Early Warning Method for Abnormal Environmental Temperature and Humidity for Coastal Power Equipment Protection Missions

Addressing the issue of power equipment structural component failures induced by high temperature and humidity environments in coastal regions, an anomaly warning method for environmental temperature and humidity was proposed based on deep neural networks and trajectory clustering algorithms. The method aims to predict potential abnormal temperature and humidity conditions through environmental observation data, enabling early deployment of equipment protection measures. First, the method utilizes a bidirectional long short-term memory network (BiLSTM) to learn from historical temperature and humidity variation data, incorporating an attention mechanism to extract key features and provide high-precision forecasts of future trends in temperature and humidity. Subsequently, historical and predicted data are concatenated to form movement trajectories within a three-dimensional data space of temperature and humidity. Finally, the DBA clustering algorithm is applied to classify these trajectories into a known trajectory library, completing pre-inspection and identification of abnormal temperature and humidity conditions in the environment. Simulation experiments demonstrate that, in the test scenarios, the proposed network achieves an average prediction loss of 0.02% for environmental temperature and humidity data, representing an 83.78<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> improvement over traditional prediction networks. The method effectively distinguishes between normal and various abnormal data fluctuations, achieving the purpose of early warning.