【华北电力大学_许建中_】2024含多类型直流链路混合的交直流系统的潮流计算方法报告_35页_7mb

Summary of the Document

Core Content

The document presents a comprehensive analysis of power flow calculation methods in hybrid AC/DC systems, focusing on the challenges and solutions related to data conversion, modeling, and algorithm development. It also introduces a novel method for non-intrusive load monitoring (NILM) using federated learning for privacy protection and discusses the optimal reactive power planning considering bus vulnerability analysis.

Main Topics

1. Brief Introduction and Research Background

• Author: Jianzhong Xu, Ph.D, Professor from Prof. Chengyong Zhao's Group at North China Electric Power University.
• Date: May 9^th, 2024.
• Location: Dali, Yunnan.
• Research Focus: Power flow calculation for hybrid AC/DC systems, particularly those incorporating modular multilevel converters (MMC) and high-voltage direct current (HVDC) systems.

2. Background of AC/DC Systems

• Hybrid AC/DC Power Grid: A new stage in power system development, characterized by unique structural and operational features, but also facing serious safety challenges.
• Southern Power Grid: Has developed an "8 AC and 11 DC lines" hybrid structure.
• Multi-type DC Converter Stations and Lines: These are distributed and embedded within the AC grid, which affects the modeling and analysis of DC data.

3. Dispatching System Data Conversion

• CIM/XML Limitations: Cannot be directly used for power flow calculations; must be converted into a specific format.
• Key Issues:
  • Topology Model Conversion: The conversion of node/switch models in CIM/XML to bus/branch models in power flow data.
  • DC System Data Reading: The need to process DC data based on the topology structure.
• Solutions:
  • Effective Device Search: Used to identify and connect physical devices.
  • Model Processing: Differentiates between LCC (with built-in transformer) and MMC (connected through a converter transformer).
  • Node Elimination: Removes unnecessary physical connection nodes, treating them as busbars in the converted topology.
• Results: A unified equivalent model of renewable energy units and converters is achieved, which simplifies the power flow calculation process.

4. Power Flow for AC/DC Hybrid Systems

• Unified Power Flow Model:
  • Node Classification: Pure AC nodes, LCC nodes, and MMC nodes.
  • AC Power Flow Model: Uses node injection power and voltage equations.
  • LCC Power Flow Model: Incorporates LCC-specific variables like DC voltage, DC current, and control angle.
  • MMC Power Flow Model: Includes MMC-specific variables like DC voltage, DC current, and modulation index.
• LCC-MMC Hybrid Converter Models:
  • Parallel Model: DC network equations remain unchanged.
  • Series Model: Represented as a "two-node" voltage source in the equivalent circuit.
• Power Flow Algorithm:
  • Unified Iterative Algorithm: Uses both AC and DC state variables as solving variables, iteratively solving the power flow equations.
  • Simulation Verification: Demonstrates the effectiveness of the algorithm through the use of CIM/XML data, showing AC and DC power flow results.

5. Non-Intrusive Load Monitoring (NILM) with Privacy Protection

• Fed-NILM Method:
  • Objective: To improve the modeling capabilities of local data owners while protecting consumer privacy.
  • Approach: Utilizes federated learning (FL) to share model parameters instead of raw data.
  • Advantages: Superior scalability and convergence compared to locally trained models, and performance close to centrally trained models.
  • Application: Useful in carbon emission reduction and energy conservation.

6. Optimal Reactive Power Planning

• Problem: Power system instability due to voltage deviation, often caused by a lack of reactive capacity.
• Approach:
  • Vulnerable Bus Identification: Uses fast voltage stability index, line stability index, and voltage collapse proximity index (VCPI).
  • Optimization Algorithms: Includes particle swarm optimization, differential evolution, whale optimization algorithm, grasshopper optimization algorithm, salp swarm algorithm, grey wolf optimization, and oppositional grey wolf optimization (OGWO).
• Results: VCPI is the best method for identifying vulnerable buses, and OGWO provides a more cost-effective solution for optimal reactive power planning.

Key Information

• Data Conversion: Critical for power flow calculations, involving topology transformation and handling of DC system data.
• Modeling: Involves converting complex CIM/XML models into simplified formats, and using equivalent models for LCC and MMC.
• Power Flow Algorithm: Unified iterative algorithm that handles both AC and DC systems, ensuring accurate and efficient calculations.
• Privacy Protection: Fed-NILM is introduced as a novel method for load monitoring, ensuring data privacy while maintaining model performance.
• Optimal Reactive Power Planning: Emphasizes the importance of identifying vulnerable buses and using advanced optimization techniques to improve system reliability and performance.

Conclusion

The document provides a detailed overview of the challenges and solutions in power flow calculation for hybrid AC/DC systems, highlighting the importance of data conversion, modeling, and algorithm development. It also introduces Fed-NILM as a promising method for NILM with privacy protection and discusses the application of advanced optimization techniques in reactive power planning. The results demonstrate the effectiveness and efficiency of the proposed methods in enhancing the reliability and performance of hybrid AC/DC power systems.