal collaborative architecture parameters, and constructs a scheduling matrix, thereby significantly improving the parallel solution efficiency and convergence stability across the entire network while ensuring local scheduling accuracy.

[0023] like Figure 1As shown in the figure, an adaptive hierarchical cooperative optimization scheduling method for smart grids based on regional coupling strength is provided by an embodiment of the present invention, which includes the following steps: S1. Power grid area division and coupling feature extraction.

[0024] Obtain the topology, line parameters, and node power generation and consumption forecast data of the smart grid to be scheduled. Model the scheduling problem of the smart grid as a large-scale global optimization problem, design the decision space of the optimization problem (including the active / reactive power output of each generator, flexible load switching, etc.), and the optimization objective is to minimize the total operating cost of the entire network system and the network loss penalty value.

[0025] Based on the physical connectivity and management affiliation of the power grid, the entire grid is divided into: The system is divided into sub-regions; then, the power flow interaction characteristics between sub-regions are extracted to quantitatively assess the regional coupling strength of the power grid. Specifically, the proportion of completely independent nodes (those that do not exchange power flow with other regions) within a region and the proportion of nodes coupled by cross-regional tie lines are calculated using a power flow sensitivity matrix.

[0026]

[0027] in, This indicates the proportion of completely independent nodes within the region. It represents a set of completely independent nodes or control variables within a region. This indicates the proportion of coupling nodes in the cross-regional tie line. This is the direct physical manifestation of the regional coupling strength. A larger value indicates a tighter electrical coupling between regions; The total dimension representing the power grid control variables (i.e., the total number of nodes or devices in the entire network). Representing the A set of variables for each subregion Representing the The set of variables for each subregion , This represents the total number of sub-regions of the entire power grid.

[0028] The power flow sensitivity matrix is ​​a mature steady-state analysis tool in the power system field. Specifically, it can be obtained by solving the inverse of the AC power flow Jacobian matrix, or quickly calculated using the nodal susceptance matrix under the DC power flow approximation. In one embodiment, the sensitivity matrix of nodal injected power and branch active power flow under the DC power flow approximation is directly used to quantify the electrical coupling strength between different regions.

[0029] The power flow interaction characteristics calculated above (including and The data is input into a pre-established architecture parameter prediction model. Because the optimal mapping relationship between collaborative control parameters and inter-regional coupling strength is not a simple linear one (e.g., when the proportion of coupling nodes in cross-regional connecting lines...), this relationship is affected. At higher levels, the collaborative window size and hierarchical coordination number (It needs to be increased in a non-linear manner to ensure sufficient cross-regional conflict resolution), therefore, the original power flow interaction features need to be extended non-linearly to improve the model's fitting accuracy.

[0030] In one embodiment, the architecture parameter prediction model is a lightweight multinomial regression model. Its input is the power flow interaction characteristics (…). and The architecture parameter prediction model first analyzes the original power flow interaction characteristics ( and Perform polynomial feature expansion to construct the feature expansion vector:

[0031] Feature extension vector It includes constant terms, first-order terms, second-order interaction terms, and square terms of the original tidal current interaction characteristics, used to capture the nonlinear interaction effects between tidal current interaction characteristics. This polynomial extension enables the architecture parameter prediction model to fit nonlinear mapping relationships while maintaining computational lightweightness (without introducing complex structures such as deep neural networks).

[0032] Subsequently, the architecture parameter prediction model is used through a weight matrix. A linear mapping is performed on the feature extension vector to output adaptive cooperative control parameters for the current power grid operating state:

[0033] in For size The weight matrix, The dimension of the feature extension vector. Weight matrix. Obtained through supervised model fitting during offline training; This represents the hierarchical coordination number.

[0034] In one embodiment, the offline training of the architecture parameter prediction model does not rely on a limited historical record of power grid operation, but is based on an active construction method of the underlying feature space. It synthesizes power grid operation scenarios with different interaction characteristics through simulation. The specific steps are as follows: Step 1: Generation of power flow simulation samples based on active construction method 1. Setting the scheduling dimension: Determine the total dimension of the power grid control variables. This corresponds to the total number of generator output setpoints to be dispatched across the entire network, flexible load peak shaving and valley filling commands, and the status of energy storage devices. This is achieved by setting different benchmark dimensions (such as...). =500, 1000, etc.), to ensure that the architecture parameter prediction model can adapt to the scheduling computing power requirements of different scales, from local microgrids to large power grids.

[0035] 2. Constructing independent features based on topological decoupling, i.e., the proportion of completely independent nodes within a region. Independent nodes are constructed using the physical connections of the power grid. In a simulation environment, by artificially disconnecting the electrical connections between nodes within a specific area and the external area (e.g., disconnecting specific branch circuit breakers), a set of nodes with dimensions of [missing information] is constructed. A completely separable set of variables, within which the node power flow distribution is only affected by the control variables of this region and does not generate cross-regional power interaction. The proportion of independent nodes is achieved by changing the number of isolated nodes. Precise and controlled generation.

[0036] 3. Based on the overlapping structure of tie lines, the coupling characteristics are constructed, i.e., the proportion of coupling nodes in cross-regional tie lines. For non-independent regions, coupling strength is constructed by controlling the overlap step size of inter-regional tie lines. In the simulation, adjacent sub-regions are partially overlapped in the electrical path design (i.e., some boundary nodes belong to the optimization subspaces of two regions simultaneously), forming an overlapping variable set.

[0037] Strongly coupled construction: reducing the electrical distance between regions or increasing the number of shared boundary nodes (i.e., short step size) makes... Increase the simulation of strongly coupled power grid operation scenarios with frequent power flow interactions.

[0038] Weakly coupled construction: increasing the electrical impedance between regions or reducing the number of overlapping nodes (i.e., long step size) makes... Reduce, thereby obtaining the desired proportion of coupled nodes. .in, This represents the total number of control variables in the entire network that belong to the overlapping variable set (i.e., overlapping nodes that cross different regional boundaries).

[0039] Step 2: Automated Acquisition of Monitoring Signals Use a script to loop through the feature space and The system combines and synthesizes hundreds of virtual power flow scenarios with varying physical structures. For each synthesized scenario, it invokes the underlying solver (such as the primal-dual interior-point method) to fully run all candidate combinations of cooperative parameters. The convergence time and total network operating cost (including network loss and over-limit penalties) for each set of parameters after being substituted into the scheduling process are recorded. The optimal parameter combination for overall evaluation is selected based on a weighted ranking of convergence speed and scheduling cost. As a supervisory label (GroundTruth) for this scenario, it forms "input-output" data pairs, and ultimately builds a logical mapping library from the physical interaction characteristics of the power grid to the optimal scheduling architecture.

[0040] Step 3: Solving for the weight matrix To eliminate differences in the numerical magnitude of different parameters (such as the size of the collaborative window) With update delay interval (Due to the different dimensions), the target label is subjected to dimensionless normalization.

[0041] The normalization process specifically converts the optimal label into a proportion relative to the upper limit of the search space:

[0042] in, , , These are the pre-set collaborative window size, hierarchical coordination number, and update delay interval, respectively, based on the power grid scale and hardware computing power. , , These are the optimal collaborative window sizes. Optimal hierarchical coordination number and optimal update delay interval After dimensionless processing, the ratio relative to its search upper limit.

[0043] The normalized labels are used to construct the target matrix. The design matrix formed by combining feature extension vectors The optimal weight matrix is ​​obtained by using the ordinary least squares (OLS) method. :

[0044] In one embodiment, for large-scale decoupled systems, if multicollinearity exists among features, a regularization term can be introduced and a ridge regression algorithm can be used to improve the model's generalization ability.

[0045] Step 4: Model Deployment and Discretization Validity Conversion During the online scheduling phase, the data extracted in real time will be... , Substitute into the model to obtain the continuous prediction vector ,in, , , These represent the continuous normalized predicted values ​​(or predicted scale values) of the architecture parameter prediction model for the current grid operating conditions, corresponding to the cooperative window size, hierarchical coordination number, and update delay interval, respectively. Since grid dispatch parameters must be discrete integers, they require legalization transformation through the following pipeline: 1. Inverse Scaling: Normalizes continuous predicted values. , , Multiply by the corresponding threshold respectively , take it from The space is restored to the order of magnitude of the original physical parameters. 2. Rounding: Performs rounding operations, converting scaled continuous values ​​into discrete integers. For example:

[0046] 3. Clipping: To ensure that parameters comply with power grid physical constraints (e.g., the window size cannot exceed the total number of sub-regions). Perform the following truncation operation:

[0047]

[0048]

[0049] in, , , These represent the original integer collaborative control parameters that have been reverse scaled and rounded to the nearest integer, but before physical boundary truncation protection has been implemented. This represents the total number of sub-regions of the entire power grid.

[0050] The final output is the optimal cooperative control parameters applicable to the current scenario. This process has extremely low computational overhead and can achieve real-time response in milliseconds.

[0051] S2. Initialization of the adaptive hierarchical collaborative architecture and construction of the scheduling matrix.

[0052] Combination Figure 2 The diagram shown illustrates the hierarchical collaborative scheduling execution timeline, based on the adaptive collaborative control parameters obtained in step S1. The size of the construction is Scheduling matrix All the divided power grid sub-regions are assigned to the scheduling matrix, and the total number of parallel scheduling batches is set. number of columns , This represents the total number of sub-regions of the entire power grid. This dispatching and scheduling matrix... It controls the computation timing of each sub-region of the entire network.

[0053] Combination Figure 2 The adaptive hierarchical collaborative architecture and scheduling matrix of the present invention Please provide a detailed explanation.

[0054] exist Figure 2 In the middle, the horizontal axis represents the time sequence of scheduling progress (corresponding to the column index of the scheduling matrix). The vertical axis visually reflects the time execution process of the algorithm; the vertical axis represents the number of hierarchical coordination layers. This indicates the depth of local reinforcement applied to active regions within the same batch. Each rectangle in the figure represents a specific sub-region of the power grid (e.g., region). ,area Optimization computation tasks (e.g., etc.).

[0055] like Figure 2 As shown, they are located in the same vertical time series (same column index). Multiple sub-regions of the algorithm will be solved in parallel by multiple computing nodes at the bottom layer, which ensures the efficiency of the algorithm.

[0056] The size of the collaborative window is marked in the figure. This determines the horizontal span of each large-batch state synchronization; the update delay interval is indicated by the dashed line with an arrow. This represents the time misalignment of the start of tasks in sub-regions between adjacent layers.

[0057] This stepped, descending scheduling structure allows each sub-region to optimize itself using staggered update delay intervals. Obtain historical boundary states of adjacent areas in advance; the "cross-regional power flow conflict areas" or "coupling areas" marked by dashed boxes in the figure are identified through multiple layers on the vertical axis (such as...). The alternating execution of the process allows sub-regions with electrical coupling to undergo multiple high-frequency state swaps and alignments locally before formally submitting a global update, thereby fundamentally resolving cross-region power flow conflicts while ensuring computational speed.

[0058] from Figure 2 The tiered scheduling structure clearly shows the scheduling matrix. The generation rule is: to generate all sub-regions Place them sequentially in each row of the matrix, while also placing the first... Shift back Each control cycle serves as a buffer delay for information transmission (e.g., ...). Figure 2Update delay interval marked in the middle (As shown). This update delay interval The introduction of this method effectively avoids power oscillations caused by instantaneous information exchange during pure parallel solution of strongly coupled power grids. Its mathematical equivalent expression is:

[0059] in, Indicates the first The first coordination level, the first The shift position corresponds to the first Each power grid sub-region This indicates that the shift position is empty. , .

[0060] Simultaneously, the maximum iterative computing power of each sub-region is allocated according to the real-time requirements of a single scheduling operation, and the total number of parallel scheduling batches is calculated. :

[0061] S3. Parallel scheduling batch division and state synchronization based on cooperative window: The scheduling tasks are executed sequentially according to the total number of parallel scheduling batches. In the initial stage of each batch, the set of currently active regions is determined based on the scheduling matrix and the size of the cooperative window, and the local boundary records of each active sub-region in this batch are synchronously initialized using the current global state of the power grid.

[0062] In one embodiment, this step specifically includes: Based on the total number of parallel scheduling batches The coordinated scheduling of the entire network is carried out in batches in sequence. (correspond Figure 2 The execution batch range is marked below the middle horizontal axis.

[0063] In each coordinated scheduling batch In the initial stage, the current batch is calculated in the scheduling matrix. The corresponding column boundary interval (i.e.) Figure 2 Window size marked in the middle (Horizontal range covered): Starting column of column boundary interval:

[0064] Terminating column:

[0065] in, Indicates the first The starting column index corresponding to each scheduling batch Indicates the first The index of the termination column corresponding to each scheduling batch. For the size of the collaboration window, This represents the number of columns in the scheduling matrix; Subsequently, the global power grid status currently recorded by the dispatch master station is extracted. (Including voltage magnitude and phase angle of all boundary nodes, tie-line switching power, etc.), and the current column boundary interval Local boundary records of all sub-regions involved Synchronous initialization to global power grid state This is to ensure that the states are aligned between batches.

[0066] S4. Alternating advancement of multi-layer local coordination and regional parallelism: Within the current scheduling batch, based on the scheduling matrix and the hierarchical coordination number, local enhanced coordination is performed layer by layer. While ensuring internal power balance, each active sub-region aligns with the local boundary record and calls the underlying solver for parallel solution, and then updates its local boundary state.

[0067] In this invention, the local enhanced coordination refers to the process of repeatedly performing boundary alignment, parallel solving, and local boundary state updates on the currently active sub-regions within the current scheduling batch, based on the scheduling matrix and the hierarchical coordination number. Specifically, each active sub-region uses the cross-regional tie line interaction power, boundary node voltage amplitude, and phase angle recorded in the local boundary record as boundary constraints. While satisfying the internal power balance, generator output constraints, line capacity constraints, and node voltage constraints of the region, it calls the underlying solver for parallel solving. After the solution is completed, the obtained local boundary state is updated to the local boundary record. By executing the above process layer by layer within the same scheduling batch, each active sub-region completes multiple local boundary alignments before the unified update of the global network operating status, thereby reducing cross-regional power flow conflicts caused by boundary state lag during parallel solving.

[0068] In one embodiment, this step specifically includes: In the current scheduling batch Inside, layer by layer Implement localized enhanced coordination (corresponding) Figure 2 The hierarchical coordination number marked on the left (The vertical levels covered). Number of hierarchical coordinations. This determines the density of local probing and alignment within each sub-region before the next global state update.

[0069] In the Layer, based on the scheduling matrix (in and Locate the currently active set of regional tasks.

[0070] Parallel solutions are initiated simultaneously for all sub-regional scheduling centers within the active regional task set. For example... Figure 2 The corresponding task boxes within each layer are shown. During the solution process, each sub-region uses the interaction power of the cross-regional connection lines as an equality boundary constraint, along with the local boundary records. Alignment is performed. In one embodiment, a standard low-level solver for power systems, such as the original dual interior-point method (PDIPM) or mixed integer linear programming (MILP), is invoked to optimize the unit output combination within the sub-region. After the solution is completed, a new local boundary state is generated and updated to the local boundary record. .

[0071] S5. Cross-regional conflict enhancement and resolution and global power grid state update: After all layer coordination in the current batch is completed, the latest conflict-free local boundary state generated by the local enhanced coordination is used to uniformly update the global operating state of the entire network, and then proceed to the next scheduling batch until the power flow of the entire network meets the convergence condition and the optimal scheduling instruction is output.

[0072] In this embodiment of the invention, the conflict-free local boundary state refers to the state in which the regional boundary state recorded in the local boundary record meets the preset verification conditions after all layers of the current scheduling batch have been coordinated and executed.

[0073] The preset verification conditions refer to the conditions used to determine whether a local boundary state can be used to update the overall network operation status. These conditions include: the interactive power deviation between the sub-regions at both ends of the cross-regional tie line does not exceed a preset threshold; the voltage amplitude and phase angle deviation of the boundary node do not exceed a preset threshold; and the line capacity constraint, node voltage constraint, generator output constraint, and regional power balance constraint have not exceeded the limits. The absence of conflict does not mean that there is no electrical coupling relationship between sub-regions, but rather that the current local boundary state meets the preset verification conditions for updating the overall network operation status.

[0074] In the current scheduling batch All After the layer coordination is completed, local boundary records are used. The latest conflict-free local boundary states of each sub-region are uniformly updated to reflect the global operational status of the entire network. and the corresponding total system operating cost .

[0075] This mechanism ensures that cross-regional power flow inconsistencies in overlapping boundary areas are kept within a preset threshold range before proceeding to the next scheduling batch, thereby reducing the risk of global state oscillations caused by state lag in parallel solution. Steps S3 to S5 are repeated until all are completed. Each scheduling batch. Determine whether the power flow of the entire network meets the convergence condition, that is, determine whether the change in the total system operating cost or network loss and over-limit penalty value of the entire network within a preset number of consecutive scheduling batches is less than the set convergence threshold. If it meets the condition, output the final globally optimal scheduling instruction.

[0076] In one embodiment, the optimal scheduling instruction includes, but is not limited to, physical control parameters such as the active and reactive power output setpoints of each conventional generator set, the grid connection output control instruction of the new energy power station, the peak shaving and valley filling response instruction of the flexible load, the charging and discharging power setting status of the energy storage device, and the power exchange plan value of the inter-regional interconnection line.

[0077] In one embodiment, such as Figure 3 As shown, an adaptive hierarchical collaborative optimization scheduling system for smart grids based on regional coupling strength is provided, including: a feature extraction and parameter prediction module, a scheduling matrix construction module, a batch partitioning and state synchronization module, a hierarchical coordination and parallel solution module, and a global update and output module. The functions of each module correspond one-to-one with steps S1 to S5 in the aforementioned method embodiments, and will not be repeated here.

[0078] In one embodiment, a computer device is provided, including a memory and a processor, the memory being electrically connected to the processor, the memory storing a computer program that, when executed by the processor, causes the processor to implement the aforementioned smart grid optimization scheduling method.

[0079] In one embodiment, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned smart grid optimization scheduling method.

[0080] In summary, the adaptive hierarchical collaborative optimization scheduling method for smart grids based on regional coupling strength proposed in this invention effectively solves the problem of convergence of cross-regional power flow conflicts during large-scale regional scheduling of power systems by establishing a unified hierarchical collaborative architecture. By introducing a collaborative window and update delay interval, it retains the efficiency of parallel computing while ensuring strict satisfaction of power flow constraints through local enhanced coordination, demonstrating broad engineering application prospects in complex and ever-changing smart grid scheduling scenarios. The system, equipment, and medium described possess the same technical effects.

[0081] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined in this invention may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart grid adaptive hierarchical cooperative optimization scheduling method based on regional coupling strength, characterized in that, Includes the following steps: The power grid is divided into regions based on its network topology, and the power flow interaction characteristics between each sub-region are extracted and input into the prediction model. The adaptive cooperative control parameters for the current power grid operating state are dynamically output. The cooperative control parameters include the cooperative window size, the number of hierarchical coordinations, and the update delay interval. A scheduling matrix is ​​constructed based on adaptive cooperative control parameters, all the divided power grid sub-regions are assigned to the scheduling matrix, and the total number of parallel scheduling batches is set. The scheduling tasks are executed sequentially according to the total number of parallel scheduling batches. At the initial stage of each batch, the set of currently active regions is determined based on the scheduling matrix and the size of the coordination window, and the local boundary records of each active sub-region within the batch are synchronously initialized using the current global state of the power grid. Within the current scheduling batch, based on the scheduling matrix and the hierarchical coordination number, local enhanced coordination is performed layer by layer. Each active sub-region aligns with the local boundary record and performs parallel solution while ensuring internal power balance, and then updates its local boundary state. After all layer coordination in the current batch is completed, the latest conflict-free local boundary state generated by the local enhanced coordination is used to uniformly update the global operating state of the entire network, and then proceed to the next scheduling batch until the power flow of the entire network meets the convergence condition and the optimal scheduling instruction is output.

2. The smart grid adaptive hierarchical cooperative optimization scheduling method based on regional coupling strength according to claim 1, characterized in that, The extraction of power flow interaction features between sub-regions specifically includes: Using the power flow sensitivity matrix, calculate the proportion of completely independent nodes within the region, and the proportion of nodes coupled by cross-regional tie lines: in, This represents the total dimension of the power grid control variables. It represents a set of completely independent nodes or control variables within a region. Indicates the first The set of control variables corresponding to each power grid sub-region Indicates the first The set of control variables corresponding to each power grid sub-region , Let be the total number of sub-regions of the entire power grid, and be an indicator function. This indicates the proportion of completely independent nodes within a region. This indicates the proportion of coupling nodes in the cross-regional connection line.

3. The smart grid adaptive hierarchical cooperative optimization scheduling method based on regional coupling strength according to claim 1, characterized in that, The process involves constructing a scheduling matrix based on adaptive cooperative control parameters, allocating all divided power grid sub-regions to the scheduling matrix, and setting the total number of parallel scheduling batches. Specifically: Build size is Scheduling matrix , where column number for: in, For the hierarchical coordination number, To update the delay interval, The total number of power grid sub-regions divided into the entire network; All power grid sub-regions are placed sequentially in the scheduling matrix. Each row, simultaneously the first Shift backward Each control cycle serves as a buffer delay for information transmission, in which ; The scheduling matrix The elements in the set satisfy: in, Represents the column index of the scheduling matrix. Indicates the first The first coordination level, the first The shift position corresponds to the first Each power grid sub-region This indicates that the shift position is empty; Based on the size of the collaboration window and the number of columns in the scheduling matrix Calculate the total number of parallel scheduling batches: in, This represents the total number of parallel scheduling batches. Set the size of the collaboration window.

4. The smart grid adaptive hierarchical cooperative optimization scheduling method based on regional coupling strength according to claim 1, characterized in that, In the initial stage of each batch, the set of currently active regions is determined based on the scheduling matrix and the size of the collaboration window, specifically as follows: Based on the total number of parallel scheduling batches The coordinated scheduling of the entire network is carried out in batches in sequence. ,in ; In each coordinated scheduling batch In the initial stage, the current batch is calculated in the scheduling matrix. The corresponding column boundary interval: in, Indicates the first The starting column index corresponding to each scheduling batch Indicates the first The index of the termination column corresponding to each scheduling batch. For the size of the collaboration window, This represents the number of columns in the scheduling matrix; The set of active regions within the current batch is determined based on the calculated column boundary intervals: in, Indicates the first The set of active regions within a scheduling batch This represents the scheduling matrix. The power grid sub-region corresponding to the non-empty element in the middle. Represents the column index of the scheduling matrix.

5. The smart grid adaptive hierarchical cooperative optimization scheduling method based on regional coupling strength according to claim 1, characterized in that, The local enhanced coordination includes: within the current scheduling batch, based on the scheduling matrix and the hierarchical coordination number, repeatedly performing boundary alignment, parallel solution, and local boundary state update for the current active sub-region; wherein, each active sub-region uses the cross-regional tie line interaction power, boundary node voltage amplitude, and phase angle recorded in the local boundary record as boundary constraints, and performs parallel solution by calling the underlying solver. The underlying solver uses the original dual interior point method or mixed integer linear programming algorithm to solve for the optimal solution of unit output and power flow distribution within each sub-region. After the solution is completed, the obtained local boundary state is updated to the local boundary record.

6. A smart grid adaptive hierarchical cooperative optimization scheduling method based on regional coupling strength according to any one of claims 1-5, characterized in that, The conflict-free local boundary state refers to the state in which the local boundary state recorded in the local boundary record meets the preset verification conditions. The preset verification conditions include: the interactive power deviation between the two sub-regions at both ends of the cross-regional tie line does not exceed the preset threshold, the voltage amplitude and phase angle deviation of the boundary node do not exceed the preset threshold, and the line capacity constraint, node voltage constraint, unit output constraint and regional power balance constraint have not exceeded the limit. The network power flow meets the convergence condition, which means that the change in the total system operating cost or network loss and limit violation penalty value of the entire network within a preset number of consecutive scheduling batches is less than the set convergence threshold.

7. A smart grid adaptive hierarchical collaborative optimization scheduling system based on regional coupling strength, characterized in that, The system for implementing the method according to any one of claims 1-6 comprises: The feature extraction and parameter prediction module is used to divide the power grid into regions according to the network topology of the power grid, extract the power flow interaction features between each sub-region and input them into the prediction model, and dynamically output adaptive cooperative control parameters for the current power grid operating state. The cooperative control parameters include cooperative window size, hierarchical coordination number and update delay interval. The scheduling matrix construction module is used to construct a scheduling matrix based on adaptive cooperative control parameters, allocate all the divided power grid sub-regions to the scheduling matrix, and set the total number of parallel scheduling batches. The batch partitioning and state synchronization module is used to execute scheduling tasks sequentially according to the total number of parallel scheduling batches. In the initial stage of each batch, the set of currently active regions is determined based on the scheduling matrix and the size of the coordination window, and the local boundary records of each active sub-region within the batch are synchronized and initialized using the current global state of the power grid. The hierarchical coordination and parallel solution module is used to perform local enhanced coordination layer by layer within the current scheduling batch based on the scheduling matrix and the hierarchical coordination number. Each active sub-region aligns with the local boundary record and performs parallel solution while ensuring internal power balance, and then updates its local boundary state. The global update and output module is used to update the global operating state of the entire network uniformly after all layer coordination in the current batch has been completed, using the latest conflict-free local boundary state generated by the local enhanced coordination, and then proceed to the next scheduling batch until the power flow of the entire network meets the convergence condition and the optimal scheduling instruction is output.

8. The smart grid adaptive hierarchical collaborative optimization scheduling system based on regional coupling strength according to claim 7, characterized in that, The prediction model in the feature extraction and parameter prediction module is a model trained offline using historical power grid operation slice data and power flow simulation samples.

9. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 6.

Images