Adaptive Neural Network Control of Three-Phase Neutral-Point-Clamped Converters

Abstract

This chapter employs radial basis function (RBF) neural network to approximate the unknown function. For a continuous function d(x) on a compact set $\bigsqcup$ and $\varepsilon _{m}$, there exists a RBF neural network $\theta ^{\mathrm {T}}\xi (x)$ such that

$$\begin{aligned} \mathop {\sup }\limits _{x \in \sqcup } \left| {d(x) - \theta ^{\mathrm {T}}\xi (x)} \right| \le {\varepsilon _m}, \end{aligned}$$

(11.1)

where $\theta =[\theta _{1},\theta _{2},...,\theta _{l}]^{\mathrm {T}}$ is the weight vector, and $\xi (x)=[\xi _{1}(x),\xi _{2}(x),...,\xi _{l}(x)]^{\mathrm {T}}$ is basis function vector commonly used Gaussian function,

$$\begin{aligned} {\xi _i}(x) = \frac{1}{{\sqrt{2\pi } {\sigma _i}}}{e^{ - \frac{1}{2}{{\left( {\frac{{{{\left\| {x - {\phi _i}} \right\| }^2}}}{{{\sigma _i}}}} \right) }^2}}}, i=1,2,...,l, \end{aligned}$$

(11.2)

where $\sigma _i$ and $\phi _i$ are the width and center of the basis function, respectively. It has been proved that in [1, 2] the RBFNN can be used to approximate any smooth nonlinear functions d(t) within arbitrary accuracy. This chapter proposes RBF neural network based control strategy for three-phase three-level neutral-point-clamped (NPC) ac/dc power converter. The control is designed in DPC mode and consists of three loops, including voltage regulation loop, power tracking control loop, and capacitor voltage balance loop. Two RBF neural network based adaptive controllers are designed for power tracking loop, to drive the active and reactive power to their reference values respectively. An adaptive controller is designed to regulate the dc-link voltage where the load is treated as external disturbance. A composite controller consisting of a reduced-order observer is developed for the voltage balance loop to ensure voltage difference between two dc-link capacitors close to zero. The effectiveness and advantage of the proposed control strategy for the NPC power converter are verified through experiments.