Developing a data-driven hydraulic excavator fuel consumption prediction system based on deep learning

With the aggravation of the global energy crisis, fuel consumption has become a key indicator for excavator manufacturers and policymakers. However, traditional experimental and theoretical approaches, which are complicated to calculate and dependent on the laboratory environment, are difficult to be applied in engineering. A data-driven prediction system consisting of data acquisition, feature decomposition, fuel consumption prediction, and performance evaluation modules, is developed to achieve excavator fuel consumption prediction during operation. The characteristic parameters of excavator energy consumption are collected through the established data acquisition module and the superfluous noise is eliminated by the feature decomposition module. A prediction model based on Informer is proposed in the fuel consumption prediction module to solve the drawback that the neural network model cannot capture key information in long series data in parallel, thus improving the prediction accuracy. Finally, the effectiveness of different prediction methods is verified by the performance prediction module.