Transmission Tower Tilt Angle On-Line Prognosis by Using Solar-Powered LoRa Sensor Node and Sliding XGBoost Predictor

This paper presents a transmission tower tilt angle prognosis method based on solar-powered LoRa sensor node and sliding XGBoost predictor. The proposed LoRa sensor node, which consists of solar panel, LoRa radio frequency chip, super-capacitors, MCU, accelerometer, and gyroscope, can measure the initial tilt angle of the transmission tower and the angular rate of the transmission tower. Then, the measuring signals of transmission tower were wirelessly transmitted to the LoRa gateway and were processed online. First, the noise of the raw angular rate is reduced by using PCA (principal components analysis) method and the tilt angle of the transmission tower can be calculated by integrating the angular rate. Second, a sliding XGBoost predictor is proposed for tilt angle prognosis, which collects the training data and trains the regression model dynamically. Third, a novel parameter optimization algorithm named DCCPSO (double chain chaos particle swarm optimization) and its execution strategy are proposed to determine the values of hyper-parameters. Finally, the experimental system and the corresponding experimental results are demonstrated and discussed in detail, which shows that the proposed method is effective in transmission tower to tilt angel on-line prognosis.