[Paper Review] Deep Learning based Model-free Robust Load Restoration to Enhance Bulk System Resilience with Wind Power Penetration

Jin Zhao, Fangxing Li+2 more|arXiv (Cornell University)|Sep 16, 2021

Electric Power System Optimization24 references4 citations

TL;DR

This paper proposes a deep learning-based, model-free robust load restoration method for power systems with high wind power penetration. Using DNN and CNN to emulate worst-case scenarios and a load pickup checklist generation (LPCG) algorithm for optimality, it enables real-time, computationally efficient on-line restoration that outperforms traditional robust optimization in speed and adaptability, validated on IEEE 30- and 118-bus systems.

ABSTRACT

This paper proposes a new deep learning (DL) based model-free robust method for bulk system on-line load restoration with high penetration of wind power. Inspired by the iterative calculation of the two-stage robust load restoration model, the deep neural network (DNN) and deep convolutional neural network (CNN) are respectively designed to find the worst-case system condition of a load pickup decision and evaluate the corresponding security. In order to find the optimal result within a limited number of checks, a load pickup checklist generation (LPCG) algorithm is developed to ensure the optimality. Then, the fast robust load restoration strategy acquisition is achieved based on the designed one-line strategy generation (OSG) algorithm. The proposed method finds the optimal result in a model-free way, holds the robustness to handle uncertainties, and provides real-time computation. It can completely replace conventional robust optimization and supports on-line robust load restoration which better satisfies the changeable restoration process. The effectiveness of the proposed method is validated using the IEEE 30-bus system and the IEEE 118-bus system, showing high computational efficiency and considerable accuracy.

Motivation & Objective

• Address the challenge of resilient load restoration in bulk power systems with high wind power variability and uncertainty.
• Overcome the computational infeasibility of conventional robust optimization for on-line applications due to high-dimensional uncertainty.
• Develop a real-time, model-free method that ensures system security under worst-case wind and load conditions.
• Enable fast, optimal load restoration that adapts dynamically to changing system conditions during restoration.
• Replace traditional robust optimization with a data-driven, scalable alternative suitable for on-line deployment.

Proposed method

• Design a deep neural network (DNN) to predict the worst-case system condition for any given load pickup decision.
• Use a deep convolutional neural network (CNN) to evaluate system security under the predicted worst-case scenario.
• Develop a load pickup checklist generation (LPCG) algorithm to systematically explore and select optimal load restoration sequences within limited checks.
• Implement a one-line strategy generation (OSG) algorithm to rapidly produce a complete, robust restoration strategy from the optimal checklist.
• Train the DNN and CNN using historical system data and worst-case scenario simulations, avoiding explicit system modeling.
• Integrate the DNN and CNN with the LPCG and OSG algorithms to enable real-time, model-free robust decision-making.

Experimental results

Research questions

• RQ1 Can a deep learning-based model-free approach achieve robust load restoration performance comparable to traditional robust optimization while enabling real-time computation?
• RQ2 How effectively can DNN and CNN models predict worst-case system conditions under high wind power uncertainty?
• RQ3 To what extent does the LPCG algorithm ensure optimality in load restoration sequences under computational constraints?
• RQ4 How does the proposed method compare to conventional robust optimization in terms of computational speed and adaptability during dynamic restoration?
• RQ5 Can the method be effectively applied to large-scale systems like the IEEE 118-bus system with high wind penetration?

Key findings

• The proposed method achieves high computational efficiency, enabling real-time on-line load restoration, which is infeasible with conventional robust optimization due to high computational burden.
• The DNN and CNN models effectively predict worst-case system conditions and assess security, with accuracy validated on both IEEE 30-bus and IEEE 118-bus systems.
• The LPCG algorithm ensures near-optimal load restoration sequences within a limited number of checks, significantly reducing search space without sacrificing robustness.
• The OSG algorithm successfully generates complete, robust restoration strategies from the optimal checklist, ensuring system security under worst-case scenarios.
• The method demonstrates considerable accuracy and resilience in systems with high wind power penetration, outperforming traditional approaches in speed and adaptability.
• Validation on the IEEE 118-bus system confirms scalability and robustness, showing the method’s readiness for practical on-line deployment in real-world power systems.