Computationally Enhanced Approach for Chance-Constrained OPF Considering Voltage Stability

TL;DR

This paper presents a novel, data-driven approach combining neural networks, adaptive polynomial chaos expansion, and dimension reduction techniques to efficiently solve chance-constrained voltage stability problems in power systems.

Contribution

It introduces a scalable, surrogate modeling framework that explicitly incorporates voltage stability constraints into chance-constrained optimal power flow using advanced uncertainty propagation methods.

Findings

Demonstrates cost-effective performance in test systems.

Effectively manages uncertainty in renewable power generation.

Enhances computational efficiency for large-scale systems.

Abstract

The effective management of stochastic characteristics of renewable power generations is vital for ensuring the stable and secure operation of power systems. This paper addresses the task of optimizing the chance-constrained voltage-stability-constrained optimal power flow (CC-VSC-OPF) problem, which is hindered by the implicit voltage stability index and intractable chance constraints Leveraging a neural network (NN)-based surrogate model, the stability constraint is explicitly formulated and directly integrated into the model. To perform uncertainty propagation without relying on presumptions or complicated transformations, an advanced data-driven method known as adaptive polynomial chaos expansion (APCE) is developed. To extend the scalability of the proposed algorithm, a partial least squares (PLS)-NN framework is designed, which enables the establishment of a parsimonious surrogate model and efficient computation of large-scale Hessian matrices. In addition, a dimensionally decomposed APCE (DD-APCE) is proposed to alleviate the "curse of dimensionality" by restricting the interaction order among random variables. Finally, the above techniques are merged into an iterative scheme to update the operation point. Simulation results reveal the cost-effective performances of the proposed method in several test systems.