Transformer-based energy-saving method for intelligent identification of electric construction machinery working cycle stage

The electrification and intelligence of construction machinery are an inevitable trend to realize green development, as they help to promote a green supply chain and sustainable development. As the core of electrification and intelligence, working condition recognition provides a basis for machine control decision-making and energy regulation mechanism by sensing equipment status in real-time, thereby serving as a foundation for realizing unmanned operation. Traditional working condition recognition methods have been gradually replaced by intelligent models due to the reliance on manual rules and poor generalization ability. Aiming at the limitation that the existing real-time working condition sensing methods for the entire machine fail to integrate the characteristics of the hydraulic system and traveling system of the electric construction machinery, this study analyzes the electro-hydraulic coupling mechanism between the two systems under cyclic working conditions. Accordingly, it proposes an intelligent recognition method for cyclic working modes based on the Transformer architecture. This method utilizes the Transformer's self-attention mechanism to capture the long-term dependencies of multi-sensor time-series data, which significantly enhances the global feature extraction, recognition accuracy, and generalization performance. This results in more efficient and accurate recognition of cyclic working conditions. The results show that the stage recognition accuracy of cyclic operation reaches 98.48 % and prediction delay of online recognition is 0.0077 ms, demonstrating good robustness and generalization ability. When applied to an electric loader for braking energy recovery and recue, this method achieves a 19.32 % energy saving in the working system during cyclic operation, while the entire vehicle achieves an 11.91 % reduction in energy consumption.