Transmission line icing thickness prediction model based on ISSA-CNN-LSTM
Zilong Han, Hongtu Lv, Zhengding Liang, Jinqiao Yi
Abstract
Abstract The prediction of transmission line ice cover thickness can effectively guide the scientific operation and maintenance of the power sector. An improved sparrow search algorithm, convolutional neural network and long short-term memory fusion ice cover prediction model is proposed in this paper. Environmental temperature, humidity, wind speed and other data are firstly normalised, CNN is used to both learn the data signature as well as input the results into LSTM, and then ISSA is used to optimise the parameters of LSTM such as the numeration of the neuron, the initial learning rate, and the regularisation coefficient, etc., and then the final output is the predicted value of the icing thickness of the transmission line cover. The simulation results show that the MSE and MAE of the ISSA-CNN-LSTM model are 0.23 and 0.37, respectively, which are much better than the prediction results of the SSA and CNN models.