A Model-Free Switching and Control Method for Three-Level Neutral Point Clamped Converter Using Deep Reinforcement Learning

Abstract

This paper presents a novel model-free switching and control method for three-level neutral point clamped (NPC) converter using deep reinforcement learning (DRL). In this method, voltage balancing, and selection of optimal switches are achieved using a reward function which is calculated based on various signals measured as observations of the DRL agent. Since the action space is discrete, a deep Q-network (DQN) agent is utilized. DQN is used due to its capability of handling high-dimensional state spaces. In order to highlight its pros and cons, the proposed method is compared with model predictive control (MPC), which is another popular non-linear control method for power electronic converters. The proposed method is evaluated and compared with the MPC method in grid-connected mode using simulations in Matlab/Simulink. To evaluate the practical performance of the DRL method, various experimental results are obtained. The simulation and experimental results demonstrate that the proposed method effectively achieves accurate voltage balancing and ensures steady operation even in the presence of various dynamic changes, including variations in the reference currents and grid voltage. Additionally, the method successfully handles uncertainties, such as sensor measurement noise, and accommodates parameter variations, such as changes in the capacity of the DC-link capacitors and line impedance. The results demonstrate that this method exhibits superior adaptability to real-time changes and uncertainties, delivering more robust performance compared to similar conventional methods like MPC. Thus, this method can be considered a promising approach for intelligent control of power electronic converters, especially when conventional methods such as MPC face challenges in performance and accuracy under severe parameter variations and uncertainties.

Reinforcement Learning-Based Control of a Power Electronic Converter

“…Theorem 2. Given that (A − BK x ) is exponentially stable, the linear output regulation problem is solvable by a static feedback controller of Equation (8) iff there exist two constant matrices X and U that solve Equation (11) with K v given by…”

“…Problem 1. We solve the static optimization problem of Equation (35) in order to find solutions to the regulator equations stated in Equation (11), thus assuring the asymptotic tracking of the reference signal.…”

“…Now, using the second equation in Equation (11) and assuming that the matrix D = 0, we can find a suitable form for the solution, X, to this equation by first selecting a constant matrix X 1 ∈ R n x ×n v such that CX 1 + F = 0. Next, we select X i ∈ R n x ×n v , i ∈ 2, 3, .…”

“…[9] provided simulation results on an LCL-coupled inverter-based distributed generation system; hence, it provides evidence of the efficacy of RL control for complex power electronic systems. More recently, a 2023 paper [11] successfully deployed a model-free, deep reinforcement learning (DRL) algorithm on a three-level neutral point clamped (NPC) converter. This paper provides irrefutable evidence of the suitability of RL-based controllers for optimal output regulation of complex, nonlinear power electronic converters.…”