A power distribution network distributed autonomous and centralized coordination planning method for massive flexible load access

Abstract

The application belongs to the technical field of power distribution network optimization planning. A kind of power distribution network distributed autonomous and centralized coordination planning method for massive flexible load access, characterized by comprising the following steps: S1, collect power distribution network planning basic data, the basic data includes the existing topological structure of power distribution network, line impedance, rated capacity and other line parameters, substation rated capacity, operating constraint parameters;S2, based on the electrical coupling strength of power distribution network node and the double index of flexible resource density;S3, construct global centralized coordination-regional distributed autonomous double-layer mixed integer programming model;S4, improved ADMM distributed iteration algorithm with adaptive parameter adjustment and integer variable projection mechanism;S5, according to the optimal decision variable obtained by iteration solution. Through multi-index dynamic partition, double-layer mixed integer modeling and adaptive distributed solution, the efficient, accurate and globally optimal planning of large-scale flexible power distribution network is realized.

Description

Technical Field

[0001] This invention belongs to the field of distribution network optimization planning technology, specifically relating to a distributed autonomous and centralized coordination planning method for distribution networks with massive flexible load access. Background Technology

[0002] With the continuous deepening of the construction of new power systems, massive flexible resources such as electric vehicles, intelligent temperature-controlled loads, and controllable industrial and commercial loads are being connected to medium and low voltage distribution networks on a large scale, completely overturning the inherent characteristics of traditional distribution networks, such as rigid loads, unidirectional power flow, and fixed operating modes. Flexible loads are characterized by time-adjustable, dynamic fluctuations, and scattered distribution, which greatly expands the resource, control, and time-series dimensions of distribution networks, placing higher demands on the refinement, efficiency, and global coordination of distribution network planning.

[0003] Existing distribution network planning models are mainly divided into two categories: centralized planning and hierarchical distributed planning. Traditional centralized planning incorporates all nodes, lines, loads, and distributed resources across the entire network into a unified optimization model, resulting in a quadratic increase in model dimensionality compared to the scale of the power grid nodes. In large-scale distribution network scenarios with densely integrated flexible loads, the computational complexity of centralized planning rises dramatically, easily leading to problems such as computational dimensionality explosion, excessively long iteration times, and solution failures, making it unsuitable for the planning needs of large-scale power grids.

[0004] To overcome the computational bottleneck of centralized planning, existing technologies have gradually developed distributed planning approaches based on hierarchical decomposition and regional partitioning. Mainstream solutions include Benders decomposition planning, Objective Cascade Analysis (ATC) planning, basic ADMM distributed planning, and community partitioning planning. Hierarchical planning methods based on Benders decomposition and ATC suffer from problems such as numerous iterations, sensitivity to initial solution parameters, and strong coupling between subproblems, resulting in poor adaptability in scenarios with multiple resources and highly fluctuating flexible loads. While distributed planning schemes based on basic ADMM can achieve parallel solutions across multiple regions, existing technologies are mostly applied to power grid operation and dispatching scenarios with purely continuous variables. For distribution network planning scenarios involving discrete 0-1 variables such as new grid construction and renovation, the basic ADMM algorithm with fixed penalty parameters is prone to iterative oscillations and unstable convergence, and the parameters require repeated manual adjustments, limiting its engineering practicality.

[0005] At the level of power grid zoning planning, existing conventional zoning methods mostly rely on administrative regions and single indicators such as electrical distance, focusing only on the electrical interconnectivity of the power grid topology and failing to consider the spatial distribution balance of massive flexible loads, distributed power sources, and energy storage resources. Some zoning schemes using the Louvain community algorithm also only use electrical coupling strength as the single optimization objective, which easily leads to polarization of flexible resource allocation in the autonomous units after zoning. This results in some units having redundant adjustment resources and others having insufficient autonomous capabilities, failing to adapt to the planning requirements of dynamic adjustment of flexible loads. Consequently, subsequent local resource planning and global network planning have poor adaptability and insufficient coordination.

[0006] In summary, existing distribution network planning technologies generally suffer from three major shortcomings: First, the zoning mechanism is not adapted to the characteristics of massive flexible resource distribution, and static, single-index zoning leads to uneven unit autonomy. Second, the hierarchical planning model fails to distinguish between discrete network planning and continuous resource optimization, resulting in ambiguous boundaries of responsibility and poor coordination between global and local planning. Third, traditional distributed solution algorithms cannot adapt to mixed-integer planning scenarios with integer variables, exhibiting insufficient convergence stability and scenario versatility. To address these technical deficiencies, there is an urgent need to propose a novel hierarchical collaborative planning method that balances the electrical characteristics of the power grid, the distribution characteristics of flexible resources, and large-scale solution efficiency to achieve optimal collaborative planning of the global distribution network structure and local flexible resources. Summary of the Invention

[0007] The purpose of this invention is to provide a distributed autonomous and centralized coordination planning method for distribution networks with massive flexible load access. Through multi-index dynamic partitioning, two-layer mixed integer modeling and adaptive distributed solution, it achieves efficient, accurate and globally optimal planning for large-scale flexible distribution networks.

[0008] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: a distributed autonomous and centralized coordination planning method for distribution networks with massive flexible load access, characterized by comprising the following steps: S1. Collect basic data for distribution network planning. The basic data includes the existing topology of the distribution network, line impedance, rated capacity and other line parameters, substation rated capacity and operating constraint parameters, benchmark load time series curves of each node (including 10kV bus nodes, distribution transformer substation nodes, flexible load and distributed power source access nodes), as well as the location of distributed power source access points and output time series characteristics, flexible load access locations and time series adjustment interval parameters, and annual load growth forecast data within the planning period. S2. Based on the dual indicators of electrical coupling strength and flexible resource density of distribution network nodes, an improved Louvain community detection algorithm is used to dynamically divide the distribution network topology into regions. Combining electrical connectivity, unit size, and tie line constraints, several (usually 10 to 30, the specific number is dynamically determined according to the total number of nodes in the distribution network and the number of unit nodes) autonomous planning units with close electrical connections and balanced distribution of flexible resources are generated, and the electrical boundaries and resource planning scope of each unit are clarified. S3. Construct a two-layer hybrid integer programming model of global centralized coordination and regional distributed autonomous planning. The upper layer is a global network coordination planning model, which coordinates the network transformation, substation expansion and unit power interaction constraint planning. The lower layer is a resource optimization planning model for each autonomous unit, which independently completes the capacity adaptation planning of distributed power sources, energy storage and flexible loads within the unit. The upper and lower layers are coupled and coordinated through unit power interaction variables. S4. An improved alternating direction multiplier method (ADMM) distributed iterative algorithm with adaptive parameter adjustment and integer variable projection mechanism is adopted to perform hierarchical decomposition and parallel iterative solution of the two-level mixed integer programming model, eliminate the optimality conflict between global network planning and local resource planning, and obtain the global optimal planning decision variables. S5. Based on the optimal decision variables obtained through iterative solution, output a complete distribution network planning scheme. The planning scheme includes a new construction and renovation scheme for the entire network structure, a substation expansion and configuration scheme, a distributed power supply and energy storage capacity configuration scheme for each autonomous unit, and a flexible load access and adaptation capacity scheme.

[0009] Furthermore, in step S2, the quantitative calculation method for node electrical coupling strength and flexible resource density is as follows: The formula for calculating the electrical coupling strength of a node is: ,in The mutual impedance between node i and node j in the distribution network node impedance matrix is ​​used to characterize the degree of electrical connection between the two nodes. The formula for calculating the flexible resource density of a node is: ,in The maximum capacity that can be connected to the distributed power source at node i. The maximum adaptable capacity for node i energy storage device. The maximum adjustable capacity of the flexible load at node i. Let i be the baseline load capacity of node i.

[0010] Furthermore, in step S2, the optimization objective function of the improved Louvain community detection algorithm is a dual objective function that balances maximizing electrical coupling and equalizing resource density: ,in, It is a dual objective function. Let k be the set of nodes of the k-th autonomous planning unit. Let be the average flexible resource density of the k-th autonomous unit. These are the weighting coefficients. The value range is 0.2 to 0.5; During the algorithm iteration process, the objective function is... The iterative criterion is to continuously increase the number of nodes. The process involves two-level iterations: node migration and community aggregation, until the objective function converges. At the same time, the entire process is constrained to ensure the electrical connectivity of each autonomous unit, the number of unit nodes is controlled at 20-30, and the number of inter-unit interconnection lines does not exceed 10% of the total number of lines in the entire network, so as to ensure the rationality and standardization of the partitioning results.

[0011] Furthermore, in step S3, the upper-level global centralized coordination planning model aims to minimize the comprehensive planning cost across the entire network and its entire lifecycle. The objective function of the upper-level model is: in, This is the objective function value of the upper-level global centralized coordination planning model, i.e., the minimum comprehensive planning cost for the entire network and its entire life cycle; Investment costs for the construction and renovation of power grid structures and the expansion of substations. The annual operation and maintenance cost of the entire power distribution equipment, The annual power loss cost for the entire network, The cost of power shortage penalties will be planned for the entire network. The decision variables of the upper-level model include discrete variables of new grid construction and renovation (0-1), continuous capacity variables of substation expansion, and upper limit variables of maximum interactive power between each autonomous unit and the main grid. The constraints include linearized distributed power flow constraints, node voltage upper and lower limit constraints, line transmission capacity constraints, substation capacity constraints, and upper limit constraints of unit power interaction. Overall, they constitute the feasible region of mixed integer convex programming, taking into account both the discrete characteristics of grid planning and the continuous constraint characteristics of power grid operation.

[0012] Furthermore, in step S3, the lower-level autonomous unit distributed resource planning model takes the minimum comprehensive cost of single-unit full-cycle planning operation as its optimization objective, and the objective function of the lower level is: in, This represents the objective function value of the distributed resource planning model for the k-th autonomous unit, i.e., the minimum comprehensive cost of full-cycle planning operation for a single autonomous unit. The investment and configuration cost of distributed power sources and energy storage devices within the k-th autonomous unit. The annual operation and maintenance cost of flexible resource equipment within the unit. The power interaction cost between the unit and the upper-level main network; The decision variables of the lower-level model include the configuration capacity of distributed power sources within the unit, the rated capacity and real-time charging and discharging power of energy storage, and the capacity for flexible load adaptation and access. The constraints include the time-series power balance constraints within the unit, the time-series constraints of distributed power source output, the constraints of energy storage capacity and upper and lower limits of charging and discharging, and the time-series adjustment range constraints of flexible load. All constraints are linear convex constraints. The lower-level sub-model is a pure convex optimization model with a unique optimal solution.

[0013] Furthermore, in step S3, both the flexible load and the distributed power source are modeled using a full-time, 8760-hour detailed model to fully reflect the time-series fluctuation characteristics of resources. The time-series adjustment constraints for the flexible load are as follows: in, Let i be the actual flexible load power at time t. , These represent the minimum and maximum adjustable power of the flexible load at node i under time series t, effectively covering the load regulation characteristics at different times throughout the year and avoiding the problem of insufficient adaptability of planning schemes caused by modeling typical days.

[0014] Furthermore, in step S4, the solution process of the improved ADMM distributed iterative algorithm includes: S41. Using the power interaction variables between each autonomous unit and the main network as global coupling variables, construct the augmented Lagrangian function of the two-level planning model, and decompose the global mixed integer planning problem into multiple autonomous unit pure convex local subproblems and a single global network mixed integer coordination subproblem. S42. Each autonomous unit solves the local resource optimization sub-problem in parallel, obtains the results of local optimal decision variables and coupling variables, and uploads them to the global coordination layer. S43. The global coordination layer receives the coupling variables of each unit, solves the global grid planning relaxation subproblem, and maps the continuous relaxation solution to a 0-1 integer feasible solution through the integer variable projection mechanism to satisfy the discrete constraints of grid planning. S44. Based on the original residual and dual residual of the iterative process, the penalty parameter of the adaptive update algorithm is used to balance the iterative convergence speed of the two-level subproblem. S45. Update the Lagrange multipliers, calculate the global iterative residual, and determine whether the preset convergence accuracy or maximum number of iterations is met. If it is met, terminate the iteration and output the global optimal planning variable; otherwise, iterate and solve the problem.

[0015] Furthermore, in step S4, the adaptive update rule for the penalty parameter is as follows: in, Let be the penalty parameter for the (t+1)th iteration. Let be the penalty parameter for the t-th iteration. Let be the original residual of the t-th iteration. Let the dual residual of the t-th iteration be a fixed parameter. , By dynamically adjusting the penalty parameter, the problem of iterative oscillation and slow convergence of fixed parameter algorithms in mixed integer scenarios is solved.

[0016] In step S4, the iterative solution process uses a preset convergence accuracy and a maximum number of iterations as dual termination conditions. After the iteration converges, the globally coordinated optimal planning variables that meet the convergence accuracy are output.

[0017] Compared with the prior art, the present invention has the following substantial technical advancements and beneficial effects: 1. This invention adopts a dynamic zoning mechanism with dual indicators of electrical coupling and flexible resource density, which breaks through the limitations of traditional single electrical indicator zoning. On the basis of ensuring close electrical connection and uniform electrical characteristics within the distribution network unit, it realizes the balanced distribution of flexible adjustment resources in each autonomous unit. It avoids the problem of excessive differences in the autonomous capabilities of units from the planning source and perfectly adapts to the operating characteristics of massive flexible load time-series fluctuations and distributed access.

[0018] 2. The two-layer hybrid integer programming model constructed in this invention accurately distinguishes the characteristics of discrete planning of the upper-layer network structure and continuous optimization of resources in the lower-layer structure, and clarifies the boundaries of responsibility between global overall planning and local autonomy. It solves the problem of computational explosion in centralized planning of large-scale distribution networks through distributed dimensionality reduction solution, and avoids the defects of local optima and global failure in pure distributed planning through global coupling constraints, thus realizing deep collaborative planning of network structure and flexible resources.

[0019] 3. This invention adopts an improved ADMM algorithm that combines parameter adaptation and integer projection, which specifically solves the technical problems of traditional distributed optimization algorithms being unable to adapt to integer planning variables of the network structure and having poor adaptability of fixed parameters. It effectively suppresses the oscillation problem in the mixed integer iteration process, improves the stability and versatility of solving complex distribution network planning, eliminates the need for manual parameter adjustment, and significantly improves engineering practicality.

[0020] 4. This invention adopts a refined time-series modeling of 8760 hours throughout the year to fully characterize the time-series dynamic characteristics of flexible loads and distributed power sources. Compared with the traditional typical daily modeling method, the planning scheme can adapt to different operating scenarios throughout the year, avoid the problems of resource configuration redundancy or insufficient capacity, and greatly improve the accuracy of distribution network planning and long-term operational adaptability. Attached Figure Description

[0021] Figure 1 This is a flowchart of the overall planning method of the present invention.

[0022] Figure 2This is a schematic diagram of the dynamic division process of the autonomous planning unit in this invention.

[0023] Figure 3 This is a schematic diagram of the hierarchical planning architecture for centralized global coordination and regional distributed autonomous planning of the power distribution network of the present invention.

[0024] Figure 4 The flowchart for the iterative solution of the improved ADMM distributed algorithm of this invention is shown. Detailed Implementation

[0025] The implementation process of this invention will be described in complete detail below, taking into account actual engineering application scenarios.

[0026] This embodiment is based on the actual planning scenario of medium-voltage distribution network in urban areas. The distribution network in this area has a wide coverage and a large number of nodes. Electric vehicle charging piles, smart air conditioners, and controllable industrial and commercial flexible loads are deployed on a large scale in the area. At the same time, distributed photovoltaic and energy storage devices are also built. It is a typical scenario of massive flexible load access. The planning period is set to ten years to guide the upgrading of the regional distribution network and the overall allocation of flexible resources.

[0027] A distributed autonomous and centralized coordination planning method for distribution networks with massive flexible load access includes the following steps: Step 1: Data Acquisition and Preprocessing. Comprehensive data collection of basic parameters for the regional distribution network is conducted, including the entire network topology, impedance and capacity parameters of various line types, rated capacity and operational constraints of existing substations, and baseline load data for each distribution node. Simultaneously, regional planning data is collected, outlining the planned access points and output characteristics of distributed photovoltaic and energy storage devices within the planning period. The distribution locations, adjustable capacities, and time-series adjustment ranges of electric vehicles, smart temperature control systems, and controllable industrial loads are statistically analyzed. Combined with regional economic development data, a ten-year load growth forecast is completed, constructing a complete time-series planning database.

[0028] Step 2: Dynamic Partitioning of Distribution Network Autonomous Planning Units. Based on the collected power grid topology and parameter data, a node impedance matrix is ​​constructed, and the electrical coupling strength between any two nodes in the entire network is calculated. Combining the flexible resource and load data of each node, the flexible resource density of each node is calculated. Objective function weighting coefficients, upper limits for unit node size, tie-line constraint thresholds, and electrical connectivity constraints are set. Through iterative optimization using an improved Louvain community discovery algorithm, node community aggregation and optimization are gradually completed until the dual objective function converges. Finally, the large-scale regional distribution network is divided into multiple autonomous planning units with balanced scale, electrical independence, and uniform distribution of flexible resources, clearly defining the physical boundaries and resource planning authority of each unit.

[0029] Step 3: Construct a two-layer hybrid integer programming model. Based on the hierarchical planning architecture, build an upper-layer global centralized coordination planning model. With the goal of minimizing the overall cost across the entire network's lifecycle, optimize the solution for new regional network infrastructure construction, old line renovation, and substation expansion. Simultaneously, set power interaction limits between each autonomous unit and the main grid to control the network's safe operation constraints and overall investment economics. Build a lower-layer distributed resource planning model for each autonomous unit. Each unit aims for optimal overall planning and operation costs, combined with 8760-hour time-series operation constraints, to optimize the configuration capacity of distributed photovoltaic and energy storage devices within the unit, adapting to the access and regulation needs of regional flexible loads, and ensuring power time-series balance and efficient resource utilization within the unit.

[0030] Step 4: Improve the ADMM algorithm for iterative solution. Initialize the algorithm's penalty parameters, convergence accuracy, maximum number of iterations, and Lagrange multiplier parameters. Each autonomous unit solves the local resource optimization subproblem in parallel and uploads the unit power interaction coupling variables to the global coordination layer. The global coordination layer solves the overall network planning subproblem, performing integer projection processing on the discrete network variables to ensure the feasibility of the solution. During the iteration process, the original residual and dual residual are calculated in real time, the penalty parameters are adaptively adjusted, the convergence speed of the two-layer model is dynamically balanced, and the multipliers and planning variables are continuously updated until the iterative residuals meet the convergence accuracy requirements, obtaining the globally optimal set of decision variables for network planning, substation expansion, and resource allocation.

[0031] Step 5: Output the implementation plan. Based on the converged optimal decision variables, a standardized distribution network planning scheme is generated. This scheme includes a list of newly built and upgraded distribution network lines in the region, substation expansion capacity parameters, precise configuration of distributed power sources and energy storage capacity in each autonomous unit, maximum adaptable access capacity for flexible loads, and timing control strategies. The scheme is fully adapted to the operational needs of long-term access to massive flexible loads in the region and can be directly used for the planning and implementation of actual power grid projects.

[0032] This embodiment fully replicates the planning process of the present invention, relying on rigorous mathematical models and optimization theories throughout. It features clear hierarchical logic, a stable solution mechanism, and a closed-loop planning logic, effectively addressing the industry pain points of high planning dimensionality, difficult solution, and poor global coordination in large-scale flexible distribution networks. The solution aligns with actual engineering needs and possesses strong practical applicability.

Claims

1. A distributed autonomous and centralized coordination planning method for distribution networks with massive flexible load access, characterized in that... Includes the following steps: S1. Collect basic data for distribution network planning. The basic data includes existing distribution network topology, line impedance, rated capacity and other line parameters, substation rated capacity, operating constraint parameters, benchmark load time sequence curves of each node, distributed power source access points and output time sequence characteristics within the planning period, flexible load access locations and time sequence adjustment range parameters, and annual load growth forecast data. S2. Based on the dual indicators of electrical coupling strength and flexible resource density of distribution network nodes, an improved Leuven community discovery algorithm is used to dynamically divide the distribution network topology into regions. Combining electrical connectivity, unit size, and tie line constraints, several autonomous planning units with close electrical connections and balanced distribution of flexible resources are generated, and the electrical boundaries and resource planning scope of each unit are clarified. S3. Construct a two-layer hybrid integer programming model of global centralized coordination and regional distributed autonomous planning. The upper layer is a global network coordination planning model, which coordinates the network transformation, substation expansion and unit power interaction constraint planning. The lower layer is a resource optimization planning model for each autonomous unit, which independently completes the capacity adaptation planning of distributed power sources, energy storage and flexible loads within the unit. The upper and lower layers are coupled and coordinated through unit power interaction variables. S4. An improved alternating direction multiplier method distributed iterative algorithm with adaptive parameter adjustment and integer variable projection mechanism is adopted to perform hierarchical decomposition and parallel iterative solution of the two-level mixed integer programming model, eliminate the optimality conflict between global network planning and local resource planning, and obtain the global optimal planning decision variables. S5. Based on the optimal decision variables obtained through iterative solution, output a complete distribution network planning scheme. The planning scheme includes a new construction and renovation scheme for the entire network structure, a substation expansion and configuration scheme, a distributed power supply and energy storage capacity configuration scheme for each autonomous unit, and a flexible load access and adaptation capacity scheme.

2. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S1, each node includes a 10kV bus node, a distribution transformer substation node, and a flexible load and distributed power supply access node.

3. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S2, several, usually 10 to 30, are generated, with the specific number dynamically determined based on the total number of nodes in the distribution network and the constraints of the number of unit nodes.

4. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S2, the quantitative calculation method for node electrical coupling strength and flexible resource density is as follows: The formula for calculating the electrical coupling strength of a node is: ,in The mutual impedance between node i and node j in the distribution network node impedance matrix is ​​used to characterize the degree of electrical connection between the two nodes. The formula for calculating the node's flexible resource density is: ,in The maximum capacity that can be connected to the distributed power source at node i. The maximum adaptable capacity for node i energy storage device. For node i, the maximum adjustable capacity of the flexible load. Let i be the baseline load capacity of node i.

5. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S2, the optimization objective function of the improved Louvain community detection algorithm is a dual objective function that balances maximizing electrical coupling and equalizing resource density: ,in, It is a dual objective function. Let k be the set of nodes of the k-th autonomous planning unit. Let be the average flexible resource density of the k-th autonomous unit. These are the weighting coefficients. The value range is 0.2 to 0.5; During the algorithm iteration process, the objective function is used The iterative criterion is to continuously increase the number of nodes. The process involves two-level iterations: node migration and community aggregation, until the objective function converges. At the same time, the entire process is constrained to ensure the electrical connectivity of each autonomous unit, the number of unit nodes is controlled at 20-30, and the number of inter-unit interconnection lines does not exceed 10% of the total number of lines in the entire network, so as to ensure the rationality and standardization of the partitioning results.

6. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S3, the upper-level global centralized coordination planning model aims to minimize the comprehensive planning cost across the entire network and its entire lifecycle. The objective function of the upper-level model is: in, This is the objective function value of the upper-level global centralized coordination planning model, i.e., the minimum comprehensive planning cost for the entire network and its entire life cycle; Investment costs for the construction and renovation of power grid structures and the expansion of substations. The annual operation and maintenance cost of the entire power distribution equipment, The annual power loss cost for the entire network, The cost of power shortage penalties will be planned for the entire network. The decision variables of the upper-level model include discrete variables of new grid construction and renovation (0-1), continuous capacity variables of substation expansion, and upper limit variables of maximum interactive power between each autonomous unit and the main grid. The constraints include linearized distributed power flow constraints, node voltage upper and lower limit constraints, line transmission capacity constraints, substation capacity constraints, and upper limit constraints of unit power interaction. Overall, they constitute the feasible region of mixed integer convex programming, taking into account both the discrete characteristics of grid planning and the continuous constraint characteristics of power grid operation.

7. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S3, the lower-level autonomous unit distributed resource planning model aims to minimize the comprehensive cost of single-unit full-cycle planning operation. The objective function of the lower level is: in, This represents the objective function value of the distributed resource planning model for the k-th autonomous unit, i.e., the minimum comprehensive cost of full-cycle planning operation for a single autonomous unit. The investment and configuration cost of distributed power sources and energy storage devices within the k-th autonomous unit. The annual operation and maintenance cost of flexible resource equipment within the unit. The power interaction cost between the unit and the upper-level main network; The decision variables of the lower-level model include the configuration capacity of distributed power sources within the unit, the rated capacity and real-time charging and discharging power of energy storage, and the capacity for flexible load adaptation and access. The constraints include the time-series power balance constraints within the unit, the time-series constraints of distributed power source output, the constraints of energy storage capacity and upper and lower limits of charging and discharging, and the time-series adjustment range constraints of flexible load. All constraints are linear convex constraints. The lower-level sub-model is a pure convex optimization model with a unique optimal solution.

8. The method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S3, both flexible loads and distributed power sources are modeled using a full-time, 8760-hour detailed model to fully reflect the temporal fluctuation characteristics of resources. The temporal adjustment constraints for flexible loads are as follows: in, Let i be the actual flexible load power at time t. , These represent the minimum and maximum adjustable power of the flexible load at node i under time series t, effectively covering the load regulation characteristics at different times throughout the year.

9. A method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S4, the solution process of the improved ADMM distributed iterative algorithm includes: S41. Using the power interaction variables between each autonomous unit and the main network as global coupling variables, construct the augmented Lagrangian function of the two-level planning model, and decompose the global mixed integer planning problem into multiple autonomous unit pure convex local subproblems and a single global network mixed integer coordination subproblem. S42. Each autonomous unit solves the local resource optimization sub-problem in parallel, obtains the results of local optimal decision variables and coupling variables, and uploads them to the global coordination layer. S43. The global coordination layer receives the coupling variables of each unit, solves the global grid planning relaxation subproblem, and maps the continuous relaxation solution to a 0-1 integer feasible solution through the integer variable projection mechanism to satisfy the discrete constraints of grid planning. S44. Based on the original residual and dual residual of the iterative process, the penalty parameter of the adaptive update algorithm is used to balance the iterative convergence speed of the two-level subproblem. S45. Update the Lagrange multipliers, calculate the global iterative residual, and determine whether the preset convergence accuracy or maximum number of iterations is met. If it is met, terminate the iteration and output the global optimal planning variable; otherwise, iterate and solve the problem. The adaptive update rule for the penalty parameter is as follows: in, Let be the penalty parameter for the (t+1)th iteration. Let be the penalty parameter for the t-th iteration. Let be the original residual of the t-th iteration. For the dual residual of the t-th iteration, with fixed parameters , .

10. A method for distributed autonomous and centralized coordination planning of distribution networks for massive flexible load access as described in claim 1, characterized in that... In step S4, the iterative solution process uses a preset convergence accuracy and a maximum number of iterations as dual termination conditions. After the iteration converges, the globally coordinated optimal planning variables that meet the convergence accuracy are output.

Images