CO2 Emission Prediction of Vehicle Fuel Consumption Based on EMD-LSTM

In this paper, a prediction model based on Empirical Mode Decomposition (EMD) combined with Long Short-Term Memory (LSTM) is proposed, which can further improve the accuracy of model learning prediction based on only using LSTM. Before model processing, it needs to find indicators that can evaluate vehicle fuel consumption, and then select the indicators that can be learned by the neural network as the initial input. The initial input is substituted into the EMD model, and then the IMF component obtained through EMD decomposition is substituted into the LSTM model. Taking the data of cars of different brands as practical examples, the test results show that the prediction error of this method is only 12.2%. But the prediction error of only LSTM network model is 15.4%, the average absolute error of the former is 7% higher than that of the latter, and the prediction accuracy is higher.