【作者】 仇志凌；

【导师】 熊蕊；

【作者基本信息】 华中科技大学 ， 电力电子与电力传动， 2004， 硕士

【摘要】 在不允许供电中断的重要用电场合,大量地使用着 UPS(不间断电源)系统。CVCF-PWM 逆变器是 UPS 系统的核心部件,要求它具有高质量的输出电压波形。为了使其即使在非线性负载情况下仍有接近正弦的输出电压,发展出了多种多样的波形控制技术。本文的研究内容就是基于谐波实时分析的 PWM 逆变器波形控制技术。主要研究工作有:将负载电流处理为可测扰动,建立了 PWM 逆变器的线性数学模型。分别给出了传递函数形式和状态方程形式的数学模型,同时指出了这两种形式的模型可以根据不同的控制方案选用。为了无差地抑制扰动,给出了本控制系统采用的结构框图,并且证明了该结构可以对扰动进行无差抑制。为了对扰动进行补偿,采用了 FFT 算法对扰动进行谐波分析。虽然 FFT 算法已经大大提高了 DFT 的计算效率,但对于实时控制来说计算量还是太大,本文提出了一种简化型的 FFT 算法,大大减少了计算量。对扰动进行谐波分析必然要对原信号进行加窗处理,这就会对频谱分析的精度造成影响。本文分析了这种影响,指出只要使分析窗长和采样频率满足一定的要求,就可以得到高精度的谐波分析结果。逆变器的开环动态特性是非常差的,具有严重的振荡倾向,有必要采用瞬时值反馈对其进行校正。本文对瞬时值反馈对系统性能的影响进行了深入的分析,指出瞬时值反馈不仅可以提高系统动态性能,对稳态精度的提高也有一定的帮助。本文还对系统采用的数字 PD 控制器的设计方法进行了探讨,总结出了一套行之有效的设计方法。FFT 得到的结果是复数形式的,并不能直接使用,需要将其转换成直接形式的幅值和初相位。在这里,角度的求取是最关键的,其精度直接影响系统性能。本文首先介绍了一种二分查表法的求解算法,该算法虽然精度极高,但计算量较大。为了提高算法效率,提出了一种基于神经网络的求解算法。该算法拥有不错的精度,且大大减少了计算量,降低编程的复杂性。最后对所采用的控制方案进行了实验验证。整个控制算法采用一片 F240 DSP 芯片进行实现。实验结果证明本文采用的控制方案不仅能在稳态时获得低 THD 的输出电压,而且在负载突变时获得了较好的动态性能。更多还原

【Abstract】 Uninterruptible Power Supply (UPS) systems are widely used for supplying critical loads which cannot afford utility power failure. A CVCF-PWM inverter is the core of a UPS system. High quality outputvoltage waveform is required for these inverters. To achieve nearly sinusoidal output voltage even withnonlinear loads, many waveform correction techniques have been proposed. This paper is focused onwaveform control technique based on real time harmonic analysis of CVCF-PWM inverters. Thefollowings are the main work: With the load current treated as a measurable disturbance input, a linear mathematical model isestablished for a PWM inverter. The mathematical model is given in form of transfet function and statesequations. In additions, it is indicated that these two forms of mathematical model can be chosen accordingto diffient control method. For surpressing disturb with no error, block diagram of this control technique is proposed and thecapability to surpress disturb with no error is demonstrated. For compansate disturb, FFT algorithm isused to analyze the harmonic of disturb. Although the FFT algorithm has dramasticly impoved calculationefficiency of the DFT algorithm, the calculation for real time control is still too complicated. A simplifiedFFT algorithm which dramasticly simplify the calculation of FFT is proposed. To analyze the disturb, adata window must be added to the original signal. This will affect the accuracy of the spectrum analysis.The affect caused by adding window is discussed, it is indicated that if the window length and sampelfrequency obey particular request, high precise result can be achieved. The transient performance is poor and has serious oscilization tendency, so transient paramentfeedback technique should be used to improve the performance. The effect to the performance of thiscontrol system caused by transient parament feedback is discussed in detail. It is indicated that transientparament feedback technique can not only improve transient performance but also make somecontribution to stable waveforme quality. The design problem for digital PD controller is also discussed, atlast, a useful set of design methord is proposed. The result of FFT algorithm is complex and can not be used directly, so it must be translated to directmagnitude and initial phase. Here, the calculation of degree is the most important problem, because thecalculation accuracy directly influences the performance of this control system. In this thesis, a kind ofcalculation algorithm based on seaching in harf table is discussed first. Though this algorithm has veryhigh lever accuracy, the calculation is very complicated. In order to improve the calculation efficiency, asolve algorithm based on neural network is proposed. This algorithm with satified accuracy dramasticlysimplify the calculation and the complication of programme implement. At last, the proposed control technique is tested. The control algorithm is implemented using a TITMS320F240 DSP. Experimental results prove that the proposed control technique can achieve not onlylow THD during steady-state operation but also good transient response subject to load step change.更多还原