264 rounds reasonably, the model can avoid over-fitting while ensuring the convergence efficiency. The design of the above model structure and training strategy focuses on constructing a stable and energy efficiency modeling framework, rather than over-fitting for a specific data set. Optimizer: Adam optimizer, with an initial learning rate of 0.001, which decays by 10% every 10 rounds to balance the convergence speed and later accuracy. Loss function: Mean Square Error (MSE), which accurately describes the deviation between the predicted value and the real fuel consumption. Early stopping strategy: If the verification set loss does not decrease for 5 consecutive rounds, the training is stopped to avoid over-fitting. Training environment: Relying on GPU to accelerate training and shorten the training time. 4.5. Model Verification and Application of Energy Efficiency A comprehensive multi-indicator evaluation method is adopted to assess the model performance, assessing the model's performance from two dimensions: prediction error and fitting ability. Among them, error-based indicators are used to characterize the deviation level between the model's predicted values and the actual fuel consumption, while goodness-of-fit indicators are employed to evaluate the model's overall ability to depict the law of fuel consumption changing with operating conditions. In this study, evaluation indicators such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²) are selected to conduct a comprehensive assessment of the model's prediction performance, thereby providing a quantitative basis for the model's applicability in energy efficiency scenarios. At the model application level, based on the trained LSTM energy efficiency model, unified prediction of fuel consumption levels under different vehicles and operating conditions can be realized, providing a standardized reference benchmark for subsequent energy efficiency analysis. Meanwhile, combined with the idea of controlled variables, perturbation analysis is conducted on key influencing factors, which can quantitatively evaluate the impact trend of changes in a single factor on fuel consumption levels. This provides model support for the classification of mining trucks energy efficiency levels, comparative analysis of typical operating conditions, and evaluation of energy-saving potential 5. CONCLUSIONS AND PROSPECTS 5.1. Summary of Main Research Results This study focuses on the refined management and control of fuel consumption of open-pit mining trucks, takes multi-source data fusion as the core, constructs a monitoring and analysis system and forms a closed-loop cost control strategy. A multi-source data collection and fusion system is built to collect fuel (second-level precision), vehicle, dispatching, positioning, road roughness and other data in modules, which realizes time-series synchronization and makes up for the deviation of road network slope, providing highquality data support. A multi-dimensional energy efficiency system is established, with fuel consumption per unit transportation volume as the core index, and the optimal mining truck models and driving habits are identified through horizontal comparison. A closed-loop cost control strategy is formed to realize the wholeprocess management and control from four aspects: proactive road maintenance, route optimization, promotion of good driving habits and assistance in fleet procurement. 5.2. Future Research Directions Combined with the research limitations and industry needs, future research can be carried out in four directions: First, expand multi-scenario data collection and model optimization, cover extreme
RkJQdWJsaXNoZXIy MTM0Mzk2