A Grid-Based Scenario Delineation Method for Distribution Networks Based on Fuzzy Comprehensive Evaluation and SNN-DPC Clustering

Liuzhu Zhu, Xin Yang, Xiuli Wang, Fan Zhou, Zhi Guan, Hejun YangProcesses 2025 · DOI

TL;DR

This paper introduces a novel grid-based method for classifying distribution network scenarios, crucial for planning in the face of increasing renewable energy and electric vehicle integration. By combining fuzzy comprehensive evaluation with an improved SNN-DPC clustering algorithm, the method effectively categorizes network characteristics, offering a more adaptable approach than traditional planning methods.

What Problem Does This Paper Address?

The increasing integration of distributed energy resources (like solar and wind) and electric vehicles is making distribution networks more dynamic and complex. Traditional, deterministic planning methods struggle to keep pace with these evolving, diversified scenarios. This complexity arises from the random nature of these new resources and the inherent limitations of existing grid structures. The paper aims to address this by developing a robust method to delineate and classify different distribution network scenarios, enabling more effective and adaptive planning.

Methods

The proposed method involves several key steps:

  1. Multi-dimensional Assessment Index System: The authors first analyze the characteristics of various flexible resources (distributed PV, EVs, energy storage, demand response) and establish a comprehensive index system. This system covers four dimensions: load characteristics, power supply capacity, grid level, and carrying capacity.
  2. Fuzzy Comprehensive Evaluation (FCE):
    • Weighting: A hybrid approach is used to determine the weights of indicators. The DEMATEL-ANP method is employed for subjective weighting, capturing interdependencies between indicators. The CRITIC method is used for objective weighting, based on data volatility and conflict. These are combined using a geometric mean to derive composite weights.
  3. Evaluation: An FCE model is then applied to calculate the assessment values for each indicator, considering expert opinions and data.
  4. SNN-DPC Clustering: The calculated assessment values are then clustered using an improved Shared Nearest Neighbor Density Peak Clustering (SNN-DPC) algorithm. This algorithm is enhanced with a Whale Optimization Algorithm (WOA) to optimize the number of shared nearest neighbors (k), a critical parameter for clustering effectiveness. This optimization aims to minimize the sum of distances from all data points to their respective cluster centers.

Datasets

The case study uses data obtained from the distribution automation system, information acquisition system, and production management system of a regional power grid. Specific details about the dataset's origin or size beyond its regional nature are not provided.

Key Insights and Ideas

  • The "Active Network" Shift: The paper highlights the transition of distribution networks from "passive" to "active" due to flexible resource integration, necessitating new planning paradigms.
  • Multi-dimensional Evaluation is Crucial: A single indicator is insufficient to capture the complex characteristics of modern distribution networks. A holistic approach considering load, supply, grid structure, and carrying capacity is essential.
  • Hybrid Weighting for Robustness: Combining subjective (DEMATEL-ANP) and objective (CRITIC) weighting methods provides a more balanced and reliable assessment of indicator importance.
  • SNN-DPC for Complex Clustering: The improved SNN-DPC algorithm, particularly with WOA optimization, demonstrates superior performance in handling density variations and identifying distinct clusters compared to traditional methods like K-means and standard DPC.
  • Scenario-Based Planning: Effectively delineating network scenarios allows for tailored planning strategies, addressing the unique challenges and opportunities presented by different grid configurations and resource mixes.

Key Results

  • The proposed method successfully classified distribution network scenarios into four categories, with results showing good consistency and robustness.
  • In the case study, the improved SNN-DPC algorithm achieved a smaller sum of distances to cluster centers (143.4096) compared to the standard DPC algorithm (147.3465) and K-means (150.3993), indicating better clustering quality.
  • The contour coefficient for the improved SNN-DPC algorithm was 0.5421, slightly outperforming DPC (0.5367) and K-means (0.5243), suggesting better cohesion and resolution in the clusters.
  • A sensitivity analysis showed that the clustering results were stable, with only minor reassignments of grids after slight adjustments to weighting parameters.

Limitations and Future Work

  • Practical Engineering Value: The authors acknowledge that while the method provides a framework for scenario delineation, its direct practical engineering value needs further demonstration. Grids within the same scenario may still have local variations that require more granular analysis.
  • Future Work: Future research should focus on providing specific guidance for grid construction within each identified scenario, addressing the local variations that the current scenario segmentation might overlook. This could involve further analysis of specific indicators within each cluster to pinpoint areas for investment and improvement.

Summary updated

Abstract

Aiming at the problems that the random probability characteristics of large-scale source and load resources lead to the ineffectiveness of deterministic planning methods, the standard grid structure is difficult to adapt to the demands of diversified scenarios. This paper proposes a grid-based scenario delineation method for distribution networks based on fuzzy comprehensive evaluation and SNN-DPC (density peak clustering based on shared-nearest-neighbors). First, analyze the response characteristics of various types of flexible resources, and establish a multi-dimensional comprehensive assessment index system that integrates operational characteristics and structural features. Second, the comprehensive weights of each index in the index layer are calculated based on the DEMATEL-ANP method and the CRITIC method, and the assessment value of the intermediate layer is calculated by the fuzzy comprehensive evaluation method. Finally, the assessment value of the intermediate layer is clustered based on the improved SNN-DPC algorithm, so as to classify the distribution grid scenarios. The results indicate that the proposed method can effectively and accurately classify distribution network scenarios.
BibTeX
@inproceedings{zhu2025grid,
  author = {Liuzhu Zhu and Xin Yang and Xiuli Wang and Fan Zhou and Zhi Guan and Hejun Yang},
  title = {{A Grid-Based Scenario Delineation Method for Distribution Networks Based on Fuzzy Comprehensive Evaluation and SNN-DPC Clustering}},
  booktitle = {Processes},
  year = {2025},
  doi = {10.3390/pr13092923},
}
Figure