Research on Intelligent Operation and Maintenance Technology Based on Health State Prediction in the Power Internet of Things

A power IoT abnormal flow warning mechanism based on wavelet decomposition and LSTM is proposed to address the issue of abnormal flow warning on the IoT management platform. Through wavelet decomposition, random and non-stationary time series can be stabilized to reduce data volatility. Through LSTM model, relevant temporal information of time series can be learned. Firstly, wavelet decomposition is performed on the time series to divide it into multiple dimensional time series. Then, LSTM models are used to predict the decomposed time series, and the predicted time series is obtained through wavelet reconstruction. By predicting the time series and inputting the time series, a sliding time window is selected to dynamically determine the warning threshold. When system flow is detected to exceed the warning threshold, relevant warnings are given. The experimental results indicate that this mechanism can effectively predict future time series and provide early warning for abnormal flow in the system.