Research on the Accurate Monitoring Model of Sound Power Level of Sound Source Equipment in +800kV Converter Station Based on Intelligent Sensor Signal Processing

A DNN-based precise monitoring model of acoustic power level is proposed, which aims at improving the precision and real time of monitoring. Firstly, a multi-channel sensor array is used to acquire the sound pressure signal, and then the original signal is preprocessed, including de-noising and normalization. Then, CNN is used to extract features from the preprocessed signal to capture the time-frequency features in the sound pressure signal. Next, LSTM is used to perform time series modeling on the extracted features to capture the time series dependency of the sound pressure signal. Finally, FC is combined to perform regression prediction of the sound power level. This model achieves high-precision estimation of the sound power level through deep learning technology. This paper builds an experimental platform on site at the ±800kV converter station, arranges a multi-channel sound pressure sensor array, and collects noise data during equipment operation. By setting different working states and environmental conditions, a variety of noise data are obtained. Experiments show that the proposed ANN is more accurate and real time than the conventional one, which proves its validity and feasibility.