Temperature Prediction Of Ship Propulsion Devices Based On Time Series

As the main carrier of world trade,shipping bears more than 80%of global trade volume,and is the main driving factor of economic globalization.Human economic activities and business development have a serious impact on marine health,and have aroused widespread concern in the world.Among them,the sustainable development of marine energy is a long-term task facing shipping at present and in the future.Saving energy and reducing pollution during navigation,accurate fault prediction of ships,and intelligent fault diagnosis are all effective means for the sustainable development of ship energy.The temperature change of the ship’s propulsion devices can usually reflect whether the system itself is faulty.To this end,this paper studies the temperature prediction of ship propulsion devices based on time series,and aims to predict the future temperature of ship propulsion devices by using the data obtained from sensor data in a data-driven manner,so as to find potential faults in time and promote ship energy saving.The main contents and work are summarized as follows:(1)Considering that the data used in this paper is time series data collected from various sensors of ships,an autoregressive distributed lag(ARDL)time series prediction model is constructed.Find the ARDL model with the best lag steps from different lag steps.According to the different properties of ship propulsion devices,all data that may affect the compressor and turbine outlet temperatures were processed for redundancy and normalization,and all features were selected for feature selection.A time series prediction model is constructed and the prediction results are obtained.By comparing and analyzing the error with the actual data,the ARDL model with the best lag steps and corresponding coefficients is obtained.(2)In order to effectively use the historical data to predict the upcoming temperature change trend of the devices,a prediction model based on the ARDL-LSTM(long short-term memory)recurrent neural network is proposed.The specific steps of the model are given,including data preparation,network input layer construction,input and output vector selection,network layer selection,hidden layer number of neurons,excitation function selection,initial weights and thresholds selection,selection of learning rate,etc.Based on this,the future temperature of the compressor air outlet and the turbine outlet are respectively trained and tested by the network,and the temperature prediction result is obtained.(3)In order to verify the effectiveness of the ARDL-LSTM prediction model mentioned in this paper,a variety of models were compared and tested.A single LSTM prediction model is used to predict the temperature of the turbine outlet and the compressor outlet,and the errors of the two models are compared and analyzed.Through comparative analysis,we can see that the ARDL-LSTM recurrent neural network model has a better prediction effect.