A Surrogate Based Computationally Efficient Method to Coordinate Damping Controllers for Enhancement of Probabilistic Small-Signal Stability

Abstract

This paper proposes a computationally efficient method based on deep neural network and a meta-heuristic optimization algorithm known as bat algorithm to coordinate power oscillation damping controllers incorporated in renewable energy stations to enhance system small signal stability considering uncertainties. The proposed method consists of three main stages: database generation, supervised learning, and optimization using the created model. A database is first created of Probabilistic small-signal stability margin calculated using the combined cumulant and Gram-Charlier expansion and the parameters of damping controllers, which is then given to different supervised machine learning algorithms such as linear regression, support vector machine, random forest, and chiefly deep neural network to create a surrogate model. The surrogate model provides an approximate relationship between the probabilistic small-signal instability margin and damping controller parameters. An optimization problem is then formulated to minimize the surrogate probabilistic instability margin with damping controller parameters acting as constraints. Finally, this optimization problem is solved using the bat algorithm to obtain the optimized parameters for power oscillation damping controllers. Our study results tested on a large IEEE 16 machines, 68 bus system show that the power oscillation damping controllers optimized using the proposed method can largely improve the system low frequency oscillatory stability margin in a very low computational time (around 19 times faster than the conventional method). This study can be used by the power system operators to tune the parameters of damping controller in a fast manner.

Keywords

• Correlation
• deep learning
• meta-heuristics
• probabilistic optimization