论文部分内容阅读
为了改善变压器有源降噪系统的性能,提出了基于多频陷波滤波和信号功率变参数的变结构滤波-X最小均方(filter-X least mean square,FXLMS)自适应有源降噪算法.该算法以变结构FXLMS算法为核心,针对变压器噪声的特点,采用多频陷波滤波技术实现了系统无参考传声器的设计,降低硬件成本并节约占地面积.采用随机噪声法对系统次级通道进行在线建模,并提出了“实时建模,延时更新”的策略,以及基于信号功率的变建模收敛系数和变随机噪声幅值相结合的方法,以降低引入的随机噪声幅值,并保证建模的准确性.同时,基于信号功率实现了对降噪收敛系数的自适应调节.仿真和理论分析表明,该算法可加快系统的收敛速度,提高系统的稳定性和降噪量.设计的软件平台与硬件设备相结合的变压器有源降噪系统,在额定运行的50MVA变压器降噪实验中,实现了误差传声器处10～18dB的降噪效果,平均声能量密度降低90.01％～98.41％,验证了算法对实际电力变压器噪声的降噪有效性.“,”In order to improve the performance of active control of transformer system,the variable structure filter-X least mean square (FXLMS) adaptive active noise control (ANC) algorithm based on the multi-frequency trap filter technology and signal power varying parameters method is proposed.Based on the variable structure FXLMS and the characteristics of transformer noise,it adopts multi-frequency trap filter technology realizing the design of the system without reference microphone,which can reduce the hardware cost and save the site area.Moreover,it uses the random noise method to on-line model the secondary path of the system.To decrease the random noise amplitude and ensure the accuracy of secondary path model,it proposes the strategy about “real-time modeling and delay updating” and the method about using the signal power to adjust the modeling convergence coefficient and the random noise amplitude.Additionally,the signal power is also adopted to adjust the noise-reducing convergence coefficient adaptively.The analysis of simulation and theory shows that this proposed algorithm can accelerate the convergence of the system and improve the system stability and noise reduction.In the rated operating 50MVA transformer noise reduction experiment,the designed active control of transformer system including software and hardware can achieve a noise reduction of 10dB-18dB and a 90.01％～98.41％ decrease in average sound energy density which verified the feasibility of the proposed algorithm to suppress the real power transformer noise.