An Active Distributed Control Method For Power Transformer Noise

Noise generated by power transformer operation has typical low frequency characteristics,capable of penetrating and diffraction and high intensity,and its management has become the focus of social attention.The use of sound insulation,sound absorption and other passive noise reduction methods for low-frequency noise control is not obvious,and the isolation leads to the main transformer heat dissipation problems,increasing the potential fire hazards.Active noise control(ANC for short)is based on the principle of acoustic phase extinction,by generating secondary acoustic waves with the same frequency and opposite phase,superimposed on the primary noise to produce phase extinction interference,to achieve active noise reduction,is the main way to solve the low-frequency noise control.Current ANC research mainly uses singlechannel,feedforward control,which cannot realize the spatial noise reduction effect of outdoor transformers under the action of disturbance.To address this problem,this paper investigates a distributed control method for power transformer noise based on hybrid control structure and secondary channel identification,and the main research work includes:First,the acoustic phase extinction and adaptive filtering principles are investigated.Considering the reliability and feasibility of system implementation,two representative classes of algorithms,namely the least mean square class algorithm and the recursive least squares class algorithm,are selected for comparison,and the filter-x LMS(Fx LMS for short)algorithm is found to be computationally small,easy to expand,and suitable for narrowband noise control.The signal band characteristics of Fx LMS are analyzed experimentally,and it is shown that Fx LMS is suitable for processing low-frequency narrowband noise.At the same time,the impact of parameters such as controller length and step size on the performance of the algorithm is analyzed to lay the foundation for algorithm extension and application.Second,the outdoor transformer channel has time-varying characteristics and there is disturbance noise,the ANC method based on feedforward cannot handle the disturbance noise,and the noise reduction effect becomes poor.To address this problem,a hybrid control structure integrating feedforward and feedback is designed.To address the coupling problem between feedforward and feedback,the hybrid control is decoupled by means of reducing and compensating the error signal,so that the influence of the feedback controller on the feedforward controller is eliminated and the feedforward control and feedback control are updated independently of each other.The decoupled hybrid control system satisfies the stability conditions and has a simple structure,reduced computational complexity and easy system implementation.Simulation experiments using actual noise data of power transformers show that the overall noise reduction of the decoupled hybrid control architecture is improved by about 3d B.The transformer secondary channel has time-varying characteristics,which leads to a reduction in noise reduction.Therefore,a hybrid secondary channel identification method is designed,using offline identification to provide initial parameters and combining with online identification to update parameters in real time,Improve the accuracy of identification.Simulation experiments using actual noise data show that the overall noise reduction is improved by about 2 d B compared with online identification alone.Third,outdoor transformer noise reduction belongs to spatial noise reduction in the free field,and single-channel ANC can only achieve single-point noise reduction.Therefore,a centralized multi-channel hybrid control algorithm is designed.The Fx LMS algorithm is extended to multi-channel ANC,and the steepest descent method is used to iteratively calculate the controller power coefficients to minimize the mean square error of all channels to realize the multi-channel Fx LMS algorithm.A centralized multi-channel transformer active noise hybrid control is implemented using a hybrid control structure based on error compensation decoupling and multi-channel hybrid identification.Taking two-channel and six-channel ANC systems as examples,simulation experiments using actual noise data of power transformers show that the overall noise reduction of two-channel ANC is about 3d B higher than that of single-channel ANC,and the overall noise reduction of six-channel ANC system is about 5d B higher than that of two-channel ANC system.Fourth,although the centralized multi-channel ANC solves the spatial noise reduction problem,the computational complexity increases quadratically with the number of channels.To solve this problem,a distributed multi-channel hybrid control algorithm is studied.The algorithm adopts the idea of distributed cooperative control,decomposes the overall spatial noise reduction into network node modules with a ring topology,and each network node sets a controller for active noise hybrid control.Using the parameter transfer principle of the controller,the controller parameters are transferred in an incremental manner.The updated parameters of each network node are passed to the next network node,thus updating the whole ANC algorithm.Distributed ANC can distribute the computational complexity of centralized ANC equally to each network node,effectively reducing the computational complexity of individual control nodes,and the cooperative strategy ensures the global noise reduction effect.Theoretical calculations illustrate that the computational effort shared by each controller of the distributed ANC increases only linearly with the number of channels,resulting in a significant improvement in computational complexity relative to the centralized ANC.Experimental comparisons using transformer noise data show that the distributed ANC improves the overall noise reduction by about 1 d B over the centralized ANC.