Model predictive control of NPC inverters for different scenarios

Abstract

Multilevel inverters have undergone a period of rapid development recently and, as a result of the progression of contemporary industry, a range of commercial applications have attained the megawatt standard, thereby meaning that there is a requirement for medium- or high-power conversion mechanisms. At the same time, several energygenerating technologies that can be renewed including solar power, bio-energy, tidal power, wind power, and geothermal power have emerged as critical areas of development in global energy production initiatives, hence creating a situation where the multilevel inverter is becoming increasingly investigated by scholars. It is possible to use multilevel inverters with renewable energy sources and, at the same time, these devices can contribute to high-power rating standards. In light of these considerations, the purpose of the present thesis is to address the control algorithms of multilevel inverters and, to be specific, the thesis will chiefly examine the neutral point clamped (NPC) inverters. The main challenge regarding the modulation and control of NPC inverters is that the entire system is nonlinear due to the utilisation of power switches. Conventional approaches are mainly based on the average state space modelling and related modulation techniques such as pulse width modulation (PWM) and sinusoidal pulse width modulation (SPWM). However, the limitations of these conventional methods are clear. The model must be linearised based on a specific operating point. The tuning procedure of the controller is complex, and this complexity increases yet further when the operating point is adjusted. Due to the development of the fast microprocessor and digital signal processors (DSP), some novel control methods such as hysteresis current control, sliding mode control and model predictive control (MPC) have been applied to NPC inverters. Among these methods, MPC stands out for its fast response and handling multiple constraints, and MPC has achieved great success in power electronics applications. Therefore, novel MPC algorithms for 3-level NPC inverters have been proposed to improve energy efficiency for different scenarios in this thesis. Firstly, an MPC algorithm with reduced switching frequency for 3-level NPC inverters is proposed. The proposed method can reduce the switching frequency through the optimisation of switching sequences while keeping the advantages of the conventional MPC method, such as its fast response and ability to handle multiple constraints. The performance of this proposed method is verified by simulation results with 3-level NPC inverters. Based on this MPC algorithm, two additional extended MPC algorithms are proposed to reduce computational burden, including a multistep model predictive control (MMPC) algorithm and a model predictive control algorithm for grid-connected NPC inverters. The effectiveness of the proposed MPC algorithms has been identified.
Furthermore, an MPC algorithm for NPC grid-connected inverters is proposed with automatic selection of weighting factors. The main objective of this proposed algorithm is to reduce switching frequency and to provide for the automatic selection of weighting factors without the need for trial and-error across different working conditions. The algorithm can also achieve active power tracking and maintain neutral point balancing. These various objectives are achieved through the use of a modified three-part cost function and the adoption of a two-dimensional fuzzy logic control scheme. The effectiveness of the proposed algorithm is verified using the results of grid-connected NPC inverters simulations, which show that the switching frequency can be reduced by at least 30% when compared to conventional MPC methods. In addition, a new proposed current slope-related control objective is added to the cost function of the proposed MPC algorithm. This control objective aims to make predicted slope values approach slope reference values. By doing this, the proposed algorithm can reduce the total harmonic distortion (THD) of the output currents and, when compared to a conventional MPC algorithm, can reduce the switching frequency by more than 30%. A case study of a grid-connected NPC inverter system is used to assess the effectiveness of the proposed MPC algorithm under conditions of nominal operation and dynamic operation (with the active power reference, P∗, being stepped down from 31kW to 24kW at 0.3s, then increased to 41kW at 0.5s) and with different degrees of parameter sensitivity (at +25%, +50%, and −75% variation in inductance).

More information

Identifiers

ORCID for Bing Chu: orcid.org/0000-0002-2711-8717

Catalogue record

Export record