An energy consumption optimal control method and device for self-driving electric mining trucks

Abstract

The present invention provides a method and device for optimal energy consumption control for self-driving electric mining trucks, relating to the technical field of electric mining trucks. The method integrates advanced sensor technology and complex simulation modeling to acquire vehicle operating status data in real time, and constructs a multi-dimensional data set in combination with environmental factors. An optimized neural network model is used to accurately predict energy consumption, and based on the comparison between the predicted results and the actual energy consumption, the control strategy is intelligently switched to ensure that the vehicle always maintains optimal energy consumption under complex operating conditions. The system not only improves the accuracy and adaptability of energy consumption control, but also enhances the reliability and stability of the system through an interactive feedback mechanism, providing strong support for the efficient operation of self-driving electric mining trucks. The system aims to solve the problems of insufficient energy consumption control accuracy and poor adaptability in the existing technology.

Description

Technical Field

[0001] The present invention relates to the technical field of electric mining trucks, and in particular to an energy consumption optimization control method and device for an autonomous driving electric mining truck. Background Art

[0002] In modern mining transportation, electric trucks are becoming a mainstream mode of transport due to their environmentally friendly and efficient features. However, with the rise of autonomous driving technology, optimizing the energy consumption of autonomous electric trucks has become a key technical bottleneck. Achieving optimal energy consumption control for autonomous electric trucks in complex mining conditions is a current research focus.

[0003] Traditional energy optimization methods primarily rely on simple data-driven or rule-based control strategies. While these methods can provide insights into energy optimization to a certain extent, their limitations are becoming increasingly apparent when applied to autonomous electric mining trucks. For example, rule-based control methods are often only adaptable to specific operating conditions and are less adaptable to the complex and changing mining environment. Data-driven methods, on the other hand, place extremely high demands on data quality and quantity, and struggle to guarantee prediction and control accuracy in practical applications.

[0004] Furthermore, existing technologies also have shortcomings in data collection and model building. Traditional energy optimization methods typically focus only on a few vehicle operating parameters, such as speed and load, while ignoring the significant impact of environmental factors (such as temperature, humidity, and slope) on energy consumption. This one-sided data collection approach results in an incomplete dataset that fails to accurately reflect the energy consumption characteristics of electric mining trucks in actual operation.

[0005] When it comes to model building, existing technologies mostly rely on single mathematical models or simple machine learning algorithms, which struggle to capture the complex time-series characteristics of electric mining trucks' energy consumption. For example, the energy consumption patterns of electric mining trucks vary in complex patterns under different terrains (uphill, downhill, and flat), loads (fully loaded, unloaded), and environmental conditions (high temperature, low temperature, and high humidity), making it difficult for traditional models to accurately predict and optimize control.

[0006] Therefore, existing technologies for optimizing the energy consumption of autonomous electric mining trucks are significantly deficient and cannot meet the requirements for optimal energy consumption control under complex mining conditions. This not only impacts the operational efficiency and economic viability of electric mining trucks, but also limits the further promotion and application of autonomous driving technology in mining transportation. Therefore, a new energy consumption optimization method is urgently needed that comprehensively considers multiple factors, enabling accurate prediction and real-time optimal control of the energy consumption of autonomous electric mining trucks, thereby promoting technological advancement and sustainable development in the mining transportation industry.

[0007] In view of this, this application is filed. Summary of the Invention

[0008] The present invention provides a method and device for optimal energy consumption control of an autonomous driving electric mining truck, which can at least partially improve the above-mentioned problems.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] An energy consumption optimization control method for an autonomous driving electric mining truck, comprising:

[0011] Acquiring a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and establishing a virtual entity corresponding to the electric mining truck based on the physical data set;

[0012] Acquiring a virtual data set of a virtual entity, fusing the virtual data set with a physical data set to obtain a driving data set based on multiple working conditions;

[0013] Call the pre-trained PSO-BiLSTM-Attention neural network model to predict energy consumption and obtain the predicted energy consumption value;

[0014] The predicted energy consumption value is compared with the actual energy consumption to obtain an energy consumption deviation, and the state parameters of the electric mining truck are adjusted according to the energy consumption deviation to ensure that the energy consumption of the electric mining truck remains within a preset optimal range.

[0015] The present invention also provides an energy consumption optimization control device for an autonomous driving electric mining truck, comprising:

[0016] a simulation unit, configured to obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and to establish a virtual entity corresponding to the electric mining truck based on the physical data set;

[0017] a fusion unit, configured to obtain a virtual data set of a virtual entity, and fuse the virtual data set with a physical data set to obtain a driving data set based on multiple working conditions;

[0018] The prediction unit is used to call the pre-trained PSO-BiLSTM-Attention neural network model to predict energy consumption and obtain the predicted energy consumption value;

[0019] The adjustment unit is used to compare the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and adjust the state parameters of the electric mining truck according to the energy consumption deviation to ensure that the energy consumption of the electric mining truck remains within a preset optimal range.

[0020] In summary, the described optimal energy consumption control method for autonomous electric mining trucks aims to address the shortcomings of existing technologies in energy consumption control by achieving accurate prediction and real-time optimal control of the energy consumption of autonomous electric mining trucks through innovative methods. This method constructs a comprehensive mapping of physical and virtual entities, uses multiple sensors to collect vehicle operating status data, and combines this with virtual data generated by simulation models to build a driving dataset based on multiple operating conditions. The simulation data is calculated using an MPC control algorithm and verified through HIL hardware-in-the-loop testing, improving the accuracy and reliability of the dataset. A PSO-BiLSTM-Attention neural network model is used to train and learn the dataset, predicting the vehicle state with optimal energy consumption. Based on the comparison of predicted energy consumption with actual energy consumption, the control mode is intelligently switched to ensure that the energy consumption of the electric mining truck remains within the optimal range.

[0021] The innovation of the described optimal energy consumption control method for autonomous electric mining trucks lies in that it comprehensively considers the impact of multiple factors on energy consumption, improves the accuracy of energy consumption prediction by combining simulation with actual data, and provides two sets of main and auxiliary control modes to achieve real-time adjustment of energy consumption. This effectively improves the operating efficiency and economy of autonomous electric mining trucks, and has important application value and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of an energy consumption optimization control method for an autonomous driving electric mining truck provided by the first embodiment of the present invention;

[0023] Figure 2 This is a flowchart of an energy consumption optimization control method for an autonomous driving electric mining truck provided by the first embodiment of the present invention;

[0024] Figure 3 This is a flowchart of constructing a driving data set based on multiple working conditions provided by an embodiment of the present invention;

[0025] Figure 4 This is a flow chart of the PSO-BiLSTM-Attention neural network model training provided by an embodiment of the present invention;

[0026] Figure 5 This is a flow chart of vehicle state control with optimal energy consumption provided by an embodiment of the present invention;

[0027] Figure 6 This is a module schematic diagram of an energy consumption optimization control device for an autonomous driving electric mining truck provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0029] refer to Figure 1 、 Figure 2 As shown, the first embodiment of the present invention discloses an energy consumption optimization control method for an autonomous electric mining truck. The method can be executed by an energy consumption optimization control device for an autonomous electric mining truck (hereinafter referred to as a control device), and in particular, by one or more processors within the control device to implement the following method:

[0030] S1, obtaining a physical data set collected by a sensor component configured on the electric mining truck to be tested, and establishing a virtual entity corresponding to the electric mining truck based on the physical data set;

[0031] Specifically, step S1 includes: obtaining a physical data set collected by a sensor component configured on the electric mining truck to be tested, constructing and processing the physical data set using AVL Cruise software to obtain a complete vehicle model of the electric mining truck;

[0032] Based on the physical data set, the dynamic equations and control system of the electric mining truck are established using Simulink software. The dynamic equations and control system are combined with the electric mining truck whole vehicle model to obtain the electric mining truck whole vehicle simulation model.

[0033] Preferably, the sensor assembly includes: a speed sensor, a SOC sensor, a tilt sensor, a mass sensor, a humidity sensor, a temperature sensor and a pedal sensor.

[0034] In this embodiment, the electric mining truck is equipped with multiple sensor components that can collect real-time vehicle operating status data to form a physical data set. These sensor components include a speed sensor, a SOC sensor, an inclination sensor, a mass sensor, a humidity sensor, a temperature sensor, and a pedal sensor, as shown in Table 1. They are used to measure key parameters such as the truck's speed, remaining battery charge, inclination angle, load mass, ambient humidity, operating temperature, and pedal opening.

[0035] Table 1 Sensor categories and their functions

[0036]

[0037] The physical data sets collected by these sensors are processed using AVL Cruise software. AVL Cruise is professional vehicle modeling and simulation software that builds a complete vehicle model corresponding to the electric mining truck based on the input physical data set. During this process, the software constructs a virtual model of each key vehicle component, such as the chassis, powertrain, and transmission, based on actual vehicle parameters such as dimensions and powertrain configuration. This creates a virtual entity that closely resembles the actual electric mining truck.

[0038] The complete electric mining truck model includes modules such as the vehicle, cab, monitor, motor, clutch, and brake, requiring mechanical connections between these modules. The vehicle structure includes numerous sensors that detect the vehicle's operating status, requiring electrical connections between the signals from each module. Taking torque as an example, to evaluate the impact of varying torques on the energy consumption of electric mining trucks, typical driving conditions of electric mining trucks were used as Cruise test cycles, and energy consumption simulations were performed at different torques. Torque values ​​were set to different values ​​and then imported into Cruise for simulation analysis. Loads were set to both unloaded and fully loaded, and Cruise was used to simulate the energy consumption of electric mining trucks at different torques.

[0039] Simultaneously, based on the same physical data set, the dynamic equations and control system for the electric mining truck were established using Simulink software. Simulink is a powerful dynamic system modeling and simulation tool. It can establish dynamic equations based on the vehicle's dynamic characteristics. These equations describe the vehicle's motion under different operating conditions, such as acceleration, deceleration, and torque output. Simulink also enables the design of the vehicle's control system, including motor control and braking control, to ensure that the vehicle operates according to the predetermined equations of motion. Specifically, appropriate modules were selected to establish the mining truck's equations of motion, sensor simulation modules, and corresponding control logic. Mathematical operation modules were used to establish the dynamic equations related to the mining truck's motion. The Sensor Simulation module in Simulink was used to simulate the outputs of the in-wheel motor's internal sensors, including motor speed and power. The Transfer Function module in Simulink was used to establish the mining truck's in-wheel motor control system, ensuring that the motor could be controlled according to the predetermined equations of motion. The "Scope" module was used to monitor the motor model's output in real time to ensure the validity of the model's equations of motion and control logic.

[0040] By combining the dynamic equations and control systems established in Simulink with the vehicle model built by AVL Cruise, a complete simulation model of the electric mining truck was created. This complete simulation model not only includes the static structural information of the vehicle but also simulates its dynamic behavior during actual operation, providing an accurate virtual platform for subsequent energy optimization control.

[0041] See also Figure 3 ,S2, obtain a virtual data set of the virtual entity, fuse the virtual data set and the physical data set to obtain a driving data set based on multiple working conditions;

[0042] Specifically, step S2 includes: calculating the virtual data set using an MPC control algorithm, and performing an in-the-loop test on the calculated data set using HIL hardware to obtain an output result;

[0043] Performing verification and comparison processing on the output result to determine whether the output result is greater than a preset value;

[0044] If not, the output is fed back to the virtual entity for improvement;

[0045] If so, the output results are combined with the physical data set to construct a driving data set based on multiple working conditions.

[0046] Preferably, the physical data set includes speed, temperature, mass, humidity, slope, pedal opening, and SOC, and the virtual data set includes torque, motor speed, power, gear, and energy consumption.

[0047] In this embodiment, after acquiring a physical dataset and establishing a virtual entity, a key step is to process the virtual dataset generated by the virtual entity and fuse it with the physical dataset to construct a multi-condition driving dataset. First, the types and functions of sensors, various characteristic parameters related to mining truck energy consumption, and different mining truck driving conditions are introduced in sequence. The sensor data is then input into a simulation model, and virtual data is output and compared with the actual mining truck data. Under various driving conditions of electric mining trucks, the physical and virtual data are fused to construct a multi-condition driving dataset. This process not only involves complex data processing but also uses specific algorithms and testing methods to ensure data accuracy and reliability, providing a solid data foundation for subsequent energy consumption optimization and control.

[0048] Specifically, first, a virtual data set is obtained from the virtual entity. These virtual data sets are generated by the simulation model and include parameters such as torque, motor speed, power, gear and energy consumption. Although these parameters are derived from the simulation model, they can reflect the theoretical operating state of the vehicle under different working conditions and are an important reference for energy consumption optimization control. The physical data and virtual data obtained by the energy consumption optimal control method for autonomous driving electric mining trucks have a total of 12 characteristic parameters, including speed, temperature, mass, humidity, slope, pedal opening, SOC, torque, motor speed, power, gear, energy consumption and other information. These parameters can reflect the driving characteristics and performance of the electric mining truck, and are all related factors that can affect the energy consumption of the mining truck, and can provide reliable data support for the energy consumption prediction of electric mining trucks. The specific data information is shown in Table 2.

[0049] Table 2 Related characteristic data information

[0050]

[0051] To improve the accuracy and practicality of this virtual data, a model predictive control (MPC) algorithm is used to calculate the virtual dataset. The MPC algorithm is an advanced control algorithm that predicts future states based on the system's dynamic characteristics and optimizes control strategies. Processing the virtual dataset using the MPC algorithm produces optimized data that better reflects actual operating conditions, thereby improving the quality of the dataset. The processed dataset is then validated through hardware-in-the-loop (HIL) testing. HIL testing is a testing method that combines actual hardware with simulation models. It verifies the accuracy and reliability of the simulation model under conditions close to the actual operating environment. HIL testing allows the data generated by the simulation model to be compared with the operating results of the actual hardware, thereby verifying the validity of the simulation data.

[0052] Next, the output results from the HIL test are verified and compared. The key to this process is determining whether the output results meet preset accuracy requirements. Specifically, the similarity between the output results and the actual operating data is determined to be greater than a preset value, typically 90%. If the output similarity is lower than the preset value, this indicates deficiencies in the current simulation model or data processing method, and the output results need to be fed back to the virtual entity for improvement. This feedback mechanism ensures continuous optimization of the simulation model and gradually improves data accuracy. Conversely, if the output similarity is greater than the preset value, it indicates that the current data processing and simulation model well reflect actual operating conditions. These verified output results are then combined with the physical dataset to construct a driving dataset based on multiple operating conditions. The physical dataset includes parameters such as speed, temperature, mass, humidity, slope, pedal opening, and SOC, which directly reflect the actual operating state of the vehicle. By integrating the virtual and physical datasets, the resulting driving dataset comprehensively reflects the operating characteristics of the electric mining truck under different operating conditions, providing comprehensive and accurate data support for subsequent energy consumption optimization and control.

[0053] Because electric mining trucks are constantly changing during operation, influenced by the working environment and road conditions, their operating conditions constantly change. To explore the impact of different operating conditions on mining truck energy consumption, it is necessary to categorize these operating conditions. This is achieved by utilizing various sensors installed on the trucks to acquire data in real time. The data collected under different operating conditions is then correlated with the operating status. The specific operating conditions are shown in Table 3.

[0054] Table 3 Electric mining truck driving conditions

[0055]

[0056] In simple terms, the fusion steps of the driving data set based on multiple working conditions can be understood as follows: the data measured by various sensors installed on the actual vehicle, including speed, temperature, mass, humidity, slope, pedal opening, SOC, a total of 7 characteristic parameters, are used as constraints and input into the Cruise simulation model and Simlink simulation model for joint simulation. The 5 characteristic parameters of torque, motor speed, power, gear, and energy consumption are used as outputs. The simulation data set is calculated using the MPC control algorithm, and the calculated data is subjected to HIL hardware-in-the-loop testing, and the test results are verified and compared. If the similarity of the output result is greater than 90%, it meets the conditions and is used as the basis for constructing the data set. If it does not meet the conditions, it returns to the joint simulation model for improvement. Under the various working conditions in Table 3, the various characteristic parameters of different working conditions are fused to map the operating status of the electric mining truck. , and construct a driving dataset based on multiple working conditions.

[0057] See also Figure 4 ,S3, calls the pre-trained PSO-BiLSTM-Attention neural network model to predict energy consumption and obtain the predicted energy consumption value;

[0058] Specifically, before calling the pre-trained PSO-BiLSTM-Attention neural network model to perform energy consumption prediction, the method further includes: obtaining energy consumption sample data, dividing the energy consumption sample data into a training set and a test set, normalizing the training set and the test set, and converting the normalized data into a cell array;

[0059] Create the function to be optimized, set the number of particles, maximum number of iterations, and particle dimensions of the PSO algorithm, and optimize the cell array based on the function to be optimized and the PSO algorithm;

[0060] Initialize the position and velocity of the particle, use the training set for training, and use the mean square error between the predicted value and the true value as the fitness function to calculate the fitness value of each particle;

[0061] Calculate the corresponding individual optimal position and global optimal position according to the fitness value, and update the position and speed of the particle according to the current position, speed, individual optimal position and global optimal position;

[0062] When it is determined that the fitness of the selected optimal solution is not less than the preset fitness error, continue to optimize according to the PSO optimization algorithm;

[0063] When it is determined that the fitness of the selected optimal solution is less than the preset fitness error, the optimal parameter combination is output, and the test set is predicted based on the optimal parameter combination to obtain the prediction result;

[0064] When the prediction result reaches a preset value, the training is terminated and a trained PSO-BiLSTM-Attention neural network model is obtained.

[0065] In this example, based on the time-series nature of electric mining truck energy consumption, a PSO-BiLSTM-Attention neural network model is proposed for energy consumption prediction. This model improves prediction accuracy by optimizing model parameters using the PSO algorithm, capturing latent information in time series data using BiLSTM, and highlighting key features through the self-attention mechanism. After obtaining a driving dataset optimized under multiple operating conditions using a co-simulation model, the prediction model uses seven characteristic parameters—speed, temperature, mass, humidity, slope, pedal opening, and SOC—as inputs, and five characteristic parameters—torque, motor speed, power, gear position, and energy consumption—as outputs.

[0066] Specifically, before using the PSO-BiLSTM-Attention neural network model for energy consumption prediction, it is necessary to first obtain energy consumption sample data. This sample data includes operating parameters of electric mining trucks under different operating conditions, such as physical data such as speed, temperature, mass, humidity, slope, pedal opening, and SOC, as well as virtual data such as torque, motor speed, power, gear position, and energy consumption. This data is divided into training and test sets for model training and validation.

[0067] To improve model training efficiency and prediction accuracy, the training and test sets are normalized. Normalization converts data of different dimensions and ranges to the same scale, preventing certain features from dominating the model training process, thereby improving model stability and accuracy. The normalized data is converted into a cell array to facilitate processing by the neural network model. Next, the function to be optimized is created, and the relevant parameters of the PSO (Particle Swarm Optimization) algorithm are set, including the number of particles, maximum number of iterations, and particle dimension. The PSO algorithm is an optimization algorithm based on swarm intelligence that simulates the foraging behavior of bird flocks to find the optimal solution. In this recipe, the PSO algorithm is used to optimize the parameters of the neural network model to improve its predictive performance.

[0068] After initializing the particle's position and velocity, the neural network model is trained using the training set. During training, the mean squared error between the predicted and true values ​​is used as the fitness function to calculate the fitness value of each particle. The fitness function reflects the difference between the model's predictions and the actual data. The smaller the fitness value, the higher the model's prediction accuracy. Based on the calculated fitness value, the individual optimal position and global optimal position of each particle are determined. The individual optimal position refers to the optimal solution found by the particle during the search process, while the global optimal position refers to the optimal solution found by all particles in the entire particle swarm. The particle's position and velocity are updated based on the current position, velocity, individual optimal position, and global optimal position. This process is iterated continuously until the optimal solution that meets the requirements is found.

[0069] When the fitness of the selected optimal solution is no less than the preset fitness error, the PSO optimization algorithm continues to optimize. This process ensures continuous optimization of model parameters and gradually improves the model's prediction accuracy. When the fitness of the optimal solution is less than the preset fitness error, the optimal parameter combination is output and used to predict the test set. If the prediction result reaches the preset value, the model has achieved high prediction accuracy, and training is terminated, resulting in a trained PSO-BiLSTM-Attention neural network model. This model can accurately predict the vehicle state with optimal energy consumption based on the input vehicle operating state parameters, providing precise data support for subsequent energy consumption optimization control.

[0070] Optimizing the parameters of the neural network model using the PSO algorithm significantly improves the model's prediction accuracy. The BiLSTM (bidirectional long short-term memory) network captures latent information in time series data, while the attention mechanism highlights key features, further enhancing model performance. This optimization and prediction method not only accurately predicts the energy consumption of electric mining trucks but also adjusts predictions in real time based on actual operating conditions, providing powerful technical support for energy consumption optimization and control. Furthermore, normalization and cell array conversion improve data processing efficiency and model training stability, ensuring efficient operation of the entire system.

[0071] See also Figure 5 , S4, compare the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and adjust the state parameters of the electric mining truck according to the energy consumption deviation to ensure that the energy consumption of the electric mining truck remains within a preset optimal range.

[0072] Specifically, step S4 includes: comparing the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and determining whether the energy consumption deviation is less than a preset deviation value;

[0073] If so, adjusting the vehicle state parameters in an end-to-end manner according to the predicted energy consumption value;

[0074] If not, a voice reminder will be given and the system will switch to auxiliary control mode based on dynamic programming to adjust the gear position, transmission ratio and motor speed.

[0075] In this embodiment, after completing energy consumption prediction using the PSO-BiLSTM-Attention neural network model, a key step is to compare the predicted energy consumption with the actual energy consumption of the electric mining truck. Based on the comparison results, the vehicle's state parameters are adjusted to ensure that the truck's energy consumption remains within a preset optimal range. This process is not a critical part of the overall method. It not only involves accurate energy consumption comparison and judgment, but also achieves real-time energy consumption optimization and adjustment through intelligent control mode switching. The energy consumption value predicted by the neural network model is compared with the energy consumption data of the electric mining truck during actual operation. This comparison process can intuitively reflect the difference between the model prediction and actual operation, thereby determining the energy consumption deviation. Energy consumption deviation is an important indicator for measuring the degree of match between prediction accuracy and actual operating status. By setting a preset deviation value (for example, 10%), it can be determined whether the current energy consumption control is in an ideal state. When the energy consumption deviation is less than the preset deviation value, it indicates that the predicted energy consumption value is close to the actual energy consumption value, and the vehicle's energy consumption control is considered to be in a relatively ideal state. In this case, the vehicle's state parameters are adjusted in an end-to-end manner based on the predicted energy consumption value. End-to-end regulation is a direct and efficient control strategy that rapidly adjusts key vehicle operating parameters, such as speed, torque, and motor speed, based on predictions, to maintain optimal energy consumption. This approach not only offers a fast response but also leverages the high precision of the prediction model to achieve refined energy control.

[0076] On the contrary, when the energy consumption deviation is greater than or equal to the preset deviation value, it means that there is a large difference between the predicted energy consumption value and the actual energy consumption, and additional control measures need to be taken. In this case, the system will issue a voice reminder to inform the operator or the autonomous driving system that there is a deviation in the current energy consumption control, and automatically switch to an auxiliary control mode based on the dynamic programming method. The dynamic programming method is a classic optimization algorithm that can dynamically adjust parameters such as gear position, transmission ratio, and motor speed according to the current vehicle status and operating conditions to achieve optimal control of energy consumption. This auxiliary control mode can provide a reliable backup plan when there is a large deviation in the prediction model to ensure that the vehicle's energy consumption does not deviate too far from the optimal range.

[0077] Simply put, step S4 inputs the mining truck's driving characteristic parameters, including speed, temperature, mass, humidity, slope, pedal opening, and SOC, into the neural network prediction model to predict the vehicle's state parameters for optimal energy consumption, including torque, motor speed, power, gear position, and energy consumption. The predicted energy consumption is then compared with the actual energy consumption of the actual vehicle. If the energy consumption deviation is within 10%, the primary control method is adopted to restore the mining truck's energy consumption to the optimal state. Otherwise, indicating that the gap between actual energy consumption and ideal energy consumption is too large, the auxiliary control method is switched to, and voice prompts are given to adjust parameters such as gear position, transmission ratio, and motor speed to keep the motor in the high-efficiency operating range, thereby maintaining energy consumption within the optimal range.

[0078] In summary, the optimal energy consumption control method for autonomous electric mining trucks uses multiple sensors to collect real-time vehicle operating status data, including physical data such as speed, temperature, mass, humidity, slope, pedal opening, and SOC. This data is combined with virtual data generated by the Cruise vehicle simulation model and the Simulink simulation model to construct a multi-condition driving dataset. This process not only captures the vehicle's actual operating parameters but also supplements theoretical operating conditions through simulation models, providing comprehensive data support for energy consumption optimization.

[0079] During the data processing phase, an MPC control algorithm was used to calculate the virtual dataset, and the data accuracy was verified through HIL hardware-in-the-loop testing. This verification process ensured the reliability and practicality of the dataset, providing a high-quality data foundation for subsequent energy consumption forecasting. By comparing the verification results with preset values, the system automatically provided feedback and improved the simulation model, further optimizing the dataset construction.

[0080] During the energy consumption prediction phase, a pre-trained PSO-BiLSTM-Attention neural network model is used. This model optimizes model parameters using the particle swarm optimization (PSO) algorithm, utilizes a bidirectional long short-term memory (BiLSTM) network to capture the potential information in time series data, and incorporates an attention mechanism to highlight key features, thereby achieving highly accurate energy consumption predictions. This process not only improves prediction accuracy but also ensures model stability and adaptability through optimization algorithms.

[0081] During the energy consumption control phase, the predicted energy consumption is compared with the actual energy consumption, and the vehicle's state parameters are adjusted based on the energy consumption deviation. When the energy consumption deviation is less than the preset value, the vehicle's state parameters are rapidly adjusted in an end-to-end manner to ensure that energy consumption remains within the optimal range. If the deviation is larger, the system will issue a voice prompt and switch to auxiliary control mode based on dynamic programming, dynamically adjusting parameters such as gear position, transmission ratio, and motor speed to further optimize energy consumption control. This dual control strategy not only improves the accuracy of energy consumption control, but also enhances the robustness and reliability of the system.

[0082] Compared with existing technologies, this optimal energy consumption control method for autonomous electric mining trucks builds a comprehensive and accurate driving dataset by integrating physical and virtual data, providing a solid data foundation for energy consumption optimization. Furthermore, by utilizing advanced neural network models and optimization algorithms, high-precision energy consumption prediction and real-time control are achieved. Furthermore, through intelligent control mode switching, optimal energy consumption can be maintained under different operating conditions, significantly improving the operating efficiency and economic efficiency of electric mining trucks. This innovative energy consumption control method is not only applicable to autonomous electric mining trucks but can also be extended to other autonomous electric vehicle fields, possessing broad application prospects and significant practical significance.

[0083] See also Figure 6 The second embodiment of the present invention provides an energy consumption optimization control device for an autonomous driving electric mining truck, which includes:

[0084] The simulation unit 201 is configured to obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and to establish a virtual entity corresponding to the electric mining truck based on the physical data set;

[0085] A fusion unit 202 is configured to obtain a virtual dataset of a virtual entity, and fuse the virtual dataset with a physical dataset to obtain a driving dataset based on multiple working conditions;

[0086] The prediction unit 203 is used to call the pre-trained PSO-BiLSTM-Attention neural network model to perform energy consumption prediction and obtain a predicted energy consumption value;

[0087] The adjustment unit 204 is configured to compare the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and adjust the state parameters of the electric mining truck according to the energy consumption deviation to ensure that the energy consumption of the electric mining truck remains within a preset optimal range.

[0088] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for optimal energy consumption control of an autonomous electric mining truck, characterized in that: include: Acquiring a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and establishing a virtual entity corresponding to the electric mining truck based on the physical data set; Acquiring a virtual data set of a virtual entity, fusing the virtual data set with a physical data set to obtain a driving data set based on multiple working conditions; Call the pre-trained PSO-BiLSTM-Attention neural network model to predict energy consumption and obtain the predicted energy consumption value; Comparing the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and adjusting the state parameters of the electric mining truck according to the energy consumption deviation to ensure that the energy consumption of the electric mining truck remains within a preset optimal range; Obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and establish a virtual entity corresponding to the electric mining truck based on the physical data set, specifically: Obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and construct and process the physical data set using AVL Cruise software to obtain a complete vehicle model of the electric mining truck; Based on the physical data set, the dynamic equations and control system of the electric mining truck are established using Simulink software, and the dynamic equations and control system are combined with the electric mining truck whole vehicle model to obtain the electric mining truck whole vehicle simulation model; Before calling the pre-trained PSO-BiLSTM-Attention neural network model for energy consumption prediction, it also includes: Acquiring energy consumption sample data, dividing the energy consumption sample data into a training set and a test set, normalizing the training set and the test set, and converting the normalized data into a cell array; Create the function to be optimized, set the number of particles, maximum number of iterations, and particle dimensions of the PSO algorithm, and optimize the cell array based on the function to be optimized and the PSO algorithm; Initialize the position and velocity of the particle, use the training set for training, and use the mean square error between the predicted value and the true value as the fitness function to calculate the fitness value of each particle; Calculate the corresponding individual optimal position and global optimal position according to the fitness value, and update the position and speed of the particle according to the current position, speed, individual optimal position and global optimal position; When it is determined that the fitness of the selected optimal solution is not less than the preset fitness error, continue to optimize according to the PSO optimization algorithm; When it is determined that the fitness of the selected optimal solution is less than the preset fitness error, the optimal parameter combination is output, and the test set is predicted based on the optimal parameter combination to obtain the prediction result; When the prediction result reaches a preset value, the training is terminated and a trained PSO-BiLSTM-Attention neural network model is obtained; The predicted energy consumption value is compared with the actual energy consumption to obtain the energy consumption deviation, and the state parameters of the electric mining truck are adjusted according to the energy consumption deviation, specifically: Comparing the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and determining whether the energy consumption deviation is less than a preset deviation value; If so, adjusting the vehicle state parameters in an end-to-end manner according to the predicted energy consumption value; If not, a voice reminder will be given and the system will switch to auxiliary control mode based on dynamic programming to adjust the gear position, transmission ratio and motor speed.

2. The energy consumption optimization control method for an autonomous driving electric mining truck according to claim 1 is characterized in that: The sensor assembly includes: a speed sensor, a SOC sensor, a tilt sensor, a mass sensor, a humidity sensor, a temperature sensor and a pedal sensor.

3. The energy consumption optimization control method for autonomous driving electric mining trucks according to claim 1 is characterized in that: The physical data set includes speed, temperature, mass, humidity, slope, pedal opening, and SOC, and the virtual data set includes torque, motor speed, power, gear, and energy consumption.

4. The energy consumption optimization control method for an autonomous driving electric mining truck according to claim 1 is characterized in that: Obtain a virtual dataset of a virtual entity, fuse the virtual dataset with the physical dataset, and obtain a driving dataset based on multiple working conditions, specifically: The virtual data set is calculated using an MPC control algorithm, and the calculated data set is tested in the loop using HIL hardware to obtain an output result; Performing verification and comparison processing on the output result to determine whether the output result is greater than a preset value; If not, the output is fed back to the virtual entity for improvement; If so, the output results are combined with the physical data set to construct a driving data set based on multiple working conditions.

5. An energy consumption optimization control device for an autonomous driving electric mining truck, characterized in that: include: a simulation unit, configured to obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and to establish a virtual entity corresponding to the electric mining truck based on the physical data set; a fusion unit, configured to obtain a virtual data set of a virtual entity, and fuse the virtual data set with a physical data set to obtain a driving data set based on multiple working conditions; The prediction unit is used to call the pre-trained PSO-BiLSTM-Attention neural network model to predict energy consumption and obtain the predicted energy consumption value; an adjustment unit, configured to compare the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and adjust the state parameters of the electric mining truck according to the energy consumption deviation to ensure that the energy consumption of the electric mining truck remains within a preset optimal range; Obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and establish a virtual entity corresponding to the electric mining truck based on the physical data set, specifically: Obtain a physical data set collected by a sensor assembly configured on the electric mining truck to be tested, and construct and process the physical data set using AVL Cruise software to obtain a complete vehicle model of the electric mining truck; Based on the physical data set, the dynamic equations and control system of the electric mining truck are established using Simulink software, and the dynamic equations and control system are combined with the electric mining truck whole vehicle model to obtain the electric mining truck whole vehicle simulation model; Before calling the pre-trained PSO-BiLSTM-Attention neural network model for energy consumption prediction, it also includes: Acquiring energy consumption sample data, dividing the energy consumption sample data into a training set and a test set, normalizing the training set and the test set, and converting the normalized data into a cell array; Create the function to be optimized, set the number of particles, maximum number of iterations, and particle dimensions of the PSO algorithm, and optimize the cell array based on the function to be optimized and the PSO algorithm; Initialize the position and velocity of the particle, use the training set for training, and use the mean square error between the predicted value and the true value as the fitness function to calculate the fitness value of each particle; Calculate the corresponding individual optimal position and global optimal position according to the fitness value, and update the position and speed of the particle according to the current position, speed, individual optimal position and global optimal position; When it is determined that the fitness of the selected optimal solution is not less than the preset fitness error, continue to optimize according to the PSO optimization algorithm; When it is determined that the fitness of the selected optimal solution is less than the preset fitness error, the optimal parameter combination is output, and the test set is predicted based on the optimal parameter combination to obtain the prediction result; When the prediction result reaches a preset value, the training is terminated and a trained PSO-BiLSTM-Attention neural network model is obtained; The predicted energy consumption value is compared with the actual energy consumption to obtain the energy consumption deviation, and the state parameters of the electric mining truck are adjusted according to the energy consumption deviation, specifically: Comparing the predicted energy consumption value with the actual energy consumption to obtain an energy consumption deviation, and determining whether the energy consumption deviation is less than a preset deviation value; If so, adjusting the vehicle state parameters in an end-to-end manner according to the predicted energy consumption value; If not, a voice reminder will be given and the system will switch to auxiliary control mode based on dynamic programming to adjust the gear position, transmission ratio and motor speed.

Images