摘   要

在对户外的变压器进行噪声测试时，经常会有周围环境噪声干扰，其会影响测试结果的准确性。因为环境噪声种类很多而且复杂，它的产生和持续时间具有不确定性，并且它也不是持续稳定的，大多数测量噪声的设备很难对这种情况进行有效处理。所以，为了准确地评估输变电工程噪声的污染程度、分析噪声信号源的强度、掌握噪声的分布规律以及变化趋势、针对性地预防和控制噪声，以便能够有效解决噪声污染的问题，提高变压器噪声测量的精度具有至关重要的意义。

本文从变压器及其外环境噪声产生机理、变压器及其外环境噪声特性分析、变压器声音信号降噪处理技术三个方面展开研究，主要工作内容如下：

1）详细解析了变压器的典型结构和组成部分，包括铁心、绕组和冷却装置等，深入探讨了铁心、绕组和冷却装置等部件对变压器噪声产生的影响机理，介绍了变压器外部环境中常见的噪声源包括虫鸣、鸟鸣、蛙鸣和车辆噪声等的产生机理。

2）介绍了声压级、计权声压级和倍频谱等声学基本参量的定义和意义。据其时频域波形，对变压器声音以及虫鸣、鸟鸣、蛙鸣和车辆等四种常见变压器外环境噪声的特性进行了分析，比较变压器声音及其外环境噪声差异。

3）提出了基于小波分析算法、小波包分析算法以及谱减法语音增强技术的变压器声音信号降噪处理方法。给定了用于评价降噪效果的指标，以量化降噪处理方法的性能和效果。通过仿真对变压器声音信号中的虫鸣、鸟鸣、蛙鸣和车辆噪声进行降噪处理，并评估不同信噪比条件下的降噪效果。对实际场景中的变压器声音信号进行采集和处理，评估降噪效果的实际应用情况。结果表明，所采用的方法能够有效降低外界环境噪声的干扰。

本文通过对变压器周围典型虫鸣、蛙鸣、鸟鸣和车辆声音的时频域特性进行分析的基础上，提出了一种基于谱减法语音增强技术的变压器声音信号降噪技术。采用小波包分解算法将变压器声信号时频分解，根据分解结果定位干扰噪声所在时刻与频段，对带噪频段小波包信号进行谱减法语音处理，利用小波包重构算法还原出纯净变压器声信号。该方法针对提高电力变压器噪声在复杂环境干扰条件下产生的监测精度、保障测量结果的有效性、基于噪声检测的变压器运行状态评价以及推动电力行业噪声监测技术发展均具有重要的理论意义和实际应用价值。

关键词：变压器噪声；环境干扰；谱减法；特性分析；降噪处理

Abstract

In the noise test of outdoor transformers, there is often ambient noise interference, which will affect the accuracy of the test results. Because there are many and complex types of environmental noise, its generation and duration are uncertain, and it is not continuous and stable, most noise measurement equipment is difficult to effectively deal with this situation. Therefore, in order to accurately evaluate the degree of noise pollution in power transmission and transformation projects, analyze the intensity of noise signal sources, master the distribution law and change trend of noise, and prevent and control noise in a targeted manner, in order to effectively solve the problem of noise pollution, it is of great significance to improve the accuracy of transformer noise measurement.

This paper studies the mechanism of transformer and its external environmental noise, the analysis of transformer and its external environmental noise characteristics, and the noise reduction technology of transformer sound signal. The main work is as follows :

1 ) The typical structure and components of the transformer are analyzed in detail, including the core, winding and cooling device. The influence mechanism of the core, winding and cooling device on the noise of the transformer is discussed in depth. The common noise sources in the external environment of the transformer are introduced, including the generation mechanism of insect noise, bird noise, frog noise and vehicle noise.

2 ) The definition and significance of basic acoustic parameters such as sound pressure level, weighted sound pressure level and frequency spectrum are introduced. According to its time-frequency domain waveform, the characteristics of transformer sound and four common transformer external environmental noises such as bugs, birds, frogs and vehicles are analyzed, and the differences between transformer sound and its external environmental noise are compared.

3 ) The noise reduction method of transformer sound signal based on wavelet analysis algorithm, wavelet packet analysis algorithm and spectral subtraction speech enhancement technology is proposed. In order to quantify the performance and effect of the noise reduction method, the indicators for evaluating the noise reduction effect are given. The noise reduction of bugs, birds, frogs and vehicle noise in transformer sound signal is carried out by simulation, and the noise reduction effect under different SNR conditions is evaluated. The transformer sound signal in the actual scene is collected and processed to evaluate the practical application of the noise reduction effect. The results show that the method can effectively reduce the interference of external environmental noise.

Based on the analysis of the time-frequency domain characteristics of typical bug, frog, bird and vehicle sounds around the transformer, this paper proposes a transformer sound signal noise reduction technology based on spectral subtraction speech enhancement technology. The wavelet packet decomposition algorithm is used to decompose the time-frequency of the transformer acoustic signal. According to the decomposition results, the time and frequency band of the interference noise are located. The wavelet packet signal with noise band is processed by spectral subtraction speech processing, and the pure transformer acoustic signal is restored by wavelet packet reconstruction algorithm. This method has important theoretical significance and practical application value for improving the monitoring accuracy of power transformer noise under complex environmental interference conditions, ensuring the effectiveness of measurement results, evaluating the operating status of transformers based on noise detection, and promoting the development of noise monitoring technology in the power industry.

Keywords: transformer noise; environmental interference; spectral subtraction; characteristics analysis; noise reduction processing