Hierarchical Distribution Network Reconfiguration With Hybrid AI Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale distribution network reconfiguration is computationally intensive due to its mixed integer nonlinear programming nature, requiring extensive calculation resources and time, especially when using intelligent algorithms that traverse all possible situations and calculate power flow distribution.
Innovation Solution
A hybrid data-and-model-driven hierarchical network reconfiguration method that sections the distribution network into loops, establishes a compression network, and uses models like the harmony search algorithm and LSTM deep learning to reduce calculation complexity, forming off-line databases for quick decompression network generation and optimal reconfiguration scheme selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If intelligent algorithms are used to solve distribution network reconfiguration, then the solution can handle large-scale networks, but the calculation time and calculation resources increase significantly
Solution Approach 1:
The distribution network is divided into multiple independent loops based on interconnection switches. Each loop is treated as a separate sub-problem with its own reconfiguration optimization, rather than solving the entire network as one large problem. This segmentation reduces the computational complexity while maintaining the ability to handle large-scale networks.
Solution Approach 2:
The problem is transformed from a single-layer network reconfiguration to a two-layer hierarchical structure: the upper layer determines optimal reconfiguration schemes for each loop independently, and the lower layer implements these schemes. This dimensional change from flat to hierarchical structure reduces calculation time while preserving adaptability.
2Reliability
If intelligent algorithms traverse all possible situations to solve distribution network reconfiguration, then the solution comprehensiveness is improved, but the calculation resources required increase significantly
Solution Approach 1:
The solution space is segmented by dividing the network into independent loops. Each loop's reconfiguration is optimized separately, which reduces the total number of combinations that need to be evaluated compared to traversing all possible network configurations. This maintains solution comprehensiveness within each loop while reducing overall computational resources.
Solution Approach 2:
Instead of traversing all possible network configurations, the method performs partial action by optimizing each loop independently with a limited number of switch states. This partial optimization approach reduces calculation resources while still achieving comprehensive coverage of relevant reconfiguration scenarios.
3Productivity
If mathematical programming algorithms are used for distribution network reconfiguration, then the calculation speed is improved, but large-scale networks become difficult to solve
Solution Approach 1:
The large-scale network is segmented into smaller loop sub-problems that can be solved using mathematical programming algorithms. Each loop's smaller size allows these algorithms to execute quickly while the collection of loop solutions handles the large-scale network overall, thus achieving both speed and adaptability.
Solution Approach 2:
The approach transitions from a single-layer mathematical programming solution to a two-layer hierarchical structure. The upper layer uses mathematical programming for fast loop-level optimization, while the lower layer aggregates these solutions for comprehensive network reconfiguration. This dimensional change enables both fast calculation and large-scale network handling.
Data Source
AI summary
A method and a system of hybrid data-and-model-driven hierarchical network reconfiguration are provided. Taking into account that the optimal operation structure of the power grid may change after a new energy is connected into the power grid in the distribution manner, this method combines the mathematical model of network reconfiguration and clustering with the deep learning model, and uses data recorded by the smart meter, to reconfigure the network in a hybrid data-and-model driving mode. This method proposes a network compression method and a hierarchical decompression method based on deep learning, and improves the efficiency of network reconfiguration by the way of distributed calculation and combination of on-line and off-line calculation.